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 <[email protected]>
  10  *  Andy Adamson   <[email protected]>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57 #include <linux/iversion.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_BITMASK_SZ         3
  75
  76 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  77 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  78
  79 /* file attributes which can be mapped to nfs attributes */
  80 #define NFS4_VALID_ATTRS (ATTR_MODE \
  81         | ATTR_UID \
  82         | ATTR_GID \
  83         | ATTR_SIZE \
  84         | ATTR_ATIME \
  85         | ATTR_MTIME \
  86         | ATTR_CTIME \
  87         | ATTR_ATIME_SET \
  88         | ATTR_MTIME_SET)
  89
  90 struct nfs4_opendata;
  91 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  92 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  93 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  94 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label, struct inode *inode);
  95 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  96 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  97                             struct nfs_fattr *fattr, struct iattr *sattr,
  98                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  99                             struct nfs4_label *olabel);
 100 #ifdef CONFIG_NFS_V4_1
 101 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 102                 const struct cred *cred,
 103                 struct nfs4_slot *slot,
 104                 bool is_privileged);
 105 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 106                 const struct cred *);
 107 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 108                 const struct cred *, bool);
 109 #endif
 110
 111 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 112 static inline struct nfs4_label *
 113 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 114         struct iattr *sattr, struct nfs4_label *label)
 115 {
 116         int err;
 117
 118         if (label == NULL)
 119                 return NULL;
 120
 121         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 122                 return NULL;
 123
 124         err = security_dentry_init_security(dentry, sattr->ia_mode,
 125                                 &dentry->d_name, (void **)&label->label, &label->len);
 126         if (err == 0)
 127                 return label;
 128
 129         return NULL;
 130 }
 131 static inline void
 132 nfs4_label_release_security(struct nfs4_label *label)
 133 {
 134         if (label)
 135                 security_release_secctx(label->label, label->len);
 136 }
 137 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 138 {
 139         if (label)
 140                 return server->attr_bitmask;
 141
 142         return server->attr_bitmask_nl;
 143 }
 144 #else
 145 static inline struct nfs4_label *
 146 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 147         struct iattr *sattr, struct nfs4_label *l)
 148 { return NULL; }
 149 static inline void
 150 nfs4_label_release_security(struct nfs4_label *label)
 151 { return; }
 152 static inline u32 *
 153 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 154 { return server->attr_bitmask; }
 155 #endif
 156
 157 /* Prevent leaks of NFSv4 errors into userland */
 158 static int nfs4_map_errors(int err)
 159 {
 160         if (err >= -1000)
 161                 return err;
 162         switch (err) {
 163         case -NFS4ERR_RESOURCE:
 164         case -NFS4ERR_LAYOUTTRYLATER:
 165         case -NFS4ERR_RECALLCONFLICT:
 166                 return -EREMOTEIO;
 167         case -NFS4ERR_WRONGSEC:
 168         case -NFS4ERR_WRONG_CRED:
 169                 return -EPERM;
 170         case -NFS4ERR_BADOWNER:
 171         case -NFS4ERR_BADNAME:
 172                 return -EINVAL;
 173         case -NFS4ERR_SHARE_DENIED:
 174                 return -EACCES;
 175         case -NFS4ERR_MINOR_VERS_MISMATCH:
 176                 return -EPROTONOSUPPORT;
 177         case -NFS4ERR_FILE_OPEN:
 178                 return -EBUSY;
 179         default:
 180                 dprintk("%s could not handle NFSv4 error %d\n",
 181                                 __func__, -err);
 182                 break;
 183         }
 184         return -EIO;
 185 }
 186
 187 /*
 188  * This is our standard bitmap for GETATTR requests.
 189  */
 190 const u32 nfs4_fattr_bitmap[3] = {
 191         FATTR4_WORD0_TYPE
 192         | FATTR4_WORD0_CHANGE
 193         | FATTR4_WORD0_SIZE
 194         | FATTR4_WORD0_FSID
 195         | FATTR4_WORD0_FILEID,
 196         FATTR4_WORD1_MODE
 197         | FATTR4_WORD1_NUMLINKS
 198         | FATTR4_WORD1_OWNER
 199         | FATTR4_WORD1_OWNER_GROUP
 200         | FATTR4_WORD1_RAWDEV
 201         | FATTR4_WORD1_SPACE_USED
 202         | FATTR4_WORD1_TIME_ACCESS
 203         | FATTR4_WORD1_TIME_METADATA
 204         | FATTR4_WORD1_TIME_MODIFY
 205         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 206 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 207         FATTR4_WORD2_SECURITY_LABEL
 208 #endif
 209 };
 210
 211 static const u32 nfs4_pnfs_open_bitmap[3] = {
 212         FATTR4_WORD0_TYPE
 213         | FATTR4_WORD0_CHANGE
 214         | FATTR4_WORD0_SIZE
 215         | FATTR4_WORD0_FSID
 216         | FATTR4_WORD0_FILEID,
 217         FATTR4_WORD1_MODE
 218         | FATTR4_WORD1_NUMLINKS
 219         | FATTR4_WORD1_OWNER
 220         | FATTR4_WORD1_OWNER_GROUP
 221         | FATTR4_WORD1_RAWDEV
 222         | FATTR4_WORD1_SPACE_USED
 223         | FATTR4_WORD1_TIME_ACCESS
 224         | FATTR4_WORD1_TIME_METADATA
 225         | FATTR4_WORD1_TIME_MODIFY,
 226         FATTR4_WORD2_MDSTHRESHOLD
 227 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 228         | FATTR4_WORD2_SECURITY_LABEL
 229 #endif
 230 };
 231
 232 static const u32 nfs4_open_noattr_bitmap[3] = {
 233         FATTR4_WORD0_TYPE
 234         | FATTR4_WORD0_FILEID,
 235 };
 236
 237 const u32 nfs4_statfs_bitmap[3] = {
 238         FATTR4_WORD0_FILES_AVAIL
 239         | FATTR4_WORD0_FILES_FREE
 240         | FATTR4_WORD0_FILES_TOTAL,
 241         FATTR4_WORD1_SPACE_AVAIL
 242         | FATTR4_WORD1_SPACE_FREE
 243         | FATTR4_WORD1_SPACE_TOTAL
 244 };
 245
 246 const u32 nfs4_pathconf_bitmap[3] = {
 247         FATTR4_WORD0_MAXLINK
 248         | FATTR4_WORD0_MAXNAME,
 249         0
 250 };
 251
 252 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 253                         | FATTR4_WORD0_MAXREAD
 254                         | FATTR4_WORD0_MAXWRITE
 255                         | FATTR4_WORD0_LEASE_TIME,
 256                         FATTR4_WORD1_TIME_DELTA
 257                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 258                         FATTR4_WORD2_LAYOUT_BLKSIZE
 259                         | FATTR4_WORD2_CLONE_BLKSIZE
 260 };
 261
 262 const u32 nfs4_fs_locations_bitmap[3] = {
 263         FATTR4_WORD0_CHANGE
 264         | FATTR4_WORD0_SIZE
 265         | FATTR4_WORD0_FSID
 266         | FATTR4_WORD0_FILEID
 267         | FATTR4_WORD0_FS_LOCATIONS,
 268         FATTR4_WORD1_OWNER
 269         | FATTR4_WORD1_OWNER_GROUP
 270         | FATTR4_WORD1_RAWDEV
 271         | FATTR4_WORD1_SPACE_USED
 272         | FATTR4_WORD1_TIME_ACCESS
 273         | FATTR4_WORD1_TIME_METADATA
 274         | FATTR4_WORD1_TIME_MODIFY
 275         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 276 };
 277
 278 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 279                 struct inode *inode)
 280 {
 281         unsigned long cache_validity;
 282
 283         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 284         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 285                 return;
 286
 287         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 288         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 289                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 290                                 | NFS_INO_INVALID_SIZE);
 291
 292         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 293                 dst[0] &= ~FATTR4_WORD0_SIZE;
 294
 295         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 296                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 297 }
 298
 299 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 300                 const __u32 *src, struct inode *inode)
 301 {
 302         nfs4_bitmap_copy_adjust(dst, src, inode);
 303 }
 304
 305 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 306                 struct nfs4_readdir_arg *readdir)
 307 {
 308         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 309         __be32 *start, *p;
 310
 311         if (cookie > 2) {
 312                 readdir->cookie = cookie;
 313                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 314                 return;
 315         }
 316
 317         readdir->cookie = 0;
 318         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 319         if (cookie == 2)
 320                 return;
 321
 322         /*
 323          * NFSv4 servers do not return entries for '.' and '..'
 324          * Therefore, we fake these entries here.  We let '.'
 325          * have cookie 0 and '..' have cookie 1.  Note that
 326          * when talking to the server, we always send cookie 0
 327          * instead of 1 or 2.
 328          */
 329         start = p = kmap_atomic(*readdir->pages);
 330
 331         if (cookie == 0) {
 332                 *p++ = xdr_one;                                  /* next */
 333                 *p++ = xdr_zero;                   /* cookie, first word */
 334                 *p++ = xdr_one;                   /* cookie, second word */
 335                 *p++ = xdr_one;                             /* entry len */
 336                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 337                 p++;
 338                 *p++ = xdr_one;                         /* bitmap length */
 339                 *p++ = htonl(attrs);                           /* bitmap */
 340                 *p++ = htonl(12);             /* attribute buffer length */
 341                 *p++ = htonl(NF4DIR);
 342                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 343         }
 344
 345         *p++ = xdr_one;                                  /* next */
 346         *p++ = xdr_zero;                   /* cookie, first word */
 347         *p++ = xdr_two;                   /* cookie, second word */
 348         *p++ = xdr_two;                             /* entry len */
 349         memcpy(p, "..\0\0", 4);                         /* entry */
 350         p++;
 351         *p++ = xdr_one;                         /* bitmap length */
 352         *p++ = htonl(attrs);                           /* bitmap */
 353         *p++ = htonl(12);             /* attribute buffer length */
 354         *p++ = htonl(NF4DIR);
 355         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 356
 357         readdir->pgbase = (char *)p - (char *)start;
 358         readdir->count -= readdir->pgbase;
 359         kunmap_atomic(start);
 360 }
 361
 362 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 363                 nfs4_stateid *stateid,
 364                 const struct cred *cred)
 365 {
 366         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 367
 368         ops->test_and_free_expired(server, stateid, cred);
 369 }
 370
 371 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 372                 nfs4_stateid *stateid,
 373                 const struct cred *cred)
 374 {
 375         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 376         nfs4_test_and_free_stateid(server, stateid, cred);
 377 }
 378
 379 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 380                 const nfs4_stateid *stateid,
 381                 const struct cred *cred)
 382 {
 383         nfs4_stateid tmp;
 384
 385         nfs4_stateid_copy(&tmp, stateid);
 386         __nfs4_free_revoked_stateid(server, &tmp, cred);
 387 }
 388
 389 static long nfs4_update_delay(long *timeout)
 390 {
 391         long ret;
 392         if (!timeout)
 393                 return NFS4_POLL_RETRY_MAX;
 394         if (*timeout <= 0)
 395                 *timeout = NFS4_POLL_RETRY_MIN;
 396         if (*timeout > NFS4_POLL_RETRY_MAX)
 397                 *timeout = NFS4_POLL_RETRY_MAX;
 398         ret = *timeout;
 399         *timeout <<= 1;
 400         return ret;
 401 }
 402
 403 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 404 {
 405         int res = 0;
 406
 407         might_sleep();
 408
 409         freezable_schedule_timeout_killable_unsafe(
 410                 nfs4_update_delay(timeout));
 411         if (fatal_signal_pending(current))
 412                 res = -ERESTARTSYS;
 413         return res;
 414 }
 415
 416 /* This is the error handling routine for processes that are allowed
 417  * to sleep.
 418  */
 419 static int nfs4_do_handle_exception(struct nfs_server *server,
 420                 int errorcode, struct nfs4_exception *exception)
 421 {
 422         struct nfs_client *clp = server->nfs_client;
 423         struct nfs4_state *state = exception->state;
 424         const nfs4_stateid *stateid = exception->stateid;
 425         struct inode *inode = exception->inode;
 426         int ret = errorcode;
 427
 428         exception->delay = 0;
 429         exception->recovering = 0;
 430         exception->retry = 0;
 431
 432         if (stateid == NULL && state != NULL)
 433                 stateid = &state->stateid;
 434
 435         switch(errorcode) {
 436                 case 0:
 437                         return 0;
 438                 case -NFS4ERR_BADHANDLE:
 439                 case -ESTALE:
 440                         if (inode != NULL && S_ISREG(inode->i_mode))
 441                                 pnfs_destroy_layout(NFS_I(inode));
 442                         break;
 443                 case -NFS4ERR_DELEG_REVOKED:
 444                 case -NFS4ERR_ADMIN_REVOKED:
 445                 case -NFS4ERR_EXPIRED:
 446                 case -NFS4ERR_BAD_STATEID:
 447                         if (inode != NULL && stateid != NULL) {
 448                                 nfs_inode_find_state_and_recover(inode,
 449                                                 stateid);
 450                                 goto wait_on_recovery;
 451                         }
 452                         /* Fall through */
 453                 case -NFS4ERR_OPENMODE:
 454                         if (inode) {
 455                                 int err;
 456
 457                                 err = nfs_async_inode_return_delegation(inode,
 458                                                 stateid);
 459                                 if (err == 0)
 460                                         goto wait_on_recovery;
 461                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 462                                         exception->retry = 1;
 463                                         break;
 464                                 }
 465                         }
 466                         if (state == NULL)
 467                                 break;
 468                         ret = nfs4_schedule_stateid_recovery(server, state);
 469                         if (ret < 0)
 470                                 break;
 471                         goto wait_on_recovery;
 472                 case -NFS4ERR_STALE_STATEID:
 473                 case -NFS4ERR_STALE_CLIENTID:
 474                         nfs4_schedule_lease_recovery(clp);
 475                         goto wait_on_recovery;
 476                 case -NFS4ERR_MOVED:
 477                         ret = nfs4_schedule_migration_recovery(server);
 478                         if (ret < 0)
 479                                 break;
 480                         goto wait_on_recovery;
 481                 case -NFS4ERR_LEASE_MOVED:
 482                         nfs4_schedule_lease_moved_recovery(clp);
 483                         goto wait_on_recovery;
 484 #if defined(CONFIG_NFS_V4_1)
 485                 case -NFS4ERR_BADSESSION:
 486                 case -NFS4ERR_BADSLOT:
 487                 case -NFS4ERR_BAD_HIGH_SLOT:
 488                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 489                 case -NFS4ERR_DEADSESSION:
 490                 case -NFS4ERR_SEQ_FALSE_RETRY:
 491                 case -NFS4ERR_SEQ_MISORDERED:
 492                         dprintk("%s ERROR: %d Reset session\n", __func__,
 493                                 errorcode);
 494                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 495                         goto wait_on_recovery;
 496 #endif /* defined(CONFIG_NFS_V4_1) */
 497                 case -NFS4ERR_FILE_OPEN:
 498                         if (exception->timeout > HZ) {
 499                                 /* We have retried a decent amount, time to
 500                                  * fail
 501                                  */
 502                                 ret = -EBUSY;
 503                                 break;
 504                         }
 505                         /* Fall through */
 506                 case -NFS4ERR_DELAY:
 507                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 508                         /* Fall through */
 509                 case -NFS4ERR_GRACE:
 510                 case -NFS4ERR_LAYOUTTRYLATER:
 511                 case -NFS4ERR_RECALLCONFLICT:
 512                         exception->delay = 1;
 513                         return 0;
 514
 515                 case -NFS4ERR_RETRY_UNCACHED_REP:
 516                 case -NFS4ERR_OLD_STATEID:
 517                         exception->retry = 1;
 518                         break;
 519                 case -NFS4ERR_BADOWNER:
 520                         /* The following works around a Linux server bug! */
 521                 case -NFS4ERR_BADNAME:
 522                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 523                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 524                                 exception->retry = 1;
 525                                 printk(KERN_WARNING "NFS: v4 server %s "
 526                                                 "does not accept raw "
 527                                                 "uid/gids. "
 528                                                 "Reenabling the idmapper.\n",
 529                                                 server->nfs_client->cl_hostname);
 530                         }
 531         }
 532         /* We failed to handle the error */
 533         return nfs4_map_errors(ret);
 534 wait_on_recovery:
 535         exception->recovering = 1;
 536         return 0;
 537 }
 538
 539 /* This is the error handling routine for processes that are allowed
 540  * to sleep.
 541  */
 542 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 543 {
 544         struct nfs_client *clp = server->nfs_client;
 545         int ret;
 546
 547         ret = nfs4_do_handle_exception(server, errorcode, exception);
 548         if (exception->delay) {
 549                 ret = nfs4_delay(server->client, &exception->timeout);
 550                 goto out_retry;
 551         }
 552         if (exception->recovering) {
 553                 ret = nfs4_wait_clnt_recover(clp);
 554                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 555                         return -EIO;
 556                 goto out_retry;
 557         }
 558         return ret;
 559 out_retry:
 560         if (ret == 0)
 561                 exception->retry = 1;
 562         return ret;
 563 }
 564
 565 static int
 566 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 567                 int errorcode, struct nfs4_exception *exception)
 568 {
 569         struct nfs_client *clp = server->nfs_client;
 570         int ret;
 571
 572         ret = nfs4_do_handle_exception(server, errorcode, exception);
 573         if (exception->delay) {
 574                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 575                 goto out_retry;
 576         }
 577         if (exception->recovering) {
 578                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 579                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 580                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 581                 goto out_retry;
 582         }
 583         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 584                 ret = -EIO;
 585         return ret;
 586 out_retry:
 587         if (ret == 0) {
 588                 exception->retry = 1;
 589                 /*
 590                  * For NFS4ERR_MOVED, the client transport will need to
 591                  * be recomputed after migration recovery has completed.
 592                  */
 593                 if (errorcode == -NFS4ERR_MOVED)
 594                         rpc_task_release_transport(task);
 595         }
 596         return ret;
 597 }
 598
 599 int
 600 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 601                         struct nfs4_state *state, long *timeout)
 602 {
 603         struct nfs4_exception exception = {
 604                 .state = state,
 605         };
 606
 607         if (task->tk_status >= 0)
 608                 return 0;
 609         if (timeout)
 610                 exception.timeout = *timeout;
 611         task->tk_status = nfs4_async_handle_exception(task, server,
 612                         task->tk_status,
 613                         &exception);
 614         if (exception.delay && timeout)
 615                 *timeout = exception.timeout;
 616         if (exception.retry)
 617                 return -EAGAIN;
 618         return 0;
 619 }
 620
 621 /*
 622  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 623  * or 'false' otherwise.
 624  */
 625 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 626 {
 627         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 628         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 629 }
 630
 631 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 632 {
 633         spin_lock(&clp->cl_lock);
 634         if (time_before(clp->cl_last_renewal,timestamp))
 635                 clp->cl_last_renewal = timestamp;
 636         spin_unlock(&clp->cl_lock);
 637 }
 638
 639 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 640 {
 641         struct nfs_client *clp = server->nfs_client;
 642
 643         if (!nfs4_has_session(clp))
 644                 do_renew_lease(clp, timestamp);
 645 }
 646
 647 struct nfs4_call_sync_data {
 648         const struct nfs_server *seq_server;
 649         struct nfs4_sequence_args *seq_args;
 650         struct nfs4_sequence_res *seq_res;
 651 };
 652
 653 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 654                         struct nfs4_sequence_res *res, int cache_reply,
 655                         int privileged)
 656 {
 657         args->sa_slot = NULL;
 658         args->sa_cache_this = cache_reply;
 659         args->sa_privileged = privileged;
 660
 661         res->sr_slot = NULL;
 662 }
 663
 664 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 665 {
 666         struct nfs4_slot *slot = res->sr_slot;
 667         struct nfs4_slot_table *tbl;
 668
 669         tbl = slot->table;
 670         spin_lock(&tbl->slot_tbl_lock);
 671         if (!nfs41_wake_and_assign_slot(tbl, slot))
 672                 nfs4_free_slot(tbl, slot);
 673         spin_unlock(&tbl->slot_tbl_lock);
 674
 675         res->sr_slot = NULL;
 676 }
 677
 678 static int nfs40_sequence_done(struct rpc_task *task,
 679                                struct nfs4_sequence_res *res)
 680 {
 681         if (res->sr_slot != NULL)
 682                 nfs40_sequence_free_slot(res);
 683         return 1;
 684 }
 685
 686 #if defined(CONFIG_NFS_V4_1)
 687
 688 static void nfs41_release_slot(struct nfs4_slot *slot)
 689 {
 690         struct nfs4_session *session;
 691         struct nfs4_slot_table *tbl;
 692         bool send_new_highest_used_slotid = false;
 693
 694         if (!slot)
 695                 return;
 696         tbl = slot->table;
 697         session = tbl->session;
 698
 699         /* Bump the slot sequence number */
 700         if (slot->seq_done)
 701                 slot->seq_nr++;
 702         slot->seq_done = 0;
 703
 704         spin_lock(&tbl->slot_tbl_lock);
 705         /* Be nice to the server: try to ensure that the last transmitted
 706          * value for highest_user_slotid <= target_highest_slotid
 707          */
 708         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 709                 send_new_highest_used_slotid = true;
 710
 711         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 712                 send_new_highest_used_slotid = false;
 713                 goto out_unlock;
 714         }
 715         nfs4_free_slot(tbl, slot);
 716
 717         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 718                 send_new_highest_used_slotid = false;
 719 out_unlock:
 720         spin_unlock(&tbl->slot_tbl_lock);
 721         if (send_new_highest_used_slotid)
 722                 nfs41_notify_server(session->clp);
 723         if (waitqueue_active(&tbl->slot_waitq))
 724                 wake_up_all(&tbl->slot_waitq);
 725 }
 726
 727 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 728 {
 729         nfs41_release_slot(res->sr_slot);
 730         res->sr_slot = NULL;
 731 }
 732
 733 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 734                 u32 seqnr)
 735 {
 736         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 737                 slot->seq_nr_highest_sent = seqnr;
 738 }
 739 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 740                 u32 seqnr)
 741 {
 742         slot->seq_nr_highest_sent = seqnr;
 743         slot->seq_nr_last_acked = seqnr;
 744 }
 745
 746 static int nfs41_sequence_process(struct rpc_task *task,
 747                 struct nfs4_sequence_res *res)
 748 {
 749         struct nfs4_session *session;
 750         struct nfs4_slot *slot = res->sr_slot;
 751         struct nfs_client *clp;
 752         int ret = 1;
 753
 754         if (slot == NULL)
 755                 goto out_noaction;
 756         /* don't increment the sequence number if the task wasn't sent */
 757         if (!RPC_WAS_SENT(task) || slot->seq_done)
 758                 goto out;
 759
 760         session = slot->table->session;
 761
 762         trace_nfs4_sequence_done(session, res);
 763         /* Check the SEQUENCE operation status */
 764         switch (res->sr_status) {
 765         case 0:
 766                 /* Mark this sequence number as having been acked */
 767                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 768                 /* Update the slot's sequence and clientid lease timer */
 769                 slot->seq_done = 1;
 770                 clp = session->clp;
 771                 do_renew_lease(clp, res->sr_timestamp);
 772                 /* Check sequence flags */
 773                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 774                                 !!slot->privileged);
 775                 nfs41_update_target_slotid(slot->table, slot, res);
 776                 break;
 777         case 1:
 778                 /*
 779                  * sr_status remains 1 if an RPC level error occurred.
 780                  * The server may or may not have processed the sequence
 781                  * operation..
 782                  */
 783                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 784                 slot->seq_done = 1;
 785                 goto out;
 786         case -NFS4ERR_DELAY:
 787                 /* The server detected a resend of the RPC call and
 788                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 789                  * of RFC5661.
 790                  */
 791                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 792                         __func__,
 793                         slot->slot_nr,
 794                         slot->seq_nr);
 795                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 796                 goto out_retry;
 797         case -NFS4ERR_RETRY_UNCACHED_REP:
 798         case -NFS4ERR_SEQ_FALSE_RETRY:
 799                 /*
 800                  * The server thinks we tried to replay a request.
 801                  * Retry the call after bumping the sequence ID.
 802                  */
 803                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 804                 goto retry_new_seq;
 805         case -NFS4ERR_BADSLOT:
 806                 /*
 807                  * The slot id we used was probably retired. Try again
 808                  * using a different slot id.
 809                  */
 810                 if (slot->slot_nr < slot->table->target_highest_slotid)
 811                         goto session_recover;
 812                 goto retry_nowait;
 813         case -NFS4ERR_SEQ_MISORDERED:
 814                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 815                 /*
 816                  * Were one or more calls using this slot interrupted?
 817                  * If the server never received the request, then our
 818                  * transmitted slot sequence number may be too high.
 819                  */
 820                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 821                         slot->seq_nr--;
 822                         goto retry_nowait;
 823                 }
 824                 /*
 825                  * RFC5661:
 826                  * A retry might be sent while the original request is
 827                  * still in progress on the replier. The replier SHOULD
 828                  * deal with the issue by returning NFS4ERR_DELAY as the
 829                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 830                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 831                  *
 832                  * Restart the search after a delay.
 833                  */
 834                 slot->seq_nr = slot->seq_nr_highest_sent;
 835                 goto out_retry;
 836         default:
 837                 /* Just update the slot sequence no. */
 838                 slot->seq_done = 1;
 839         }
 840 out:
 841         /* The session may be reset by one of the error handlers. */
 842         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 843 out_noaction:
 844         return ret;
 845 session_recover:
 846         nfs4_schedule_session_recovery(session, res->sr_status);
 847         goto retry_nowait;
 848 retry_new_seq:
 849         ++slot->seq_nr;
 850 retry_nowait:
 851         if (rpc_restart_call_prepare(task)) {
 852                 nfs41_sequence_free_slot(res);
 853                 task->tk_status = 0;
 854                 ret = 0;
 855         }
 856         goto out;
 857 out_retry:
 858         if (!rpc_restart_call(task))
 859                 goto out;
 860         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 861         return 0;
 862 }
 863
 864 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 865 {
 866         if (!nfs41_sequence_process(task, res))
 867                 return 0;
 868         if (res->sr_slot != NULL)
 869                 nfs41_sequence_free_slot(res);
 870         return 1;
 871
 872 }
 873 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 874
 875 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 876 {
 877         if (res->sr_slot == NULL)
 878                 return 1;
 879         if (res->sr_slot->table->session != NULL)
 880                 return nfs41_sequence_process(task, res);
 881         return nfs40_sequence_done(task, res);
 882 }
 883
 884 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 885 {
 886         if (res->sr_slot != NULL) {
 887                 if (res->sr_slot->table->session != NULL)
 888                         nfs41_sequence_free_slot(res);
 889                 else
 890                         nfs40_sequence_free_slot(res);
 891         }
 892 }
 893
 894 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 895 {
 896         if (res->sr_slot == NULL)
 897                 return 1;
 898         if (!res->sr_slot->table->session)
 899                 return nfs40_sequence_done(task, res);
 900         return nfs41_sequence_done(task, res);
 901 }
 902 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 903
 904 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 905 {
 906         struct nfs4_call_sync_data *data = calldata;
 907
 908         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 909
 910         nfs4_setup_sequence(data->seq_server->nfs_client,
 911                             data->seq_args, data->seq_res, task);
 912 }
 913
 914 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 915 {
 916         struct nfs4_call_sync_data *data = calldata;
 917
 918         nfs41_sequence_done(task, data->seq_res);
 919 }
 920
 921 static const struct rpc_call_ops nfs41_call_sync_ops = {
 922         .rpc_call_prepare = nfs41_call_sync_prepare,
 923         .rpc_call_done = nfs41_call_sync_done,
 924 };
 925
 926 #else   /* !CONFIG_NFS_V4_1 */
 927
 928 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 929 {
 930         return nfs40_sequence_done(task, res);
 931 }
 932
 933 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 934 {
 935         if (res->sr_slot != NULL)
 936                 nfs40_sequence_free_slot(res);
 937 }
 938
 939 int nfs4_sequence_done(struct rpc_task *task,
 940                        struct nfs4_sequence_res *res)
 941 {
 942         return nfs40_sequence_done(task, res);
 943 }
 944 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 945
 946 #endif  /* !CONFIG_NFS_V4_1 */
 947
 948 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
 949 {
 950         res->sr_timestamp = jiffies;
 951         res->sr_status_flags = 0;
 952         res->sr_status = 1;
 953 }
 954
 955 static
 956 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
 957                 struct nfs4_sequence_res *res,
 958                 struct nfs4_slot *slot)
 959 {
 960         if (!slot)
 961                 return;
 962         slot->privileged = args->sa_privileged ? 1 : 0;
 963         args->sa_slot = slot;
 964
 965         res->sr_slot = slot;
 966 }
 967
 968 int nfs4_setup_sequence(struct nfs_client *client,
 969                         struct nfs4_sequence_args *args,
 970                         struct nfs4_sequence_res *res,
 971                         struct rpc_task *task)
 972 {
 973         struct nfs4_session *session = nfs4_get_session(client);
 974         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
 975         struct nfs4_slot *slot;
 976
 977         /* slot already allocated? */
 978         if (res->sr_slot != NULL)
 979                 goto out_start;
 980
 981         if (session) {
 982                 tbl = &session->fc_slot_table;
 983                 task->tk_timeout = 0;
 984         }
 985
 986         spin_lock(&tbl->slot_tbl_lock);
 987         /* The state manager will wait until the slot table is empty */
 988         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 989                 goto out_sleep;
 990
 991         slot = nfs4_alloc_slot(tbl);
 992         if (IS_ERR(slot)) {
 993                 /* Try again in 1/4 second */
 994                 if (slot == ERR_PTR(-ENOMEM))
 995                         task->tk_timeout = HZ >> 2;
 996                 goto out_sleep;
 997         }
 998         spin_unlock(&tbl->slot_tbl_lock);
 999
1000         nfs4_sequence_attach_slot(args, res, slot);
1001
1002         trace_nfs4_setup_sequence(session, args);
1003 out_start:
1004         nfs41_sequence_res_init(res);
1005         rpc_call_start(task);
1006         return 0;
1007
1008 out_sleep:
1009         if (args->sa_privileged)
1010                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1011                                 NULL, RPC_PRIORITY_PRIVILEGED);
1012         else
1013                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1014         spin_unlock(&tbl->slot_tbl_lock);
1015         return -EAGAIN;
1016 }
1017 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1018
1019 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1020 {
1021         struct nfs4_call_sync_data *data = calldata;
1022         nfs4_setup_sequence(data->seq_server->nfs_client,
1023                                 data->seq_args, data->seq_res, task);
1024 }
1025
1026 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1027 {
1028         struct nfs4_call_sync_data *data = calldata;
1029         nfs4_sequence_done(task, data->seq_res);
1030 }
1031
1032 static const struct rpc_call_ops nfs40_call_sync_ops = {
1033         .rpc_call_prepare = nfs40_call_sync_prepare,
1034         .rpc_call_done = nfs40_call_sync_done,
1035 };
1036
1037 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1038                                    struct nfs_server *server,
1039                                    struct rpc_message *msg,
1040                                    struct nfs4_sequence_args *args,
1041                                    struct nfs4_sequence_res *res)
1042 {
1043         int ret;
1044         struct rpc_task *task;
1045         struct nfs_client *clp = server->nfs_client;
1046         struct nfs4_call_sync_data data = {
1047                 .seq_server = server,
1048                 .seq_args = args,
1049                 .seq_res = res,
1050         };
1051         struct rpc_task_setup task_setup = {
1052                 .rpc_client = clnt,
1053                 .rpc_message = msg,
1054                 .callback_ops = clp->cl_mvops->call_sync_ops,
1055                 .callback_data = &data
1056         };
1057
1058         task = rpc_run_task(&task_setup);
1059         if (IS_ERR(task))
1060                 ret = PTR_ERR(task);
1061         else {
1062                 ret = task->tk_status;
1063                 rpc_put_task(task);
1064         }
1065         return ret;
1066 }
1067
1068 int nfs4_call_sync(struct rpc_clnt *clnt,
1069                    struct nfs_server *server,
1070                    struct rpc_message *msg,
1071                    struct nfs4_sequence_args *args,
1072                    struct nfs4_sequence_res *res,
1073                    int cache_reply)
1074 {
1075         nfs4_init_sequence(args, res, cache_reply, 0);
1076         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1077 }
1078
1079 static void
1080 nfs4_inc_nlink_locked(struct inode *inode)
1081 {
1082         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1083         inc_nlink(inode);
1084 }
1085
1086 static void
1087 nfs4_dec_nlink_locked(struct inode *inode)
1088 {
1089         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1090         drop_nlink(inode);
1091 }
1092
1093 static void
1094 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1095                 unsigned long timestamp, unsigned long cache_validity)
1096 {
1097         struct nfs_inode *nfsi = NFS_I(dir);
1098
1099         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1100                 | NFS_INO_INVALID_MTIME
1101                 | NFS_INO_INVALID_DATA
1102                 | cache_validity;
1103         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1104                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1105                 nfsi->attrtimeo_timestamp = jiffies;
1106         } else {
1107                 nfs_force_lookup_revalidate(dir);
1108                 if (cinfo->before != inode_peek_iversion_raw(dir))
1109                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1110                                 NFS_INO_INVALID_ACL;
1111         }
1112         inode_set_iversion_raw(dir, cinfo->after);
1113         nfsi->read_cache_jiffies = timestamp;
1114         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1115         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1116         nfs_fscache_invalidate(dir);
1117 }
1118
1119 static void
1120 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1121                 unsigned long timestamp, unsigned long cache_validity)
1122 {
1123         spin_lock(&dir->i_lock);
1124         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1125         spin_unlock(&dir->i_lock);
1126 }
1127
1128 struct nfs4_open_createattrs {
1129         struct nfs4_label *label;
1130         struct iattr *sattr;
1131         const __u32 verf[2];
1132 };
1133
1134 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1135                 int err, struct nfs4_exception *exception)
1136 {
1137         if (err != -EINVAL)
1138                 return false;
1139         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1140                 return false;
1141         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1142         exception->retry = 1;
1143         return true;
1144 }
1145
1146 static u32
1147 nfs4_map_atomic_open_share(struct nfs_server *server,
1148                 fmode_t fmode, int openflags)
1149 {
1150         u32 res = 0;
1151
1152         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1153         case FMODE_READ:
1154                 res = NFS4_SHARE_ACCESS_READ;
1155                 break;
1156         case FMODE_WRITE:
1157                 res = NFS4_SHARE_ACCESS_WRITE;
1158                 break;
1159         case FMODE_READ|FMODE_WRITE:
1160                 res = NFS4_SHARE_ACCESS_BOTH;
1161         }
1162         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1163                 goto out;
1164         /* Want no delegation if we're using O_DIRECT */
1165         if (openflags & O_DIRECT)
1166                 res |= NFS4_SHARE_WANT_NO_DELEG;
1167 out:
1168         return res;
1169 }
1170
1171 static enum open_claim_type4
1172 nfs4_map_atomic_open_claim(struct nfs_server *server,
1173                 enum open_claim_type4 claim)
1174 {
1175         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1176                 return claim;
1177         switch (claim) {
1178         default:
1179                 return claim;
1180         case NFS4_OPEN_CLAIM_FH:
1181                 return NFS4_OPEN_CLAIM_NULL;
1182         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1183                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1184         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1185                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1186         }
1187 }
1188
1189 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1190 {
1191         p->o_res.f_attr = &p->f_attr;
1192         p->o_res.f_label = p->f_label;
1193         p->o_res.seqid = p->o_arg.seqid;
1194         p->c_res.seqid = p->c_arg.seqid;
1195         p->o_res.server = p->o_arg.server;
1196         p->o_res.access_request = p->o_arg.access;
1197         nfs_fattr_init(&p->f_attr);
1198         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1199 }
1200
1201 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1202                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1203                 const struct nfs4_open_createattrs *c,
1204                 enum open_claim_type4 claim,
1205                 gfp_t gfp_mask)
1206 {
1207         struct dentry *parent = dget_parent(dentry);
1208         struct inode *dir = d_inode(parent);
1209         struct nfs_server *server = NFS_SERVER(dir);
1210         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1211         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1212         struct nfs4_opendata *p;
1213
1214         p = kzalloc(sizeof(*p), gfp_mask);
1215         if (p == NULL)
1216                 goto err;
1217
1218         p->f_label = nfs4_label_alloc(server, gfp_mask);
1219         if (IS_ERR(p->f_label))
1220                 goto err_free_p;
1221
1222         p->a_label = nfs4_label_alloc(server, gfp_mask);
1223         if (IS_ERR(p->a_label))
1224                 goto err_free_f;
1225
1226         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1227         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1228         if (IS_ERR(p->o_arg.seqid))
1229                 goto err_free_label;
1230         nfs_sb_active(dentry->d_sb);
1231         p->dentry = dget(dentry);
1232         p->dir = parent;
1233         p->owner = sp;
1234         atomic_inc(&sp->so_count);
1235         p->o_arg.open_flags = flags;
1236         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1237         p->o_arg.umask = current_umask();
1238         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1239         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1240                         fmode, flags);
1241         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1242          * will return permission denied for all bits until close */
1243         if (!(flags & O_EXCL)) {
1244                 /* ask server to check for all possible rights as results
1245                  * are cached */
1246                 switch (p->o_arg.claim) {
1247                 default:
1248                         break;
1249                 case NFS4_OPEN_CLAIM_NULL:
1250                 case NFS4_OPEN_CLAIM_FH:
1251                         p->o_arg.access = NFS4_ACCESS_READ |
1252                                 NFS4_ACCESS_MODIFY |
1253                                 NFS4_ACCESS_EXTEND |
1254                                 NFS4_ACCESS_EXECUTE;
1255                 }
1256         }
1257         p->o_arg.clientid = server->nfs_client->cl_clientid;
1258         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1259         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1260         p->o_arg.name = &dentry->d_name;
1261         p->o_arg.server = server;
1262         p->o_arg.bitmask = nfs4_bitmask(server, label);
1263         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1264         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1265         switch (p->o_arg.claim) {
1266         case NFS4_OPEN_CLAIM_NULL:
1267         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1268         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1269                 p->o_arg.fh = NFS_FH(dir);
1270                 break;
1271         case NFS4_OPEN_CLAIM_PREVIOUS:
1272         case NFS4_OPEN_CLAIM_FH:
1273         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1274         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1275                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1276         }
1277         if (c != NULL && c->sattr != NULL && c->sattr->ia_valid != 0) {
1278                 p->o_arg.u.attrs = &p->attrs;
1279                 memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1280
1281                 memcpy(p->o_arg.u.verifier.data, c->verf,
1282                                 sizeof(p->o_arg.u.verifier.data));
1283         }
1284         p->c_arg.fh = &p->o_res.fh;
1285         p->c_arg.stateid = &p->o_res.stateid;
1286         p->c_arg.seqid = p->o_arg.seqid;
1287         nfs4_init_opendata_res(p);
1288         kref_init(&p->kref);
1289         return p;
1290
1291 err_free_label:
1292         nfs4_label_free(p->a_label);
1293 err_free_f:
1294         nfs4_label_free(p->f_label);
1295 err_free_p:
1296         kfree(p);
1297 err:
1298         dput(parent);
1299         return NULL;
1300 }
1301
1302 static void nfs4_opendata_free(struct kref *kref)
1303 {
1304         struct nfs4_opendata *p = container_of(kref,
1305                         struct nfs4_opendata, kref);
1306         struct super_block *sb = p->dentry->d_sb;
1307
1308         nfs4_lgopen_release(p->lgp);
1309         nfs_free_seqid(p->o_arg.seqid);
1310         nfs4_sequence_free_slot(&p->o_res.seq_res);
1311         if (p->state != NULL)
1312                 nfs4_put_open_state(p->state);
1313         nfs4_put_state_owner(p->owner);
1314
1315         nfs4_label_free(p->a_label);
1316         nfs4_label_free(p->f_label);
1317
1318         dput(p->dir);
1319         dput(p->dentry);
1320         nfs_sb_deactive(sb);
1321         nfs_fattr_free_names(&p->f_attr);
1322         kfree(p->f_attr.mdsthreshold);
1323         kfree(p);
1324 }
1325
1326 static void nfs4_opendata_put(struct nfs4_opendata *p)
1327 {
1328         if (p != NULL)
1329                 kref_put(&p->kref, nfs4_opendata_free);
1330 }
1331
1332 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1333                 fmode_t fmode)
1334 {
1335         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1336         case FMODE_READ|FMODE_WRITE:
1337                 return state->n_rdwr != 0;
1338         case FMODE_WRITE:
1339                 return state->n_wronly != 0;
1340         case FMODE_READ:
1341                 return state->n_rdonly != 0;
1342         }
1343         WARN_ON_ONCE(1);
1344         return false;
1345 }
1346
1347 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1348                 int open_mode, enum open_claim_type4 claim)
1349 {
1350         int ret = 0;
1351
1352         if (open_mode & (O_EXCL|O_TRUNC))
1353                 goto out;
1354         switch (claim) {
1355         case NFS4_OPEN_CLAIM_NULL:
1356         case NFS4_OPEN_CLAIM_FH:
1357                 goto out;
1358         default:
1359                 break;
1360         }
1361         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1362                 case FMODE_READ:
1363                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1364                                 && state->n_rdonly != 0;
1365                         break;
1366                 case FMODE_WRITE:
1367                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1368                                 && state->n_wronly != 0;
1369                         break;
1370                 case FMODE_READ|FMODE_WRITE:
1371                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1372                                 && state->n_rdwr != 0;
1373         }
1374 out:
1375         return ret;
1376 }
1377
1378 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1379                 enum open_claim_type4 claim)
1380 {
1381         if (delegation == NULL)
1382                 return 0;
1383         if ((delegation->type & fmode) != fmode)
1384                 return 0;
1385         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1386                 return 0;
1387         switch (claim) {
1388         case NFS4_OPEN_CLAIM_NULL:
1389         case NFS4_OPEN_CLAIM_FH:
1390                 break;
1391         case NFS4_OPEN_CLAIM_PREVIOUS:
1392                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1393                         break;
1394                 /* Fall through */
1395         default:
1396                 return 0;
1397         }
1398         nfs_mark_delegation_referenced(delegation);
1399         return 1;
1400 }
1401
1402 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1403 {
1404         switch (fmode) {
1405                 case FMODE_WRITE:
1406                         state->n_wronly++;
1407                         break;
1408                 case FMODE_READ:
1409                         state->n_rdonly++;
1410                         break;
1411                 case FMODE_READ|FMODE_WRITE:
1412                         state->n_rdwr++;
1413         }
1414         nfs4_state_set_mode_locked(state, state->state | fmode);
1415 }
1416
1417 #ifdef CONFIG_NFS_V4_1
1418 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1419 {
1420         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1421                 return true;
1422         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1423                 return true;
1424         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1425                 return true;
1426         return false;
1427 }
1428 #endif /* CONFIG_NFS_V4_1 */
1429
1430 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1431 {
1432         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1433                 wake_up_all(&state->waitq);
1434 }
1435
1436 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1437                 const nfs4_stateid *stateid)
1438 {
1439         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1440         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1441
1442         if (stateid_seqid == state_seqid + 1U ||
1443             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1444                 nfs_state_log_update_open_stateid(state);
1445         else
1446                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1447 }
1448
1449 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1450 {
1451         struct nfs_client *clp = state->owner->so_server->nfs_client;
1452         bool need_recover = false;
1453
1454         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1455                 need_recover = true;
1456         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1457                 need_recover = true;
1458         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1459                 need_recover = true;
1460         if (need_recover)
1461                 nfs4_state_mark_reclaim_nograce(clp, state);
1462 }
1463
1464 /*
1465  * Check for whether or not the caller may update the open stateid
1466  * to the value passed in by stateid.
1467  *
1468  * Note: This function relies heavily on the server implementing
1469  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1470  * correctly.
1471  * i.e. The stateid seqids have to be initialised to 1, and
1472  * are then incremented on every state transition.
1473  */
1474 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1475                 const nfs4_stateid *stateid)
1476 {
1477         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1478             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1479                 if (stateid->seqid == cpu_to_be32(1))
1480                         nfs_state_log_update_open_stateid(state);
1481                 else
1482                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1483                 return true;
1484         }
1485
1486         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1487                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1488                 return true;
1489         }
1490         return false;
1491 }
1492
1493 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1494 {
1495         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1496                 return;
1497         if (state->n_wronly)
1498                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1499         if (state->n_rdonly)
1500                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1501         if (state->n_rdwr)
1502                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1503         set_bit(NFS_OPEN_STATE, &state->flags);
1504 }
1505
1506 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1507                 nfs4_stateid *stateid, fmode_t fmode)
1508 {
1509         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1510         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1511         case FMODE_WRITE:
1512                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1513                 break;
1514         case FMODE_READ:
1515                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1516                 break;
1517         case 0:
1518                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1519                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1520                 clear_bit(NFS_OPEN_STATE, &state->flags);
1521         }
1522         if (stateid == NULL)
1523                 return;
1524         /* Handle OPEN+OPEN_DOWNGRADE races */
1525         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1526             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1527                 nfs_resync_open_stateid_locked(state);
1528                 goto out;
1529         }
1530         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1531                 nfs4_stateid_copy(&state->stateid, stateid);
1532         nfs4_stateid_copy(&state->open_stateid, stateid);
1533         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1534 out:
1535         nfs_state_log_update_open_stateid(state);
1536 }
1537
1538 static void nfs_clear_open_stateid(struct nfs4_state *state,
1539         nfs4_stateid *arg_stateid,
1540         nfs4_stateid *stateid, fmode_t fmode)
1541 {
1542         write_seqlock(&state->seqlock);
1543         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1544         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1545                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1546         write_sequnlock(&state->seqlock);
1547         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1548                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1549 }
1550
1551 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1552                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1553         __must_hold(&state->owner->so_lock)
1554         __must_hold(&state->seqlock)
1555         __must_hold(RCU)
1556
1557 {
1558         DEFINE_WAIT(wait);
1559         int status = 0;
1560         for (;;) {
1561
1562                 if (!nfs_need_update_open_stateid(state, stateid))
1563                         return;
1564                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1565                         break;
1566                 if (status)
1567                         break;
1568                 /* Rely on seqids for serialisation with NFSv4.0 */
1569                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1570                         break;
1571
1572                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1573                 /*
1574                  * Ensure we process the state changes in the same order
1575                  * in which the server processed them by delaying the
1576                  * update of the stateid until we are in sequence.
1577                  */
1578                 write_sequnlock(&state->seqlock);
1579                 spin_unlock(&state->owner->so_lock);
1580                 rcu_read_unlock();
1581                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1582                 if (!signal_pending(current)) {
1583                         if (schedule_timeout(5*HZ) == 0)
1584                                 status = -EAGAIN;
1585                         else
1586                                 status = 0;
1587                 } else
1588                         status = -EINTR;
1589                 finish_wait(&state->waitq, &wait);
1590                 rcu_read_lock();
1591                 spin_lock(&state->owner->so_lock);
1592                 write_seqlock(&state->seqlock);
1593         }
1594
1595         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1596             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1597                 nfs4_stateid_copy(freeme, &state->open_stateid);
1598                 nfs_test_and_clear_all_open_stateid(state);
1599         }
1600
1601         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1602                 nfs4_stateid_copy(&state->stateid, stateid);
1603         nfs4_stateid_copy(&state->open_stateid, stateid);
1604         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1605         nfs_state_log_update_open_stateid(state);
1606 }
1607
1608 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1609                 const nfs4_stateid *open_stateid,
1610                 fmode_t fmode,
1611                 nfs4_stateid *freeme)
1612 {
1613         /*
1614          * Protect the call to nfs4_state_set_mode_locked and
1615          * serialise the stateid update
1616          */
1617         write_seqlock(&state->seqlock);
1618         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1619         switch (fmode) {
1620         case FMODE_READ:
1621                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1622                 break;
1623         case FMODE_WRITE:
1624                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1625                 break;
1626         case FMODE_READ|FMODE_WRITE:
1627                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1628         }
1629         set_bit(NFS_OPEN_STATE, &state->flags);
1630         write_sequnlock(&state->seqlock);
1631 }
1632
1633 static void nfs_state_set_delegation(struct nfs4_state *state,
1634                 const nfs4_stateid *deleg_stateid,
1635                 fmode_t fmode)
1636 {
1637         /*
1638          * Protect the call to nfs4_state_set_mode_locked and
1639          * serialise the stateid update
1640          */
1641         write_seqlock(&state->seqlock);
1642         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1643         set_bit(NFS_DELEGATED_STATE, &state->flags);
1644         write_sequnlock(&state->seqlock);
1645 }
1646
1647 static void nfs_state_clear_delegation(struct nfs4_state *state)
1648 {
1649         write_seqlock(&state->seqlock);
1650         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1651         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1652         write_sequnlock(&state->seqlock);
1653 }
1654
1655 static int update_open_stateid(struct nfs4_state *state,
1656                 const nfs4_stateid *open_stateid,
1657                 const nfs4_stateid *delegation,
1658                 fmode_t fmode)
1659 {
1660         struct nfs_server *server = NFS_SERVER(state->inode);
1661         struct nfs_client *clp = server->nfs_client;
1662         struct nfs_inode *nfsi = NFS_I(state->inode);
1663         struct nfs_delegation *deleg_cur;
1664         nfs4_stateid freeme = { };
1665         int ret = 0;
1666
1667         fmode &= (FMODE_READ|FMODE_WRITE);
1668
1669         rcu_read_lock();
1670         spin_lock(&state->owner->so_lock);
1671         if (open_stateid != NULL) {
1672                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1673                 ret = 1;
1674         }
1675
1676         deleg_cur = rcu_dereference(nfsi->delegation);
1677         if (deleg_cur == NULL)
1678                 goto no_delegation;
1679
1680         spin_lock(&deleg_cur->lock);
1681         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1682            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1683             (deleg_cur->type & fmode) != fmode)
1684                 goto no_delegation_unlock;
1685
1686         if (delegation == NULL)
1687                 delegation = &deleg_cur->stateid;
1688         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1689                 goto no_delegation_unlock;
1690
1691         nfs_mark_delegation_referenced(deleg_cur);
1692         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1693         ret = 1;
1694 no_delegation_unlock:
1695         spin_unlock(&deleg_cur->lock);
1696 no_delegation:
1697         if (ret)
1698                 update_open_stateflags(state, fmode);
1699         spin_unlock(&state->owner->so_lock);
1700         rcu_read_unlock();
1701
1702         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1703                 nfs4_schedule_state_manager(clp);
1704         if (freeme.type != 0)
1705                 nfs4_test_and_free_stateid(server, &freeme,
1706                                 state->owner->so_cred);
1707
1708         return ret;
1709 }
1710
1711 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1712                 const nfs4_stateid *stateid)
1713 {
1714         struct nfs4_state *state = lsp->ls_state;
1715         bool ret = false;
1716
1717         spin_lock(&state->state_lock);
1718         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1719                 goto out_noupdate;
1720         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1721                 goto out_noupdate;
1722         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1723         ret = true;
1724 out_noupdate:
1725         spin_unlock(&state->state_lock);
1726         return ret;
1727 }
1728
1729 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1730 {
1731         struct nfs_delegation *delegation;
1732
1733         fmode &= FMODE_READ|FMODE_WRITE;
1734         rcu_read_lock();
1735         delegation = rcu_dereference(NFS_I(inode)->delegation);
1736         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1737                 rcu_read_unlock();
1738                 return;
1739         }
1740         rcu_read_unlock();
1741         nfs4_inode_return_delegation(inode);
1742 }
1743
1744 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1745 {
1746         struct nfs4_state *state = opendata->state;
1747         struct nfs_inode *nfsi = NFS_I(state->inode);
1748         struct nfs_delegation *delegation;
1749         int open_mode = opendata->o_arg.open_flags;
1750         fmode_t fmode = opendata->o_arg.fmode;
1751         enum open_claim_type4 claim = opendata->o_arg.claim;
1752         nfs4_stateid stateid;
1753         int ret = -EAGAIN;
1754
1755         for (;;) {
1756                 spin_lock(&state->owner->so_lock);
1757                 if (can_open_cached(state, fmode, open_mode, claim)) {
1758                         update_open_stateflags(state, fmode);
1759                         spin_unlock(&state->owner->so_lock);
1760                         goto out_return_state;
1761                 }
1762                 spin_unlock(&state->owner->so_lock);
1763                 rcu_read_lock();
1764                 delegation = rcu_dereference(nfsi->delegation);
1765                 if (!can_open_delegated(delegation, fmode, claim)) {
1766                         rcu_read_unlock();
1767                         break;
1768                 }
1769                 /* Save the delegation */
1770                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1771                 rcu_read_unlock();
1772                 nfs_release_seqid(opendata->o_arg.seqid);
1773                 if (!opendata->is_recover) {
1774                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1775                         if (ret != 0)
1776                                 goto out;
1777                 }
1778                 ret = -EAGAIN;
1779
1780                 /* Try to update the stateid using the delegation */
1781                 if (update_open_stateid(state, NULL, &stateid, fmode))
1782                         goto out_return_state;
1783         }
1784 out:
1785         return ERR_PTR(ret);
1786 out_return_state:
1787         refcount_inc(&state->count);
1788         return state;
1789 }
1790
1791 static void
1792 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1793 {
1794         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1795         struct nfs_delegation *delegation;
1796         int delegation_flags = 0;
1797
1798         rcu_read_lock();
1799         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1800         if (delegation)
1801                 delegation_flags = delegation->flags;
1802         rcu_read_unlock();
1803         switch (data->o_arg.claim) {
1804         default:
1805                 break;
1806         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1807         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1808                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1809                                    "returning a delegation for "
1810                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1811                                    clp->cl_hostname);
1812                 return;
1813         }
1814         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1815                 nfs_inode_set_delegation(state->inode,
1816                                 data->owner->so_cred,
1817                                 data->o_res.delegation_type,
1818                                 &data->o_res.delegation,
1819                                 data->o_res.pagemod_limit);
1820         else
1821                 nfs_inode_reclaim_delegation(state->inode,
1822                                 data->owner->so_cred,
1823                                 data->o_res.delegation_type,
1824                                 &data->o_res.delegation,
1825                                 data->o_res.pagemod_limit);
1826
1827         if (data->o_res.do_recall)
1828                 nfs_async_inode_return_delegation(state->inode,
1829                                                   &data->o_res.delegation);
1830 }
1831
1832 /*
1833  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1834  * and update the nfs4_state.
1835  */
1836 static struct nfs4_state *
1837 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1838 {
1839         struct inode *inode = data->state->inode;
1840         struct nfs4_state *state = data->state;
1841         int ret;
1842
1843         if (!data->rpc_done) {
1844                 if (data->rpc_status)
1845                         return ERR_PTR(data->rpc_status);
1846                 /* cached opens have already been processed */
1847                 goto update;
1848         }
1849
1850         ret = nfs_refresh_inode(inode, &data->f_attr);
1851         if (ret)
1852                 return ERR_PTR(ret);
1853
1854         if (data->o_res.delegation_type != 0)
1855                 nfs4_opendata_check_deleg(data, state);
1856 update:
1857         update_open_stateid(state, &data->o_res.stateid, NULL,
1858                             data->o_arg.fmode);
1859         refcount_inc(&state->count);
1860
1861         return state;
1862 }
1863
1864 static struct inode *
1865 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1866 {
1867         struct inode *inode;
1868
1869         switch (data->o_arg.claim) {
1870         case NFS4_OPEN_CLAIM_NULL:
1871         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1872         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1873                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1874                         return ERR_PTR(-EAGAIN);
1875                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1876                                 &data->f_attr, data->f_label);
1877                 break;
1878         default:
1879                 inode = d_inode(data->dentry);
1880                 ihold(inode);
1881                 nfs_refresh_inode(inode, &data->f_attr);
1882         }
1883         return inode;
1884 }
1885
1886 static struct nfs4_state *
1887 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1888 {
1889         struct nfs4_state *state;
1890         struct inode *inode;
1891
1892         inode = nfs4_opendata_get_inode(data);
1893         if (IS_ERR(inode))
1894                 return ERR_CAST(inode);
1895         if (data->state != NULL && data->state->inode == inode) {
1896                 state = data->state;
1897                 refcount_inc(&state->count);
1898         } else
1899                 state = nfs4_get_open_state(inode, data->owner);
1900         iput(inode);
1901         if (state == NULL)
1902                 state = ERR_PTR(-ENOMEM);
1903         return state;
1904 }
1905
1906 static struct nfs4_state *
1907 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1908 {
1909         struct nfs4_state *state;
1910
1911         if (!data->rpc_done) {
1912                 state = nfs4_try_open_cached(data);
1913                 trace_nfs4_cached_open(data->state);
1914                 goto out;
1915         }
1916
1917         state = nfs4_opendata_find_nfs4_state(data);
1918         if (IS_ERR(state))
1919                 goto out;
1920
1921         if (data->o_res.delegation_type != 0)
1922                 nfs4_opendata_check_deleg(data, state);
1923         update_open_stateid(state, &data->o_res.stateid, NULL,
1924                         data->o_arg.fmode);
1925 out:
1926         nfs_release_seqid(data->o_arg.seqid);
1927         return state;
1928 }
1929
1930 static struct nfs4_state *
1931 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1932 {
1933         struct nfs4_state *ret;
1934
1935         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1936                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1937         else
1938                 ret = _nfs4_opendata_to_nfs4_state(data);
1939         nfs4_sequence_free_slot(&data->o_res.seq_res);
1940         return ret;
1941 }
1942
1943 static struct nfs_open_context *
1944 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
1945 {
1946         struct nfs_inode *nfsi = NFS_I(state->inode);
1947         struct nfs_open_context *ctx;
1948
1949         rcu_read_lock();
1950         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
1951                 if (ctx->state != state)
1952                         continue;
1953                 if ((ctx->mode & mode) != mode)
1954                         continue;
1955                 if (!get_nfs_open_context(ctx))
1956                         continue;
1957                 rcu_read_unlock();
1958                 return ctx;
1959         }
1960         rcu_read_unlock();
1961         return ERR_PTR(-ENOENT);
1962 }
1963
1964 static struct nfs_open_context *
1965 nfs4_state_find_open_context(struct nfs4_state *state)
1966 {
1967         struct nfs_open_context *ctx;
1968
1969         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
1970         if (!IS_ERR(ctx))
1971                 return ctx;
1972         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
1973         if (!IS_ERR(ctx))
1974                 return ctx;
1975         return nfs4_state_find_open_context_mode(state, FMODE_READ);
1976 }
1977
1978 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1979                 struct nfs4_state *state, enum open_claim_type4 claim)
1980 {
1981         struct nfs4_opendata *opendata;
1982
1983         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1984                         NULL, claim, GFP_NOFS);
1985         if (opendata == NULL)
1986                 return ERR_PTR(-ENOMEM);
1987         opendata->state = state;
1988         refcount_inc(&state->count);
1989         return opendata;
1990 }
1991
1992 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1993                 fmode_t fmode)
1994 {
1995         struct nfs4_state *newstate;
1996         int ret;
1997
1998         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1999                 return 0;
2000         opendata->o_arg.open_flags = 0;
2001         opendata->o_arg.fmode = fmode;
2002         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2003                         NFS_SB(opendata->dentry->d_sb),
2004                         fmode, 0);
2005         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2006         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2007         nfs4_init_opendata_res(opendata);
2008         ret = _nfs4_recover_proc_open(opendata);
2009         if (ret != 0)
2010                 return ret;
2011         newstate = nfs4_opendata_to_nfs4_state(opendata);
2012         if (IS_ERR(newstate))
2013                 return PTR_ERR(newstate);
2014         if (newstate != opendata->state)
2015                 ret = -ESTALE;
2016         nfs4_close_state(newstate, fmode);
2017         return ret;
2018 }
2019
2020 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2021 {
2022         int ret;
2023
2024         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
2025         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2026         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2027         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2028         /* memory barrier prior to reading state->n_* */
2029         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2030         clear_bit(NFS_OPEN_STATE, &state->flags);
2031         smp_rmb();
2032         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2033         if (ret != 0)
2034                 return ret;
2035         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2036         if (ret != 0)
2037                 return ret;
2038         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2039         if (ret != 0)
2040                 return ret;
2041         /*
2042          * We may have performed cached opens for all three recoveries.
2043          * Check if we need to update the current stateid.
2044          */
2045         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2046             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2047                 write_seqlock(&state->seqlock);
2048                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2049                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2050                 write_sequnlock(&state->seqlock);
2051         }
2052         return 0;
2053 }
2054
2055 /*
2056  * OPEN_RECLAIM:
2057  *      reclaim state on the server after a reboot.
2058  */
2059 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2060 {
2061         struct nfs_delegation *delegation;
2062         struct nfs4_opendata *opendata;
2063         fmode_t delegation_type = 0;
2064         int status;
2065
2066         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2067                         NFS4_OPEN_CLAIM_PREVIOUS);
2068         if (IS_ERR(opendata))
2069                 return PTR_ERR(opendata);
2070         rcu_read_lock();
2071         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2072         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2073                 delegation_type = delegation->type;
2074         rcu_read_unlock();
2075         opendata->o_arg.u.delegation_type = delegation_type;
2076         status = nfs4_open_recover(opendata, state);
2077         nfs4_opendata_put(opendata);
2078         return status;
2079 }
2080
2081 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2082 {
2083         struct nfs_server *server = NFS_SERVER(state->inode);
2084         struct nfs4_exception exception = { };
2085         int err;
2086         do {
2087                 err = _nfs4_do_open_reclaim(ctx, state);
2088                 trace_nfs4_open_reclaim(ctx, 0, err);
2089                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2090                         continue;
2091                 if (err != -NFS4ERR_DELAY)
2092                         break;
2093                 nfs4_handle_exception(server, err, &exception);
2094         } while (exception.retry);
2095         return err;
2096 }
2097
2098 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2099 {
2100         struct nfs_open_context *ctx;
2101         int ret;
2102
2103         ctx = nfs4_state_find_open_context(state);
2104         if (IS_ERR(ctx))
2105                 return -EAGAIN;
2106         ret = nfs4_do_open_reclaim(ctx, state);
2107         put_nfs_open_context(ctx);
2108         return ret;
2109 }
2110
2111 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)
2112 {
2113         switch (err) {
2114                 default:
2115                         printk(KERN_ERR "NFS: %s: unhandled error "
2116                                         "%d.\n", __func__, err);
2117                 case 0:
2118                 case -ENOENT:
2119                 case -EAGAIN:
2120                 case -ESTALE:
2121                         break;
2122                 case -NFS4ERR_BADSESSION:
2123                 case -NFS4ERR_BADSLOT:
2124                 case -NFS4ERR_BAD_HIGH_SLOT:
2125                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2126                 case -NFS4ERR_DEADSESSION:
2127                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2128                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
2129                         return -EAGAIN;
2130                 case -NFS4ERR_STALE_CLIENTID:
2131                 case -NFS4ERR_STALE_STATEID:
2132                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2133                         /* Don't recall a delegation if it was lost */
2134                         nfs4_schedule_lease_recovery(server->nfs_client);
2135                         return -EAGAIN;
2136                 case -NFS4ERR_MOVED:
2137                         nfs4_schedule_migration_recovery(server);
2138                         return -EAGAIN;
2139                 case -NFS4ERR_LEASE_MOVED:
2140                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2141                         return -EAGAIN;
2142                 case -NFS4ERR_DELEG_REVOKED:
2143                 case -NFS4ERR_ADMIN_REVOKED:
2144                 case -NFS4ERR_EXPIRED:
2145                 case -NFS4ERR_BAD_STATEID:
2146                 case -NFS4ERR_OPENMODE:
2147                         nfs_inode_find_state_and_recover(state->inode,
2148                                         stateid);
2149                         nfs4_schedule_stateid_recovery(server, state);
2150                         return -EAGAIN;
2151                 case -NFS4ERR_DELAY:
2152                 case -NFS4ERR_GRACE:
2153                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2154                         ssleep(1);
2155                         return -EAGAIN;
2156                 case -ENOMEM:
2157                 case -NFS4ERR_DENIED:
2158                         if (fl) {
2159                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2160                                 if (lsp)
2161                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2162                         }
2163                         return 0;
2164         }
2165         return err;
2166 }
2167
2168 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2169                 struct nfs4_state *state, const nfs4_stateid *stateid,
2170                 fmode_t type)
2171 {
2172         struct nfs_server *server = NFS_SERVER(state->inode);
2173         struct nfs4_opendata *opendata;
2174         int err = 0;
2175
2176         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2177                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2178         if (IS_ERR(opendata))
2179                 return PTR_ERR(opendata);
2180         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2181         nfs_state_clear_delegation(state);
2182         switch (type & (FMODE_READ|FMODE_WRITE)) {
2183         case FMODE_READ|FMODE_WRITE:
2184         case FMODE_WRITE:
2185                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2186                 if (err)
2187                         break;
2188                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2189                 if (err)
2190                         break;
2191                 /* Fall through */
2192         case FMODE_READ:
2193                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2194         }
2195         nfs4_opendata_put(opendata);
2196         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2197 }
2198
2199 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2200 {
2201         struct nfs4_opendata *data = calldata;
2202
2203         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2204                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2205 }
2206
2207 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2208 {
2209         struct nfs4_opendata *data = calldata;
2210
2211         nfs40_sequence_done(task, &data->c_res.seq_res);
2212
2213         data->rpc_status = task->tk_status;
2214         if (data->rpc_status == 0) {
2215                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2216                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2217                 renew_lease(data->o_res.server, data->timestamp);
2218                 data->rpc_done = true;
2219         }
2220 }
2221
2222 static void nfs4_open_confirm_release(void *calldata)
2223 {
2224         struct nfs4_opendata *data = calldata;
2225         struct nfs4_state *state = NULL;
2226
2227         /* If this request hasn't been cancelled, do nothing */
2228         if (!data->cancelled)
2229                 goto out_free;
2230         /* In case of error, no cleanup! */
2231         if (!data->rpc_done)
2232                 goto out_free;
2233         state = nfs4_opendata_to_nfs4_state(data);
2234         if (!IS_ERR(state))
2235                 nfs4_close_state(state, data->o_arg.fmode);
2236 out_free:
2237         nfs4_opendata_put(data);
2238 }
2239
2240 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2241         .rpc_call_prepare = nfs4_open_confirm_prepare,
2242         .rpc_call_done = nfs4_open_confirm_done,
2243         .rpc_release = nfs4_open_confirm_release,
2244 };
2245
2246 /*
2247  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2248  */
2249 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2250 {
2251         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2252         struct rpc_task *task;
2253         struct  rpc_message msg = {
2254                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2255                 .rpc_argp = &data->c_arg,
2256                 .rpc_resp = &data->c_res,
2257                 .rpc_cred = data->owner->so_cred,
2258         };
2259         struct rpc_task_setup task_setup_data = {
2260                 .rpc_client = server->client,
2261                 .rpc_message = &msg,
2262                 .callback_ops = &nfs4_open_confirm_ops,
2263                 .callback_data = data,
2264                 .workqueue = nfsiod_workqueue,
2265                 .flags = RPC_TASK_ASYNC,
2266         };
2267         int status;
2268
2269         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2270                                 data->is_recover);
2271         kref_get(&data->kref);
2272         data->rpc_done = false;
2273         data->rpc_status = 0;
2274         data->timestamp = jiffies;
2275         task = rpc_run_task(&task_setup_data);
2276         if (IS_ERR(task))
2277                 return PTR_ERR(task);
2278         status = rpc_wait_for_completion_task(task);
2279         if (status != 0) {
2280                 data->cancelled = true;
2281                 smp_wmb();
2282         } else
2283                 status = data->rpc_status;
2284         rpc_put_task(task);
2285         return status;
2286 }
2287
2288 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2289 {
2290         struct nfs4_opendata *data = calldata;
2291         struct nfs4_state_owner *sp = data->owner;
2292         struct nfs_client *clp = sp->so_server->nfs_client;
2293         enum open_claim_type4 claim = data->o_arg.claim;
2294
2295         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2296                 goto out_wait;
2297         /*
2298          * Check if we still need to send an OPEN call, or if we can use
2299          * a delegation instead.
2300          */
2301         if (data->state != NULL) {
2302                 struct nfs_delegation *delegation;
2303
2304                 if (can_open_cached(data->state, data->o_arg.fmode,
2305                                         data->o_arg.open_flags, claim))
2306                         goto out_no_action;
2307                 rcu_read_lock();
2308                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2309                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2310                         goto unlock_no_action;
2311                 rcu_read_unlock();
2312         }
2313         /* Update client id. */
2314         data->o_arg.clientid = clp->cl_clientid;
2315         switch (claim) {
2316         default:
2317                 break;
2318         case NFS4_OPEN_CLAIM_PREVIOUS:
2319         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2320         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2321                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2322                 /* Fall through */
2323         case NFS4_OPEN_CLAIM_FH:
2324                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2325         }
2326         data->timestamp = jiffies;
2327         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2328                                 &data->o_arg.seq_args,
2329                                 &data->o_res.seq_res,
2330                                 task) != 0)
2331                 nfs_release_seqid(data->o_arg.seqid);
2332
2333         /* Set the create mode (note dependency on the session type) */
2334         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2335         if (data->o_arg.open_flags & O_EXCL) {
2336                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2337                 if (nfs4_has_persistent_session(clp))
2338                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2339                 else if (clp->cl_mvops->minor_version > 0)
2340                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2341         }
2342         return;
2343 unlock_no_action:
2344         trace_nfs4_cached_open(data->state);
2345         rcu_read_unlock();
2346 out_no_action:
2347         task->tk_action = NULL;
2348 out_wait:
2349         nfs4_sequence_done(task, &data->o_res.seq_res);
2350 }
2351
2352 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2353 {
2354         struct nfs4_opendata *data = calldata;
2355
2356         data->rpc_status = task->tk_status;
2357
2358         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2359                 return;
2360
2361         if (task->tk_status == 0) {
2362                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2363                         switch (data->o_res.f_attr->mode & S_IFMT) {
2364                         case S_IFREG:
2365                                 break;
2366                         case S_IFLNK:
2367                                 data->rpc_status = -ELOOP;
2368                                 break;
2369                         case S_IFDIR:
2370                                 data->rpc_status = -EISDIR;
2371                                 break;
2372                         default:
2373                                 data->rpc_status = -ENOTDIR;
2374                         }
2375                 }
2376                 renew_lease(data->o_res.server, data->timestamp);
2377                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2378                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2379         }
2380         data->rpc_done = true;
2381 }
2382
2383 static void nfs4_open_release(void *calldata)
2384 {
2385         struct nfs4_opendata *data = calldata;
2386         struct nfs4_state *state = NULL;
2387
2388         /* If this request hasn't been cancelled, do nothing */
2389         if (!data->cancelled)
2390                 goto out_free;
2391         /* In case of error, no cleanup! */
2392         if (data->rpc_status != 0 || !data->rpc_done)
2393                 goto out_free;
2394         /* In case we need an open_confirm, no cleanup! */
2395         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2396                 goto out_free;
2397         state = nfs4_opendata_to_nfs4_state(data);
2398         if (!IS_ERR(state))
2399                 nfs4_close_state(state, data->o_arg.fmode);
2400 out_free:
2401         nfs4_opendata_put(data);
2402 }
2403
2404 static const struct rpc_call_ops nfs4_open_ops = {
2405         .rpc_call_prepare = nfs4_open_prepare,
2406         .rpc_call_done = nfs4_open_done,
2407         .rpc_release = nfs4_open_release,
2408 };
2409
2410 static int nfs4_run_open_task(struct nfs4_opendata *data,
2411                               struct nfs_open_context *ctx)
2412 {
2413         struct inode *dir = d_inode(data->dir);
2414         struct nfs_server *server = NFS_SERVER(dir);
2415         struct nfs_openargs *o_arg = &data->o_arg;
2416         struct nfs_openres *o_res = &data->o_res;
2417         struct rpc_task *task;
2418         struct rpc_message msg = {
2419                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2420                 .rpc_argp = o_arg,
2421                 .rpc_resp = o_res,
2422                 .rpc_cred = data->owner->so_cred,
2423         };
2424         struct rpc_task_setup task_setup_data = {
2425                 .rpc_client = server->client,
2426                 .rpc_message = &msg,
2427                 .callback_ops = &nfs4_open_ops,
2428                 .callback_data = data,
2429                 .workqueue = nfsiod_workqueue,
2430                 .flags = RPC_TASK_ASYNC,
2431         };
2432         int status;
2433
2434         kref_get(&data->kref);
2435         data->rpc_done = false;
2436         data->rpc_status = 0;
2437         data->cancelled = false;
2438         data->is_recover = false;
2439         if (!ctx) {
2440                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2441                 data->is_recover = true;
2442         } else {
2443                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2444                 pnfs_lgopen_prepare(data, ctx);
2445         }
2446         task = rpc_run_task(&task_setup_data);
2447         if (IS_ERR(task))
2448                 return PTR_ERR(task);
2449         status = rpc_wait_for_completion_task(task);
2450         if (status != 0) {
2451                 data->cancelled = true;
2452                 smp_wmb();
2453         } else
2454                 status = data->rpc_status;
2455         rpc_put_task(task);
2456
2457         return status;
2458 }
2459
2460 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2461 {
2462         struct inode *dir = d_inode(data->dir);
2463         struct nfs_openres *o_res = &data->o_res;
2464         int status;
2465
2466         status = nfs4_run_open_task(data, NULL);
2467         if (status != 0 || !data->rpc_done)
2468                 return status;
2469
2470         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2471
2472         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2473                 status = _nfs4_proc_open_confirm(data);
2474
2475         return status;
2476 }
2477
2478 /*
2479  * Additional permission checks in order to distinguish between an
2480  * open for read, and an open for execute. This works around the
2481  * fact that NFSv4 OPEN treats read and execute permissions as being
2482  * the same.
2483  * Note that in the non-execute case, we want to turn off permission
2484  * checking if we just created a new file (POSIX open() semantics).
2485  */
2486 static int nfs4_opendata_access(const struct cred *cred,
2487                                 struct nfs4_opendata *opendata,
2488                                 struct nfs4_state *state, fmode_t fmode,
2489                                 int openflags)
2490 {
2491         struct nfs_access_entry cache;
2492         u32 mask, flags;
2493
2494         /* access call failed or for some reason the server doesn't
2495          * support any access modes -- defer access call until later */
2496         if (opendata->o_res.access_supported == 0)
2497                 return 0;
2498
2499         mask = 0;
2500         /*
2501          * Use openflags to check for exec, because fmode won't
2502          * always have FMODE_EXEC set when file open for exec.
2503          */
2504         if (openflags & __FMODE_EXEC) {
2505                 /* ONLY check for exec rights */
2506                 if (S_ISDIR(state->inode->i_mode))
2507                         mask = NFS4_ACCESS_LOOKUP;
2508                 else
2509                         mask = NFS4_ACCESS_EXECUTE;
2510         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2511                 mask = NFS4_ACCESS_READ;
2512
2513         cache.cred = cred;
2514         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2515         nfs_access_add_cache(state->inode, &cache);
2516
2517         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2518         if ((mask & ~cache.mask & flags) == 0)
2519                 return 0;
2520
2521         return -EACCES;
2522 }
2523
2524 /*
2525  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2526  */
2527 static int _nfs4_proc_open(struct nfs4_opendata *data,
2528                            struct nfs_open_context *ctx)
2529 {
2530         struct inode *dir = d_inode(data->dir);
2531         struct nfs_server *server = NFS_SERVER(dir);
2532         struct nfs_openargs *o_arg = &data->o_arg;
2533         struct nfs_openres *o_res = &data->o_res;
2534         int status;
2535
2536         status = nfs4_run_open_task(data, ctx);
2537         if (!data->rpc_done)
2538                 return status;
2539         if (status != 0) {
2540                 if (status == -NFS4ERR_BADNAME &&
2541                                 !(o_arg->open_flags & O_CREAT))
2542                         return -ENOENT;
2543                 return status;
2544         }
2545
2546         nfs_fattr_map_and_free_names(server, &data->f_attr);
2547
2548         if (o_arg->open_flags & O_CREAT) {
2549                 if (o_arg->open_flags & O_EXCL)
2550                         data->file_created = true;
2551                 else if (o_res->cinfo.before != o_res->cinfo.after)
2552                         data->file_created = true;
2553                 if (data->file_created ||
2554                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2555                         update_changeattr(dir, &o_res->cinfo,
2556                                         o_res->f_attr->time_start, 0);
2557         }
2558         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2559                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2560         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2561                 status = _nfs4_proc_open_confirm(data);
2562                 if (status != 0)
2563                         return status;
2564         }
2565         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2566                 nfs4_sequence_free_slot(&o_res->seq_res);
2567                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2568                                 o_res->f_label, NULL);
2569         }
2570         return 0;
2571 }
2572
2573 /*
2574  * OPEN_EXPIRED:
2575  *      reclaim state on the server after a network partition.
2576  *      Assumes caller holds the appropriate lock
2577  */
2578 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2579 {
2580         struct nfs4_opendata *opendata;
2581         int ret;
2582
2583         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2584                         NFS4_OPEN_CLAIM_FH);
2585         if (IS_ERR(opendata))
2586                 return PTR_ERR(opendata);
2587         ret = nfs4_open_recover(opendata, state);
2588         if (ret == -ESTALE)
2589                 d_drop(ctx->dentry);
2590         nfs4_opendata_put(opendata);
2591         return ret;
2592 }
2593
2594 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2595 {
2596         struct nfs_server *server = NFS_SERVER(state->inode);
2597         struct nfs4_exception exception = { };
2598         int err;
2599
2600         do {
2601                 err = _nfs4_open_expired(ctx, state);
2602                 trace_nfs4_open_expired(ctx, 0, err);
2603                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2604                         continue;
2605                 switch (err) {
2606                 default:
2607                         goto out;
2608                 case -NFS4ERR_GRACE:
2609                 case -NFS4ERR_DELAY:
2610                         nfs4_handle_exception(server, err, &exception);
2611                         err = 0;
2612                 }
2613         } while (exception.retry);
2614 out:
2615         return err;
2616 }
2617
2618 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2619 {
2620         struct nfs_open_context *ctx;
2621         int ret;
2622
2623         ctx = nfs4_state_find_open_context(state);
2624         if (IS_ERR(ctx))
2625                 return -EAGAIN;
2626         ret = nfs4_do_open_expired(ctx, state);
2627         put_nfs_open_context(ctx);
2628         return ret;
2629 }
2630
2631 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2632                 const nfs4_stateid *stateid)
2633 {
2634         nfs_remove_bad_delegation(state->inode, stateid);
2635         nfs_state_clear_delegation(state);
2636 }
2637
2638 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2639 {
2640         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2641                 nfs_finish_clear_delegation_stateid(state, NULL);
2642 }
2643
2644 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2645 {
2646         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2647         nfs40_clear_delegation_stateid(state);
2648         return nfs4_open_expired(sp, state);
2649 }
2650
2651 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2652                 nfs4_stateid *stateid,
2653                 const struct cred *cred)
2654 {
2655         return -NFS4ERR_BAD_STATEID;
2656 }
2657
2658 #if defined(CONFIG_NFS_V4_1)
2659 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2660                 nfs4_stateid *stateid,
2661                 const struct cred *cred)
2662 {
2663         int status;
2664
2665         switch (stateid->type) {
2666         default:
2667                 break;
2668         case NFS4_INVALID_STATEID_TYPE:
2669         case NFS4_SPECIAL_STATEID_TYPE:
2670                 return -NFS4ERR_BAD_STATEID;
2671         case NFS4_REVOKED_STATEID_TYPE:
2672                 goto out_free;
2673         }
2674
2675         status = nfs41_test_stateid(server, stateid, cred);
2676         switch (status) {
2677         case -NFS4ERR_EXPIRED:
2678         case -NFS4ERR_ADMIN_REVOKED:
2679         case -NFS4ERR_DELEG_REVOKED:
2680                 break;
2681         default:
2682                 return status;
2683         }
2684 out_free:
2685         /* Ack the revoked state to the server */
2686         nfs41_free_stateid(server, stateid, cred, true);
2687         return -NFS4ERR_EXPIRED;
2688 }
2689
2690 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2691 {
2692         struct nfs_server *server = NFS_SERVER(state->inode);
2693         nfs4_stateid stateid;
2694         struct nfs_delegation *delegation;
2695         const struct cred *cred = NULL;
2696         int status;
2697
2698         /* Get the delegation credential for use by test/free_stateid */
2699         rcu_read_lock();
2700         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2701         if (delegation == NULL) {
2702                 rcu_read_unlock();
2703                 nfs_state_clear_delegation(state);
2704                 return;
2705         }
2706
2707         nfs4_stateid_copy(&stateid, &delegation->stateid);
2708         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2709                 rcu_read_unlock();
2710                 nfs_state_clear_delegation(state);
2711                 return;
2712         }
2713
2714         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2715                                 &delegation->flags)) {
2716                 rcu_read_unlock();
2717                 return;
2718         }
2719
2720         if (delegation->cred)
2721                 cred = get_cred(delegation->cred);
2722         rcu_read_unlock();
2723         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2724         trace_nfs4_test_delegation_stateid(state, NULL, status);
2725         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2726                 nfs_finish_clear_delegation_stateid(state, &stateid);
2727
2728         if (delegation->cred)
2729                 put_cred(cred);
2730 }
2731
2732 /**
2733  * nfs41_check_expired_locks - possibly free a lock stateid
2734  *
2735  * @state: NFSv4 state for an inode
2736  *
2737  * Returns NFS_OK if recovery for this stateid is now finished.
2738  * Otherwise a negative NFS4ERR value is returned.
2739  */
2740 static int nfs41_check_expired_locks(struct nfs4_state *state)
2741 {
2742         int status, ret = NFS_OK;
2743         struct nfs4_lock_state *lsp, *prev = NULL;
2744         struct nfs_server *server = NFS_SERVER(state->inode);
2745
2746         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2747                 goto out;
2748
2749         spin_lock(&state->state_lock);
2750         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2751                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2752                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2753
2754                         refcount_inc(&lsp->ls_count);
2755                         spin_unlock(&state->state_lock);
2756
2757                         nfs4_put_lock_state(prev);
2758                         prev = lsp;
2759
2760                         status = nfs41_test_and_free_expired_stateid(server,
2761                                         &lsp->ls_stateid,
2762                                         cred);
2763                         trace_nfs4_test_lock_stateid(state, lsp, status);
2764                         if (status == -NFS4ERR_EXPIRED ||
2765                             status == -NFS4ERR_BAD_STATEID) {
2766                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2767                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2768                                 if (!recover_lost_locks)
2769                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2770                         } else if (status != NFS_OK) {
2771                                 ret = status;
2772                                 nfs4_put_lock_state(prev);
2773                                 goto out;
2774                         }
2775                         spin_lock(&state->state_lock);
2776                 }
2777         }
2778         spin_unlock(&state->state_lock);
2779         nfs4_put_lock_state(prev);
2780 out:
2781         return ret;
2782 }
2783
2784 /**
2785  * nfs41_check_open_stateid - possibly free an open stateid
2786  *
2787  * @state: NFSv4 state for an inode
2788  *
2789  * Returns NFS_OK if recovery for this stateid is now finished.
2790  * Otherwise a negative NFS4ERR value is returned.
2791  */
2792 static int nfs41_check_open_stateid(struct nfs4_state *state)
2793 {
2794         struct nfs_server *server = NFS_SERVER(state->inode);
2795         nfs4_stateid *stateid = &state->open_stateid;
2796         const struct cred *cred = state->owner->so_cred;
2797         int status;
2798
2799         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2800                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2801                         if (nfs4_have_delegation(state->inode, state->state))
2802                                 return NFS_OK;
2803                         return -NFS4ERR_OPENMODE;
2804                 }
2805                 return -NFS4ERR_BAD_STATEID;
2806         }
2807         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2808         trace_nfs4_test_open_stateid(state, NULL, status);
2809         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2810                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2811                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2812                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2813                 clear_bit(NFS_OPEN_STATE, &state->flags);
2814                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2815                 return status;
2816         }
2817         if (nfs_open_stateid_recover_openmode(state))
2818                 return -NFS4ERR_OPENMODE;
2819         return NFS_OK;
2820 }
2821
2822 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2823 {
2824         int status;
2825
2826         nfs41_check_delegation_stateid(state);
2827         status = nfs41_check_expired_locks(state);
2828         if (status != NFS_OK)
2829                 return status;
2830         status = nfs41_check_open_stateid(state);
2831         if (status != NFS_OK)
2832                 status = nfs4_open_expired(sp, state);
2833         return status;
2834 }
2835 #endif
2836
2837 /*
2838  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2839  * fields corresponding to attributes that were used to store the verifier.
2840  * Make sure we clobber those fields in the later setattr call
2841  */
2842 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2843                                 struct iattr *sattr, struct nfs4_label **label)
2844 {
2845         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2846         __u32 attrset[3];
2847         unsigned ret;
2848         unsigned i;
2849
2850         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2851                 attrset[i] = opendata->o_res.attrset[i];
2852                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2853                         attrset[i] &= ~bitmask[i];
2854         }
2855
2856         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2857                 sattr->ia_valid : 0;
2858
2859         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2860                 if (sattr->ia_valid & ATTR_ATIME_SET)
2861                         ret |= ATTR_ATIME_SET;
2862                 else
2863                         ret |= ATTR_ATIME;
2864         }
2865
2866         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2867                 if (sattr->ia_valid & ATTR_MTIME_SET)
2868                         ret |= ATTR_MTIME_SET;
2869                 else
2870                         ret |= ATTR_MTIME;
2871         }
2872
2873         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2874                 *label = NULL;
2875         return ret;
2876 }
2877
2878 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2879                 fmode_t fmode,
2880                 int flags,
2881                 struct nfs_open_context *ctx)
2882 {
2883         struct nfs4_state_owner *sp = opendata->owner;
2884         struct nfs_server *server = sp->so_server;
2885         struct dentry *dentry;
2886         struct nfs4_state *state;
2887         unsigned int seq;
2888         int ret;
2889
2890         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2891
2892         ret = _nfs4_proc_open(opendata, ctx);
2893         if (ret != 0)
2894                 goto out;
2895
2896         state = _nfs4_opendata_to_nfs4_state(opendata);
2897         ret = PTR_ERR(state);
2898         if (IS_ERR(state))
2899                 goto out;
2900         ctx->state = state;
2901         if (server->caps & NFS_CAP_POSIX_LOCK)
2902                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2903         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2904                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2905
2906         dentry = opendata->dentry;
2907         if (d_really_is_negative(dentry)) {
2908                 struct dentry *alias;
2909                 d_drop(dentry);
2910                 alias = d_exact_alias(dentry, state->inode);
2911                 if (!alias)
2912                         alias = d_splice_alias(igrab(state->inode), dentry);
2913                 /* d_splice_alias() can't fail here - it's a non-directory */
2914                 if (alias) {
2915                         dput(ctx->dentry);
2916                         ctx->dentry = dentry = alias;
2917                 }
2918                 nfs_set_verifier(dentry,
2919                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2920         }
2921
2922         /* Parse layoutget results before we check for access */
2923         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
2924
2925         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2926         if (ret != 0)
2927                 goto out;
2928
2929         if (d_inode(dentry) == state->inode) {
2930                 nfs_inode_attach_open_context(ctx);
2931                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2932                         nfs4_schedule_stateid_recovery(server, state);
2933         }
2934
2935 out:
2936         nfs4_sequence_free_slot(&opendata->o_res.seq_res);
2937         return ret;
2938 }
2939
2940 /*
2941  * Returns a referenced nfs4_state
2942  */
2943 static int _nfs4_do_open(struct inode *dir,
2944                         struct nfs_open_context *ctx,
2945                         int flags,
2946                         const struct nfs4_open_createattrs *c,
2947                         int *opened)
2948 {
2949         struct nfs4_state_owner  *sp;
2950         struct nfs4_state     *state = NULL;
2951         struct nfs_server       *server = NFS_SERVER(dir);
2952         struct nfs4_opendata *opendata;
2953         struct dentry *dentry = ctx->dentry;
2954         const struct cred *cred = ctx->cred;
2955         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2956         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2957         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2958         struct iattr *sattr = c->sattr;
2959         struct nfs4_label *label = c->label;
2960         struct nfs4_label *olabel = NULL;
2961         int status;
2962
2963         /* Protect against reboot recovery conflicts */
2964         status = -ENOMEM;
2965         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2966         if (sp == NULL) {
2967                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2968                 goto out_err;
2969         }
2970         status = nfs4_client_recover_expired_lease(server->nfs_client);
2971         if (status != 0)
2972                 goto err_put_state_owner;
2973         if (d_really_is_positive(dentry))
2974                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2975         status = -ENOMEM;
2976         if (d_really_is_positive(dentry))
2977                 claim = NFS4_OPEN_CLAIM_FH;
2978         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
2979                         c, claim, GFP_KERNEL);
2980         if (opendata == NULL)
2981                 goto err_put_state_owner;
2982
2983         if (label) {
2984                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2985                 if (IS_ERR(olabel)) {
2986                         status = PTR_ERR(olabel);
2987                         goto err_opendata_put;
2988                 }
2989         }
2990
2991         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2992                 if (!opendata->f_attr.mdsthreshold) {
2993                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2994                         if (!opendata->f_attr.mdsthreshold)
2995                                 goto err_free_label;
2996                 }
2997                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2998         }
2999         if (d_really_is_positive(dentry))
3000                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3001
3002         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
3003         if (status != 0)
3004                 goto err_free_label;
3005         state = ctx->state;
3006
3007         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3008             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3009                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3010                 /*
3011                  * send create attributes which was not set by open
3012                  * with an extra setattr.
3013                  */
3014                 if (attrs || label) {
3015                         unsigned ia_old = sattr->ia_valid;
3016
3017                         sattr->ia_valid = attrs;
3018                         nfs_fattr_init(opendata->o_res.f_attr);
3019                         status = nfs4_do_setattr(state->inode, cred,
3020                                         opendata->o_res.f_attr, sattr,
3021                                         ctx, label, olabel);
3022                         if (status == 0) {
3023                                 nfs_setattr_update_inode(state->inode, sattr,
3024                                                 opendata->o_res.f_attr);
3025                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3026                         }
3027                         sattr->ia_valid = ia_old;
3028                 }
3029         }
3030         if (opened && opendata->file_created)
3031                 *opened = 1;
3032
3033         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3034                 *ctx_th = opendata->f_attr.mdsthreshold;
3035                 opendata->f_attr.mdsthreshold = NULL;
3036         }
3037
3038         nfs4_label_free(olabel);
3039
3040         nfs4_opendata_put(opendata);
3041         nfs4_put_state_owner(sp);
3042         return 0;
3043 err_free_label:
3044         nfs4_label_free(olabel);
3045 err_opendata_put:
3046         nfs4_opendata_put(opendata);
3047 err_put_state_owner:
3048         nfs4_put_state_owner(sp);
3049 out_err:
3050         return status;
3051 }
3052
3053
3054 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3055                                         struct nfs_open_context *ctx,
3056                                         int flags,
3057                                         struct iattr *sattr,
3058                                         struct nfs4_label *label,
3059                                         int *opened)
3060 {
3061         struct nfs_server *server = NFS_SERVER(dir);
3062         struct nfs4_exception exception = { };
3063         struct nfs4_state *res;
3064         struct nfs4_open_createattrs c = {
3065                 .label = label,
3066                 .sattr = sattr,
3067                 .verf = {
3068                         [0] = (__u32)jiffies,
3069                         [1] = (__u32)current->pid,
3070                 },
3071         };
3072         int status;
3073
3074         do {
3075                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3076                 res = ctx->state;
3077                 trace_nfs4_open_file(ctx, flags, status);
3078                 if (status == 0)
3079                         break;
3080                 /* NOTE: BAD_SEQID means the server and client disagree about the
3081                  * book-keeping w.r.t. state-changing operations
3082                  * (OPEN/CLOSE/LOCK/LOCKU...)
3083                  * It is actually a sign of a bug on the client or on the server.
3084                  *
3085                  * If we receive a BAD_SEQID error in the particular case of
3086                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3087                  * have unhashed the old state_owner for us, and that we can
3088                  * therefore safely retry using a new one. We should still warn
3089                  * the user though...
3090                  */
3091                 if (status == -NFS4ERR_BAD_SEQID) {
3092                         pr_warn_ratelimited("NFS: v4 server %s "
3093                                         " returned a bad sequence-id error!\n",
3094                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3095                         exception.retry = 1;
3096                         continue;
3097                 }
3098                 /*
3099                  * BAD_STATEID on OPEN means that the server cancelled our
3100                  * state before it received the OPEN_CONFIRM.
3101                  * Recover by retrying the request as per the discussion
3102                  * on Page 181 of RFC3530.
3103                  */
3104                 if (status == -NFS4ERR_BAD_STATEID) {
3105                         exception.retry = 1;
3106                         continue;
3107                 }
3108                 if (status == -EAGAIN) {
3109                         /* We must have found a delegation */
3110                         exception.retry = 1;
3111                         continue;
3112                 }
3113                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3114                         continue;
3115                 res = ERR_PTR(nfs4_handle_exception(server,
3116                                         status, &exception));
3117         } while (exception.retry);
3118         return res;
3119 }
3120
3121 static int _nfs4_do_setattr(struct inode *inode,
3122                             struct nfs_setattrargs *arg,
3123                             struct nfs_setattrres *res,
3124                             const struct cred *cred,
3125                             struct nfs_open_context *ctx)
3126 {
3127         struct nfs_server *server = NFS_SERVER(inode);
3128         struct rpc_message msg = {
3129                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3130                 .rpc_argp       = arg,
3131                 .rpc_resp       = res,
3132                 .rpc_cred       = cred,
3133         };
3134         const struct cred *delegation_cred = NULL;
3135         unsigned long timestamp = jiffies;
3136         bool truncate;
3137         int status;
3138
3139         nfs_fattr_init(res->fattr);
3140
3141         /* Servers should only apply open mode checks for file size changes */
3142         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3143         if (!truncate)
3144                 goto zero_stateid;
3145
3146         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3147                 /* Use that stateid */
3148         } else if (ctx != NULL) {
3149                 struct nfs_lock_context *l_ctx;
3150                 if (!nfs4_valid_open_stateid(ctx->state))
3151                         return -EBADF;
3152                 l_ctx = nfs_get_lock_context(ctx);
3153                 if (IS_ERR(l_ctx))
3154                         return PTR_ERR(l_ctx);
3155                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3156                                                 &arg->stateid, &delegation_cred);
3157                 nfs_put_lock_context(l_ctx);
3158                 if (status == -EIO)
3159                         return -EBADF;
3160         } else {
3161 zero_stateid:
3162                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3163         }
3164         if (delegation_cred)
3165                 msg.rpc_cred = delegation_cred;
3166
3167         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3168
3169         put_cred(delegation_cred);
3170         if (status == 0 && ctx != NULL)
3171                 renew_lease(server, timestamp);
3172         trace_nfs4_setattr(inode, &arg->stateid, status);
3173         return status;
3174 }
3175
3176 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3177                            struct nfs_fattr *fattr, struct iattr *sattr,
3178                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3179                            struct nfs4_label *olabel)
3180 {
3181         struct nfs_server *server = NFS_SERVER(inode);
3182         __u32 bitmask[NFS4_BITMASK_SZ];
3183         struct nfs4_state *state = ctx ? ctx->state : NULL;
3184         struct nfs_setattrargs  arg = {
3185                 .fh             = NFS_FH(inode),
3186                 .iap            = sattr,
3187                 .server         = server,
3188                 .bitmask = bitmask,
3189                 .label          = ilabel,
3190         };
3191         struct nfs_setattrres  res = {
3192                 .fattr          = fattr,
3193                 .label          = olabel,
3194                 .server         = server,
3195         };
3196         struct nfs4_exception exception = {
3197                 .state = state,
3198                 .inode = inode,
3199                 .stateid = &arg.stateid,
3200         };
3201         int err;
3202
3203         do {
3204                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3205                                 nfs4_bitmask(server, olabel),
3206                                 inode);
3207
3208                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3209                 switch (err) {
3210                 case -NFS4ERR_OPENMODE:
3211                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3212                                 pr_warn_once("NFSv4: server %s is incorrectly "
3213                                                 "applying open mode checks to "
3214                                                 "a SETATTR that is not "
3215                                                 "changing file size.\n",
3216                                                 server->nfs_client->cl_hostname);
3217                         }
3218                         if (state && !(state->state & FMODE_WRITE)) {
3219                                 err = -EBADF;
3220                                 if (sattr->ia_valid & ATTR_OPEN)
3221                                         err = -EACCES;
3222                                 goto out;
3223                         }
3224                 }
3225                 err = nfs4_handle_exception(server, err, &exception);
3226         } while (exception.retry);
3227 out:
3228         return err;
3229 }
3230
3231 static bool
3232 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3233 {
3234         if (inode == NULL || !nfs_have_layout(inode))
3235                 return false;
3236
3237         return pnfs_wait_on_layoutreturn(inode, task);
3238 }
3239
3240 struct nfs4_closedata {
3241         struct inode *inode;
3242         struct nfs4_state *state;
3243         struct nfs_closeargs arg;
3244         struct nfs_closeres res;
3245         struct {
3246                 struct nfs4_layoutreturn_args arg;
3247                 struct nfs4_layoutreturn_res res;
3248                 struct nfs4_xdr_opaque_data ld_private;
3249                 u32 roc_barrier;
3250                 bool roc;
3251         } lr;
3252         struct nfs_fattr fattr;
3253         unsigned long timestamp;
3254 };
3255
3256 static void nfs4_free_closedata(void *data)
3257 {
3258         struct nfs4_closedata *calldata = data;
3259         struct nfs4_state_owner *sp = calldata->state->owner;
3260         struct super_block *sb = calldata->state->inode->i_sb;
3261
3262         if (calldata->lr.roc)
3263                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3264                                 calldata->res.lr_ret);
3265         nfs4_put_open_state(calldata->state);
3266         nfs_free_seqid(calldata->arg.seqid);
3267         nfs4_put_state_owner(sp);
3268         nfs_sb_deactive(sb);
3269         kfree(calldata);
3270 }
3271
3272 static void nfs4_close_done(struct rpc_task *task, void *data)
3273 {
3274         struct nfs4_closedata *calldata = data;
3275         struct nfs4_state *state = calldata->state;
3276         struct nfs_server *server = NFS_SERVER(calldata->inode);
3277         nfs4_stateid *res_stateid = NULL;
3278         struct nfs4_exception exception = {
3279                 .state = state,
3280                 .inode = calldata->inode,
3281                 .stateid = &calldata->arg.stateid,
3282         };
3283
3284         dprintk("%s: begin!\n", __func__);
3285         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3286                 return;
3287         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3288
3289         /* Handle Layoutreturn errors */
3290         if (calldata->arg.lr_args && task->tk_status != 0) {
3291                 switch (calldata->res.lr_ret) {
3292                 default:
3293                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3294                         break;
3295                 case 0:
3296                         calldata->arg.lr_args = NULL;
3297                         calldata->res.lr_res = NULL;
3298                         break;
3299                 case -NFS4ERR_OLD_STATEID:
3300                         if (nfs4_layoutreturn_refresh_stateid(&calldata->arg.lr_args->stateid,
3301                                                 &calldata->arg.lr_args->range,
3302                                                 calldata->inode))
3303                                 goto lr_restart;
3304                         /* Fallthrough */
3305                 case -NFS4ERR_ADMIN_REVOKED:
3306                 case -NFS4ERR_DELEG_REVOKED:
3307                 case -NFS4ERR_EXPIRED:
3308                 case -NFS4ERR_BAD_STATEID:
3309                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3310                 case -NFS4ERR_WRONG_CRED:
3311                         calldata->arg.lr_args = NULL;
3312                         calldata->res.lr_res = NULL;
3313                         goto lr_restart;
3314                 }
3315         }
3316
3317         /* hmm. we are done with the inode, and in the process of freeing
3318          * the state_owner. we keep this around to process errors
3319          */
3320         switch (task->tk_status) {
3321                 case 0:
3322                         res_stateid = &calldata->res.stateid;
3323                         renew_lease(server, calldata->timestamp);
3324                         break;
3325                 case -NFS4ERR_ACCESS:
3326                         if (calldata->arg.bitmask != NULL) {
3327                                 calldata->arg.bitmask = NULL;
3328                                 calldata->res.fattr = NULL;
3329                                 goto out_restart;
3330
3331                         }
3332                         break;
3333                 case -NFS4ERR_OLD_STATEID:
3334                         /* Did we race with OPEN? */
3335                         if (nfs4_refresh_open_stateid(&calldata->arg.stateid,
3336                                                 state))
3337                                 goto out_restart;
3338                         goto out_release;
3339                 case -NFS4ERR_ADMIN_REVOKED:
3340                 case -NFS4ERR_STALE_STATEID:
3341                 case -NFS4ERR_EXPIRED:
3342                         nfs4_free_revoked_stateid(server,
3343                                         &calldata->arg.stateid,
3344                                         task->tk_msg.rpc_cred);
3345                         /* Fallthrough */
3346                 case -NFS4ERR_BAD_STATEID:
3347                         break;
3348                 default:
3349                         task->tk_status = nfs4_async_handle_exception(task,
3350                                         server, task->tk_status, &exception);
3351                         if (exception.retry)
3352                                 goto out_restart;
3353         }
3354         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3355                         res_stateid, calldata->arg.fmode);
3356 out_release:
3357         task->tk_status = 0;
3358         nfs_release_seqid(calldata->arg.seqid);
3359         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3360         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3361         return;
3362 lr_restart:
3363         calldata->res.lr_ret = 0;
3364 out_restart:
3365         task->tk_status = 0;
3366         rpc_restart_call_prepare(task);
3367         goto out_release;
3368 }
3369
3370 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3371 {
3372         struct nfs4_closedata *calldata = data;
3373         struct nfs4_state *state = calldata->state;
3374         struct inode *inode = calldata->inode;
3375         struct pnfs_layout_hdr *lo;
3376         bool is_rdonly, is_wronly, is_rdwr;
3377         int call_close = 0;
3378
3379         dprintk("%s: begin!\n", __func__);
3380         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3381                 goto out_wait;
3382
3383         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3384         spin_lock(&state->owner->so_lock);
3385         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3386         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3387         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3388         /* Calculate the change in open mode */
3389         calldata->arg.fmode = 0;
3390         if (state->n_rdwr == 0) {
3391                 if (state->n_rdonly == 0)
3392                         call_close |= is_rdonly;
3393                 else if (is_rdonly)
3394                         calldata->arg.fmode |= FMODE_READ;
3395                 if (state->n_wronly == 0)
3396                         call_close |= is_wronly;
3397                 else if (is_wronly)
3398                         calldata->arg.fmode |= FMODE_WRITE;
3399                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3400                         call_close |= is_rdwr;
3401         } else if (is_rdwr)
3402                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3403
3404         if (!nfs4_valid_open_stateid(state) ||
3405             !nfs4_refresh_open_stateid(&calldata->arg.stateid, state))
3406                 call_close = 0;
3407         spin_unlock(&state->owner->so_lock);
3408
3409         if (!call_close) {
3410                 /* Note: exit _without_ calling nfs4_close_done */
3411                 goto out_no_action;
3412         }
3413
3414         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3415                 nfs_release_seqid(calldata->arg.seqid);
3416                 goto out_wait;
3417         }
3418
3419         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3420         if (lo && !pnfs_layout_is_valid(lo)) {
3421                 calldata->arg.lr_args = NULL;
3422                 calldata->res.lr_res = NULL;
3423         }
3424
3425         if (calldata->arg.fmode == 0)
3426                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3427
3428         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3429                 /* Close-to-open cache consistency revalidation */
3430                 if (!nfs4_have_delegation(inode, FMODE_READ))
3431                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3432                 else
3433                         calldata->arg.bitmask = NULL;
3434         }
3435
3436         calldata->arg.share_access =
3437                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3438                                 calldata->arg.fmode, 0);
3439
3440         if (calldata->res.fattr == NULL)
3441                 calldata->arg.bitmask = NULL;
3442         else if (calldata->arg.bitmask == NULL)
3443                 calldata->res.fattr = NULL;
3444         calldata->timestamp = jiffies;
3445         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3446                                 &calldata->arg.seq_args,
3447                                 &calldata->res.seq_res,
3448                                 task) != 0)
3449                 nfs_release_seqid(calldata->arg.seqid);
3450         dprintk("%s: done!\n", __func__);
3451         return;
3452 out_no_action:
3453         task->tk_action = NULL;
3454 out_wait:
3455         nfs4_sequence_done(task, &calldata->res.seq_res);
3456 }
3457
3458 static const struct rpc_call_ops nfs4_close_ops = {
3459         .rpc_call_prepare = nfs4_close_prepare,
3460         .rpc_call_done = nfs4_close_done,
3461         .rpc_release = nfs4_free_closedata,
3462 };
3463
3464 /*
3465  * It is possible for data to be read/written from a mem-mapped file
3466  * after the sys_close call (which hits the vfs layer as a flush).
3467  * This means that we can't safely call nfsv4 close on a file until
3468  * the inode is cleared. This in turn means that we are not good
3469  * NFSv4 citizens - we do not indicate to the server to update the file's
3470  * share state even when we are done with one of the three share
3471  * stateid's in the inode.
3472  *
3473  * NOTE: Caller must be holding the sp->so_owner semaphore!
3474  */
3475 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3476 {
3477         struct nfs_server *server = NFS_SERVER(state->inode);
3478         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3479         struct nfs4_closedata *calldata;
3480         struct nfs4_state_owner *sp = state->owner;
3481         struct rpc_task *task;
3482         struct rpc_message msg = {
3483                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3484                 .rpc_cred = state->owner->so_cred,
3485         };
3486         struct rpc_task_setup task_setup_data = {
3487                 .rpc_client = server->client,
3488                 .rpc_message = &msg,
3489                 .callback_ops = &nfs4_close_ops,
3490                 .workqueue = nfsiod_workqueue,
3491                 .flags = RPC_TASK_ASYNC,
3492         };
3493         int status = -ENOMEM;
3494
3495         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3496                 &task_setup_data.rpc_client, &msg);
3497
3498         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3499         if (calldata == NULL)
3500                 goto out;
3501         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3502         calldata->inode = state->inode;
3503         calldata->state = state;
3504         calldata->arg.fh = NFS_FH(state->inode);
3505         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3506                 goto out_free_calldata;
3507         /* Serialization for the sequence id */
3508         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3509         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3510         if (IS_ERR(calldata->arg.seqid))
3511                 goto out_free_calldata;
3512         nfs_fattr_init(&calldata->fattr);
3513         calldata->arg.fmode = 0;
3514         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3515         calldata->res.fattr = &calldata->fattr;
3516         calldata->res.seqid = calldata->arg.seqid;
3517         calldata->res.server = server;
3518         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3519         calldata->lr.roc = pnfs_roc(state->inode,
3520                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3521         if (calldata->lr.roc) {
3522                 calldata->arg.lr_args = &calldata->lr.arg;
3523                 calldata->res.lr_res = &calldata->lr.res;
3524         }
3525         nfs_sb_active(calldata->inode->i_sb);
3526
3527         msg.rpc_argp = &calldata->arg;
3528         msg.rpc_resp = &calldata->res;
3529         task_setup_data.callback_data = calldata;
3530         task = rpc_run_task(&task_setup_data);
3531         if (IS_ERR(task))
3532                 return PTR_ERR(task);
3533         status = 0;
3534         if (wait)
3535                 status = rpc_wait_for_completion_task(task);
3536         rpc_put_task(task);
3537         return status;
3538 out_free_calldata:
3539         kfree(calldata);
3540 out:
3541         nfs4_put_open_state(state);
3542         nfs4_put_state_owner(sp);
3543         return status;
3544 }
3545
3546 static struct inode *
3547 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3548                 int open_flags, struct iattr *attr, int *opened)
3549 {
3550         struct nfs4_state *state;
3551         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3552
3553         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3554
3555         /* Protect against concurrent sillydeletes */
3556         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3557
3558         nfs4_label_release_security(label);
3559
3560         if (IS_ERR(state))
3561                 return ERR_CAST(state);
3562         return state->inode;
3563 }
3564
3565 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3566 {
3567         if (ctx->state == NULL)
3568                 return;
3569         if (is_sync)
3570                 nfs4_close_sync(ctx->state, ctx->mode);
3571         else
3572                 nfs4_close_state(ctx->state, ctx->mode);
3573 }
3574
3575 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3576 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3577 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3578
3579 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3580 {
3581         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3582         struct nfs4_server_caps_arg args = {
3583                 .fhandle = fhandle,
3584                 .bitmask = bitmask,
3585         };
3586         struct nfs4_server_caps_res res = {};
3587         struct rpc_message msg = {
3588                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3589                 .rpc_argp = &args,
3590                 .rpc_resp = &res,
3591         };
3592         int status;
3593         int i;
3594
3595         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3596                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3597                      FATTR4_WORD0_LINK_SUPPORT |
3598                      FATTR4_WORD0_SYMLINK_SUPPORT |
3599                      FATTR4_WORD0_ACLSUPPORT;
3600         if (minorversion)
3601                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3602
3603         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3604         if (status == 0) {
3605                 /* Sanity check the server answers */
3606                 switch (minorversion) {
3607                 case 0:
3608                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3609                         res.attr_bitmask[2] = 0;
3610                         break;
3611                 case 1:
3612                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3613                         break;
3614                 case 2:
3615                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3616                 }
3617                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3618                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3619                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3620                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3621                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3622                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3623                                 NFS_CAP_SECURITY_LABEL);
3624                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3625                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3626                         server->caps |= NFS_CAP_ACLS;
3627                 if (res.has_links != 0)
3628                         server->caps |= NFS_CAP_HARDLINKS;
3629                 if (res.has_symlinks != 0)
3630                         server->caps |= NFS_CAP_SYMLINKS;
3631                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3632                         server->caps |= NFS_CAP_FILEID;
3633                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3634                         server->caps |= NFS_CAP_MODE;
3635                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3636                         server->caps |= NFS_CAP_NLINK;
3637                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3638                         server->caps |= NFS_CAP_OWNER;
3639                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3640                         server->caps |= NFS_CAP_OWNER_GROUP;
3641                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3642                         server->caps |= NFS_CAP_ATIME;
3643                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3644                         server->caps |= NFS_CAP_CTIME;
3645                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3646                         server->caps |= NFS_CAP_MTIME;
3647 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3648                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3649                         server->caps |= NFS_CAP_SECURITY_LABEL;
3650 #endif
3651                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3652                                 sizeof(server->attr_bitmask));
3653                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3654
3655                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3656                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3657                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3658                 server->cache_consistency_bitmask[2] = 0;
3659
3660                 /* Avoid a regression due to buggy server */
3661                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3662                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3663                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3664                         sizeof(server->exclcreat_bitmask));
3665
3666                 server->acl_bitmask = res.acl_bitmask;
3667                 server->fh_expire_type = res.fh_expire_type;
3668         }
3669
3670         return status;
3671 }
3672
3673 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3674 {
3675         struct nfs4_exception exception = { };
3676         int err;
3677         do {
3678                 err = nfs4_handle_exception(server,
3679                                 _nfs4_server_capabilities(server, fhandle),
3680                                 &exception);
3681         } while (exception.retry);
3682         return err;
3683 }
3684
3685 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3686                 struct nfs_fsinfo *info)
3687 {
3688         u32 bitmask[3];
3689         struct nfs4_lookup_root_arg args = {
3690                 .bitmask = bitmask,
3691         };
3692         struct nfs4_lookup_res res = {
3693                 .server = server,
3694                 .fattr = info->fattr,
3695                 .fh = fhandle,
3696         };
3697         struct rpc_message msg = {
3698                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3699                 .rpc_argp = &args,
3700                 .rpc_resp = &res,
3701         };
3702
3703         bitmask[0] = nfs4_fattr_bitmap[0];
3704         bitmask[1] = nfs4_fattr_bitmap[1];
3705         /*
3706          * Process the label in the upcoming getfattr
3707          */
3708         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3709
3710         nfs_fattr_init(info->fattr);
3711         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3712 }
3713
3714 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3715                 struct nfs_fsinfo *info)
3716 {
3717         struct nfs4_exception exception = { };
3718         int err;
3719         do {
3720                 err = _nfs4_lookup_root(server, fhandle, info);
3721                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3722                 switch (err) {
3723                 case 0:
3724                 case -NFS4ERR_WRONGSEC:
3725                         goto out;
3726                 default:
3727                         err = nfs4_handle_exception(server, err, &exception);
3728                 }
3729         } while (exception.retry);
3730 out:
3731         return err;
3732 }
3733
3734 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3735                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3736 {
3737         struct rpc_auth_create_args auth_args = {
3738                 .pseudoflavor = flavor,
3739         };
3740         struct rpc_auth *auth;
3741
3742         auth = rpcauth_create(&auth_args, server->client);
3743         if (IS_ERR(auth))
3744                 return -EACCES;
3745         return nfs4_lookup_root(server, fhandle, info);
3746 }
3747
3748 /*
3749  * Retry pseudoroot lookup with various security flavors.  We do this when:
3750  *
3751  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3752  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3753  *
3754  * Returns zero on success, or a negative NFS4ERR value, or a
3755  * negative errno value.
3756  */
3757 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3758                               struct nfs_fsinfo *info)
3759 {
3760         /* Per 3530bis 15.33.5 */
3761         static const rpc_authflavor_t flav_array[] = {
3762                 RPC_AUTH_GSS_KRB5P,
3763                 RPC_AUTH_GSS_KRB5I,
3764                 RPC_AUTH_GSS_KRB5,
3765                 RPC_AUTH_UNIX,                  /* courtesy */
3766                 RPC_AUTH_NULL,
3767         };
3768         int status = -EPERM;
3769         size_t i;
3770
3771         if (server->auth_info.flavor_len > 0) {
3772                 /* try each flavor specified by user */
3773                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3774                         status = nfs4_lookup_root_sec(server, fhandle, info,
3775                                                 server->auth_info.flavors[i]);
3776                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3777                                 continue;
3778                         break;
3779                 }
3780         } else {
3781                 /* no flavors specified by user, try default list */
3782                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3783                         status = nfs4_lookup_root_sec(server, fhandle, info,
3784                                                       flav_array[i]);
3785                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3786                                 continue;
3787                         break;
3788                 }
3789         }
3790
3791         /*
3792          * -EACCES could mean that the user doesn't have correct permissions
3793          * to access the mount.  It could also mean that we tried to mount
3794          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3795          * existing mount programs don't handle -EACCES very well so it should
3796          * be mapped to -EPERM instead.
3797          */
3798         if (status == -EACCES)
3799                 status = -EPERM;
3800         return status;
3801 }
3802
3803 /**
3804  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3805  * @server: initialized nfs_server handle
3806  * @fhandle: we fill in the pseudo-fs root file handle
3807  * @info: we fill in an FSINFO struct
3808  * @auth_probe: probe the auth flavours
3809  *
3810  * Returns zero on success, or a negative errno.
3811  */
3812 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3813                          struct nfs_fsinfo *info,
3814                          bool auth_probe)
3815 {
3816         int status = 0;
3817
3818         if (!auth_probe)
3819                 status = nfs4_lookup_root(server, fhandle, info);
3820
3821         if (auth_probe || status == NFS4ERR_WRONGSEC)
3822                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3823                                 fhandle, info);
3824
3825         if (status == 0)
3826                 status = nfs4_server_capabilities(server, fhandle);
3827         if (status == 0)
3828                 status = nfs4_do_fsinfo(server, fhandle, info);
3829
3830         return nfs4_map_errors(status);
3831 }
3832
3833 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3834                               struct nfs_fsinfo *info)
3835 {
3836         int error;
3837         struct nfs_fattr *fattr = info->fattr;
3838         struct nfs4_label *label = NULL;
3839
3840         error = nfs4_server_capabilities(server, mntfh);
3841         if (error < 0) {
3842                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3843                 return error;
3844         }
3845
3846         label = nfs4_label_alloc(server, GFP_KERNEL);
3847         if (IS_ERR(label))
3848                 return PTR_ERR(label);
3849
3850         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
3851         if (error < 0) {
3852                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3853                 goto err_free_label;
3854         }
3855
3856         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3857             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3858                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3859
3860 err_free_label:
3861         nfs4_label_free(label);
3862
3863         return error;
3864 }
3865
3866 /*
3867  * Get locations and (maybe) other attributes of a referral.
3868  * Note that we'll actually follow the referral later when
3869  * we detect fsid mismatch in inode revalidation
3870  */
3871 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3872                              const struct qstr *name, struct nfs_fattr *fattr,
3873                              struct nfs_fh *fhandle)
3874 {
3875         int status = -ENOMEM;
3876         struct page *page = NULL;
3877         struct nfs4_fs_locations *locations = NULL;
3878
3879         page = alloc_page(GFP_KERNEL);
3880         if (page == NULL)
3881                 goto out;
3882         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3883         if (locations == NULL)
3884                 goto out;
3885
3886         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3887         if (status != 0)
3888                 goto out;
3889
3890         /*
3891          * If the fsid didn't change, this is a migration event, not a
3892          * referral.  Cause us to drop into the exception handler, which
3893          * will kick off migration recovery.
3894          */
3895         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3896                 dprintk("%s: server did not return a different fsid for"
3897                         " a referral at %s\n", __func__, name->name);
3898                 status = -NFS4ERR_MOVED;
3899                 goto out;
3900         }
3901         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3902         nfs_fixup_referral_attributes(&locations->fattr);
3903
3904         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3905         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3906         memset(fhandle, 0, sizeof(struct nfs_fh));
3907 out:
3908         if (page)
3909                 __free_page(page);
3910         kfree(locations);
3911         return status;
3912 }
3913
3914 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3915                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3916                                 struct inode *inode)
3917 {
3918         __u32 bitmask[NFS4_BITMASK_SZ];
3919         struct nfs4_getattr_arg args = {
3920                 .fh = fhandle,
3921                 .bitmask = bitmask,
3922         };
3923         struct nfs4_getattr_res res = {
3924                 .fattr = fattr,
3925                 .label = label,
3926                 .server = server,
3927         };
3928         struct rpc_message msg = {
3929                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3930                 .rpc_argp = &args,
3931                 .rpc_resp = &res,
3932         };
3933
3934         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
3935
3936         nfs_fattr_init(fattr);
3937         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3938 }
3939
3940 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3941                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3942                                 struct inode *inode)
3943 {
3944         struct nfs4_exception exception = { };
3945         int err;
3946         do {
3947                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
3948                 trace_nfs4_getattr(server, fhandle, fattr, err);
3949                 err = nfs4_handle_exception(server, err,
3950                                 &exception);
3951         } while (exception.retry);
3952         return err;
3953 }
3954
3955 /*
3956  * The file is not closed if it is opened due to the a request to change
3957  * the size of the file. The open call will not be needed once the
3958  * VFS layer lookup-intents are implemented.
3959  *
3960  * Close is called when the inode is destroyed.
3961  * If we haven't opened the file for O_WRONLY, we
3962  * need to in the size_change case to obtain a stateid.
3963  *
3964  * Got race?
3965  * Because OPEN is always done by name in nfsv4, it is
3966  * possible that we opened a different file by the same
3967  * name.  We can recognize this race condition, but we
3968  * can't do anything about it besides returning an error.
3969  *
3970  * This will be fixed with VFS changes (lookup-intent).
3971  */
3972 static int
3973 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3974                   struct iattr *sattr)
3975 {
3976         struct inode *inode = d_inode(dentry);
3977         const struct cred *cred = NULL;
3978         struct nfs_open_context *ctx = NULL;
3979         struct nfs4_label *label = NULL;
3980         int status;
3981
3982         if (pnfs_ld_layoutret_on_setattr(inode) &&
3983             sattr->ia_valid & ATTR_SIZE &&
3984             sattr->ia_size < i_size_read(inode))
3985                 pnfs_commit_and_return_layout(inode);
3986
3987         nfs_fattr_init(fattr);
3988
3989         /* Deal with open(O_TRUNC) */
3990         if (sattr->ia_valid & ATTR_OPEN)
3991                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3992
3993         /* Optimization: if the end result is no change, don't RPC */
3994         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3995                 return 0;
3996
3997         /* Search for an existing open(O_WRITE) file */
3998         if (sattr->ia_valid & ATTR_FILE) {
3999
4000                 ctx = nfs_file_open_context(sattr->ia_file);
4001                 if (ctx)
4002                         cred = ctx->cred;
4003         }
4004
4005         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4006         if (IS_ERR(label))
4007                 return PTR_ERR(label);
4008
4009         /* Return any delegations if we're going to change ACLs */
4010         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4011                 nfs4_inode_make_writeable(inode);
4012
4013         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4014         if (status == 0) {
4015                 nfs_setattr_update_inode(inode, sattr, fattr);
4016                 nfs_setsecurity(inode, fattr, label);
4017         }
4018         nfs4_label_free(label);
4019         return status;
4020 }
4021
4022 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4023                 const struct qstr *name, struct nfs_fh *fhandle,
4024                 struct nfs_fattr *fattr, struct nfs4_label *label)
4025 {
4026         struct nfs_server *server = NFS_SERVER(dir);
4027         int                    status;
4028         struct nfs4_lookup_arg args = {
4029                 .bitmask = server->attr_bitmask,
4030                 .dir_fh = NFS_FH(dir),
4031                 .name = name,
4032         };
4033         struct nfs4_lookup_res res = {
4034                 .server = server,
4035                 .fattr = fattr,
4036                 .label = label,
4037                 .fh = fhandle,
4038         };
4039         struct rpc_message msg = {
4040                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4041                 .rpc_argp = &args,
4042                 .rpc_resp = &res,
4043         };
4044
4045         args.bitmask = nfs4_bitmask(server, label);
4046
4047         nfs_fattr_init(fattr);
4048
4049         dprintk("NFS call  lookup %s\n", name->name);
4050         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
4051         dprintk("NFS reply lookup: %d\n", status);
4052         return status;
4053 }
4054
4055 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4056 {
4057         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4058                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4059         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4060         fattr->nlink = 2;
4061 }
4062
4063 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4064                                    const struct qstr *name, struct nfs_fh *fhandle,
4065                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4066 {
4067         struct nfs4_exception exception = { };
4068         struct rpc_clnt *client = *clnt;
4069         int err;
4070         do {
4071                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
4072                 trace_nfs4_lookup(dir, name, err);
4073                 switch (err) {
4074                 case -NFS4ERR_BADNAME:
4075                         err = -ENOENT;
4076                         goto out;
4077                 case -NFS4ERR_MOVED:
4078                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4079                         if (err == -NFS4ERR_MOVED)
4080                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4081                         goto out;
4082                 case -NFS4ERR_WRONGSEC:
4083                         err = -EPERM;
4084                         if (client != *clnt)
4085                                 goto out;
4086                         client = nfs4_negotiate_security(client, dir, name);
4087                         if (IS_ERR(client))
4088                                 return PTR_ERR(client);
4089
4090                         exception.retry = 1;
4091                         break;
4092                 default:
4093                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4094                 }
4095         } while (exception.retry);
4096
4097 out:
4098         if (err == 0)
4099                 *clnt = client;
4100         else if (client != *clnt)
4101                 rpc_shutdown_client(client);
4102
4103         return err;
4104 }
4105
4106 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
4107                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4108                             struct nfs4_label *label)
4109 {
4110         int status;
4111         struct rpc_clnt *client = NFS_CLIENT(dir);
4112
4113         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
4114         if (client != NFS_CLIENT(dir)) {
4115                 rpc_shutdown_client(client);
4116                 nfs_fixup_secinfo_attributes(fattr);
4117         }
4118         return status;
4119 }
4120
4121 struct rpc_clnt *
4122 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
4123                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4124 {
4125         struct rpc_clnt *client = NFS_CLIENT(dir);
4126         int status;
4127
4128         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
4129         if (status < 0)
4130                 return ERR_PTR(status);
4131         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4132 }
4133
4134 static int _nfs4_proc_lookupp(struct inode *inode,
4135                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4136                 struct nfs4_label *label)
4137 {
4138         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4139         struct nfs_server *server = NFS_SERVER(inode);
4140         int                    status;
4141         struct nfs4_lookupp_arg args = {
4142                 .bitmask = server->attr_bitmask,
4143                 .fh = NFS_FH(inode),
4144         };
4145         struct nfs4_lookupp_res res = {
4146                 .server = server,
4147                 .fattr = fattr,
4148                 .label = label,
4149                 .fh = fhandle,
4150         };
4151         struct rpc_message msg = {
4152                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4153                 .rpc_argp = &args,
4154                 .rpc_resp = &res,
4155         };
4156
4157         args.bitmask = nfs4_bitmask(server, label);
4158
4159         nfs_fattr_init(fattr);
4160
4161         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4162         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4163                                 &res.seq_res, 0);
4164         dprintk("NFS reply lookupp: %d\n", status);
4165         return status;
4166 }
4167
4168 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4169                              struct nfs_fattr *fattr, struct nfs4_label *label)
4170 {
4171         struct nfs4_exception exception = { };
4172         int err;
4173         do {
4174                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4175                 trace_nfs4_lookupp(inode, err);
4176                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4177                                 &exception);
4178         } while (exception.retry);
4179         return err;
4180 }
4181
4182 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4183 {
4184         struct nfs_server *server = NFS_SERVER(inode);
4185         struct nfs4_accessargs args = {
4186                 .fh = NFS_FH(inode),
4187                 .access = entry->mask,
4188         };
4189         struct nfs4_accessres res = {
4190                 .server = server,
4191         };
4192         struct rpc_message msg = {
4193                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4194                 .rpc_argp = &args,
4195                 .rpc_resp = &res,
4196                 .rpc_cred = entry->cred,
4197         };
4198         int status = 0;
4199
4200         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4201                 res.fattr = nfs_alloc_fattr();
4202                 if (res.fattr == NULL)
4203                         return -ENOMEM;
4204                 args.bitmask = server->cache_consistency_bitmask;
4205         }
4206         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4207         if (!status) {
4208                 nfs_access_set_mask(entry, res.access);
4209                 if (res.fattr)
4210                         nfs_refresh_inode(inode, res.fattr);
4211         }
4212         nfs_free_fattr(res.fattr);
4213         return status;
4214 }
4215
4216 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4217 {
4218         struct nfs4_exception exception = { };
4219         int err;
4220         do {
4221                 err = _nfs4_proc_access(inode, entry);
4222                 trace_nfs4_access(inode, err);
4223                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4224                                 &exception);
4225         } while (exception.retry);
4226         return err;
4227 }
4228
4229 /*
4230  * TODO: For the time being, we don't try to get any attributes
4231  * along with any of the zero-copy operations READ, READDIR,
4232  * READLINK, WRITE.
4233  *
4234  * In the case of the first three, we want to put the GETATTR
4235  * after the read-type operation -- this is because it is hard
4236  * to predict the length of a GETATTR response in v4, and thus
4237  * align the READ data correctly.  This means that the GETATTR
4238  * may end up partially falling into the page cache, and we should
4239  * shift it into the 'tail' of the xdr_buf before processing.
4240  * To do this efficiently, we need to know the total length
4241  * of data received, which doesn't seem to be available outside
4242  * of the RPC layer.
4243  *
4244  * In the case of WRITE, we also want to put the GETATTR after
4245  * the operation -- in this case because we want to make sure
4246  * we get the post-operation mtime and size.
4247  *
4248  * Both of these changes to the XDR layer would in fact be quite
4249  * minor, but I decided to leave them for a subsequent patch.
4250  */
4251 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4252                 unsigned int pgbase, unsigned int pglen)
4253 {
4254         struct nfs4_readlink args = {
4255                 .fh       = NFS_FH(inode),
4256                 .pgbase   = pgbase,
4257                 .pglen    = pglen,
4258                 .pages    = &page,
4259         };
4260         struct nfs4_readlink_res res;
4261         struct rpc_message msg = {
4262                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4263                 .rpc_argp = &args,
4264                 .rpc_resp = &res,
4265         };
4266
4267         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4268 }
4269
4270 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4271                 unsigned int pgbase, unsigned int pglen)
4272 {
4273         struct nfs4_exception exception = { };
4274         int err;
4275         do {
4276                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4277                 trace_nfs4_readlink(inode, err);
4278                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4279                                 &exception);
4280         } while (exception.retry);
4281         return err;
4282 }
4283
4284 /*
4285  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4286  */
4287 static int
4288 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4289                  int flags)
4290 {
4291         struct nfs_server *server = NFS_SERVER(dir);
4292         struct nfs4_label l, *ilabel = NULL;
4293         struct nfs_open_context *ctx;
4294         struct nfs4_state *state;
4295         int status = 0;
4296
4297         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4298         if (IS_ERR(ctx))
4299                 return PTR_ERR(ctx);
4300
4301         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4302
4303         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4304                 sattr->ia_mode &= ~current_umask();
4305         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4306         if (IS_ERR(state)) {
4307                 status = PTR_ERR(state);
4308                 goto out;
4309         }
4310 out:
4311         nfs4_label_release_security(ilabel);
4312         put_nfs_open_context(ctx);
4313         return status;
4314 }
4315
4316 static int
4317 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4318 {
4319         struct nfs_server *server = NFS_SERVER(dir);
4320         struct nfs_removeargs args = {
4321                 .fh = NFS_FH(dir),
4322                 .name = *name,
4323         };
4324         struct nfs_removeres res = {
4325                 .server = server,
4326         };
4327         struct rpc_message msg = {
4328                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4329                 .rpc_argp = &args,
4330                 .rpc_resp = &res,
4331         };
4332         unsigned long timestamp = jiffies;
4333         int status;
4334
4335         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4336         if (status == 0) {
4337                 spin_lock(&dir->i_lock);
4338                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4339                 /* Removing a directory decrements nlink in the parent */
4340                 if (ftype == NF4DIR && dir->i_nlink > 2)
4341                         nfs4_dec_nlink_locked(dir);
4342                 spin_unlock(&dir->i_lock);
4343         }
4344         return status;
4345 }
4346
4347 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4348 {
4349         struct nfs4_exception exception = { };
4350         struct inode *inode = d_inode(dentry);
4351         int err;
4352
4353         if (inode) {
4354                 if (inode->i_nlink == 1)
4355                         nfs4_inode_return_delegation(inode);
4356                 else
4357                         nfs4_inode_make_writeable(inode);
4358         }
4359         do {
4360                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4361                 trace_nfs4_remove(dir, &dentry->d_name, err);
4362                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4363                                 &exception);
4364         } while (exception.retry);
4365         return err;
4366 }
4367
4368 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4369 {
4370         struct nfs4_exception exception = { };
4371         int err;
4372
4373         do {
4374                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4375                 trace_nfs4_remove(dir, name, err);
4376                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4377                                 &exception);
4378         } while (exception.retry);
4379         return err;
4380 }
4381
4382 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4383                 struct dentry *dentry,
4384                 struct inode *inode)
4385 {
4386         struct nfs_removeargs *args = msg->rpc_argp;
4387         struct nfs_removeres *res = msg->rpc_resp;
4388
4389         res->server = NFS_SB(dentry->d_sb);
4390         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4391         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4392
4393         nfs_fattr_init(res->dir_attr);
4394
4395         if (inode)
4396                 nfs4_inode_return_delegation(inode);
4397 }
4398
4399 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4400 {
4401         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4402                         &data->args.seq_args,
4403                         &data->res.seq_res,
4404                         task);
4405 }
4406
4407 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4408 {
4409         struct nfs_unlinkdata *data = task->tk_calldata;
4410         struct nfs_removeres *res = &data->res;
4411
4412         if (!nfs4_sequence_done(task, &res->seq_res))
4413                 return 0;
4414         if (nfs4_async_handle_error(task, res->server, NULL,
4415                                     &data->timeout) == -EAGAIN)
4416                 return 0;
4417         if (task->tk_status == 0)
4418                 update_changeattr(dir, &res->cinfo,
4419                                 res->dir_attr->time_start, 0);
4420         return 1;
4421 }
4422
4423 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4424                 struct dentry *old_dentry,
4425                 struct dentry *new_dentry)
4426 {
4427         struct nfs_renameargs *arg = msg->rpc_argp;
4428         struct nfs_renameres *res = msg->rpc_resp;
4429         struct inode *old_inode = d_inode(old_dentry);
4430         struct inode *new_inode = d_inode(new_dentry);
4431
4432         if (old_inode)
4433                 nfs4_inode_make_writeable(old_inode);
4434         if (new_inode)
4435                 nfs4_inode_return_delegation(new_inode);
4436         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4437         res->server = NFS_SB(old_dentry->d_sb);
4438         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4439 }
4440
4441 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4442 {
4443         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4444                         &data->args.seq_args,
4445                         &data->res.seq_res,
4446                         task);
4447 }
4448
4449 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4450                                  struct inode *new_dir)
4451 {
4452         struct nfs_renamedata *data = task->tk_calldata;
4453         struct nfs_renameres *res = &data->res;
4454
4455         if (!nfs4_sequence_done(task, &res->seq_res))
4456                 return 0;
4457         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4458                 return 0;
4459
4460         if (task->tk_status == 0) {
4461                 if (new_dir != old_dir) {
4462                         /* Note: If we moved a directory, nlink will change */
4463                         update_changeattr(old_dir, &res->old_cinfo,
4464                                         res->old_fattr->time_start,
4465                                         NFS_INO_INVALID_OTHER);
4466                         update_changeattr(new_dir, &res->new_cinfo,
4467                                         res->new_fattr->time_start,
4468                                         NFS_INO_INVALID_OTHER);
4469                 } else
4470                         update_changeattr(old_dir, &res->old_cinfo,
4471                                         res->old_fattr->time_start,
4472                                         0);
4473         }
4474         return 1;
4475 }
4476
4477 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4478 {
4479         struct nfs_server *server = NFS_SERVER(inode);
4480         __u32 bitmask[NFS4_BITMASK_SZ];
4481         struct nfs4_link_arg arg = {
4482                 .fh     = NFS_FH(inode),
4483                 .dir_fh = NFS_FH(dir),
4484                 .name   = name,
4485                 .bitmask = bitmask,
4486         };
4487         struct nfs4_link_res res = {
4488                 .server = server,
4489                 .label = NULL,
4490         };
4491         struct rpc_message msg = {
4492                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4493                 .rpc_argp = &arg,
4494                 .rpc_resp = &res,
4495         };
4496         int status = -ENOMEM;
4497
4498         res.fattr = nfs_alloc_fattr();
4499         if (res.fattr == NULL)
4500                 goto out;
4501
4502         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4503         if (IS_ERR(res.label)) {
4504                 status = PTR_ERR(res.label);
4505                 goto out;
4506         }
4507
4508         nfs4_inode_make_writeable(inode);
4509         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4510
4511         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4512         if (!status) {
4513                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4514                 status = nfs_post_op_update_inode(inode, res.fattr);
4515                 if (!status)
4516                         nfs_setsecurity(inode, res.fattr, res.label);
4517         }
4518
4519
4520         nfs4_label_free(res.label);
4521
4522 out:
4523         nfs_free_fattr(res.fattr);
4524         return status;
4525 }
4526
4527 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4528 {
4529         struct nfs4_exception exception = { };
4530         int err;
4531         do {
4532                 err = nfs4_handle_exception(NFS_SERVER(inode),
4533                                 _nfs4_proc_link(inode, dir, name),
4534                                 &exception);
4535         } while (exception.retry);
4536         return err;
4537 }
4538
4539 struct nfs4_createdata {
4540         struct rpc_message msg;
4541         struct nfs4_create_arg arg;
4542         struct nfs4_create_res res;
4543         struct nfs_fh fh;
4544         struct nfs_fattr fattr;
4545         struct nfs4_label *label;
4546 };
4547
4548 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4549                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4550 {
4551         struct nfs4_createdata *data;
4552
4553         data = kzalloc(sizeof(*data), GFP_KERNEL);
4554         if (data != NULL) {
4555                 struct nfs_server *server = NFS_SERVER(dir);
4556
4557                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4558                 if (IS_ERR(data->label))
4559                         goto out_free;
4560
4561                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4562                 data->msg.rpc_argp = &data->arg;
4563                 data->msg.rpc_resp = &data->res;
4564                 data->arg.dir_fh = NFS_FH(dir);
4565                 data->arg.server = server;
4566                 data->arg.name = name;
4567                 data->arg.attrs = sattr;
4568                 data->arg.ftype = ftype;
4569                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4570                 data->arg.umask = current_umask();
4571                 data->res.server = server;
4572                 data->res.fh = &data->fh;
4573                 data->res.fattr = &data->fattr;
4574                 data->res.label = data->label;
4575                 nfs_fattr_init(data->res.fattr);
4576         }
4577         return data;
4578 out_free:
4579         kfree(data);
4580         return NULL;
4581 }
4582
4583 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4584 {
4585         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4586                                     &data->arg.seq_args, &data->res.seq_res, 1);
4587         if (status == 0) {
4588                 spin_lock(&dir->i_lock);
4589                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4590                                 data->res.fattr->time_start, 0);
4591                 /* Creating a directory bumps nlink in the parent */
4592                 if (data->arg.ftype == NF4DIR)
4593                         nfs4_inc_nlink_locked(dir);
4594                 spin_unlock(&dir->i_lock);
4595                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4596         }
4597         return status;
4598 }
4599
4600 static void nfs4_free_createdata(struct nfs4_createdata *data)
4601 {
4602         nfs4_label_free(data->label);
4603         kfree(data);
4604 }
4605
4606 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4607                 struct page *page, unsigned int len, struct iattr *sattr,
4608                 struct nfs4_label *label)
4609 {
4610         struct nfs4_createdata *data;
4611         int status = -ENAMETOOLONG;
4612
4613         if (len > NFS4_MAXPATHLEN)
4614                 goto out;
4615
4616         status = -ENOMEM;
4617         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4618         if (data == NULL)
4619                 goto out;
4620
4621         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4622         data->arg.u.symlink.pages = &page;
4623         data->arg.u.symlink.len = len;
4624         data->arg.label = label;
4625
4626         status = nfs4_do_create(dir, dentry, data);
4627
4628         nfs4_free_createdata(data);
4629 out:
4630         return status;
4631 }
4632
4633 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4634                 struct page *page, unsigned int len, struct iattr *sattr)
4635 {
4636         struct nfs4_exception exception = { };
4637         struct nfs4_label l, *label = NULL;
4638         int err;
4639
4640         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4641
4642         do {
4643                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4644                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4645                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4646                                 &exception);
4647         } while (exception.retry);
4648
4649         nfs4_label_release_security(label);
4650         return err;
4651 }
4652
4653 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4654                 struct iattr *sattr, struct nfs4_label *label)
4655 {
4656         struct nfs4_createdata *data;
4657         int status = -ENOMEM;
4658
4659         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4660         if (data == NULL)
4661                 goto out;
4662
4663         data->arg.label = label;
4664         status = nfs4_do_create(dir, dentry, data);
4665
4666         nfs4_free_createdata(data);
4667 out:
4668         return status;
4669 }
4670
4671 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4672                 struct iattr *sattr)
4673 {
4674         struct nfs_server *server = NFS_SERVER(dir);
4675         struct nfs4_exception exception = { };
4676         struct nfs4_label l, *label = NULL;
4677         int err;
4678
4679         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4680
4681         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4682                 sattr->ia_mode &= ~current_umask();
4683         do {
4684                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4685                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4686                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4687                                 &exception);
4688         } while (exception.retry);
4689         nfs4_label_release_security(label);
4690
4691         return err;
4692 }
4693
4694 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4695                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4696 {
4697         struct inode            *dir = d_inode(dentry);
4698         struct nfs4_readdir_arg args = {
4699                 .fh = NFS_FH(dir),
4700                 .pages = pages,
4701                 .pgbase = 0,
4702                 .count = count,
4703                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4704                 .plus = plus,
4705         };
4706         struct nfs4_readdir_res res;
4707         struct rpc_message msg = {
4708                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4709                 .rpc_argp = &args,
4710                 .rpc_resp = &res,
4711                 .rpc_cred = cred,
4712         };
4713         int                     status;
4714
4715         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4716                         dentry,
4717                         (unsigned long long)cookie);
4718         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4719         res.pgbase = args.pgbase;
4720         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4721         if (status >= 0) {
4722                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4723                 status += args.pgbase;
4724         }
4725
4726         nfs_invalidate_atime(dir);
4727
4728         dprintk("%s: returns %d\n", __func__, status);
4729         return status;
4730 }
4731
4732 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4733                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4734 {
4735         struct nfs4_exception exception = { };
4736         int err;
4737         do {
4738                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4739                                 pages, count, plus);
4740                 trace_nfs4_readdir(d_inode(dentry), err);
4741                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4742                                 &exception);
4743         } while (exception.retry);
4744         return err;
4745 }
4746
4747 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4748                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4749 {
4750         struct nfs4_createdata *data;
4751         int mode = sattr->ia_mode;
4752         int status = -ENOMEM;
4753
4754         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4755         if (data == NULL)
4756                 goto out;
4757
4758         if (S_ISFIFO(mode))
4759                 data->arg.ftype = NF4FIFO;
4760         else if (S_ISBLK(mode)) {
4761                 data->arg.ftype = NF4BLK;
4762                 data->arg.u.device.specdata1 = MAJOR(rdev);
4763                 data->arg.u.device.specdata2 = MINOR(rdev);
4764         }
4765         else if (S_ISCHR(mode)) {
4766                 data->arg.ftype = NF4CHR;
4767                 data->arg.u.device.specdata1 = MAJOR(rdev);
4768                 data->arg.u.device.specdata2 = MINOR(rdev);
4769         } else if (!S_ISSOCK(mode)) {
4770                 status = -EINVAL;
4771                 goto out_free;
4772         }
4773
4774         data->arg.label = label;
4775         status = nfs4_do_create(dir, dentry, data);
4776 out_free:
4777         nfs4_free_createdata(data);
4778 out:
4779         return status;
4780 }
4781
4782 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4783                 struct iattr *sattr, dev_t rdev)
4784 {
4785         struct nfs_server *server = NFS_SERVER(dir);
4786         struct nfs4_exception exception = { };
4787         struct nfs4_label l, *label = NULL;
4788         int err;
4789
4790         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4791
4792         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4793                 sattr->ia_mode &= ~current_umask();
4794         do {
4795                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4796                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4797                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4798                                 &exception);
4799         } while (exception.retry);
4800
4801         nfs4_label_release_security(label);
4802
4803         return err;
4804 }
4805
4806 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4807                  struct nfs_fsstat *fsstat)
4808 {
4809         struct nfs4_statfs_arg args = {
4810                 .fh = fhandle,
4811                 .bitmask = server->attr_bitmask,
4812         };
4813         struct nfs4_statfs_res res = {
4814                 .fsstat = fsstat,
4815         };
4816         struct rpc_message msg = {
4817                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4818                 .rpc_argp = &args,
4819                 .rpc_resp = &res,
4820         };
4821
4822         nfs_fattr_init(fsstat->fattr);
4823         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4824 }
4825
4826 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4827 {
4828         struct nfs4_exception exception = { };
4829         int err;
4830         do {
4831                 err = nfs4_handle_exception(server,
4832                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4833                                 &exception);
4834         } while (exception.retry);
4835         return err;
4836 }
4837
4838 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4839                 struct nfs_fsinfo *fsinfo)
4840 {
4841         struct nfs4_fsinfo_arg args = {
4842                 .fh = fhandle,
4843                 .bitmask = server->attr_bitmask,
4844         };
4845         struct nfs4_fsinfo_res res = {
4846                 .fsinfo = fsinfo,
4847         };
4848         struct rpc_message msg = {
4849                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4850                 .rpc_argp = &args,
4851                 .rpc_resp = &res,
4852         };
4853
4854         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4855 }
4856
4857 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4858 {
4859         struct nfs4_exception exception = { };
4860         unsigned long now = jiffies;
4861         int err;
4862
4863         do {
4864                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4865                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4866                 if (err == 0) {
4867                         nfs4_set_lease_period(server->nfs_client,
4868                                         fsinfo->lease_time * HZ,
4869                                         now);
4870                         break;
4871                 }
4872                 err = nfs4_handle_exception(server, err, &exception);
4873         } while (exception.retry);
4874         return err;
4875 }
4876
4877 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4878 {
4879         int error;
4880
4881         nfs_fattr_init(fsinfo->fattr);
4882         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4883         if (error == 0) {
4884                 /* block layout checks this! */
4885                 server->pnfs_blksize = fsinfo->blksize;
4886                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4887         }
4888
4889         return error;
4890 }
4891
4892 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4893                 struct nfs_pathconf *pathconf)
4894 {
4895         struct nfs4_pathconf_arg args = {
4896                 .fh = fhandle,
4897                 .bitmask = server->attr_bitmask,
4898         };
4899         struct nfs4_pathconf_res res = {
4900                 .pathconf = pathconf,
4901         };
4902         struct rpc_message msg = {
4903                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4904                 .rpc_argp = &args,
4905                 .rpc_resp = &res,
4906         };
4907
4908         /* None of the pathconf attributes are mandatory to implement */
4909         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4910                 memset(pathconf, 0, sizeof(*pathconf));
4911                 return 0;
4912         }
4913
4914         nfs_fattr_init(pathconf->fattr);
4915         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4916 }
4917
4918 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4919                 struct nfs_pathconf *pathconf)
4920 {
4921         struct nfs4_exception exception = { };
4922         int err;
4923
4924         do {
4925                 err = nfs4_handle_exception(server,
4926                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4927                                 &exception);
4928         } while (exception.retry);
4929         return err;
4930 }
4931
4932 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4933                 const struct nfs_open_context *ctx,
4934                 const struct nfs_lock_context *l_ctx,
4935                 fmode_t fmode)
4936 {
4937         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4938 }
4939 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4940
4941 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4942                 const struct nfs_open_context *ctx,
4943                 const struct nfs_lock_context *l_ctx,
4944                 fmode_t fmode)
4945 {
4946         nfs4_stateid current_stateid;
4947
4948         /* If the current stateid represents a lost lock, then exit */
4949         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4950                 return true;
4951         return nfs4_stateid_match(stateid, &current_stateid);
4952 }
4953
4954 static bool nfs4_error_stateid_expired(int err)
4955 {
4956         switch (err) {
4957         case -NFS4ERR_DELEG_REVOKED:
4958         case -NFS4ERR_ADMIN_REVOKED:
4959         case -NFS4ERR_BAD_STATEID:
4960         case -NFS4ERR_STALE_STATEID:
4961         case -NFS4ERR_OLD_STATEID:
4962         case -NFS4ERR_OPENMODE:
4963         case -NFS4ERR_EXPIRED:
4964                 return true;
4965         }
4966         return false;
4967 }
4968
4969 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4970 {
4971         struct nfs_server *server = NFS_SERVER(hdr->inode);
4972
4973         trace_nfs4_read(hdr, task->tk_status);
4974         if (task->tk_status < 0) {
4975                 struct nfs4_exception exception = {
4976                         .inode = hdr->inode,
4977                         .state = hdr->args.context->state,
4978                         .stateid = &hdr->args.stateid,
4979                 };
4980                 task->tk_status = nfs4_async_handle_exception(task,
4981                                 server, task->tk_status, &exception);
4982                 if (exception.retry) {
4983                         rpc_restart_call_prepare(task);
4984                         return -EAGAIN;
4985                 }
4986         }
4987
4988         if (task->tk_status > 0)
4989                 renew_lease(server, hdr->timestamp);
4990         return 0;
4991 }
4992
4993 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4994                 struct nfs_pgio_args *args)
4995 {
4996
4997         if (!nfs4_error_stateid_expired(task->tk_status) ||
4998                 nfs4_stateid_is_current(&args->stateid,
4999                                 args->context,
5000                                 args->lock_context,
5001                                 FMODE_READ))
5002                 return false;
5003         rpc_restart_call_prepare(task);
5004         return true;
5005 }
5006
5007 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5008 {
5009
5010         dprintk("--> %s\n", __func__);
5011
5012         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5013                 return -EAGAIN;
5014         if (nfs4_read_stateid_changed(task, &hdr->args))
5015                 return -EAGAIN;
5016         if (task->tk_status > 0)
5017                 nfs_invalidate_atime(hdr->inode);
5018         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5019                                     nfs4_read_done_cb(task, hdr);
5020 }
5021
5022 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5023                                  struct rpc_message *msg)
5024 {
5025         hdr->timestamp   = jiffies;
5026         if (!hdr->pgio_done_cb)
5027                 hdr->pgio_done_cb = nfs4_read_done_cb;
5028         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5029         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5030 }
5031
5032 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5033                                       struct nfs_pgio_header *hdr)
5034 {
5035         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5036                         &hdr->args.seq_args,
5037                         &hdr->res.seq_res,
5038                         task))
5039                 return 0;
5040         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5041                                 hdr->args.lock_context,
5042                                 hdr->rw_mode) == -EIO)
5043                 return -EIO;
5044         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5045                 return -EIO;
5046         return 0;
5047 }
5048
5049 static int nfs4_write_done_cb(struct rpc_task *task,
5050                               struct nfs_pgio_header *hdr)
5051 {
5052         struct inode *inode = hdr->inode;
5053
5054         trace_nfs4_write(hdr, task->tk_status);
5055         if (task->tk_status < 0) {
5056                 struct nfs4_exception exception = {
5057                         .inode = hdr->inode,
5058                         .state = hdr->args.context->state,
5059                         .stateid = &hdr->args.stateid,
5060                 };
5061                 task->tk_status = nfs4_async_handle_exception(task,
5062                                 NFS_SERVER(inode), task->tk_status,
5063                                 &exception);
5064                 if (exception.retry) {
5065                         rpc_restart_call_prepare(task);
5066                         return -EAGAIN;
5067                 }
5068         }
5069         if (task->tk_status >= 0) {
5070                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5071                 nfs_writeback_update_inode(hdr);
5072         }
5073         return 0;
5074 }
5075
5076 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5077                 struct nfs_pgio_args *args)
5078 {
5079
5080         if (!nfs4_error_stateid_expired(task->tk_status) ||
5081                 nfs4_stateid_is_current(&args->stateid,
5082                                 args->context,
5083                                 args->lock_context,
5084                                 FMODE_WRITE))
5085                 return false;
5086         rpc_restart_call_prepare(task);
5087         return true;
5088 }
5089
5090 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5091 {
5092         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5093                 return -EAGAIN;
5094         if (nfs4_write_stateid_changed(task, &hdr->args))
5095                 return -EAGAIN;
5096         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5097                 nfs4_write_done_cb(task, hdr);
5098 }
5099
5100 static
5101 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5102 {
5103         /* Don't request attributes for pNFS or O_DIRECT writes */
5104         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5105                 return false;
5106         /* Otherwise, request attributes if and only if we don't hold
5107          * a delegation
5108          */
5109         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5110 }
5111
5112 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5113                                   struct rpc_message *msg,
5114                                   struct rpc_clnt **clnt)
5115 {
5116         struct nfs_server *server = NFS_SERVER(hdr->inode);
5117
5118         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5119                 hdr->args.bitmask = NULL;
5120                 hdr->res.fattr = NULL;
5121         } else
5122                 hdr->args.bitmask = server->cache_consistency_bitmask;
5123
5124         if (!hdr->pgio_done_cb)
5125                 hdr->pgio_done_cb = nfs4_write_done_cb;
5126         hdr->res.server = server;
5127         hdr->timestamp   = jiffies;
5128
5129         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5130         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1, 0);
5131         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5132 }
5133
5134 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5135 {
5136         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5137                         &data->args.seq_args,
5138                         &data->res.seq_res,
5139                         task);
5140 }
5141
5142 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5143 {
5144         struct inode *inode = data->inode;
5145
5146         trace_nfs4_commit(data, task->tk_status);
5147         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5148                                     NULL, NULL) == -EAGAIN) {
5149                 rpc_restart_call_prepare(task);
5150                 return -EAGAIN;
5151         }
5152         return 0;
5153 }
5154
5155 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5156 {
5157         if (!nfs4_sequence_done(task, &data->res.seq_res))
5158                 return -EAGAIN;
5159         return data->commit_done_cb(task, data);
5160 }
5161
5162 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5163                                    struct rpc_clnt **clnt)
5164 {
5165         struct nfs_server *server = NFS_SERVER(data->inode);
5166
5167         if (data->commit_done_cb == NULL)
5168                 data->commit_done_cb = nfs4_commit_done_cb;
5169         data->res.server = server;
5170         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5171         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5172         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5173 }
5174
5175 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5176                                 struct nfs_commitres *res)
5177 {
5178         struct inode *dst_inode = file_inode(dst);
5179         struct nfs_server *server = NFS_SERVER(dst_inode);
5180         struct rpc_message msg = {
5181                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5182                 .rpc_argp = args,
5183                 .rpc_resp = res,
5184         };
5185
5186         args->fh = NFS_FH(dst_inode);
5187         return nfs4_call_sync(server->client, server, &msg,
5188                         &args->seq_args, &res->seq_res, 1);
5189 }
5190
5191 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5192 {
5193         struct nfs_commitargs args = {
5194                 .offset = offset,
5195                 .count = count,
5196         };
5197         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5198         struct nfs4_exception exception = { };
5199         int status;
5200
5201         do {
5202                 status = _nfs4_proc_commit(dst, &args, res);
5203                 status = nfs4_handle_exception(dst_server, status, &exception);
5204         } while (exception.retry);
5205
5206         return status;
5207 }
5208
5209 struct nfs4_renewdata {
5210         struct nfs_client       *client;
5211         unsigned long           timestamp;
5212 };
5213
5214 /*
5215  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5216  * standalone procedure for queueing an asynchronous RENEW.
5217  */
5218 static void nfs4_renew_release(void *calldata)
5219 {
5220         struct nfs4_renewdata *data = calldata;
5221         struct nfs_client *clp = data->client;
5222
5223         if (refcount_read(&clp->cl_count) > 1)
5224                 nfs4_schedule_state_renewal(clp);
5225         nfs_put_client(clp);
5226         kfree(data);
5227 }
5228
5229 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5230 {
5231         struct nfs4_renewdata *data = calldata;
5232         struct nfs_client *clp = data->client;
5233         unsigned long timestamp = data->timestamp;
5234
5235         trace_nfs4_renew_async(clp, task->tk_status);
5236         switch (task->tk_status) {
5237         case 0:
5238                 break;
5239         case -NFS4ERR_LEASE_MOVED:
5240                 nfs4_schedule_lease_moved_recovery(clp);
5241                 break;
5242         default:
5243                 /* Unless we're shutting down, schedule state recovery! */
5244                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5245                         return;
5246                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5247                         nfs4_schedule_lease_recovery(clp);
5248                         return;
5249                 }
5250                 nfs4_schedule_path_down_recovery(clp);
5251         }
5252         do_renew_lease(clp, timestamp);
5253 }
5254
5255 static const struct rpc_call_ops nfs4_renew_ops = {
5256         .rpc_call_done = nfs4_renew_done,
5257         .rpc_release = nfs4_renew_release,
5258 };
5259
5260 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5261 {
5262         struct rpc_message msg = {
5263                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5264                 .rpc_argp       = clp,
5265                 .rpc_cred       = cred,
5266         };
5267         struct nfs4_renewdata *data;
5268
5269         if (renew_flags == 0)
5270                 return 0;
5271         if (!refcount_inc_not_zero(&clp->cl_count))
5272                 return -EIO;
5273         data = kmalloc(sizeof(*data), GFP_NOFS);
5274         if (data == NULL) {
5275                 nfs_put_client(clp);
5276                 return -ENOMEM;
5277         }
5278         data->client = clp;
5279         data->timestamp = jiffies;
5280         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5281                         &nfs4_renew_ops, data);
5282 }
5283
5284 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5285 {
5286         struct rpc_message msg = {
5287                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5288                 .rpc_argp       = clp,
5289                 .rpc_cred       = cred,
5290         };
5291         unsigned long now = jiffies;
5292         int status;
5293
5294         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5295         if (status < 0)
5296                 return status;
5297         do_renew_lease(clp, now);
5298         return 0;
5299 }
5300
5301 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5302 {
5303         return server->caps & NFS_CAP_ACLS;
5304 }
5305
5306 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5307  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5308  * the stack.
5309  */
5310 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5311
5312 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5313                 struct page **pages)
5314 {
5315         struct page *newpage, **spages;
5316         int rc = 0;
5317         size_t len;
5318         spages = pages;
5319
5320         do {
5321                 len = min_t(size_t, PAGE_SIZE, buflen);
5322                 newpage = alloc_page(GFP_KERNEL);
5323
5324                 if (newpage == NULL)
5325                         goto unwind;
5326                 memcpy(page_address(newpage), buf, len);
5327                 buf += len;
5328                 buflen -= len;
5329                 *pages++ = newpage;
5330                 rc++;
5331         } while (buflen != 0);
5332
5333         return rc;
5334
5335 unwind:
5336         for(; rc > 0; rc--)
5337                 __free_page(spages[rc-1]);
5338         return -ENOMEM;
5339 }
5340
5341 struct nfs4_cached_acl {
5342         int cached;
5343         size_t len;
5344         char data[0];
5345 };
5346
5347 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5348 {
5349         struct nfs_inode *nfsi = NFS_I(inode);
5350
5351         spin_lock(&inode->i_lock);
5352         kfree(nfsi->nfs4_acl);
5353         nfsi->nfs4_acl = acl;
5354         spin_unlock(&inode->i_lock);
5355 }
5356
5357 static void nfs4_zap_acl_attr(struct inode *inode)
5358 {
5359         nfs4_set_cached_acl(inode, NULL);
5360 }
5361
5362 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5363 {
5364         struct nfs_inode *nfsi = NFS_I(inode);
5365         struct nfs4_cached_acl *acl;
5366         int ret = -ENOENT;
5367
5368         spin_lock(&inode->i_lock);
5369         acl = nfsi->nfs4_acl;
5370         if (acl == NULL)
5371                 goto out;
5372         if (buf == NULL) /* user is just asking for length */
5373                 goto out_len;
5374         if (acl->cached == 0)
5375                 goto out;
5376         ret = -ERANGE; /* see getxattr(2) man page */
5377         if (acl->len > buflen)
5378                 goto out;
5379         memcpy(buf, acl->data, acl->len);
5380 out_len:
5381         ret = acl->len;
5382 out:
5383         spin_unlock(&inode->i_lock);
5384         return ret;
5385 }
5386
5387 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5388 {
5389         struct nfs4_cached_acl *acl;
5390         size_t buflen = sizeof(*acl) + acl_len;
5391
5392         if (buflen <= PAGE_SIZE) {
5393                 acl = kmalloc(buflen, GFP_KERNEL);
5394                 if (acl == NULL)
5395                         goto out;
5396                 acl->cached = 1;
5397                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5398         } else {
5399                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5400                 if (acl == NULL)
5401                         goto out;
5402                 acl->cached = 0;
5403         }
5404         acl->len = acl_len;
5405 out:
5406         nfs4_set_cached_acl(inode, acl);
5407 }
5408
5409 /*
5410  * The getxattr API returns the required buffer length when called with a
5411  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5412  * the required buf.  On a NULL buf, we send a page of data to the server
5413  * guessing that the ACL request can be serviced by a page. If so, we cache
5414  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5415  * the cache. If not so, we throw away the page, and cache the required
5416  * length. The next getxattr call will then produce another round trip to
5417  * the server, this time with the input buf of the required size.
5418  */
5419 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5420 {
5421         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5422         struct nfs_getaclargs args = {
5423                 .fh = NFS_FH(inode),
5424                 .acl_pages = pages,
5425                 .acl_len = buflen,
5426         };
5427         struct nfs_getaclres res = {
5428                 .acl_len = buflen,
5429         };
5430         struct rpc_message msg = {
5431                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5432                 .rpc_argp = &args,
5433                 .rpc_resp = &res,
5434         };
5435         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5436         int ret = -ENOMEM, i;
5437
5438         if (npages > ARRAY_SIZE(pages))
5439                 return -ERANGE;
5440
5441         for (i = 0; i < npages; i++) {
5442                 pages[i] = alloc_page(GFP_KERNEL);
5443                 if (!pages[i])
5444                         goto out_free;
5445         }
5446
5447         /* for decoding across pages */
5448         res.acl_scratch = alloc_page(GFP_KERNEL);
5449         if (!res.acl_scratch)
5450                 goto out_free;
5451
5452         args.acl_len = npages * PAGE_SIZE;
5453
5454         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5455                 __func__, buf, buflen, npages, args.acl_len);
5456         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5457                              &msg, &args.seq_args, &res.seq_res, 0);
5458         if (ret)
5459                 goto out_free;
5460
5461         /* Handle the case where the passed-in buffer is too short */
5462         if (res.acl_flags & NFS4_ACL_TRUNC) {
5463                 /* Did the user only issue a request for the acl length? */
5464                 if (buf == NULL)
5465                         goto out_ok;
5466                 ret = -ERANGE;
5467                 goto out_free;
5468         }
5469         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5470         if (buf) {
5471                 if (res.acl_len > buflen) {
5472                         ret = -ERANGE;
5473                         goto out_free;
5474                 }
5475                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5476         }
5477 out_ok:
5478         ret = res.acl_len;
5479 out_free:
5480         for (i = 0; i < npages; i++)
5481                 if (pages[i])
5482                         __free_page(pages[i]);
5483         if (res.acl_scratch)
5484                 __free_page(res.acl_scratch);
5485         return ret;
5486 }
5487
5488 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5489 {
5490         struct nfs4_exception exception = { };
5491         ssize_t ret;
5492         do {
5493                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5494                 trace_nfs4_get_acl(inode, ret);
5495                 if (ret >= 0)
5496                         break;
5497                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5498         } while (exception.retry);
5499         return ret;
5500 }
5501
5502 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5503 {
5504         struct nfs_server *server = NFS_SERVER(inode);
5505         int ret;
5506
5507         if (!nfs4_server_supports_acls(server))
5508                 return -EOPNOTSUPP;
5509         ret = nfs_revalidate_inode(server, inode);
5510         if (ret < 0)
5511                 return ret;
5512         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5513                 nfs_zap_acl_cache(inode);
5514         ret = nfs4_read_cached_acl(inode, buf, buflen);
5515         if (ret != -ENOENT)
5516                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5517                  * but no cached acl data, just the acl length */
5518                 return ret;
5519         return nfs4_get_acl_uncached(inode, buf, buflen);
5520 }
5521
5522 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5523 {
5524         struct nfs_server *server = NFS_SERVER(inode);
5525         struct page *pages[NFS4ACL_MAXPAGES];
5526         struct nfs_setaclargs arg = {
5527                 .fh             = NFS_FH(inode),
5528                 .acl_pages      = pages,
5529                 .acl_len        = buflen,
5530         };
5531         struct nfs_setaclres res;
5532         struct rpc_message msg = {
5533                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5534                 .rpc_argp       = &arg,
5535                 .rpc_resp       = &res,
5536         };
5537         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5538         int ret, i;
5539
5540         if (!nfs4_server_supports_acls(server))
5541                 return -EOPNOTSUPP;
5542         if (npages > ARRAY_SIZE(pages))
5543                 return -ERANGE;
5544         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5545         if (i < 0)
5546                 return i;
5547         nfs4_inode_make_writeable(inode);
5548         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5549
5550         /*
5551          * Free each page after tx, so the only ref left is
5552          * held by the network stack
5553          */
5554         for (; i > 0; i--)
5555                 put_page(pages[i-1]);
5556
5557         /*
5558          * Acl update can result in inode attribute update.
5559          * so mark the attribute cache invalid.
5560          */
5561         spin_lock(&inode->i_lock);
5562         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5563                 | NFS_INO_INVALID_CTIME
5564                 | NFS_INO_REVAL_FORCED;
5565         spin_unlock(&inode->i_lock);
5566         nfs_access_zap_cache(inode);
5567         nfs_zap_acl_cache(inode);
5568         return ret;
5569 }
5570
5571 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5572 {
5573         struct nfs4_exception exception = { };
5574         int err;
5575         do {
5576                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5577                 trace_nfs4_set_acl(inode, err);
5578                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5579                                 &exception);
5580         } while (exception.retry);
5581         return err;
5582 }
5583
5584 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5585 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5586                                         size_t buflen)
5587 {
5588         struct nfs_server *server = NFS_SERVER(inode);
5589         struct nfs_fattr fattr;
5590         struct nfs4_label label = {0, 0, buflen, buf};
5591
5592         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5593         struct nfs4_getattr_arg arg = {
5594                 .fh             = NFS_FH(inode),
5595                 .bitmask        = bitmask,
5596         };
5597         struct nfs4_getattr_res res = {
5598                 .fattr          = &fattr,
5599                 .label          = &label,
5600                 .server         = server,
5601         };
5602         struct rpc_message msg = {
5603                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5604                 .rpc_argp       = &arg,
5605                 .rpc_resp       = &res,
5606         };
5607         int ret;
5608
5609         nfs_fattr_init(&fattr);
5610
5611         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5612         if (ret)
5613                 return ret;
5614         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5615                 return -ENOENT;
5616         if (buflen < label.len)
5617                 return -ERANGE;
5618         return 0;
5619 }
5620
5621 static int nfs4_get_security_label(struct inode *inode, void *buf,
5622                                         size_t buflen)
5623 {
5624         struct nfs4_exception exception = { };
5625         int err;
5626
5627         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5628                 return -EOPNOTSUPP;
5629
5630         do {
5631                 err = _nfs4_get_security_label(inode, buf, buflen);
5632                 trace_nfs4_get_security_label(inode, err);
5633                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5634                                 &exception);
5635         } while (exception.retry);
5636         return err;
5637 }
5638
5639 static int _nfs4_do_set_security_label(struct inode *inode,
5640                 struct nfs4_label *ilabel,
5641                 struct nfs_fattr *fattr,
5642                 struct nfs4_label *olabel)
5643 {
5644
5645         struct iattr sattr = {0};
5646         struct nfs_server *server = NFS_SERVER(inode);
5647         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5648         struct nfs_setattrargs arg = {
5649                 .fh             = NFS_FH(inode),
5650                 .iap            = &sattr,
5651                 .server         = server,
5652                 .bitmask        = bitmask,
5653                 .label          = ilabel,
5654         };
5655         struct nfs_setattrres res = {
5656                 .fattr          = fattr,
5657                 .label          = olabel,
5658                 .server         = server,
5659         };
5660         struct rpc_message msg = {
5661                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5662                 .rpc_argp       = &arg,
5663                 .rpc_resp       = &res,
5664         };
5665         int status;
5666
5667         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5668
5669         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5670         if (status)
5671                 dprintk("%s failed: %d\n", __func__, status);
5672
5673         return status;
5674 }
5675
5676 static int nfs4_do_set_security_label(struct inode *inode,
5677                 struct nfs4_label *ilabel,
5678                 struct nfs_fattr *fattr,
5679                 struct nfs4_label *olabel)
5680 {
5681         struct nfs4_exception exception = { };
5682         int err;
5683
5684         do {
5685                 err = _nfs4_do_set_security_label(inode, ilabel,
5686                                 fattr, olabel);
5687                 trace_nfs4_set_security_label(inode, err);
5688                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5689                                 &exception);
5690         } while (exception.retry);
5691         return err;
5692 }
5693
5694 static int
5695 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5696 {
5697         struct nfs4_label ilabel, *olabel = NULL;
5698         struct nfs_fattr fattr;
5699         int status;
5700
5701         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5702                 return -EOPNOTSUPP;
5703
5704         nfs_fattr_init(&fattr);
5705
5706         ilabel.pi = 0;
5707         ilabel.lfs = 0;
5708         ilabel.label = (char *)buf;
5709         ilabel.len = buflen;
5710
5711         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5712         if (IS_ERR(olabel)) {
5713                 status = -PTR_ERR(olabel);
5714                 goto out;
5715         }
5716
5717         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5718         if (status == 0)
5719                 nfs_setsecurity(inode, &fattr, olabel);
5720
5721         nfs4_label_free(olabel);
5722 out:
5723         return status;
5724 }
5725 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5726
5727
5728 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5729                                     nfs4_verifier *bootverf)
5730 {
5731         __be32 verf[2];
5732
5733         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5734                 /* An impossible timestamp guarantees this value
5735                  * will never match a generated boot time. */
5736                 verf[0] = cpu_to_be32(U32_MAX);
5737                 verf[1] = cpu_to_be32(U32_MAX);
5738         } else {
5739                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5740                 u64 ns = ktime_to_ns(nn->boot_time);
5741
5742                 verf[0] = cpu_to_be32(ns >> 32);
5743                 verf[1] = cpu_to_be32(ns);
5744         }
5745         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5746 }
5747
5748 static int
5749 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5750 {
5751         size_t len;
5752         char *str;
5753
5754         if (clp->cl_owner_id != NULL)
5755                 return 0;
5756
5757         rcu_read_lock();
5758         len = 14 +
5759                 strlen(clp->cl_rpcclient->cl_nodename) +
5760                 1 +
5761                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5762                 1;
5763         rcu_read_unlock();
5764         if (nfs4_client_id_uniquifier[0] != '\0')
5765                 len += strlen(nfs4_client_id_uniquifier) + 1;
5766         if (len > NFS4_OPAQUE_LIMIT + 1)
5767                 return -EINVAL;
5768
5769         /*
5770          * Since this string is allocated at mount time, and held until the
5771          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5772          * about a memory-reclaim deadlock.
5773          */
5774         str = kmalloc(len, GFP_KERNEL);
5775         if (!str)
5776                 return -ENOMEM;
5777
5778         rcu_read_lock();
5779         if (nfs4_client_id_uniquifier[0] != '\0')
5780                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5781                           clp->cl_rpcclient->cl_nodename,
5782                           nfs4_client_id_uniquifier,
5783                           rpc_peeraddr2str(clp->cl_rpcclient,
5784                                            RPC_DISPLAY_ADDR));
5785         else
5786                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
5787                           clp->cl_rpcclient->cl_nodename,
5788                           rpc_peeraddr2str(clp->cl_rpcclient,
5789                                            RPC_DISPLAY_ADDR));
5790         rcu_read_unlock();
5791
5792         clp->cl_owner_id = str;
5793         return 0;
5794 }
5795
5796 static int
5797 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5798 {
5799         size_t len;
5800         char *str;
5801
5802         len = 10 + 10 + 1 + 10 + 1 +
5803                 strlen(nfs4_client_id_uniquifier) + 1 +
5804                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5805
5806         if (len > NFS4_OPAQUE_LIMIT + 1)
5807                 return -EINVAL;
5808
5809         /*
5810          * Since this string is allocated at mount time, and held until the
5811          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5812          * about a memory-reclaim deadlock.
5813          */
5814         str = kmalloc(len, GFP_KERNEL);
5815         if (!str)
5816                 return -ENOMEM;
5817
5818         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5819                         clp->rpc_ops->version, clp->cl_minorversion,
5820                         nfs4_client_id_uniquifier,
5821                         clp->cl_rpcclient->cl_nodename);
5822         clp->cl_owner_id = str;
5823         return 0;
5824 }
5825
5826 static int
5827 nfs4_init_uniform_client_string(struct nfs_client *clp)
5828 {
5829         size_t len;
5830         char *str;
5831
5832         if (clp->cl_owner_id != NULL)
5833                 return 0;
5834
5835         if (nfs4_client_id_uniquifier[0] != '\0')
5836                 return nfs4_init_uniquifier_client_string(clp);
5837
5838         len = 10 + 10 + 1 + 10 + 1 +
5839                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5840
5841         if (len > NFS4_OPAQUE_LIMIT + 1)
5842                 return -EINVAL;
5843
5844         /*
5845          * Since this string is allocated at mount time, and held until the
5846          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5847          * about a memory-reclaim deadlock.
5848          */
5849         str = kmalloc(len, GFP_KERNEL);
5850         if (!str)
5851                 return -ENOMEM;
5852
5853         scnprintf(str, len, "Linux NFSv%u.%u %s",
5854                         clp->rpc_ops->version, clp->cl_minorversion,
5855                         clp->cl_rpcclient->cl_nodename);
5856         clp->cl_owner_id = str;
5857         return 0;
5858 }
5859
5860 /*
5861  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5862  * services.  Advertise one based on the address family of the
5863  * clientaddr.
5864  */
5865 static unsigned int
5866 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5867 {
5868         if (strchr(clp->cl_ipaddr, ':') != NULL)
5869                 return scnprintf(buf, len, "tcp6");
5870         else
5871                 return scnprintf(buf, len, "tcp");
5872 }
5873
5874 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5875 {
5876         struct nfs4_setclientid *sc = calldata;
5877
5878         if (task->tk_status == 0)
5879                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5880 }
5881
5882 static const struct rpc_call_ops nfs4_setclientid_ops = {
5883         .rpc_call_done = nfs4_setclientid_done,
5884 };
5885
5886 /**
5887  * nfs4_proc_setclientid - Negotiate client ID
5888  * @clp: state data structure
5889  * @program: RPC program for NFSv4 callback service
5890  * @port: IP port number for NFS4 callback service
5891  * @cred: credential to use for this call
5892  * @res: where to place the result
5893  *
5894  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5895  */
5896 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5897                 unsigned short port, const struct cred *cred,
5898                 struct nfs4_setclientid_res *res)
5899 {
5900         nfs4_verifier sc_verifier;
5901         struct nfs4_setclientid setclientid = {
5902                 .sc_verifier = &sc_verifier,
5903                 .sc_prog = program,
5904                 .sc_clnt = clp,
5905         };
5906         struct rpc_message msg = {
5907                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5908                 .rpc_argp = &setclientid,
5909                 .rpc_resp = res,
5910                 .rpc_cred = cred,
5911         };
5912         struct rpc_task *task;
5913         struct rpc_task_setup task_setup_data = {
5914                 .rpc_client = clp->cl_rpcclient,
5915                 .rpc_message = &msg,
5916                 .callback_ops = &nfs4_setclientid_ops,
5917                 .callback_data = &setclientid,
5918                 .flags = RPC_TASK_TIMEOUT,
5919         };
5920         int status;
5921
5922         /* nfs_client_id4 */
5923         nfs4_init_boot_verifier(clp, &sc_verifier);
5924
5925         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5926                 status = nfs4_init_uniform_client_string(clp);
5927         else
5928                 status = nfs4_init_nonuniform_client_string(clp);
5929
5930         if (status)
5931                 goto out;
5932
5933         /* cb_client4 */
5934         setclientid.sc_netid_len =
5935                                 nfs4_init_callback_netid(clp,
5936                                                 setclientid.sc_netid,
5937                                                 sizeof(setclientid.sc_netid));
5938         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5939                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5940                                 clp->cl_ipaddr, port >> 8, port & 255);
5941
5942         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5943                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5944                 clp->cl_owner_id);
5945         task = rpc_run_task(&task_setup_data);
5946         if (IS_ERR(task)) {
5947                 status = PTR_ERR(task);
5948                 goto out;
5949         }
5950         status = task->tk_status;
5951         if (setclientid.sc_cred) {
5952                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5953                 put_rpccred(setclientid.sc_cred);
5954         }
5955         rpc_put_task(task);
5956 out:
5957         trace_nfs4_setclientid(clp, status);
5958         dprintk("NFS reply setclientid: %d\n", status);
5959         return status;
5960 }
5961
5962 /**
5963  * nfs4_proc_setclientid_confirm - Confirm client ID
5964  * @clp: state data structure
5965  * @arg: result of a previous SETCLIENTID
5966  * @cred: credential to use for this call
5967  *
5968  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5969  */
5970 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5971                 struct nfs4_setclientid_res *arg,
5972                 const struct cred *cred)
5973 {
5974         struct rpc_message msg = {
5975                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5976                 .rpc_argp = arg,
5977                 .rpc_cred = cred,
5978         };
5979         int status;
5980
5981         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5982                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5983                 clp->cl_clientid);
5984         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5985         trace_nfs4_setclientid_confirm(clp, status);
5986         dprintk("NFS reply setclientid_confirm: %d\n", status);
5987         return status;
5988 }
5989
5990 struct nfs4_delegreturndata {
5991         struct nfs4_delegreturnargs args;
5992         struct nfs4_delegreturnres res;
5993         struct nfs_fh fh;
5994         nfs4_stateid stateid;
5995         unsigned long timestamp;
5996         struct {
5997                 struct nfs4_layoutreturn_args arg;
5998                 struct nfs4_layoutreturn_res res;
5999                 struct nfs4_xdr_opaque_data ld_private;
6000                 u32 roc_barrier;
6001                 bool roc;
6002         } lr;
6003         struct nfs_fattr fattr;
6004         int rpc_status;
6005         struct inode *inode;
6006 };
6007
6008 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6009 {
6010         struct nfs4_delegreturndata *data = calldata;
6011         struct nfs4_exception exception = {
6012                 .inode = data->inode,
6013                 .stateid = &data->stateid,
6014         };
6015
6016         if (!nfs4_sequence_done(task, &data->res.seq_res))
6017                 return;
6018
6019         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6020
6021         /* Handle Layoutreturn errors */
6022         if (data->args.lr_args && task->tk_status != 0) {
6023                 switch(data->res.lr_ret) {
6024                 default:
6025                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6026                         break;
6027                 case 0:
6028                         data->args.lr_args = NULL;
6029                         data->res.lr_res = NULL;
6030                         break;
6031                 case -NFS4ERR_OLD_STATEID:
6032                         if (nfs4_layoutreturn_refresh_stateid(&data->args.lr_args->stateid,
6033                                                 &data->args.lr_args->range,
6034                                                 data->inode))
6035                                 goto lr_restart;
6036                         /* Fallthrough */
6037                 case -NFS4ERR_ADMIN_REVOKED:
6038                 case -NFS4ERR_DELEG_REVOKED:
6039                 case -NFS4ERR_EXPIRED:
6040                 case -NFS4ERR_BAD_STATEID:
6041                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
6042                 case -NFS4ERR_WRONG_CRED:
6043                         data->args.lr_args = NULL;
6044                         data->res.lr_res = NULL;
6045                         goto lr_restart;
6046                 }
6047         }
6048
6049         switch (task->tk_status) {
6050         case 0:
6051                 renew_lease(data->res.server, data->timestamp);
6052                 break;
6053         case -NFS4ERR_ADMIN_REVOKED:
6054         case -NFS4ERR_DELEG_REVOKED:
6055         case -NFS4ERR_EXPIRED:
6056                 nfs4_free_revoked_stateid(data->res.server,
6057                                 data->args.stateid,
6058                                 task->tk_msg.rpc_cred);
6059                 /* Fallthrough */
6060         case -NFS4ERR_BAD_STATEID:
6061         case -NFS4ERR_STALE_STATEID:
6062                 task->tk_status = 0;
6063                 break;
6064         case -NFS4ERR_OLD_STATEID:
6065                 if (nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6066                         goto out_restart;
6067                 task->tk_status = 0;
6068                 break;
6069         case -NFS4ERR_ACCESS:
6070                 if (data->args.bitmask) {
6071                         data->args.bitmask = NULL;
6072                         data->res.fattr = NULL;
6073                         goto out_restart;
6074                 }
6075                 /* Fallthrough */
6076         default:
6077                 task->tk_status = nfs4_async_handle_exception(task,
6078                                 data->res.server, task->tk_status,
6079                                 &exception);
6080                 if (exception.retry)
6081                         goto out_restart;
6082         }
6083         data->rpc_status = task->tk_status;
6084         return;
6085 lr_restart:
6086         data->res.lr_ret = 0;
6087 out_restart:
6088         task->tk_status = 0;
6089         rpc_restart_call_prepare(task);
6090 }
6091
6092 static void nfs4_delegreturn_release(void *calldata)
6093 {
6094         struct nfs4_delegreturndata *data = calldata;
6095         struct inode *inode = data->inode;
6096
6097         if (inode) {
6098                 if (data->lr.roc)
6099                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6100                                         data->res.lr_ret);
6101                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6102                 nfs_iput_and_deactive(inode);
6103         }
6104         kfree(calldata);
6105 }
6106
6107 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6108 {
6109         struct nfs4_delegreturndata *d_data;
6110         struct pnfs_layout_hdr *lo;
6111
6112         d_data = (struct nfs4_delegreturndata *)data;
6113
6114         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
6115                 return;
6116
6117         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6118         if (lo && !pnfs_layout_is_valid(lo)) {
6119                 d_data->args.lr_args = NULL;
6120                 d_data->res.lr_res = NULL;
6121         }
6122
6123         nfs4_setup_sequence(d_data->res.server->nfs_client,
6124                         &d_data->args.seq_args,
6125                         &d_data->res.seq_res,
6126                         task);
6127 }
6128
6129 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6130         .rpc_call_prepare = nfs4_delegreturn_prepare,
6131         .rpc_call_done = nfs4_delegreturn_done,
6132         .rpc_release = nfs4_delegreturn_release,
6133 };
6134
6135 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6136 {
6137         struct nfs4_delegreturndata *data;
6138         struct nfs_server *server = NFS_SERVER(inode);
6139         struct rpc_task *task;
6140         struct rpc_message msg = {
6141                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6142                 .rpc_cred = cred,
6143         };
6144         struct rpc_task_setup task_setup_data = {
6145                 .rpc_client = server->client,
6146                 .rpc_message = &msg,
6147                 .callback_ops = &nfs4_delegreturn_ops,
6148                 .flags = RPC_TASK_ASYNC,
6149         };
6150         int status = 0;
6151
6152         data = kzalloc(sizeof(*data), GFP_NOFS);
6153         if (data == NULL)
6154                 return -ENOMEM;
6155         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6156
6157         nfs4_state_protect(server->nfs_client,
6158                         NFS_SP4_MACH_CRED_CLEANUP,
6159                         &task_setup_data.rpc_client, &msg);
6160
6161         data->args.fhandle = &data->fh;
6162         data->args.stateid = &data->stateid;
6163         data->args.bitmask = server->cache_consistency_bitmask;
6164         nfs_copy_fh(&data->fh, NFS_FH(inode));
6165         nfs4_stateid_copy(&data->stateid, stateid);
6166         data->res.fattr = &data->fattr;
6167         data->res.server = server;
6168         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6169         data->lr.arg.ld_private = &data->lr.ld_private;
6170         nfs_fattr_init(data->res.fattr);
6171         data->timestamp = jiffies;
6172         data->rpc_status = 0;
6173         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6174         data->inode = nfs_igrab_and_active(inode);
6175         if (data->inode) {
6176                 if (data->lr.roc) {
6177                         data->args.lr_args = &data->lr.arg;
6178                         data->res.lr_res = &data->lr.res;
6179                 }
6180         } else if (data->lr.roc) {
6181                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6182                 data->lr.roc = false;
6183         }
6184
6185         task_setup_data.callback_data = data;
6186         msg.rpc_argp = &data->args;
6187         msg.rpc_resp = &data->res;
6188         task = rpc_run_task(&task_setup_data);
6189         if (IS_ERR(task))
6190                 return PTR_ERR(task);
6191         if (!issync)
6192                 goto out;
6193         status = rpc_wait_for_completion_task(task);
6194         if (status != 0)
6195                 goto out;
6196         status = data->rpc_status;
6197 out:
6198         rpc_put_task(task);
6199         return status;
6200 }
6201
6202 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6203 {
6204         struct nfs_server *server = NFS_SERVER(inode);
6205         struct nfs4_exception exception = { };
6206         int err;
6207         do {
6208                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6209                 trace_nfs4_delegreturn(inode, stateid, err);
6210                 switch (err) {
6211                         case -NFS4ERR_STALE_STATEID:
6212                         case -NFS4ERR_EXPIRED:
6213                         case 0:
6214                                 return 0;
6215                 }
6216                 err = nfs4_handle_exception(server, err, &exception);
6217         } while (exception.retry);
6218         return err;
6219 }
6220
6221 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6222 {
6223         struct inode *inode = state->inode;
6224         struct nfs_server *server = NFS_SERVER(inode);
6225         struct nfs_client *clp = server->nfs_client;
6226         struct nfs_lockt_args arg = {
6227                 .fh = NFS_FH(inode),
6228                 .fl = request,
6229         };
6230         struct nfs_lockt_res res = {
6231                 .denied = request,
6232         };
6233         struct rpc_message msg = {
6234                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6235                 .rpc_argp       = &arg,
6236                 .rpc_resp       = &res,
6237                 .rpc_cred       = state->owner->so_cred,
6238         };
6239         struct nfs4_lock_state *lsp;
6240         int status;
6241
6242         arg.lock_owner.clientid = clp->cl_clientid;
6243         status = nfs4_set_lock_state(state, request);
6244         if (status != 0)
6245                 goto out;
6246         lsp = request->fl_u.nfs4_fl.owner;
6247         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6248         arg.lock_owner.s_dev = server->s_dev;
6249         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6250         switch (status) {
6251                 case 0:
6252                         request->fl_type = F_UNLCK;
6253                         break;
6254                 case -NFS4ERR_DENIED:
6255                         status = 0;
6256         }
6257         request->fl_ops->fl_release_private(request);
6258         request->fl_ops = NULL;
6259 out:
6260         return status;
6261 }
6262
6263 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6264 {
6265         struct nfs4_exception exception = { };
6266         int err;
6267
6268         do {
6269                 err = _nfs4_proc_getlk(state, cmd, request);
6270                 trace_nfs4_get_lock(request, state, cmd, err);
6271                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6272                                 &exception);
6273         } while (exception.retry);
6274         return err;
6275 }
6276
6277 struct nfs4_unlockdata {
6278         struct nfs_locku_args arg;
6279         struct nfs_locku_res res;
6280         struct nfs4_lock_state *lsp;
6281         struct nfs_open_context *ctx;
6282         struct nfs_lock_context *l_ctx;
6283         struct file_lock fl;
6284         struct nfs_server *server;
6285         unsigned long timestamp;
6286 };
6287
6288 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6289                 struct nfs_open_context *ctx,
6290                 struct nfs4_lock_state *lsp,
6291                 struct nfs_seqid *seqid)
6292 {
6293         struct nfs4_unlockdata *p;
6294         struct inode *inode = lsp->ls_state->inode;
6295
6296         p = kzalloc(sizeof(*p), GFP_NOFS);
6297         if (p == NULL)
6298                 return NULL;
6299         p->arg.fh = NFS_FH(inode);
6300         p->arg.fl = &p->fl;
6301         p->arg.seqid = seqid;
6302         p->res.seqid = seqid;
6303         p->lsp = lsp;
6304         refcount_inc(&lsp->ls_count);
6305         /* Ensure we don't close file until we're done freeing locks! */
6306         p->ctx = get_nfs_open_context(ctx);
6307         p->l_ctx = nfs_get_lock_context(ctx);
6308         locks_init_lock(&p->fl);
6309         locks_copy_lock(&p->fl, fl);
6310         p->server = NFS_SERVER(inode);
6311         return p;
6312 }
6313
6314 static void nfs4_locku_release_calldata(void *data)
6315 {
6316         struct nfs4_unlockdata *calldata = data;
6317         nfs_free_seqid(calldata->arg.seqid);
6318         nfs4_put_lock_state(calldata->lsp);
6319         nfs_put_lock_context(calldata->l_ctx);
6320         put_nfs_open_context(calldata->ctx);
6321         kfree(calldata);
6322 }
6323
6324 static void nfs4_locku_done(struct rpc_task *task, void *data)
6325 {
6326         struct nfs4_unlockdata *calldata = data;
6327         struct nfs4_exception exception = {
6328                 .inode = calldata->lsp->ls_state->inode,
6329                 .stateid = &calldata->arg.stateid,
6330         };
6331
6332         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6333                 return;
6334         switch (task->tk_status) {
6335                 case 0:
6336                         renew_lease(calldata->server, calldata->timestamp);
6337                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6338                         if (nfs4_update_lock_stateid(calldata->lsp,
6339                                         &calldata->res.stateid))
6340                                 break;
6341                         /* Fall through */
6342                 case -NFS4ERR_ADMIN_REVOKED:
6343                 case -NFS4ERR_EXPIRED:
6344                         nfs4_free_revoked_stateid(calldata->server,
6345                                         &calldata->arg.stateid,
6346                                         task->tk_msg.rpc_cred);
6347                         /* Fall through */
6348                 case -NFS4ERR_BAD_STATEID:
6349                 case -NFS4ERR_OLD_STATEID:
6350                 case -NFS4ERR_STALE_STATEID:
6351                         if (!nfs4_stateid_match(&calldata->arg.stateid,
6352                                                 &calldata->lsp->ls_stateid))
6353                                 rpc_restart_call_prepare(task);
6354                         break;
6355                 default:
6356                         task->tk_status = nfs4_async_handle_exception(task,
6357                                         calldata->server, task->tk_status,
6358                                         &exception);
6359                         if (exception.retry)
6360                                 rpc_restart_call_prepare(task);
6361         }
6362         nfs_release_seqid(calldata->arg.seqid);
6363 }
6364
6365 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6366 {
6367         struct nfs4_unlockdata *calldata = data;
6368
6369         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6370                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6371                 return;
6372
6373         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6374                 goto out_wait;
6375         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
6376         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6377                 /* Note: exit _without_ running nfs4_locku_done */
6378                 goto out_no_action;
6379         }
6380         calldata->timestamp = jiffies;
6381         if (nfs4_setup_sequence(calldata->server->nfs_client,
6382                                 &calldata->arg.seq_args,
6383                                 &calldata->res.seq_res,
6384                                 task) != 0)
6385                 nfs_release_seqid(calldata->arg.seqid);
6386         return;
6387 out_no_action:
6388         task->tk_action = NULL;
6389 out_wait:
6390         nfs4_sequence_done(task, &calldata->res.seq_res);
6391 }
6392
6393 static const struct rpc_call_ops nfs4_locku_ops = {
6394         .rpc_call_prepare = nfs4_locku_prepare,
6395         .rpc_call_done = nfs4_locku_done,
6396         .rpc_release = nfs4_locku_release_calldata,
6397 };
6398
6399 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6400                 struct nfs_open_context *ctx,
6401                 struct nfs4_lock_state *lsp,
6402                 struct nfs_seqid *seqid)
6403 {
6404         struct nfs4_unlockdata *data;
6405         struct rpc_message msg = {
6406                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6407                 .rpc_cred = ctx->cred,
6408         };
6409         struct rpc_task_setup task_setup_data = {
6410                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6411                 .rpc_message = &msg,
6412                 .callback_ops = &nfs4_locku_ops,
6413                 .workqueue = nfsiod_workqueue,
6414                 .flags = RPC_TASK_ASYNC,
6415         };
6416
6417         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6418                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6419
6420         /* Ensure this is an unlock - when canceling a lock, the
6421          * canceled lock is passed in, and it won't be an unlock.
6422          */
6423         fl->fl_type = F_UNLCK;
6424         if (fl->fl_flags & FL_CLOSE)
6425                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6426
6427         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6428         if (data == NULL) {
6429                 nfs_free_seqid(seqid);
6430                 return ERR_PTR(-ENOMEM);
6431         }
6432
6433         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6434         msg.rpc_argp = &data->arg;
6435         msg.rpc_resp = &data->res;
6436         task_setup_data.callback_data = data;
6437         return rpc_run_task(&task_setup_data);
6438 }
6439
6440 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6441 {
6442         struct inode *inode = state->inode;
6443         struct nfs4_state_owner *sp = state->owner;
6444         struct nfs_inode *nfsi = NFS_I(inode);
6445         struct nfs_seqid *seqid;
6446         struct nfs4_lock_state *lsp;
6447         struct rpc_task *task;
6448         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6449         int status = 0;
6450         unsigned char fl_flags = request->fl_flags;
6451
6452         status = nfs4_set_lock_state(state, request);
6453         /* Unlock _before_ we do the RPC call */
6454         request->fl_flags |= FL_EXISTS;
6455         /* Exclude nfs_delegation_claim_locks() */
6456         mutex_lock(&sp->so_delegreturn_mutex);
6457         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6458         down_read(&nfsi->rwsem);
6459         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6460                 up_read(&nfsi->rwsem);
6461                 mutex_unlock(&sp->so_delegreturn_mutex);
6462                 goto out;
6463         }
6464         up_read(&nfsi->rwsem);
6465         mutex_unlock(&sp->so_delegreturn_mutex);
6466         if (status != 0)
6467                 goto out;
6468         /* Is this a delegated lock? */
6469         lsp = request->fl_u.nfs4_fl.owner;
6470         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6471                 goto out;
6472         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6473         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6474         status = -ENOMEM;
6475         if (IS_ERR(seqid))
6476                 goto out;
6477         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6478         status = PTR_ERR(task);
6479         if (IS_ERR(task))
6480                 goto out;
6481         status = rpc_wait_for_completion_task(task);
6482         rpc_put_task(task);
6483 out:
6484         request->fl_flags = fl_flags;
6485         trace_nfs4_unlock(request, state, F_SETLK, status);
6486         return status;
6487 }
6488
6489 struct nfs4_lockdata {
6490         struct nfs_lock_args arg;
6491         struct nfs_lock_res res;
6492         struct nfs4_lock_state *lsp;
6493         struct nfs_open_context *ctx;
6494         struct file_lock fl;
6495         unsigned long timestamp;
6496         int rpc_status;
6497         int cancelled;
6498         struct nfs_server *server;
6499 };
6500
6501 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6502                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6503                 gfp_t gfp_mask)
6504 {
6505         struct nfs4_lockdata *p;
6506         struct inode *inode = lsp->ls_state->inode;
6507         struct nfs_server *server = NFS_SERVER(inode);
6508         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6509
6510         p = kzalloc(sizeof(*p), gfp_mask);
6511         if (p == NULL)
6512                 return NULL;
6513
6514         p->arg.fh = NFS_FH(inode);
6515         p->arg.fl = &p->fl;
6516         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6517         if (IS_ERR(p->arg.open_seqid))
6518                 goto out_free;
6519         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6520         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6521         if (IS_ERR(p->arg.lock_seqid))
6522                 goto out_free_seqid;
6523         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6524         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6525         p->arg.lock_owner.s_dev = server->s_dev;
6526         p->res.lock_seqid = p->arg.lock_seqid;
6527         p->lsp = lsp;
6528         p->server = server;
6529         refcount_inc(&lsp->ls_count);
6530         p->ctx = get_nfs_open_context(ctx);
6531         locks_init_lock(&p->fl);
6532         locks_copy_lock(&p->fl, fl);
6533         return p;
6534 out_free_seqid:
6535         nfs_free_seqid(p->arg.open_seqid);
6536 out_free:
6537         kfree(p);
6538         return NULL;
6539 }
6540
6541 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6542 {
6543         struct nfs4_lockdata *data = calldata;
6544         struct nfs4_state *state = data->lsp->ls_state;
6545
6546         dprintk("%s: begin!\n", __func__);
6547         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6548                 goto out_wait;
6549         /* Do we need to do an open_to_lock_owner? */
6550         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6551                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6552                         goto out_release_lock_seqid;
6553                 }
6554                 nfs4_stateid_copy(&data->arg.open_stateid,
6555                                 &state->open_stateid);
6556                 data->arg.new_lock_owner = 1;
6557                 data->res.open_seqid = data->arg.open_seqid;
6558         } else {
6559                 data->arg.new_lock_owner = 0;
6560                 nfs4_stateid_copy(&data->arg.lock_stateid,
6561                                 &data->lsp->ls_stateid);
6562         }
6563         if (!nfs4_valid_open_stateid(state)) {
6564                 data->rpc_status = -EBADF;
6565                 task->tk_action = NULL;
6566                 goto out_release_open_seqid;
6567         }
6568         data->timestamp = jiffies;
6569         if (nfs4_setup_sequence(data->server->nfs_client,
6570                                 &data->arg.seq_args,
6571                                 &data->res.seq_res,
6572                                 task) == 0)
6573                 return;
6574 out_release_open_seqid:
6575         nfs_release_seqid(data->arg.open_seqid);
6576 out_release_lock_seqid:
6577         nfs_release_seqid(data->arg.lock_seqid);
6578 out_wait:
6579         nfs4_sequence_done(task, &data->res.seq_res);
6580         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6581 }
6582
6583 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6584 {
6585         struct nfs4_lockdata *data = calldata;
6586         struct nfs4_lock_state *lsp = data->lsp;
6587
6588         dprintk("%s: begin!\n", __func__);
6589
6590         if (!nfs4_sequence_done(task, &data->res.seq_res))
6591                 return;
6592
6593         data->rpc_status = task->tk_status;
6594         switch (task->tk_status) {
6595         case 0:
6596                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6597                                 data->timestamp);
6598                 if (data->arg.new_lock && !data->cancelled) {
6599                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6600                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6601                                 goto out_restart;
6602                 }
6603                 if (data->arg.new_lock_owner != 0) {
6604                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6605                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6606                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6607                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6608                         goto out_restart;
6609                 break;
6610         case -NFS4ERR_BAD_STATEID:
6611         case -NFS4ERR_OLD_STATEID:
6612         case -NFS4ERR_STALE_STATEID:
6613         case -NFS4ERR_EXPIRED:
6614                 if (data->arg.new_lock_owner != 0) {
6615                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6616                                                 &lsp->ls_state->open_stateid))
6617                                 goto out_restart;
6618                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6619                                                 &lsp->ls_stateid))
6620                                 goto out_restart;
6621         }
6622 out_done:
6623         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6624         return;
6625 out_restart:
6626         if (!data->cancelled)
6627                 rpc_restart_call_prepare(task);
6628         goto out_done;
6629 }
6630
6631 static void nfs4_lock_release(void *calldata)
6632 {
6633         struct nfs4_lockdata *data = calldata;
6634
6635         dprintk("%s: begin!\n", __func__);
6636         nfs_free_seqid(data->arg.open_seqid);
6637         if (data->cancelled && data->rpc_status == 0) {
6638                 struct rpc_task *task;
6639                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6640                                 data->arg.lock_seqid);
6641                 if (!IS_ERR(task))
6642                         rpc_put_task_async(task);
6643                 dprintk("%s: cancelling lock!\n", __func__);
6644         } else
6645                 nfs_free_seqid(data->arg.lock_seqid);
6646         nfs4_put_lock_state(data->lsp);
6647         put_nfs_open_context(data->ctx);
6648         kfree(data);
6649         dprintk("%s: done!\n", __func__);
6650 }
6651
6652 static const struct rpc_call_ops nfs4_lock_ops = {
6653         .rpc_call_prepare = nfs4_lock_prepare,
6654         .rpc_call_done = nfs4_lock_done,
6655         .rpc_release = nfs4_lock_release,
6656 };
6657
6658 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6659 {
6660         switch (error) {
6661         case -NFS4ERR_ADMIN_REVOKED:
6662         case -NFS4ERR_EXPIRED:
6663         case -NFS4ERR_BAD_STATEID:
6664                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6665                 if (new_lock_owner != 0 ||
6666                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6667                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6668                 break;
6669         case -NFS4ERR_STALE_STATEID:
6670                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6671                 nfs4_schedule_lease_recovery(server->nfs_client);
6672         };
6673 }
6674
6675 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6676 {
6677         struct nfs4_lockdata *data;
6678         struct rpc_task *task;
6679         struct rpc_message msg = {
6680                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6681                 .rpc_cred = state->owner->so_cred,
6682         };
6683         struct rpc_task_setup task_setup_data = {
6684                 .rpc_client = NFS_CLIENT(state->inode),
6685                 .rpc_message = &msg,
6686                 .callback_ops = &nfs4_lock_ops,
6687                 .workqueue = nfsiod_workqueue,
6688                 .flags = RPC_TASK_ASYNC,
6689         };
6690         int ret;
6691
6692         dprintk("%s: begin!\n", __func__);
6693         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6694                         fl->fl_u.nfs4_fl.owner,
6695                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6696         if (data == NULL)
6697                 return -ENOMEM;
6698         if (IS_SETLKW(cmd))
6699                 data->arg.block = 1;
6700         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6701                                 recovery_type > NFS_LOCK_NEW);
6702         msg.rpc_argp = &data->arg;
6703         msg.rpc_resp = &data->res;
6704         task_setup_data.callback_data = data;
6705         if (recovery_type > NFS_LOCK_NEW) {
6706                 if (recovery_type == NFS_LOCK_RECLAIM)
6707                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6708         } else
6709                 data->arg.new_lock = 1;
6710         task = rpc_run_task(&task_setup_data);
6711         if (IS_ERR(task))
6712                 return PTR_ERR(task);
6713         ret = rpc_wait_for_completion_task(task);
6714         if (ret == 0) {
6715                 ret = data->rpc_status;
6716                 if (ret)
6717                         nfs4_handle_setlk_error(data->server, data->lsp,
6718                                         data->arg.new_lock_owner, ret);
6719         } else
6720                 data->cancelled = true;
6721         rpc_put_task(task);
6722         dprintk("%s: done, ret = %d!\n", __func__, ret);
6723         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6724         return ret;
6725 }
6726
6727 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6728 {
6729         struct nfs_server *server = NFS_SERVER(state->inode);
6730         struct nfs4_exception exception = {
6731                 .inode = state->inode,
6732         };
6733         int err;
6734
6735         do {
6736                 /* Cache the lock if possible... */
6737                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6738                         return 0;
6739                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6740                 if (err != -NFS4ERR_DELAY)
6741                         break;
6742                 nfs4_handle_exception(server, err, &exception);
6743         } while (exception.retry);
6744         return err;
6745 }
6746
6747 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6748 {
6749         struct nfs_server *server = NFS_SERVER(state->inode);
6750         struct nfs4_exception exception = {
6751                 .inode = state->inode,
6752         };
6753         int err;
6754
6755         err = nfs4_set_lock_state(state, request);
6756         if (err != 0)
6757                 return err;
6758         if (!recover_lost_locks) {
6759                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6760                 return 0;
6761         }
6762         do {
6763                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6764                         return 0;
6765                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6766                 switch (err) {
6767                 default:
6768                         goto out;
6769                 case -NFS4ERR_GRACE:
6770                 case -NFS4ERR_DELAY:
6771                         nfs4_handle_exception(server, err, &exception);
6772                         err = 0;
6773                 }
6774         } while (exception.retry);
6775 out:
6776         return err;
6777 }
6778
6779 #if defined(CONFIG_NFS_V4_1)
6780 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6781 {
6782         struct nfs4_lock_state *lsp;
6783         int status;
6784
6785         status = nfs4_set_lock_state(state, request);
6786         if (status != 0)
6787                 return status;
6788         lsp = request->fl_u.nfs4_fl.owner;
6789         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6790             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6791                 return 0;
6792         return nfs4_lock_expired(state, request);
6793 }
6794 #endif
6795
6796 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6797 {
6798         struct nfs_inode *nfsi = NFS_I(state->inode);
6799         struct nfs4_state_owner *sp = state->owner;
6800         unsigned char fl_flags = request->fl_flags;
6801         int status;
6802
6803         request->fl_flags |= FL_ACCESS;
6804         status = locks_lock_inode_wait(state->inode, request);
6805         if (status < 0)
6806                 goto out;
6807         mutex_lock(&sp->so_delegreturn_mutex);
6808         down_read(&nfsi->rwsem);
6809         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6810                 /* Yes: cache locks! */
6811                 /* ...but avoid races with delegation recall... */
6812                 request->fl_flags = fl_flags & ~FL_SLEEP;
6813                 status = locks_lock_inode_wait(state->inode, request);
6814                 up_read(&nfsi->rwsem);
6815                 mutex_unlock(&sp->so_delegreturn_mutex);
6816                 goto out;
6817         }
6818         up_read(&nfsi->rwsem);
6819         mutex_unlock(&sp->so_delegreturn_mutex);
6820         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6821 out:
6822         request->fl_flags = fl_flags;
6823         return status;
6824 }
6825
6826 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6827 {
6828         struct nfs4_exception exception = {
6829                 .state = state,
6830                 .inode = state->inode,
6831         };
6832         int err;
6833
6834         do {
6835                 err = _nfs4_proc_setlk(state, cmd, request);
6836                 if (err == -NFS4ERR_DENIED)
6837                         err = -EAGAIN;
6838                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6839                                 err, &exception);
6840         } while (exception.retry);
6841         return err;
6842 }
6843
6844 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6845 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6846
6847 static int
6848 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6849                         struct file_lock *request)
6850 {
6851         int             status = -ERESTARTSYS;
6852         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6853
6854         while(!signalled()) {
6855                 status = nfs4_proc_setlk(state, cmd, request);
6856                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6857                         break;
6858                 freezable_schedule_timeout_interruptible(timeout);
6859                 timeout *= 2;
6860                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6861                 status = -ERESTARTSYS;
6862         }
6863         return status;
6864 }
6865
6866 #ifdef CONFIG_NFS_V4_1
6867 struct nfs4_lock_waiter {
6868         struct task_struct      *task;
6869         struct inode            *inode;
6870         struct nfs_lowner       *owner;
6871         bool                    notified;
6872 };
6873
6874 static int
6875 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6876 {
6877         int ret;
6878         struct nfs4_lock_waiter *waiter = wait->private;
6879
6880         /* NULL key means to wake up everyone */
6881         if (key) {
6882                 struct cb_notify_lock_args      *cbnl = key;
6883                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
6884                                                 *wowner = waiter->owner;
6885
6886                 /* Only wake if the callback was for the same owner. */
6887                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
6888                         return 0;
6889
6890                 /* Make sure it's for the right inode */
6891                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6892                         return 0;
6893
6894                 waiter->notified = true;
6895         }
6896
6897         /* override "private" so we can use default_wake_function */
6898         wait->private = waiter->task;
6899         ret = autoremove_wake_function(wait, mode, flags, key);
6900         wait->private = waiter;
6901         return ret;
6902 }
6903
6904 static int
6905 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6906 {
6907         int status = -ERESTARTSYS;
6908         unsigned long flags;
6909         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6910         struct nfs_server *server = NFS_SERVER(state->inode);
6911         struct nfs_client *clp = server->nfs_client;
6912         wait_queue_head_t *q = &clp->cl_lock_waitq;
6913         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6914                                     .id = lsp->ls_seqid.owner_id,
6915                                     .s_dev = server->s_dev };
6916         struct nfs4_lock_waiter waiter = { .task  = current,
6917                                            .inode = state->inode,
6918                                            .owner = &owner,
6919                                            .notified = false };
6920         wait_queue_entry_t wait;
6921
6922         /* Don't bother with waitqueue if we don't expect a callback */
6923         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6924                 return nfs4_retry_setlk_simple(state, cmd, request);
6925
6926         init_wait(&wait);
6927         wait.private = &waiter;
6928         wait.func = nfs4_wake_lock_waiter;
6929         add_wait_queue(q, &wait);
6930
6931         while(!signalled()) {
6932                 waiter.notified = false;
6933                 status = nfs4_proc_setlk(state, cmd, request);
6934                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6935                         break;
6936
6937                 status = -ERESTARTSYS;
6938                 spin_lock_irqsave(&q->lock, flags);
6939                 if (waiter.notified) {
6940                         spin_unlock_irqrestore(&q->lock, flags);
6941                         continue;
6942                 }
6943                 set_current_state(TASK_INTERRUPTIBLE);
6944                 spin_unlock_irqrestore(&q->lock, flags);
6945
6946                 freezable_schedule_timeout(NFS4_LOCK_MAXTIMEOUT);
6947         }
6948
6949         finish_wait(q, &wait);
6950         return status;
6951 }
6952 #else /* !CONFIG_NFS_V4_1 */
6953 static inline int
6954 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6955 {
6956         return nfs4_retry_setlk_simple(state, cmd, request);
6957 }
6958 #endif
6959
6960 static int
6961 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6962 {
6963         struct nfs_open_context *ctx;
6964         struct nfs4_state *state;
6965         int status;
6966
6967         /* verify open state */
6968         ctx = nfs_file_open_context(filp);
6969         state = ctx->state;
6970
6971         if (IS_GETLK(cmd)) {
6972                 if (state != NULL)
6973                         return nfs4_proc_getlk(state, F_GETLK, request);
6974                 return 0;
6975         }
6976
6977         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6978                 return -EINVAL;
6979
6980         if (request->fl_type == F_UNLCK) {
6981                 if (state != NULL)
6982                         return nfs4_proc_unlck(state, cmd, request);
6983                 return 0;
6984         }
6985
6986         if (state == NULL)
6987                 return -ENOLCK;
6988
6989         if ((request->fl_flags & FL_POSIX) &&
6990             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6991                 return -ENOLCK;
6992
6993         /*
6994          * Don't rely on the VFS having checked the file open mode,
6995          * since it won't do this for flock() locks.
6996          */
6997         switch (request->fl_type) {
6998         case F_RDLCK:
6999                 if (!(filp->f_mode & FMODE_READ))
7000                         return -EBADF;
7001                 break;
7002         case F_WRLCK:
7003                 if (!(filp->f_mode & FMODE_WRITE))
7004                         return -EBADF;
7005         }
7006
7007         status = nfs4_set_lock_state(state, request);
7008         if (status != 0)
7009                 return status;
7010
7011         return nfs4_retry_setlk(state, cmd, request);
7012 }
7013
7014 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7015 {
7016         struct nfs_server *server = NFS_SERVER(state->inode);
7017         int err;
7018
7019         err = nfs4_set_lock_state(state, fl);
7020         if (err != 0)
7021                 return err;
7022         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7023         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7024 }
7025
7026 struct nfs_release_lockowner_data {
7027         struct nfs4_lock_state *lsp;
7028         struct nfs_server *server;
7029         struct nfs_release_lockowner_args args;
7030         struct nfs_release_lockowner_res res;
7031         unsigned long timestamp;
7032 };
7033
7034 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7035 {
7036         struct nfs_release_lockowner_data *data = calldata;
7037         struct nfs_server *server = data->server;
7038         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7039                            &data->res.seq_res, task);
7040         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7041         data->timestamp = jiffies;
7042 }
7043
7044 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7045 {
7046         struct nfs_release_lockowner_data *data = calldata;
7047         struct nfs_server *server = data->server;
7048
7049         nfs40_sequence_done(task, &data->res.seq_res);
7050
7051         switch (task->tk_status) {
7052         case 0:
7053                 renew_lease(server, data->timestamp);
7054                 break;
7055         case -NFS4ERR_STALE_CLIENTID:
7056         case -NFS4ERR_EXPIRED:
7057                 nfs4_schedule_lease_recovery(server->nfs_client);
7058                 break;
7059         case -NFS4ERR_LEASE_MOVED:
7060         case -NFS4ERR_DELAY:
7061                 if (nfs4_async_handle_error(task, server,
7062                                             NULL, NULL) == -EAGAIN)
7063                         rpc_restart_call_prepare(task);
7064         }
7065 }
7066
7067 static void nfs4_release_lockowner_release(void *calldata)
7068 {
7069         struct nfs_release_lockowner_data *data = calldata;
7070         nfs4_free_lock_state(data->server, data->lsp);
7071         kfree(calldata);
7072 }
7073
7074 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7075         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7076         .rpc_call_done = nfs4_release_lockowner_done,
7077         .rpc_release = nfs4_release_lockowner_release,
7078 };
7079
7080 static void
7081 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7082 {
7083         struct nfs_release_lockowner_data *data;
7084         struct rpc_message msg = {
7085                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7086         };
7087
7088         if (server->nfs_client->cl_mvops->minor_version != 0)
7089                 return;
7090
7091         data = kmalloc(sizeof(*data), GFP_NOFS);
7092         if (!data)
7093                 return;
7094         data->lsp = lsp;
7095         data->server = server;
7096         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7097         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7098         data->args.lock_owner.s_dev = server->s_dev;
7099
7100         msg.rpc_argp = &data->args;
7101         msg.rpc_resp = &data->res;
7102         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7103         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7104 }
7105
7106 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7107
7108 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7109                                    struct dentry *unused, struct inode *inode,
7110                                    const char *key, const void *buf,
7111                                    size_t buflen, int flags)
7112 {
7113         return nfs4_proc_set_acl(inode, buf, buflen);
7114 }
7115
7116 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7117                                    struct dentry *unused, struct inode *inode,
7118                                    const char *key, void *buf, size_t buflen)
7119 {
7120         return nfs4_proc_get_acl(inode, buf, buflen);
7121 }
7122
7123 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7124 {
7125         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7126 }
7127
7128 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7129
7130 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7131                                      struct dentry *unused, struct inode *inode,
7132                                      const char *key, const void *buf,
7133                                      size_t buflen, int flags)
7134 {
7135         if (security_ismaclabel(key))
7136                 return nfs4_set_security_label(inode, buf, buflen);
7137
7138         return -EOPNOTSUPP;
7139 }
7140
7141 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7142                                      struct dentry *unused, struct inode *inode,
7143                                      const char *key, void *buf, size_t buflen)
7144 {
7145         if (security_ismaclabel(key))
7146                 return nfs4_get_security_label(inode, buf, buflen);
7147         return -EOPNOTSUPP;
7148 }
7149
7150 static ssize_t
7151 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7152 {
7153         int len = 0;
7154
7155         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7156                 len = security_inode_listsecurity(inode, list, list_len);
7157                 if (list_len && len > list_len)
7158                         return -ERANGE;
7159         }
7160         return len;
7161 }
7162
7163 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7164         .prefix = XATTR_SECURITY_PREFIX,
7165         .get    = nfs4_xattr_get_nfs4_label,
7166         .set    = nfs4_xattr_set_nfs4_label,
7167 };
7168
7169 #else
7170
7171 static ssize_t
7172 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7173 {
7174         return 0;
7175 }
7176
7177 #endif
7178
7179 /*
7180  * nfs_fhget will use either the mounted_on_fileid or the fileid
7181  */
7182 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7183 {
7184         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7185                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7186               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7187               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7188                 return;
7189
7190         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7191                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7192         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7193         fattr->nlink = 2;
7194 }
7195
7196 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7197                                    const struct qstr *name,
7198                                    struct nfs4_fs_locations *fs_locations,
7199                                    struct page *page)
7200 {
7201         struct nfs_server *server = NFS_SERVER(dir);
7202         u32 bitmask[3];
7203         struct nfs4_fs_locations_arg args = {
7204                 .dir_fh = NFS_FH(dir),
7205                 .name = name,
7206                 .page = page,
7207                 .bitmask = bitmask,
7208         };
7209         struct nfs4_fs_locations_res res = {
7210                 .fs_locations = fs_locations,
7211         };
7212         struct rpc_message msg = {
7213                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7214                 .rpc_argp = &args,
7215                 .rpc_resp = &res,
7216         };
7217         int status;
7218
7219         dprintk("%s: start\n", __func__);
7220
7221         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7222         bitmask[1] = nfs4_fattr_bitmap[1];
7223
7224         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7225          * is not supported */
7226         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7227                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7228         else
7229                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7230
7231         nfs_fattr_init(&fs_locations->fattr);
7232         fs_locations->server = server;
7233         fs_locations->nlocations = 0;
7234         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7235         dprintk("%s: returned status = %d\n", __func__, status);
7236         return status;
7237 }
7238
7239 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7240                            const struct qstr *name,
7241                            struct nfs4_fs_locations *fs_locations,
7242                            struct page *page)
7243 {
7244         struct nfs4_exception exception = { };
7245         int err;
7246         do {
7247                 err = _nfs4_proc_fs_locations(client, dir, name,
7248                                 fs_locations, page);
7249                 trace_nfs4_get_fs_locations(dir, name, err);
7250                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7251                                 &exception);
7252         } while (exception.retry);
7253         return err;
7254 }
7255
7256 /*
7257  * This operation also signals the server that this client is
7258  * performing migration recovery.  The server can stop returning
7259  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7260  * appended to this compound to identify the client ID which is
7261  * performing recovery.
7262  */
7263 static int _nfs40_proc_get_locations(struct inode *inode,
7264                                      struct nfs4_fs_locations *locations,
7265                                      struct page *page, const struct cred *cred)
7266 {
7267         struct nfs_server *server = NFS_SERVER(inode);
7268         struct rpc_clnt *clnt = server->client;
7269         u32 bitmask[2] = {
7270                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7271         };
7272         struct nfs4_fs_locations_arg args = {
7273                 .clientid       = server->nfs_client->cl_clientid,
7274                 .fh             = NFS_FH(inode),
7275                 .page           = page,
7276                 .bitmask        = bitmask,
7277                 .migration      = 1,            /* skip LOOKUP */
7278                 .renew          = 1,            /* append RENEW */
7279         };
7280         struct nfs4_fs_locations_res res = {
7281                 .fs_locations   = locations,
7282                 .migration      = 1,
7283                 .renew          = 1,
7284         };
7285         struct rpc_message msg = {
7286                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7287                 .rpc_argp       = &args,
7288                 .rpc_resp       = &res,
7289                 .rpc_cred       = cred,
7290         };
7291         unsigned long now = jiffies;
7292         int status;
7293
7294         nfs_fattr_init(&locations->fattr);
7295         locations->server = server;
7296         locations->nlocations = 0;
7297
7298         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7299         status = nfs4_call_sync_sequence(clnt, server, &msg,
7300                                         &args.seq_args, &res.seq_res);
7301         if (status)
7302                 return status;
7303
7304         renew_lease(server, now);
7305         return 0;
7306 }
7307
7308 #ifdef CONFIG_NFS_V4_1
7309
7310 /*
7311  * This operation also signals the server that this client is
7312  * performing migration recovery.  The server can stop asserting
7313  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7314  * performing this operation is identified in the SEQUENCE
7315  * operation in this compound.
7316  *
7317  * When the client supports GETATTR(fs_locations_info), it can
7318  * be plumbed in here.
7319  */
7320 static int _nfs41_proc_get_locations(struct inode *inode,
7321                                      struct nfs4_fs_locations *locations,
7322                                      struct page *page, const struct cred *cred)
7323 {
7324         struct nfs_server *server = NFS_SERVER(inode);
7325         struct rpc_clnt *clnt = server->client;
7326         u32 bitmask[2] = {
7327                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7328         };
7329         struct nfs4_fs_locations_arg args = {
7330                 .fh             = NFS_FH(inode),
7331                 .page           = page,
7332                 .bitmask        = bitmask,
7333                 .migration      = 1,            /* skip LOOKUP */
7334         };
7335         struct nfs4_fs_locations_res res = {
7336                 .fs_locations   = locations,
7337                 .migration      = 1,
7338         };
7339         struct rpc_message msg = {
7340                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7341                 .rpc_argp       = &args,
7342                 .rpc_resp       = &res,
7343                 .rpc_cred       = cred,
7344         };
7345         int status;
7346
7347         nfs_fattr_init(&locations->fattr);
7348         locations->server = server;
7349         locations->nlocations = 0;
7350
7351         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7352         status = nfs4_call_sync_sequence(clnt, server, &msg,
7353                                         &args.seq_args, &res.seq_res);
7354         if (status == NFS4_OK &&
7355             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7356                 status = -NFS4ERR_LEASE_MOVED;
7357         return status;
7358 }
7359
7360 #endif  /* CONFIG_NFS_V4_1 */
7361
7362 /**
7363  * nfs4_proc_get_locations - discover locations for a migrated FSID
7364  * @inode: inode on FSID that is migrating
7365  * @locations: result of query
7366  * @page: buffer
7367  * @cred: credential to use for this operation
7368  *
7369  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7370  * operation failed, or a negative errno if a local error occurred.
7371  *
7372  * On success, "locations" is filled in, but if the server has
7373  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7374  * asserted.
7375  *
7376  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7377  * from this client that require migration recovery.
7378  */
7379 int nfs4_proc_get_locations(struct inode *inode,
7380                             struct nfs4_fs_locations *locations,
7381                             struct page *page, const struct cred *cred)
7382 {
7383         struct nfs_server *server = NFS_SERVER(inode);
7384         struct nfs_client *clp = server->nfs_client;
7385         const struct nfs4_mig_recovery_ops *ops =
7386                                         clp->cl_mvops->mig_recovery_ops;
7387         struct nfs4_exception exception = { };
7388         int status;
7389
7390         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7391                 (unsigned long long)server->fsid.major,
7392                 (unsigned long long)server->fsid.minor,
7393                 clp->cl_hostname);
7394         nfs_display_fhandle(NFS_FH(inode), __func__);
7395
7396         do {
7397                 status = ops->get_locations(inode, locations, page, cred);
7398                 if (status != -NFS4ERR_DELAY)
7399                         break;
7400                 nfs4_handle_exception(server, status, &exception);
7401         } while (exception.retry);
7402         return status;
7403 }
7404
7405 /*
7406  * This operation also signals the server that this client is
7407  * performing "lease moved" recovery.  The server can stop
7408  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7409  * is appended to this compound to identify the client ID which is
7410  * performing recovery.
7411  */
7412 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7413 {
7414         struct nfs_server *server = NFS_SERVER(inode);
7415         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7416         struct rpc_clnt *clnt = server->client;
7417         struct nfs4_fsid_present_arg args = {
7418                 .fh             = NFS_FH(inode),
7419                 .clientid       = clp->cl_clientid,
7420                 .renew          = 1,            /* append RENEW */
7421         };
7422         struct nfs4_fsid_present_res res = {
7423                 .renew          = 1,
7424         };
7425         struct rpc_message msg = {
7426                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7427                 .rpc_argp       = &args,
7428                 .rpc_resp       = &res,
7429                 .rpc_cred       = cred,
7430         };
7431         unsigned long now = jiffies;
7432         int status;
7433
7434         res.fh = nfs_alloc_fhandle();
7435         if (res.fh == NULL)
7436                 return -ENOMEM;
7437
7438         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7439         status = nfs4_call_sync_sequence(clnt, server, &msg,
7440                                                 &args.seq_args, &res.seq_res);
7441         nfs_free_fhandle(res.fh);
7442         if (status)
7443                 return status;
7444
7445         do_renew_lease(clp, now);
7446         return 0;
7447 }
7448
7449 #ifdef CONFIG_NFS_V4_1
7450
7451 /*
7452  * This operation also signals the server that this client is
7453  * performing "lease moved" recovery.  The server can stop asserting
7454  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7455  * this operation is identified in the SEQUENCE operation in this
7456  * compound.
7457  */
7458 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7459 {
7460         struct nfs_server *server = NFS_SERVER(inode);
7461         struct rpc_clnt *clnt = server->client;
7462         struct nfs4_fsid_present_arg args = {
7463                 .fh             = NFS_FH(inode),
7464         };
7465         struct nfs4_fsid_present_res res = {
7466         };
7467         struct rpc_message msg = {
7468                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7469                 .rpc_argp       = &args,
7470                 .rpc_resp       = &res,
7471                 .rpc_cred       = cred,
7472         };
7473         int status;
7474
7475         res.fh = nfs_alloc_fhandle();
7476         if (res.fh == NULL)
7477                 return -ENOMEM;
7478
7479         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7480         status = nfs4_call_sync_sequence(clnt, server, &msg,
7481                                                 &args.seq_args, &res.seq_res);
7482         nfs_free_fhandle(res.fh);
7483         if (status == NFS4_OK &&
7484             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7485                 status = -NFS4ERR_LEASE_MOVED;
7486         return status;
7487 }
7488
7489 #endif  /* CONFIG_NFS_V4_1 */
7490
7491 /**
7492  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7493  * @inode: inode on FSID to check
7494  * @cred: credential to use for this operation
7495  *
7496  * Server indicates whether the FSID is present, moved, or not
7497  * recognized.  This operation is necessary to clear a LEASE_MOVED
7498  * condition for this client ID.
7499  *
7500  * Returns NFS4_OK if the FSID is present on this server,
7501  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7502  *  NFS4ERR code if some error occurred on the server, or a
7503  *  negative errno if a local failure occurred.
7504  */
7505 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
7506 {
7507         struct nfs_server *server = NFS_SERVER(inode);
7508         struct nfs_client *clp = server->nfs_client;
7509         const struct nfs4_mig_recovery_ops *ops =
7510                                         clp->cl_mvops->mig_recovery_ops;
7511         struct nfs4_exception exception = { };
7512         int status;
7513
7514         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7515                 (unsigned long long)server->fsid.major,
7516                 (unsigned long long)server->fsid.minor,
7517                 clp->cl_hostname);
7518         nfs_display_fhandle(NFS_FH(inode), __func__);
7519
7520         do {
7521                 status = ops->fsid_present(inode, cred);
7522                 if (status != -NFS4ERR_DELAY)
7523                         break;
7524                 nfs4_handle_exception(server, status, &exception);
7525         } while (exception.retry);
7526         return status;
7527 }
7528
7529 /*
7530  * If 'use_integrity' is true and the state managment nfs_client
7531  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7532  * and the machine credential as per RFC3530bis and RFC5661 Security
7533  * Considerations sections. Otherwise, just use the user cred with the
7534  * filesystem's rpc_client.
7535  */
7536 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7537 {
7538         int status;
7539         struct nfs4_secinfo_arg args = {
7540                 .dir_fh = NFS_FH(dir),
7541                 .name   = name,
7542         };
7543         struct nfs4_secinfo_res res = {
7544                 .flavors     = flavors,
7545         };
7546         struct rpc_message msg = {
7547                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7548                 .rpc_argp = &args,
7549                 .rpc_resp = &res,
7550         };
7551         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7552         const struct cred *cred = NULL;
7553
7554         if (use_integrity) {
7555                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7556                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7557                 msg.rpc_cred = cred;
7558         }
7559
7560         dprintk("NFS call  secinfo %s\n", name->name);
7561
7562         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7563                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7564
7565         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7566                                 &res.seq_res, 0);
7567         dprintk("NFS reply  secinfo: %d\n", status);
7568
7569         put_cred(cred);
7570
7571         return status;
7572 }
7573
7574 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7575                       struct nfs4_secinfo_flavors *flavors)
7576 {
7577         struct nfs4_exception exception = { };
7578         int err;
7579         do {
7580                 err = -NFS4ERR_WRONGSEC;
7581
7582                 /* try to use integrity protection with machine cred */
7583                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7584                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7585
7586                 /*
7587                  * if unable to use integrity protection, or SECINFO with
7588                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7589                  * disallowed by spec, but exists in deployed servers) use
7590                  * the current filesystem's rpc_client and the user cred.
7591                  */
7592                 if (err == -NFS4ERR_WRONGSEC)
7593                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7594
7595                 trace_nfs4_secinfo(dir, name, err);
7596                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7597                                 &exception);
7598         } while (exception.retry);
7599         return err;
7600 }
7601
7602 #ifdef CONFIG_NFS_V4_1
7603 /*
7604  * Check the exchange flags returned by the server for invalid flags, having
7605  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7606  * DS flags set.
7607  */
7608 static int nfs4_check_cl_exchange_flags(u32 flags)
7609 {
7610         if (flags & ~EXCHGID4_FLAG_MASK_R)
7611                 goto out_inval;
7612         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7613             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7614                 goto out_inval;
7615         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7616                 goto out_inval;
7617         return NFS_OK;
7618 out_inval:
7619         return -NFS4ERR_INVAL;
7620 }
7621
7622 static bool
7623 nfs41_same_server_scope(struct nfs41_server_scope *a,
7624                         struct nfs41_server_scope *b)
7625 {
7626         if (a->server_scope_sz != b->server_scope_sz)
7627                 return false;
7628         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7629 }
7630
7631 static void
7632 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7633 {
7634 }
7635
7636 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7637         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7638 };
7639
7640 /*
7641  * nfs4_proc_bind_one_conn_to_session()
7642  *
7643  * The 4.1 client currently uses the same TCP connection for the
7644  * fore and backchannel.
7645  */
7646 static
7647 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7648                 struct rpc_xprt *xprt,
7649                 struct nfs_client *clp,
7650                 const struct cred *cred)
7651 {
7652         int status;
7653         struct nfs41_bind_conn_to_session_args args = {
7654                 .client = clp,
7655                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7656         };
7657         struct nfs41_bind_conn_to_session_res res;
7658         struct rpc_message msg = {
7659                 .rpc_proc =
7660                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7661                 .rpc_argp = &args,
7662                 .rpc_resp = &res,
7663                 .rpc_cred = cred,
7664         };
7665         struct rpc_task_setup task_setup_data = {
7666                 .rpc_client = clnt,
7667                 .rpc_xprt = xprt,
7668                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7669                 .rpc_message = &msg,
7670                 .flags = RPC_TASK_TIMEOUT,
7671         };
7672         struct rpc_task *task;
7673
7674         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7675         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7676                 args.dir = NFS4_CDFC4_FORE;
7677
7678         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7679         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7680                 args.dir = NFS4_CDFC4_FORE;
7681
7682         task = rpc_run_task(&task_setup_data);
7683         if (!IS_ERR(task)) {
7684                 status = task->tk_status;
7685                 rpc_put_task(task);
7686         } else
7687                 status = PTR_ERR(task);
7688         trace_nfs4_bind_conn_to_session(clp, status);
7689         if (status == 0) {
7690                 if (memcmp(res.sessionid.data,
7691                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7692                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7693                         return -EIO;
7694                 }
7695                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7696                         dprintk("NFS: %s: Unexpected direction from server\n",
7697                                 __func__);
7698                         return -EIO;
7699                 }
7700                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7701                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7702                                 __func__);
7703                         return -EIO;
7704                 }
7705         }
7706
7707         return status;
7708 }
7709
7710 struct rpc_bind_conn_calldata {
7711         struct nfs_client *clp;
7712         const struct cred *cred;
7713 };
7714
7715 static int
7716 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7717                 struct rpc_xprt *xprt,
7718                 void *calldata)
7719 {
7720         struct rpc_bind_conn_calldata *p = calldata;
7721
7722         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7723 }
7724
7725 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
7726 {
7727         struct rpc_bind_conn_calldata data = {
7728                 .clp = clp,
7729                 .cred = cred,
7730         };
7731         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7732                         nfs4_proc_bind_conn_to_session_callback, &data);
7733 }
7734
7735 /*
7736  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7737  * and operations we'd like to see to enable certain features in the allow map
7738  */
7739 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7740         .how = SP4_MACH_CRED,
7741         .enforce.u.words = {
7742                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7743                       1 << (OP_EXCHANGE_ID - 32) |
7744                       1 << (OP_CREATE_SESSION - 32) |
7745                       1 << (OP_DESTROY_SESSION - 32) |
7746                       1 << (OP_DESTROY_CLIENTID - 32)
7747         },
7748         .allow.u.words = {
7749                 [0] = 1 << (OP_CLOSE) |
7750                       1 << (OP_OPEN_DOWNGRADE) |
7751                       1 << (OP_LOCKU) |
7752                       1 << (OP_DELEGRETURN) |
7753                       1 << (OP_COMMIT),
7754                 [1] = 1 << (OP_SECINFO - 32) |
7755                       1 << (OP_SECINFO_NO_NAME - 32) |
7756                       1 << (OP_LAYOUTRETURN - 32) |
7757                       1 << (OP_TEST_STATEID - 32) |
7758                       1 << (OP_FREE_STATEID - 32) |
7759                       1 << (OP_WRITE - 32)
7760         }
7761 };
7762
7763 /*
7764  * Select the state protection mode for client `clp' given the server results
7765  * from exchange_id in `sp'.
7766  *
7767  * Returns 0 on success, negative errno otherwise.
7768  */
7769 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7770                                  struct nfs41_state_protection *sp)
7771 {
7772         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7773                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7774                       1 << (OP_EXCHANGE_ID - 32) |
7775                       1 << (OP_CREATE_SESSION - 32) |
7776                       1 << (OP_DESTROY_SESSION - 32) |
7777                       1 << (OP_DESTROY_CLIENTID - 32)
7778         };
7779         unsigned long flags = 0;
7780         unsigned int i;
7781         int ret = 0;
7782
7783         if (sp->how == SP4_MACH_CRED) {
7784                 /* Print state protect result */
7785                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7786                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7787                         if (test_bit(i, sp->enforce.u.longs))
7788                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7789                         if (test_bit(i, sp->allow.u.longs))
7790                                 dfprintk(MOUNT, "  allow op %d\n", i);
7791                 }
7792
7793                 /* make sure nothing is on enforce list that isn't supported */
7794                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7795                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7796                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7797                                 ret = -EINVAL;
7798                                 goto out;
7799                         }
7800                 }
7801
7802                 /*
7803                  * Minimal mode - state operations are allowed to use machine
7804                  * credential.  Note this already happens by default, so the
7805                  * client doesn't have to do anything more than the negotiation.
7806                  *
7807                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7808                  *       we're already using the machine cred for exchange_id
7809                  *       and will never use a different cred.
7810                  */
7811                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7812                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7813                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7814                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7815                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7816                         dfprintk(MOUNT, "  minimal mode enabled\n");
7817                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
7818                 } else {
7819                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7820                         ret = -EINVAL;
7821                         goto out;
7822                 }
7823
7824                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7825                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7826                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7827                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7828                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7829                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
7830                 }
7831
7832                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7833                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7834                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
7835                 }
7836
7837                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7838                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7839                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7840                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
7841                 }
7842
7843                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7844                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7845                         dfprintk(MOUNT, "  stateid mode enabled\n");
7846                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
7847                 }
7848
7849                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7850                         dfprintk(MOUNT, "  write mode enabled\n");
7851                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
7852                 }
7853
7854                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7855                         dfprintk(MOUNT, "  commit mode enabled\n");
7856                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
7857                 }
7858         }
7859 out:
7860         clp->cl_sp4_flags = flags;
7861         return ret;
7862 }
7863
7864 struct nfs41_exchange_id_data {
7865         struct nfs41_exchange_id_res res;
7866         struct nfs41_exchange_id_args args;
7867 };
7868
7869 static void nfs4_exchange_id_release(void *data)
7870 {
7871         struct nfs41_exchange_id_data *cdata =
7872                                         (struct nfs41_exchange_id_data *)data;
7873
7874         nfs_put_client(cdata->args.client);
7875         kfree(cdata->res.impl_id);
7876         kfree(cdata->res.server_scope);
7877         kfree(cdata->res.server_owner);
7878         kfree(cdata);
7879 }
7880
7881 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7882         .rpc_release = nfs4_exchange_id_release,
7883 };
7884
7885 /*
7886  * _nfs4_proc_exchange_id()
7887  *
7888  * Wrapper for EXCHANGE_ID operation.
7889  */
7890 static struct rpc_task *
7891 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
7892                         u32 sp4_how, struct rpc_xprt *xprt)
7893 {
7894         struct rpc_message msg = {
7895                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7896                 .rpc_cred = cred,
7897         };
7898         struct rpc_task_setup task_setup_data = {
7899                 .rpc_client = clp->cl_rpcclient,
7900                 .callback_ops = &nfs4_exchange_id_call_ops,
7901                 .rpc_message = &msg,
7902                 .flags = RPC_TASK_TIMEOUT,
7903         };
7904         struct nfs41_exchange_id_data *calldata;
7905         int status;
7906
7907         if (!refcount_inc_not_zero(&clp->cl_count))
7908                 return ERR_PTR(-EIO);
7909
7910         status = -ENOMEM;
7911         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7912         if (!calldata)
7913                 goto out;
7914
7915         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
7916
7917         status = nfs4_init_uniform_client_string(clp);
7918         if (status)
7919                 goto out_calldata;
7920
7921         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7922                                                 GFP_NOFS);
7923         status = -ENOMEM;
7924         if (unlikely(calldata->res.server_owner == NULL))
7925                 goto out_calldata;
7926
7927         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7928                                         GFP_NOFS);
7929         if (unlikely(calldata->res.server_scope == NULL))
7930                 goto out_server_owner;
7931
7932         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7933         if (unlikely(calldata->res.impl_id == NULL))
7934                 goto out_server_scope;
7935
7936         switch (sp4_how) {
7937         case SP4_NONE:
7938                 calldata->args.state_protect.how = SP4_NONE;
7939                 break;
7940
7941         case SP4_MACH_CRED:
7942                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7943                 break;
7944
7945         default:
7946                 /* unsupported! */
7947                 WARN_ON_ONCE(1);
7948                 status = -EINVAL;
7949                 goto out_impl_id;
7950         }
7951         if (xprt) {
7952                 task_setup_data.rpc_xprt = xprt;
7953                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
7954                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
7955                                 sizeof(calldata->args.verifier.data));
7956         }
7957         calldata->args.client = clp;
7958         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7959         EXCHGID4_FLAG_BIND_PRINC_STATEID;
7960 #ifdef CONFIG_NFS_V4_1_MIGRATION
7961         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
7962 #endif
7963         msg.rpc_argp = &calldata->args;
7964         msg.rpc_resp = &calldata->res;
7965         task_setup_data.callback_data = calldata;
7966
7967         return rpc_run_task(&task_setup_data);
7968
7969 out_impl_id:
7970         kfree(calldata->res.impl_id);
7971 out_server_scope:
7972         kfree(calldata->res.server_scope);
7973 out_server_owner:
7974         kfree(calldata->res.server_owner);
7975 out_calldata:
7976         kfree(calldata);
7977 out:
7978         nfs_put_client(clp);
7979         return ERR_PTR(status);
7980 }
7981
7982 /*
7983  * _nfs4_proc_exchange_id()
7984  *
7985  * Wrapper for EXCHANGE_ID operation.
7986  */
7987 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
7988                         u32 sp4_how)
7989 {
7990         struct rpc_task *task;
7991         struct nfs41_exchange_id_args *argp;
7992         struct nfs41_exchange_id_res *resp;
7993         int status;
7994
7995         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
7996         if (IS_ERR(task))
7997                 return PTR_ERR(task);
7998
7999         argp = task->tk_msg.rpc_argp;
8000         resp = task->tk_msg.rpc_resp;
8001         status = task->tk_status;
8002         if (status  != 0)
8003                 goto out;
8004
8005         status = nfs4_check_cl_exchange_flags(resp->flags);
8006         if (status  != 0)
8007                 goto out;
8008
8009         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8010         if (status != 0)
8011                 goto out;
8012
8013         clp->cl_clientid = resp->clientid;
8014         clp->cl_exchange_flags = resp->flags;
8015         clp->cl_seqid = resp->seqid;
8016         /* Client ID is not confirmed */
8017         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8018                 clear_bit(NFS4_SESSION_ESTABLISHED,
8019                           &clp->cl_session->session_state);
8020
8021         if (clp->cl_serverscope != NULL &&
8022             !nfs41_same_server_scope(clp->cl_serverscope,
8023                                 resp->server_scope)) {
8024                 dprintk("%s: server_scope mismatch detected\n",
8025                         __func__);
8026                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8027         }
8028
8029         swap(clp->cl_serverowner, resp->server_owner);
8030         swap(clp->cl_serverscope, resp->server_scope);
8031         swap(clp->cl_implid, resp->impl_id);
8032
8033         /* Save the EXCHANGE_ID verifier session trunk tests */
8034         memcpy(clp->cl_confirm.data, argp->verifier.data,
8035                sizeof(clp->cl_confirm.data));
8036 out:
8037         trace_nfs4_exchange_id(clp, status);
8038         rpc_put_task(task);
8039         return status;
8040 }
8041
8042 /*
8043  * nfs4_proc_exchange_id()
8044  *
8045  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8046  *
8047  * Since the clientid has expired, all compounds using sessions
8048  * associated with the stale clientid will be returning
8049  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8050  * be in some phase of session reset.
8051  *
8052  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8053  */
8054 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8055 {
8056         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8057         int status;
8058
8059         /* try SP4_MACH_CRED if krb5i/p */
8060         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8061             authflavor == RPC_AUTH_GSS_KRB5P) {
8062                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8063                 if (!status)
8064                         return 0;
8065         }
8066
8067         /* try SP4_NONE */
8068         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8069 }
8070
8071 /**
8072  * nfs4_test_session_trunk
8073  *
8074  * This is an add_xprt_test() test function called from
8075  * rpc_clnt_setup_test_and_add_xprt.
8076  *
8077  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8078  * and is dereferrenced in nfs4_exchange_id_release
8079  *
8080  * Upon success, add the new transport to the rpc_clnt
8081  *
8082  * @clnt: struct rpc_clnt to get new transport
8083  * @xprt: the rpc_xprt to test
8084  * @data: call data for _nfs4_proc_exchange_id.
8085  */
8086 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8087                             void *data)
8088 {
8089         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8090         struct rpc_task *task;
8091         int status;
8092
8093         u32 sp4_how;
8094
8095         dprintk("--> %s try %s\n", __func__,
8096                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8097
8098         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8099
8100         /* Test connection for session trunking. Async exchange_id call */
8101         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8102         if (IS_ERR(task))
8103                 return;
8104
8105         status = task->tk_status;
8106         if (status == 0)
8107                 status = nfs4_detect_session_trunking(adata->clp,
8108                                 task->tk_msg.rpc_resp, xprt);
8109
8110         if (status == 0)
8111                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8112
8113         rpc_put_task(task);
8114 }
8115 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8116
8117 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8118                 const struct cred *cred)
8119 {
8120         struct rpc_message msg = {
8121                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8122                 .rpc_argp = clp,
8123                 .rpc_cred = cred,
8124         };
8125         int status;
8126
8127         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8128         trace_nfs4_destroy_clientid(clp, status);
8129         if (status)
8130                 dprintk("NFS: Got error %d from the server %s on "
8131                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8132         return status;
8133 }
8134
8135 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8136                 const struct cred *cred)
8137 {
8138         unsigned int loop;
8139         int ret;
8140
8141         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8142                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8143                 switch (ret) {
8144                 case -NFS4ERR_DELAY:
8145                 case -NFS4ERR_CLIENTID_BUSY:
8146                         ssleep(1);
8147                         break;
8148                 default:
8149                         return ret;
8150                 }
8151         }
8152         return 0;
8153 }
8154
8155 int nfs4_destroy_clientid(struct nfs_client *clp)
8156 {
8157         const struct cred *cred;
8158         int ret = 0;
8159
8160         if (clp->cl_mvops->minor_version < 1)
8161                 goto out;
8162         if (clp->cl_exchange_flags == 0)
8163                 goto out;
8164         if (clp->cl_preserve_clid)
8165                 goto out;
8166         cred = nfs4_get_clid_cred(clp);
8167         ret = nfs4_proc_destroy_clientid(clp, cred);
8168         put_cred(cred);
8169         switch (ret) {
8170         case 0:
8171         case -NFS4ERR_STALE_CLIENTID:
8172                 clp->cl_exchange_flags = 0;
8173         }
8174 out:
8175         return ret;
8176 }
8177
8178 struct nfs4_get_lease_time_data {
8179         struct nfs4_get_lease_time_args *args;
8180         struct nfs4_get_lease_time_res *res;
8181         struct nfs_client *clp;
8182 };
8183
8184 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8185                                         void *calldata)
8186 {
8187         struct nfs4_get_lease_time_data *data =
8188                         (struct nfs4_get_lease_time_data *)calldata;
8189
8190         dprintk("--> %s\n", __func__);
8191         /* just setup sequence, do not trigger session recovery
8192            since we're invoked within one */
8193         nfs4_setup_sequence(data->clp,
8194                         &data->args->la_seq_args,
8195                         &data->res->lr_seq_res,
8196                         task);
8197         dprintk("<-- %s\n", __func__);
8198 }
8199
8200 /*
8201  * Called from nfs4_state_manager thread for session setup, so don't recover
8202  * from sequence operation or clientid errors.
8203  */
8204 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8205 {
8206         struct nfs4_get_lease_time_data *data =
8207                         (struct nfs4_get_lease_time_data *)calldata;
8208
8209         dprintk("--> %s\n", __func__);
8210         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
8211                 return;
8212         switch (task->tk_status) {
8213         case -NFS4ERR_DELAY:
8214         case -NFS4ERR_GRACE:
8215                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8216                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8217                 task->tk_status = 0;
8218                 /* fall through */
8219         case -NFS4ERR_RETRY_UNCACHED_REP:
8220                 rpc_restart_call_prepare(task);
8221                 return;
8222         }
8223         dprintk("<-- %s\n", __func__);
8224 }
8225
8226 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8227         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8228         .rpc_call_done = nfs4_get_lease_time_done,
8229 };
8230
8231 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8232 {
8233         struct rpc_task *task;
8234         struct nfs4_get_lease_time_args args;
8235         struct nfs4_get_lease_time_res res = {
8236                 .lr_fsinfo = fsinfo,
8237         };
8238         struct nfs4_get_lease_time_data data = {
8239                 .args = &args,
8240                 .res = &res,
8241                 .clp = clp,
8242         };
8243         struct rpc_message msg = {
8244                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8245                 .rpc_argp = &args,
8246                 .rpc_resp = &res,
8247         };
8248         struct rpc_task_setup task_setup = {
8249                 .rpc_client = clp->cl_rpcclient,
8250                 .rpc_message = &msg,
8251                 .callback_ops = &nfs4_get_lease_time_ops,
8252                 .callback_data = &data,
8253                 .flags = RPC_TASK_TIMEOUT,
8254         };
8255         int status;
8256
8257         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8258         task = rpc_run_task(&task_setup);
8259
8260         if (IS_ERR(task))
8261                 return PTR_ERR(task);
8262
8263         status = task->tk_status;
8264         rpc_put_task(task);
8265         return status;
8266 }
8267
8268 /*
8269  * Initialize the values to be used by the client in CREATE_SESSION
8270  * If nfs4_init_session set the fore channel request and response sizes,
8271  * use them.
8272  *
8273  * Set the back channel max_resp_sz_cached to zero to force the client to
8274  * always set csa_cachethis to FALSE because the current implementation
8275  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8276  */
8277 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8278                                     struct rpc_clnt *clnt)
8279 {
8280         unsigned int max_rqst_sz, max_resp_sz;
8281         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8282
8283         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8284         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8285
8286         /* Fore channel attributes */
8287         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8288         args->fc_attrs.max_resp_sz = max_resp_sz;
8289         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8290         args->fc_attrs.max_reqs = max_session_slots;
8291
8292         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8293                 "max_ops=%u max_reqs=%u\n",
8294                 __func__,
8295                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8296                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8297
8298         /* Back channel attributes */
8299         args->bc_attrs.max_rqst_sz = max_bc_payload;
8300         args->bc_attrs.max_resp_sz = max_bc_payload;
8301         args->bc_attrs.max_resp_sz_cached = 0;
8302         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8303         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8304
8305         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8306                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8307                 __func__,
8308                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8309                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8310                 args->bc_attrs.max_reqs);
8311 }
8312
8313 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8314                 struct nfs41_create_session_res *res)
8315 {
8316         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8317         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8318
8319         if (rcvd->max_resp_sz > sent->max_resp_sz)
8320                 return -EINVAL;
8321         /*
8322          * Our requested max_ops is the minimum we need; we're not
8323          * prepared to break up compounds into smaller pieces than that.
8324          * So, no point even trying to continue if the server won't
8325          * cooperate:
8326          */
8327         if (rcvd->max_ops < sent->max_ops)
8328                 return -EINVAL;
8329         if (rcvd->max_reqs == 0)
8330                 return -EINVAL;
8331         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8332                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8333         return 0;
8334 }
8335
8336 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8337                 struct nfs41_create_session_res *res)
8338 {
8339         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8340         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8341
8342         if (!(res->flags & SESSION4_BACK_CHAN))
8343                 goto out;
8344         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8345                 return -EINVAL;
8346         if (rcvd->max_resp_sz < sent->max_resp_sz)
8347                 return -EINVAL;
8348         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8349                 return -EINVAL;
8350         if (rcvd->max_ops > sent->max_ops)
8351                 return -EINVAL;
8352         if (rcvd->max_reqs > sent->max_reqs)
8353                 return -EINVAL;
8354 out:
8355         return 0;
8356 }
8357
8358 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8359                                      struct nfs41_create_session_res *res)
8360 {
8361         int ret;
8362
8363         ret = nfs4_verify_fore_channel_attrs(args, res);
8364         if (ret)
8365                 return ret;
8366         return nfs4_verify_back_channel_attrs(args, res);
8367 }
8368
8369 static void nfs4_update_session(struct nfs4_session *session,
8370                 struct nfs41_create_session_res *res)
8371 {
8372         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8373         /* Mark client id and session as being confirmed */
8374         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8375         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8376         session->flags = res->flags;
8377         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8378         if (res->flags & SESSION4_BACK_CHAN)
8379                 memcpy(&session->bc_attrs, &res->bc_attrs,
8380                                 sizeof(session->bc_attrs));
8381 }
8382
8383 static int _nfs4_proc_create_session(struct nfs_client *clp,
8384                 const struct cred *cred)
8385 {
8386         struct nfs4_session *session = clp->cl_session;
8387         struct nfs41_create_session_args args = {
8388                 .client = clp,
8389                 .clientid = clp->cl_clientid,
8390                 .seqid = clp->cl_seqid,
8391                 .cb_program = NFS4_CALLBACK,
8392         };
8393         struct nfs41_create_session_res res;
8394
8395         struct rpc_message msg = {
8396                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8397                 .rpc_argp = &args,
8398                 .rpc_resp = &res,
8399                 .rpc_cred = cred,
8400         };
8401         int status;
8402
8403         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8404         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8405
8406         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8407         trace_nfs4_create_session(clp, status);
8408
8409         switch (status) {
8410         case -NFS4ERR_STALE_CLIENTID:
8411         case -NFS4ERR_DELAY:
8412         case -ETIMEDOUT:
8413         case -EACCES:
8414         case -EAGAIN:
8415                 goto out;
8416         };
8417
8418         clp->cl_seqid++;
8419         if (!status) {
8420                 /* Verify the session's negotiated channel_attrs values */
8421                 status = nfs4_verify_channel_attrs(&args, &res);
8422                 /* Increment the clientid slot sequence id */
8423                 if (status)
8424                         goto out;
8425                 nfs4_update_session(session, &res);
8426         }
8427 out:
8428         return status;
8429 }
8430
8431 /*
8432  * Issues a CREATE_SESSION operation to the server.
8433  * It is the responsibility of the caller to verify the session is
8434  * expired before calling this routine.
8435  */
8436 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8437 {
8438         int status;
8439         unsigned *ptr;
8440         struct nfs4_session *session = clp->cl_session;
8441
8442         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8443
8444         status = _nfs4_proc_create_session(clp, cred);
8445         if (status)
8446                 goto out;
8447
8448         /* Init or reset the session slot tables */
8449         status = nfs4_setup_session_slot_tables(session);
8450         dprintk("slot table setup returned %d\n", status);
8451         if (status)
8452                 goto out;
8453
8454         ptr = (unsigned *)&session->sess_id.data[0];
8455         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8456                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8457 out:
8458         dprintk("<-- %s\n", __func__);
8459         return status;
8460 }
8461
8462 /*
8463  * Issue the over-the-wire RPC DESTROY_SESSION.
8464  * The caller must serialize access to this routine.
8465  */
8466 int nfs4_proc_destroy_session(struct nfs4_session *session,
8467                 const struct cred *cred)
8468 {
8469         struct rpc_message msg = {
8470                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8471                 .rpc_argp = session,
8472                 .rpc_cred = cred,
8473         };
8474         int status = 0;
8475
8476         dprintk("--> nfs4_proc_destroy_session\n");
8477
8478         /* session is still being setup */
8479         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8480                 return 0;
8481
8482         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8483         trace_nfs4_destroy_session(session->clp, status);
8484
8485         if (status)
8486                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8487                         "Session has been destroyed regardless...\n", status);
8488
8489         dprintk("<-- nfs4_proc_destroy_session\n");
8490         return status;
8491 }
8492
8493 /*
8494  * Renew the cl_session lease.
8495  */
8496 struct nfs4_sequence_data {
8497         struct nfs_client *clp;
8498         struct nfs4_sequence_args args;
8499         struct nfs4_sequence_res res;
8500 };
8501
8502 static void nfs41_sequence_release(void *data)
8503 {
8504         struct nfs4_sequence_data *calldata = data;
8505         struct nfs_client *clp = calldata->clp;
8506
8507         if (refcount_read(&clp->cl_count) > 1)
8508                 nfs4_schedule_state_renewal(clp);
8509         nfs_put_client(clp);
8510         kfree(calldata);
8511 }
8512
8513 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8514 {
8515         switch(task->tk_status) {
8516         case -NFS4ERR_DELAY:
8517                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8518                 return -EAGAIN;
8519         default:
8520                 nfs4_schedule_lease_recovery(clp);
8521         }
8522         return 0;
8523 }
8524
8525 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8526 {
8527         struct nfs4_sequence_data *calldata = data;
8528         struct nfs_client *clp = calldata->clp;
8529
8530         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8531                 return;
8532
8533         trace_nfs4_sequence(clp, task->tk_status);
8534         if (task->tk_status < 0) {
8535                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8536                 if (refcount_read(&clp->cl_count) == 1)
8537                         goto out;
8538
8539                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8540                         rpc_restart_call_prepare(task);
8541                         return;
8542                 }
8543         }
8544         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8545 out:
8546         dprintk("<-- %s\n", __func__);
8547 }
8548
8549 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8550 {
8551         struct nfs4_sequence_data *calldata = data;
8552         struct nfs_client *clp = calldata->clp;
8553         struct nfs4_sequence_args *args;
8554         struct nfs4_sequence_res *res;
8555
8556         args = task->tk_msg.rpc_argp;
8557         res = task->tk_msg.rpc_resp;
8558
8559         nfs4_setup_sequence(clp, args, res, task);
8560 }
8561
8562 static const struct rpc_call_ops nfs41_sequence_ops = {
8563         .rpc_call_done = nfs41_sequence_call_done,
8564         .rpc_call_prepare = nfs41_sequence_prepare,
8565         .rpc_release = nfs41_sequence_release,
8566 };
8567
8568 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8569                 const struct cred *cred,
8570                 struct nfs4_slot *slot,
8571                 bool is_privileged)
8572 {
8573         struct nfs4_sequence_data *calldata;
8574         struct rpc_message msg = {
8575                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8576                 .rpc_cred = cred,
8577         };
8578         struct rpc_task_setup task_setup_data = {
8579                 .rpc_client = clp->cl_rpcclient,
8580                 .rpc_message = &msg,
8581                 .callback_ops = &nfs41_sequence_ops,
8582                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8583         };
8584         struct rpc_task *ret;
8585
8586         ret = ERR_PTR(-EIO);
8587         if (!refcount_inc_not_zero(&clp->cl_count))
8588                 goto out_err;
8589
8590         ret = ERR_PTR(-ENOMEM);
8591         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8592         if (calldata == NULL)
8593                 goto out_put_clp;
8594         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8595         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8596         msg.rpc_argp = &calldata->args;
8597         msg.rpc_resp = &calldata->res;
8598         calldata->clp = clp;
8599         task_setup_data.callback_data = calldata;
8600
8601         ret = rpc_run_task(&task_setup_data);
8602         if (IS_ERR(ret))
8603                 goto out_err;
8604         return ret;
8605 out_put_clp:
8606         nfs_put_client(clp);
8607 out_err:
8608         nfs41_release_slot(slot);
8609         return ret;
8610 }
8611
8612 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
8613 {
8614         struct rpc_task *task;
8615         int ret = 0;
8616
8617         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8618                 return -EAGAIN;
8619         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8620         if (IS_ERR(task))
8621                 ret = PTR_ERR(task);
8622         else
8623                 rpc_put_task_async(task);
8624         dprintk("<-- %s status=%d\n", __func__, ret);
8625         return ret;
8626 }
8627
8628 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
8629 {
8630         struct rpc_task *task;
8631         int ret;
8632
8633         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8634         if (IS_ERR(task)) {
8635                 ret = PTR_ERR(task);
8636                 goto out;
8637         }
8638         ret = rpc_wait_for_completion_task(task);
8639         if (!ret)
8640                 ret = task->tk_status;
8641         rpc_put_task(task);
8642 out:
8643         dprintk("<-- %s status=%d\n", __func__, ret);
8644         return ret;
8645 }
8646
8647 struct nfs4_reclaim_complete_data {
8648         struct nfs_client *clp;
8649         struct nfs41_reclaim_complete_args arg;
8650         struct nfs41_reclaim_complete_res res;
8651 };
8652
8653 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8654 {
8655         struct nfs4_reclaim_complete_data *calldata = data;
8656
8657         nfs4_setup_sequence(calldata->clp,
8658                         &calldata->arg.seq_args,
8659                         &calldata->res.seq_res,
8660                         task);
8661 }
8662
8663 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8664 {
8665         switch(task->tk_status) {
8666         case 0:
8667                 wake_up_all(&clp->cl_lock_waitq);
8668                 /* Fallthrough */
8669         case -NFS4ERR_COMPLETE_ALREADY:
8670         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8671                 break;
8672         case -NFS4ERR_DELAY:
8673                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8674                 /* fall through */
8675         case -NFS4ERR_RETRY_UNCACHED_REP:
8676                 return -EAGAIN;
8677         case -NFS4ERR_BADSESSION:
8678         case -NFS4ERR_DEADSESSION:
8679         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8680                 nfs4_schedule_session_recovery(clp->cl_session,
8681                                 task->tk_status);
8682                 break;
8683         default:
8684                 nfs4_schedule_lease_recovery(clp);
8685         }
8686         return 0;
8687 }
8688
8689 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8690 {
8691         struct nfs4_reclaim_complete_data *calldata = data;
8692         struct nfs_client *clp = calldata->clp;
8693         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8694
8695         dprintk("--> %s\n", __func__);
8696         if (!nfs41_sequence_done(task, res))
8697                 return;
8698
8699         trace_nfs4_reclaim_complete(clp, task->tk_status);
8700         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8701                 rpc_restart_call_prepare(task);
8702                 return;
8703         }
8704         dprintk("<-- %s\n", __func__);
8705 }
8706
8707 static void nfs4_free_reclaim_complete_data(void *data)
8708 {
8709         struct nfs4_reclaim_complete_data *calldata = data;
8710
8711         kfree(calldata);
8712 }
8713
8714 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8715         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8716         .rpc_call_done = nfs4_reclaim_complete_done,
8717         .rpc_release = nfs4_free_reclaim_complete_data,
8718 };
8719
8720 /*
8721  * Issue a global reclaim complete.
8722  */
8723 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8724                 const struct cred *cred)
8725 {
8726         struct nfs4_reclaim_complete_data *calldata;
8727         struct rpc_task *task;
8728         struct rpc_message msg = {
8729                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8730                 .rpc_cred = cred,
8731         };
8732         struct rpc_task_setup task_setup_data = {
8733                 .rpc_client = clp->cl_rpcclient,
8734                 .rpc_message = &msg,
8735                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8736                 .flags = RPC_TASK_ASYNC,
8737         };
8738         int status = -ENOMEM;
8739
8740         dprintk("--> %s\n", __func__);
8741         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8742         if (calldata == NULL)
8743                 goto out;
8744         calldata->clp = clp;
8745         calldata->arg.one_fs = 0;
8746
8747         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
8748         msg.rpc_argp = &calldata->arg;
8749         msg.rpc_resp = &calldata->res;
8750         task_setup_data.callback_data = calldata;
8751         task = rpc_run_task(&task_setup_data);
8752         if (IS_ERR(task)) {
8753                 status = PTR_ERR(task);
8754                 goto out;
8755         }
8756         status = rpc_wait_for_completion_task(task);
8757         if (status == 0)
8758                 status = task->tk_status;
8759         rpc_put_task(task);
8760 out:
8761         dprintk("<-- %s status=%d\n", __func__, status);
8762         return status;
8763 }
8764
8765 static void
8766 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8767 {
8768         struct nfs4_layoutget *lgp = calldata;
8769         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8770
8771         dprintk("--> %s\n", __func__);
8772         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8773                                 &lgp->res.seq_res, task);
8774         dprintk("<-- %s\n", __func__);
8775 }
8776
8777 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8778 {
8779         struct nfs4_layoutget *lgp = calldata;
8780
8781         dprintk("--> %s\n", __func__);
8782         nfs41_sequence_process(task, &lgp->res.seq_res);
8783         dprintk("<-- %s\n", __func__);
8784 }
8785
8786 static int
8787 nfs4_layoutget_handle_exception(struct rpc_task *task,
8788                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8789 {
8790         struct inode *inode = lgp->args.inode;
8791         struct nfs_server *server = NFS_SERVER(inode);
8792         struct pnfs_layout_hdr *lo;
8793         int nfs4err = task->tk_status;
8794         int err, status = 0;
8795         LIST_HEAD(head);
8796
8797         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8798
8799         nfs4_sequence_free_slot(&lgp->res.seq_res);
8800
8801         switch (nfs4err) {
8802         case 0:
8803                 goto out;
8804
8805         /*
8806          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8807          * on the file. set tk_status to -ENODATA to tell upper layer to
8808          * retry go inband.
8809          */
8810         case -NFS4ERR_LAYOUTUNAVAILABLE:
8811                 status = -ENODATA;
8812                 goto out;
8813         /*
8814          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8815          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8816          */
8817         case -NFS4ERR_BADLAYOUT:
8818                 status = -EOVERFLOW;
8819                 goto out;
8820         /*
8821          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8822          * (or clients) writing to the same RAID stripe except when
8823          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8824          *
8825          * Treat it like we would RECALLCONFLICT -- we retry for a little
8826          * while, and then eventually give up.
8827          */
8828         case -NFS4ERR_LAYOUTTRYLATER:
8829                 if (lgp->args.minlength == 0) {
8830                         status = -EOVERFLOW;
8831                         goto out;
8832                 }
8833                 status = -EBUSY;
8834                 break;
8835         case -NFS4ERR_RECALLCONFLICT:
8836                 status = -ERECALLCONFLICT;
8837                 break;
8838         case -NFS4ERR_DELEG_REVOKED:
8839         case -NFS4ERR_ADMIN_REVOKED:
8840         case -NFS4ERR_EXPIRED:
8841         case -NFS4ERR_BAD_STATEID:
8842                 exception->timeout = 0;
8843                 spin_lock(&inode->i_lock);
8844                 lo = NFS_I(inode)->layout;
8845                 /* If the open stateid was bad, then recover it. */
8846                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8847                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
8848                         spin_unlock(&inode->i_lock);
8849                         exception->state = lgp->args.ctx->state;
8850                         exception->stateid = &lgp->args.stateid;
8851                         break;
8852                 }
8853
8854                 /*
8855                  * Mark the bad layout state as invalid, then retry
8856                  */
8857                 pnfs_mark_layout_stateid_invalid(lo, &head);
8858                 spin_unlock(&inode->i_lock);
8859                 nfs_commit_inode(inode, 0);
8860                 pnfs_free_lseg_list(&head);
8861                 status = -EAGAIN;
8862                 goto out;
8863         }
8864
8865         err = nfs4_handle_exception(server, nfs4err, exception);
8866         if (!status) {
8867                 if (exception->retry)
8868                         status = -EAGAIN;
8869                 else
8870                         status = err;
8871         }
8872 out:
8873         dprintk("<-- %s\n", __func__);
8874         return status;
8875 }
8876
8877 size_t max_response_pages(struct nfs_server *server)
8878 {
8879         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8880         return nfs_page_array_len(0, max_resp_sz);
8881 }
8882
8883 static void nfs4_layoutget_release(void *calldata)
8884 {
8885         struct nfs4_layoutget *lgp = calldata;
8886
8887         dprintk("--> %s\n", __func__);
8888         nfs4_sequence_free_slot(&lgp->res.seq_res);
8889         pnfs_layoutget_free(lgp);
8890         dprintk("<-- %s\n", __func__);
8891 }
8892
8893 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8894         .rpc_call_prepare = nfs4_layoutget_prepare,
8895         .rpc_call_done = nfs4_layoutget_done,
8896         .rpc_release = nfs4_layoutget_release,
8897 };
8898
8899 struct pnfs_layout_segment *
8900 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
8901 {
8902         struct inode *inode = lgp->args.inode;
8903         struct nfs_server *server = NFS_SERVER(inode);
8904         struct rpc_task *task;
8905         struct rpc_message msg = {
8906                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8907                 .rpc_argp = &lgp->args,
8908                 .rpc_resp = &lgp->res,
8909                 .rpc_cred = lgp->cred,
8910         };
8911         struct rpc_task_setup task_setup_data = {
8912                 .rpc_client = server->client,
8913                 .rpc_message = &msg,
8914                 .callback_ops = &nfs4_layoutget_call_ops,
8915                 .callback_data = lgp,
8916                 .flags = RPC_TASK_ASYNC,
8917         };
8918         struct pnfs_layout_segment *lseg = NULL;
8919         struct nfs4_exception exception = {
8920                 .inode = inode,
8921                 .timeout = *timeout,
8922         };
8923         int status = 0;
8924
8925         dprintk("--> %s\n", __func__);
8926
8927         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8928         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8929
8930         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
8931
8932         task = rpc_run_task(&task_setup_data);
8933         if (IS_ERR(task))
8934                 return ERR_CAST(task);
8935         status = rpc_wait_for_completion_task(task);
8936         if (status != 0)
8937                 goto out;
8938
8939         if (task->tk_status < 0) {
8940                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8941                 *timeout = exception.timeout;
8942         } else if (lgp->res.layoutp->len == 0) {
8943                 status = -EAGAIN;
8944                 *timeout = nfs4_update_delay(&exception.timeout);
8945         } else
8946                 lseg = pnfs_layout_process(lgp);
8947 out:
8948         trace_nfs4_layoutget(lgp->args.ctx,
8949                         &lgp->args.range,
8950                         &lgp->res.range,
8951                         &lgp->res.stateid,
8952                         status);
8953
8954         rpc_put_task(task);
8955         dprintk("<-- %s status=%d\n", __func__, status);
8956         if (status)
8957                 return ERR_PTR(status);
8958         return lseg;
8959 }
8960
8961 static void
8962 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8963 {
8964         struct nfs4_layoutreturn *lrp = calldata;
8965
8966         dprintk("--> %s\n", __func__);
8967         nfs4_setup_sequence(lrp->clp,
8968                         &lrp->args.seq_args,
8969                         &lrp->res.seq_res,
8970                         task);
8971         if (!pnfs_layout_is_valid(lrp->args.layout))
8972                 rpc_exit(task, 0);
8973 }
8974
8975 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8976 {
8977         struct nfs4_layoutreturn *lrp = calldata;
8978         struct nfs_server *server;
8979
8980         dprintk("--> %s\n", __func__);
8981
8982         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8983                 return;
8984
8985         server = NFS_SERVER(lrp->args.inode);
8986         switch (task->tk_status) {
8987         case -NFS4ERR_OLD_STATEID:
8988                 if (nfs4_layoutreturn_refresh_stateid(&lrp->args.stateid,
8989                                         &lrp->args.range,
8990                                         lrp->args.inode))
8991                         goto out_restart;
8992                 /* Fallthrough */
8993         default:
8994                 task->tk_status = 0;
8995                 /* Fallthrough */
8996         case 0:
8997                 break;
8998         case -NFS4ERR_DELAY:
8999                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9000                         break;
9001                 goto out_restart;
9002         }
9003         dprintk("<-- %s\n", __func__);
9004         return;
9005 out_restart:
9006         task->tk_status = 0;
9007         nfs4_sequence_free_slot(&lrp->res.seq_res);
9008         rpc_restart_call_prepare(task);
9009 }
9010
9011 static void nfs4_layoutreturn_release(void *calldata)
9012 {
9013         struct nfs4_layoutreturn *lrp = calldata;
9014         struct pnfs_layout_hdr *lo = lrp->args.layout;
9015
9016         dprintk("--> %s\n", __func__);
9017         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9018                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9019         nfs4_sequence_free_slot(&lrp->res.seq_res);
9020         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9021                 lrp->ld_private.ops->free(&lrp->ld_private);
9022         pnfs_put_layout_hdr(lrp->args.layout);
9023         nfs_iput_and_deactive(lrp->inode);
9024         kfree(calldata);
9025         dprintk("<-- %s\n", __func__);
9026 }
9027
9028 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9029         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9030         .rpc_call_done = nfs4_layoutreturn_done,
9031         .rpc_release = nfs4_layoutreturn_release,
9032 };
9033
9034 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9035 {
9036         struct rpc_task *task;
9037         struct rpc_message msg = {
9038                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9039                 .rpc_argp = &lrp->args,
9040                 .rpc_resp = &lrp->res,
9041                 .rpc_cred = lrp->cred,
9042         };
9043         struct rpc_task_setup task_setup_data = {
9044                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9045                 .rpc_message = &msg,
9046                 .callback_ops = &nfs4_layoutreturn_call_ops,
9047                 .callback_data = lrp,
9048         };
9049         int status = 0;
9050
9051         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9052                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9053                         &task_setup_data.rpc_client, &msg);
9054
9055         dprintk("--> %s\n", __func__);
9056         if (!sync) {
9057                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9058                 if (!lrp->inode) {
9059                         nfs4_layoutreturn_release(lrp);
9060                         return -EAGAIN;
9061                 }
9062                 task_setup_data.flags |= RPC_TASK_ASYNC;
9063         }
9064         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9065         task = rpc_run_task(&task_setup_data);
9066         if (IS_ERR(task))
9067                 return PTR_ERR(task);
9068         if (sync)
9069                 status = task->tk_status;
9070         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9071         dprintk("<-- %s status=%d\n", __func__, status);
9072         rpc_put_task(task);
9073         return status;
9074 }
9075
9076 static int
9077 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9078                 struct pnfs_device *pdev,
9079                 const struct cred *cred)
9080 {
9081         struct nfs4_getdeviceinfo_args args = {
9082                 .pdev = pdev,
9083                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9084                         NOTIFY_DEVICEID4_DELETE,
9085         };
9086         struct nfs4_getdeviceinfo_res res = {
9087                 .pdev = pdev,
9088         };
9089         struct rpc_message msg = {
9090                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9091                 .rpc_argp = &args,
9092                 .rpc_resp = &res,
9093                 .rpc_cred = cred,
9094         };
9095         int status;
9096
9097         dprintk("--> %s\n", __func__);
9098         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9099         if (res.notification & ~args.notify_types)
9100                 dprintk("%s: unsupported notification\n", __func__);
9101         if (res.notification != args.notify_types)
9102                 pdev->nocache = 1;
9103
9104         dprintk("<-- %s status=%d\n", __func__, status);
9105
9106         return status;
9107 }
9108
9109 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9110                 struct pnfs_device *pdev,
9111                 const struct cred *cred)
9112 {
9113         struct nfs4_exception exception = { };
9114         int err;
9115
9116         do {
9117                 err = nfs4_handle_exception(server,
9118                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9119                                         &exception);
9120         } while (exception.retry);
9121         return err;
9122 }
9123 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9124
9125 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9126 {
9127         struct nfs4_layoutcommit_data *data = calldata;
9128         struct nfs_server *server = NFS_SERVER(data->args.inode);
9129
9130         nfs4_setup_sequence(server->nfs_client,
9131                         &data->args.seq_args,
9132                         &data->res.seq_res,
9133                         task);
9134 }
9135
9136 static void
9137 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9138 {
9139         struct nfs4_layoutcommit_data *data = calldata;
9140         struct nfs_server *server = NFS_SERVER(data->args.inode);
9141
9142         if (!nfs41_sequence_done(task, &data->res.seq_res))
9143                 return;
9144
9145         switch (task->tk_status) { /* Just ignore these failures */
9146         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9147         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9148         case -NFS4ERR_BADLAYOUT:     /* no layout */
9149         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9150                 task->tk_status = 0;
9151         case 0:
9152                 break;
9153         default:
9154                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9155                         rpc_restart_call_prepare(task);
9156                         return;
9157                 }
9158         }
9159 }
9160
9161 static void nfs4_layoutcommit_release(void *calldata)
9162 {
9163         struct nfs4_layoutcommit_data *data = calldata;
9164
9165         pnfs_cleanup_layoutcommit(data);
9166         nfs_post_op_update_inode_force_wcc(data->args.inode,
9167                                            data->res.fattr);
9168         put_cred(data->cred);
9169         nfs_iput_and_deactive(data->inode);
9170         kfree(data);
9171 }
9172
9173 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9174         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9175         .rpc_call_done = nfs4_layoutcommit_done,
9176         .rpc_release = nfs4_layoutcommit_release,
9177 };
9178
9179 int
9180 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9181 {
9182         struct rpc_message msg = {
9183                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9184                 .rpc_argp = &data->args,
9185                 .rpc_resp = &data->res,
9186                 .rpc_cred = data->cred,
9187         };
9188         struct rpc_task_setup task_setup_data = {
9189                 .task = &data->task,
9190                 .rpc_client = NFS_CLIENT(data->args.inode),
9191                 .rpc_message = &msg,
9192                 .callback_ops = &nfs4_layoutcommit_ops,
9193                 .callback_data = data,
9194         };
9195         struct rpc_task *task;
9196         int status = 0;
9197
9198         dprintk("NFS: initiating layoutcommit call. sync %d "
9199                 "lbw: %llu inode %lu\n", sync,
9200                 data->args.lastbytewritten,
9201                 data->args.inode->i_ino);
9202
9203         if (!sync) {
9204                 data->inode = nfs_igrab_and_active(data->args.inode);
9205                 if (data->inode == NULL) {
9206                         nfs4_layoutcommit_release(data);
9207                         return -EAGAIN;
9208                 }
9209                 task_setup_data.flags = RPC_TASK_ASYNC;
9210         }
9211         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9212         task = rpc_run_task(&task_setup_data);
9213         if (IS_ERR(task))
9214                 return PTR_ERR(task);
9215         if (sync)
9216                 status = task->tk_status;
9217         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9218         dprintk("%s: status %d\n", __func__, status);
9219         rpc_put_task(task);
9220         return status;
9221 }
9222
9223 /*
9224  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9225  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9226  */
9227 static int
9228 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9229                     struct nfs_fsinfo *info,
9230                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9231 {
9232         struct nfs41_secinfo_no_name_args args = {
9233                 .style = SECINFO_STYLE_CURRENT_FH,
9234         };
9235         struct nfs4_secinfo_res res = {
9236                 .flavors = flavors,
9237         };
9238         struct rpc_message msg = {
9239                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9240                 .rpc_argp = &args,
9241                 .rpc_resp = &res,
9242         };
9243         struct rpc_clnt *clnt = server->client;
9244         const struct cred *cred = NULL;
9245         int status;
9246
9247         if (use_integrity) {
9248                 clnt = server->nfs_client->cl_rpcclient;
9249                 cred = nfs4_get_clid_cred(server->nfs_client);
9250                 msg.rpc_cred = cred;
9251         }
9252
9253         dprintk("--> %s\n", __func__);
9254         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
9255                                 &res.seq_res, 0);
9256         dprintk("<-- %s status=%d\n", __func__, status);
9257
9258         put_cred(cred);
9259
9260         return status;
9261 }
9262
9263 static int
9264 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9265                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9266 {
9267         struct nfs4_exception exception = { };
9268         int err;
9269         do {
9270                 /* first try using integrity protection */
9271                 err = -NFS4ERR_WRONGSEC;
9272
9273                 /* try to use integrity protection with machine cred */
9274                 if (_nfs4_is_integrity_protected(server->nfs_client))
9275                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9276                                                           flavors, true);
9277
9278                 /*
9279                  * if unable to use integrity protection, or SECINFO with
9280                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9281                  * disallowed by spec, but exists in deployed servers) use
9282                  * the current filesystem's rpc_client and the user cred.
9283                  */
9284                 if (err == -NFS4ERR_WRONGSEC)
9285                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9286                                                           flavors, false);
9287
9288                 switch (err) {
9289                 case 0:
9290                 case -NFS4ERR_WRONGSEC:
9291                 case -ENOTSUPP:
9292                         goto out;
9293                 default:
9294                         err = nfs4_handle_exception(server, err, &exception);
9295                 }
9296         } while (exception.retry);
9297 out:
9298         return err;
9299 }
9300
9301 static int
9302 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9303                     struct nfs_fsinfo *info)
9304 {
9305         int err;
9306         struct page *page;
9307         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9308         struct nfs4_secinfo_flavors *flavors;
9309         struct nfs4_secinfo4 *secinfo;
9310         int i;
9311
9312         page = alloc_page(GFP_KERNEL);
9313         if (!page) {
9314                 err = -ENOMEM;
9315                 goto out;
9316         }
9317
9318         flavors = page_address(page);
9319         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9320
9321         /*
9322          * Fall back on "guess and check" method if
9323          * the server doesn't support SECINFO_NO_NAME
9324          */
9325         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9326                 err = nfs4_find_root_sec(server, fhandle, info);
9327                 goto out_freepage;
9328         }
9329         if (err)
9330                 goto out_freepage;
9331
9332         for (i = 0; i < flavors->num_flavors; i++) {
9333                 secinfo = &flavors->flavors[i];
9334
9335                 switch (secinfo->flavor) {
9336                 case RPC_AUTH_NULL:
9337                 case RPC_AUTH_UNIX:
9338                 case RPC_AUTH_GSS:
9339                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9340                                         &secinfo->flavor_info);
9341                         break;
9342                 default:
9343                         flavor = RPC_AUTH_MAXFLAVOR;
9344                         break;
9345                 }
9346
9347                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9348                         flavor = RPC_AUTH_MAXFLAVOR;
9349
9350                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9351                         err = nfs4_lookup_root_sec(server, fhandle,
9352                                                    info, flavor);
9353                         if (!err)
9354                                 break;
9355                 }
9356         }
9357
9358         if (flavor == RPC_AUTH_MAXFLAVOR)
9359                 err = -EPERM;
9360
9361 out_freepage:
9362         put_page(page);
9363         if (err == -EACCES)
9364                 return -EPERM;
9365 out:
9366         return err;
9367 }
9368
9369 static int _nfs41_test_stateid(struct nfs_server *server,
9370                 nfs4_stateid *stateid,
9371                 const struct cred *cred)
9372 {
9373         int status;
9374         struct nfs41_test_stateid_args args = {
9375                 .stateid = stateid,
9376         };
9377         struct nfs41_test_stateid_res res;
9378         struct rpc_message msg = {
9379                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9380                 .rpc_argp = &args,
9381                 .rpc_resp = &res,
9382                 .rpc_cred = cred,
9383         };
9384         struct rpc_clnt *rpc_client = server->client;
9385
9386         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9387                 &rpc_client, &msg);
9388
9389         dprintk("NFS call  test_stateid %p\n", stateid);
9390         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9391         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9392                         &args.seq_args, &res.seq_res);
9393         if (status != NFS_OK) {
9394                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9395                 return status;
9396         }
9397         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9398         return -res.status;
9399 }
9400
9401 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9402                 int err, struct nfs4_exception *exception)
9403 {
9404         exception->retry = 0;
9405         switch(err) {
9406         case -NFS4ERR_DELAY:
9407         case -NFS4ERR_RETRY_UNCACHED_REP:
9408                 nfs4_handle_exception(server, err, exception);
9409                 break;
9410         case -NFS4ERR_BADSESSION:
9411         case -NFS4ERR_BADSLOT:
9412         case -NFS4ERR_BAD_HIGH_SLOT:
9413         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9414         case -NFS4ERR_DEADSESSION:
9415                 nfs4_do_handle_exception(server, err, exception);
9416         }
9417 }
9418
9419 /**
9420  * nfs41_test_stateid - perform a TEST_STATEID operation
9421  *
9422  * @server: server / transport on which to perform the operation
9423  * @stateid: state ID to test
9424  * @cred: credential
9425  *
9426  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9427  * Otherwise a negative NFS4ERR value is returned if the operation
9428  * failed or the state ID is not currently valid.
9429  */
9430 static int nfs41_test_stateid(struct nfs_server *server,
9431                 nfs4_stateid *stateid,
9432                 const struct cred *cred)
9433 {
9434         struct nfs4_exception exception = { };
9435         int err;
9436         do {
9437                 err = _nfs41_test_stateid(server, stateid, cred);
9438                 nfs4_handle_delay_or_session_error(server, err, &exception);
9439         } while (exception.retry);
9440         return err;
9441 }
9442
9443 struct nfs_free_stateid_data {
9444         struct nfs_server *server;
9445         struct nfs41_free_stateid_args args;
9446         struct nfs41_free_stateid_res res;
9447 };
9448
9449 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9450 {
9451         struct nfs_free_stateid_data *data = calldata;
9452         nfs4_setup_sequence(data->server->nfs_client,
9453                         &data->args.seq_args,
9454                         &data->res.seq_res,
9455                         task);
9456 }
9457
9458 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9459 {
9460         struct nfs_free_stateid_data *data = calldata;
9461
9462         nfs41_sequence_done(task, &data->res.seq_res);
9463
9464         switch (task->tk_status) {
9465         case -NFS4ERR_DELAY:
9466                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9467                         rpc_restart_call_prepare(task);
9468         }
9469 }
9470
9471 static void nfs41_free_stateid_release(void *calldata)
9472 {
9473         kfree(calldata);
9474 }
9475
9476 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9477         .rpc_call_prepare = nfs41_free_stateid_prepare,
9478         .rpc_call_done = nfs41_free_stateid_done,
9479         .rpc_release = nfs41_free_stateid_release,
9480 };
9481
9482 /**
9483  * nfs41_free_stateid - perform a FREE_STATEID operation
9484  *
9485  * @server: server / transport on which to perform the operation
9486  * @stateid: state ID to release
9487  * @cred: credential
9488  * @privileged: set to true if this call needs to be privileged
9489  *
9490  * Note: this function is always asynchronous.
9491  */
9492 static int nfs41_free_stateid(struct nfs_server *server,
9493                 const nfs4_stateid *stateid,
9494                 const struct cred *cred,
9495                 bool privileged)
9496 {
9497         struct rpc_message msg = {
9498                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9499                 .rpc_cred = cred,
9500         };
9501         struct rpc_task_setup task_setup = {
9502                 .rpc_client = server->client,
9503                 .rpc_message = &msg,
9504                 .callback_ops = &nfs41_free_stateid_ops,
9505                 .flags = RPC_TASK_ASYNC,
9506         };
9507         struct nfs_free_stateid_data *data;
9508         struct rpc_task *task;
9509
9510         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9511                 &task_setup.rpc_client, &msg);
9512
9513         dprintk("NFS call  free_stateid %p\n", stateid);
9514         data = kmalloc(sizeof(*data), GFP_NOFS);
9515         if (!data)
9516                 return -ENOMEM;
9517         data->server = server;
9518         nfs4_stateid_copy(&data->args.stateid, stateid);
9519
9520         task_setup.callback_data = data;
9521
9522         msg.rpc_argp = &data->args;
9523         msg.rpc_resp = &data->res;
9524         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9525         task = rpc_run_task(&task_setup);
9526         if (IS_ERR(task))
9527                 return PTR_ERR(task);
9528         rpc_put_task(task);
9529         return 0;
9530 }
9531
9532 static void
9533 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9534 {
9535         const struct cred *cred = lsp->ls_state->owner->so_cred;
9536
9537         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9538         nfs4_free_lock_state(server, lsp);
9539 }
9540
9541 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9542                 const nfs4_stateid *s2)
9543 {
9544         if (s1->type != s2->type)
9545                 return false;
9546
9547         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9548                 return false;
9549
9550         if (s1->seqid == s2->seqid)
9551                 return true;
9552
9553         return s1->seqid == 0 || s2->seqid == 0;
9554 }
9555
9556 #endif /* CONFIG_NFS_V4_1 */
9557
9558 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9559                 const nfs4_stateid *s2)
9560 {
9561         return nfs4_stateid_match(s1, s2);
9562 }
9563
9564
9565 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9566         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9567         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9568         .recover_open   = nfs4_open_reclaim,
9569         .recover_lock   = nfs4_lock_reclaim,
9570         .establish_clid = nfs4_init_clientid,
9571         .detect_trunking = nfs40_discover_server_trunking,
9572 };
9573
9574 #if defined(CONFIG_NFS_V4_1)
9575 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9576         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9577         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9578         .recover_open   = nfs4_open_reclaim,
9579         .recover_lock   = nfs4_lock_reclaim,
9580         .establish_clid = nfs41_init_clientid,
9581         .reclaim_complete = nfs41_proc_reclaim_complete,
9582         .detect_trunking = nfs41_discover_server_trunking,
9583 };
9584 #endif /* CONFIG_NFS_V4_1 */
9585
9586 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9587         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9588         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9589         .recover_open   = nfs40_open_expired,
9590         .recover_lock   = nfs4_lock_expired,
9591         .establish_clid = nfs4_init_clientid,
9592 };
9593
9594 #if defined(CONFIG_NFS_V4_1)
9595 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9596         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9597         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9598         .recover_open   = nfs41_open_expired,
9599         .recover_lock   = nfs41_lock_expired,
9600         .establish_clid = nfs41_init_clientid,
9601 };
9602 #endif /* CONFIG_NFS_V4_1 */
9603
9604 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9605         .sched_state_renewal = nfs4_proc_async_renew,
9606         .get_state_renewal_cred = nfs4_get_renew_cred,
9607         .renew_lease = nfs4_proc_renew,
9608 };
9609
9610 #if defined(CONFIG_NFS_V4_1)
9611 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9612         .sched_state_renewal = nfs41_proc_async_sequence,
9613         .get_state_renewal_cred = nfs4_get_machine_cred,
9614         .renew_lease = nfs4_proc_sequence,
9615 };
9616 #endif
9617
9618 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9619         .get_locations = _nfs40_proc_get_locations,
9620         .fsid_present = _nfs40_proc_fsid_present,
9621 };
9622
9623 #if defined(CONFIG_NFS_V4_1)
9624 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9625         .get_locations = _nfs41_proc_get_locations,
9626         .fsid_present = _nfs41_proc_fsid_present,
9627 };
9628 #endif  /* CONFIG_NFS_V4_1 */
9629
9630 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9631         .minor_version = 0,
9632         .init_caps = NFS_CAP_READDIRPLUS
9633                 | NFS_CAP_ATOMIC_OPEN
9634                 | NFS_CAP_POSIX_LOCK,
9635         .init_client = nfs40_init_client,
9636         .shutdown_client = nfs40_shutdown_client,
9637         .match_stateid = nfs4_match_stateid,
9638         .find_root_sec = nfs4_find_root_sec,
9639         .free_lock_state = nfs4_release_lockowner,
9640         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9641         .alloc_seqid = nfs_alloc_seqid,
9642         .call_sync_ops = &nfs40_call_sync_ops,
9643         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9644         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9645         .state_renewal_ops = &nfs40_state_renewal_ops,
9646         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9647 };
9648
9649 #if defined(CONFIG_NFS_V4_1)
9650 static struct nfs_seqid *
9651 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9652 {
9653         return NULL;
9654 }
9655
9656 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9657         .minor_version = 1,
9658         .init_caps = NFS_CAP_READDIRPLUS
9659                 | NFS_CAP_ATOMIC_OPEN
9660                 | NFS_CAP_POSIX_LOCK
9661                 | NFS_CAP_STATEID_NFSV41
9662                 | NFS_CAP_ATOMIC_OPEN_V1
9663                 | NFS_CAP_LGOPEN,
9664         .init_client = nfs41_init_client,
9665         .shutdown_client = nfs41_shutdown_client,
9666         .match_stateid = nfs41_match_stateid,
9667         .find_root_sec = nfs41_find_root_sec,
9668         .free_lock_state = nfs41_free_lock_state,
9669         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9670         .alloc_seqid = nfs_alloc_no_seqid,
9671         .session_trunk = nfs4_test_session_trunk,
9672         .call_sync_ops = &nfs41_call_sync_ops,
9673         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9674         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9675         .state_renewal_ops = &nfs41_state_renewal_ops,
9676         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9677 };
9678 #endif
9679
9680 #if defined(CONFIG_NFS_V4_2)
9681 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9682         .minor_version = 2,
9683         .init_caps = NFS_CAP_READDIRPLUS
9684                 | NFS_CAP_ATOMIC_OPEN
9685                 | NFS_CAP_POSIX_LOCK
9686                 | NFS_CAP_STATEID_NFSV41
9687                 | NFS_CAP_ATOMIC_OPEN_V1
9688                 | NFS_CAP_LGOPEN
9689                 | NFS_CAP_ALLOCATE
9690                 | NFS_CAP_COPY
9691                 | NFS_CAP_OFFLOAD_CANCEL
9692                 | NFS_CAP_DEALLOCATE
9693                 | NFS_CAP_SEEK
9694                 | NFS_CAP_LAYOUTSTATS
9695                 | NFS_CAP_CLONE
9696                 | NFS_CAP_LAYOUTERROR,
9697         .init_client = nfs41_init_client,
9698         .shutdown_client = nfs41_shutdown_client,
9699         .match_stateid = nfs41_match_stateid,
9700         .find_root_sec = nfs41_find_root_sec,
9701         .free_lock_state = nfs41_free_lock_state,
9702         .call_sync_ops = &nfs41_call_sync_ops,
9703         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9704         .alloc_seqid = nfs_alloc_no_seqid,
9705         .session_trunk = nfs4_test_session_trunk,
9706         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9707         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9708         .state_renewal_ops = &nfs41_state_renewal_ops,
9709         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9710 };
9711 #endif
9712
9713 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9714         [0] = &nfs_v4_0_minor_ops,
9715 #if defined(CONFIG_NFS_V4_1)
9716         [1] = &nfs_v4_1_minor_ops,
9717 #endif
9718 #if defined(CONFIG_NFS_V4_2)
9719         [2] = &nfs_v4_2_minor_ops,
9720 #endif
9721 };
9722
9723 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9724 {
9725         ssize_t error, error2;
9726
9727         error = generic_listxattr(dentry, list, size);
9728         if (error < 0)
9729                 return error;
9730         if (list) {
9731                 list += error;
9732                 size -= error;
9733         }
9734
9735         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9736         if (error2 < 0)
9737                 return error2;
9738         return error + error2;
9739 }
9740
9741 static const struct inode_operations nfs4_dir_inode_operations = {
9742         .create         = nfs_create,
9743         .lookup         = nfs_lookup,
9744         .atomic_open    = nfs_atomic_open,
9745         .link           = nfs_link,
9746         .unlink         = nfs_unlink,
9747         .symlink        = nfs_symlink,
9748         .mkdir          = nfs_mkdir,
9749         .rmdir          = nfs_rmdir,
9750         .mknod          = nfs_mknod,
9751         .rename         = nfs_rename,
9752         .permission     = nfs_permission,
9753         .getattr        = nfs_getattr,
9754         .setattr        = nfs_setattr,
9755         .listxattr      = nfs4_listxattr,
9756 };
9757
9758 static const struct inode_operations nfs4_file_inode_operations = {
9759         .permission     = nfs_permission,
9760         .getattr        = nfs_getattr,
9761         .setattr        = nfs_setattr,
9762         .listxattr      = nfs4_listxattr,
9763 };
9764
9765 const struct nfs_rpc_ops nfs_v4_clientops = {
9766         .version        = 4,                    /* protocol version */
9767         .dentry_ops     = &nfs4_dentry_operations,
9768         .dir_inode_ops  = &nfs4_dir_inode_operations,
9769         .file_inode_ops = &nfs4_file_inode_operations,
9770         .file_ops       = &nfs4_file_operations,
9771         .getroot        = nfs4_proc_get_root,
9772         .submount       = nfs4_submount,
9773         .try_mount      = nfs4_try_mount,
9774         .getattr        = nfs4_proc_getattr,
9775         .setattr        = nfs4_proc_setattr,
9776         .lookup         = nfs4_proc_lookup,
9777         .lookupp        = nfs4_proc_lookupp,
9778         .access         = nfs4_proc_access,
9779         .readlink       = nfs4_proc_readlink,
9780         .create         = nfs4_proc_create,
9781         .remove         = nfs4_proc_remove,
9782         .unlink_setup   = nfs4_proc_unlink_setup,
9783         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9784         .unlink_done    = nfs4_proc_unlink_done,
9785         .rename_setup   = nfs4_proc_rename_setup,
9786         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9787         .rename_done    = nfs4_proc_rename_done,
9788         .link           = nfs4_proc_link,
9789         .symlink        = nfs4_proc_symlink,
9790         .mkdir          = nfs4_proc_mkdir,
9791         .rmdir          = nfs4_proc_rmdir,
9792         .readdir        = nfs4_proc_readdir,
9793         .mknod          = nfs4_proc_mknod,
9794         .statfs         = nfs4_proc_statfs,
9795         .fsinfo         = nfs4_proc_fsinfo,
9796         .pathconf       = nfs4_proc_pathconf,
9797         .set_capabilities = nfs4_server_capabilities,
9798         .decode_dirent  = nfs4_decode_dirent,
9799         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9800         .read_setup     = nfs4_proc_read_setup,
9801         .read_done      = nfs4_read_done,
9802         .write_setup    = nfs4_proc_write_setup,
9803         .write_done     = nfs4_write_done,
9804         .commit_setup   = nfs4_proc_commit_setup,
9805         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9806         .commit_done    = nfs4_commit_done,
9807         .lock           = nfs4_proc_lock,
9808         .clear_acl_cache = nfs4_zap_acl_attr,
9809         .close_context  = nfs4_close_context,
9810         .open_context   = nfs4_atomic_open,
9811         .have_delegation = nfs4_have_delegation,
9812         .alloc_client   = nfs4_alloc_client,
9813         .init_client    = nfs4_init_client,
9814         .free_client    = nfs4_free_client,
9815         .create_server  = nfs4_create_server,
9816         .clone_server   = nfs_clone_server,
9817 };
9818
9819 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9820         .name   = XATTR_NAME_NFSV4_ACL,
9821         .list   = nfs4_xattr_list_nfs4_acl,
9822         .get    = nfs4_xattr_get_nfs4_acl,
9823         .set    = nfs4_xattr_set_nfs4_acl,
9824 };
9825
9826 const struct xattr_handler *nfs4_xattr_handlers[] = {
9827         &nfs4_xattr_nfs4_acl_handler,
9828 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9829         &nfs4_xattr_nfs4_label_handler,
9830 #endif
9831         NULL
9832 };
9833
9834 /*
9835  * Local variables:
9836  *  c-basic-offset: 8
9837  * End:
9838  */