fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57 #include <linux/iversion.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_BITMASK_SZ         3
  75
  76 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  77 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  78
  79 /* file attributes which can be mapped to nfs attributes */
  80 #define NFS4_VALID_ATTRS (ATTR_MODE \
  81         | ATTR_UID \
  82         | ATTR_GID \
  83         | ATTR_SIZE \
  84         | ATTR_ATIME \
  85         | ATTR_MTIME \
  86         | ATTR_CTIME \
  87         | ATTR_ATIME_SET \
  88         | ATTR_MTIME_SET)
  89
  90 struct nfs4_opendata;
  91 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  92 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  93 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  94 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label, struct inode *inode);
  95 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  96 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  97                             struct nfs_fattr *fattr, struct iattr *sattr,
  98                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  99                             struct nfs4_label *olabel);
 100 #ifdef CONFIG_NFS_V4_1
 101 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 102                 struct rpc_cred *cred,
 103                 struct nfs4_slot *slot,
 104                 bool is_privileged);
 105 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 106                 struct rpc_cred *);
 107 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 108                 struct rpc_cred *, bool);
 109 #endif
 110
 111 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 112 static inline struct nfs4_label *
 113 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 114         struct iattr *sattr, struct nfs4_label *label)
 115 {
 116         int err;
 117
 118         if (label == NULL)
 119                 return NULL;
 120
 121         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 122                 return NULL;
 123
 124         err = security_dentry_init_security(dentry, sattr->ia_mode,
 125                                 &dentry->d_name, (void **)&label->label, &label->len);
 126         if (err == 0)
 127                 return label;
 128
 129         return NULL;
 130 }
 131 static inline void
 132 nfs4_label_release_security(struct nfs4_label *label)
 133 {
 134         if (label)
 135                 security_release_secctx(label->label, label->len);
 136 }
 137 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 138 {
 139         if (label)
 140                 return server->attr_bitmask;
 141
 142         return server->attr_bitmask_nl;
 143 }
 144 #else
 145 static inline struct nfs4_label *
 146 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 147         struct iattr *sattr, struct nfs4_label *l)
 148 { return NULL; }
 149 static inline void
 150 nfs4_label_release_security(struct nfs4_label *label)
 151 { return; }
 152 static inline u32 *
 153 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 154 { return server->attr_bitmask; }
 155 #endif
 156
 157 /* Prevent leaks of NFSv4 errors into userland */
 158 static int nfs4_map_errors(int err)
 159 {
 160         if (err >= -1000)
 161                 return err;
 162         switch (err) {
 163         case -NFS4ERR_RESOURCE:
 164         case -NFS4ERR_LAYOUTTRYLATER:
 165         case -NFS4ERR_RECALLCONFLICT:
 166                 return -EREMOTEIO;
 167         case -NFS4ERR_WRONGSEC:
 168         case -NFS4ERR_WRONG_CRED:
 169                 return -EPERM;
 170         case -NFS4ERR_BADOWNER:
 171         case -NFS4ERR_BADNAME:
 172                 return -EINVAL;
 173         case -NFS4ERR_SHARE_DENIED:
 174                 return -EACCES;
 175         case -NFS4ERR_MINOR_VERS_MISMATCH:
 176                 return -EPROTONOSUPPORT;
 177         case -NFS4ERR_FILE_OPEN:
 178                 return -EBUSY;
 179         default:
 180                 dprintk("%s could not handle NFSv4 error %d\n",
 181                                 __func__, -err);
 182                 break;
 183         }
 184         return -EIO;
 185 }
 186
 187 /*
 188  * This is our standard bitmap for GETATTR requests.
 189  */
 190 const u32 nfs4_fattr_bitmap[3] = {
 191         FATTR4_WORD0_TYPE
 192         | FATTR4_WORD0_CHANGE
 193         | FATTR4_WORD0_SIZE
 194         | FATTR4_WORD0_FSID
 195         | FATTR4_WORD0_FILEID,
 196         FATTR4_WORD1_MODE
 197         | FATTR4_WORD1_NUMLINKS
 198         | FATTR4_WORD1_OWNER
 199         | FATTR4_WORD1_OWNER_GROUP
 200         | FATTR4_WORD1_RAWDEV
 201         | FATTR4_WORD1_SPACE_USED
 202         | FATTR4_WORD1_TIME_ACCESS
 203         | FATTR4_WORD1_TIME_METADATA
 204         | FATTR4_WORD1_TIME_MODIFY
 205         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 206 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 207         FATTR4_WORD2_SECURITY_LABEL
 208 #endif
 209 };
 210
 211 static const u32 nfs4_pnfs_open_bitmap[3] = {
 212         FATTR4_WORD0_TYPE
 213         | FATTR4_WORD0_CHANGE
 214         | FATTR4_WORD0_SIZE
 215         | FATTR4_WORD0_FSID
 216         | FATTR4_WORD0_FILEID,
 217         FATTR4_WORD1_MODE
 218         | FATTR4_WORD1_NUMLINKS
 219         | FATTR4_WORD1_OWNER
 220         | FATTR4_WORD1_OWNER_GROUP
 221         | FATTR4_WORD1_RAWDEV
 222         | FATTR4_WORD1_SPACE_USED
 223         | FATTR4_WORD1_TIME_ACCESS
 224         | FATTR4_WORD1_TIME_METADATA
 225         | FATTR4_WORD1_TIME_MODIFY,
 226         FATTR4_WORD2_MDSTHRESHOLD
 227 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 228         | FATTR4_WORD2_SECURITY_LABEL
 229 #endif
 230 };
 231
 232 static const u32 nfs4_open_noattr_bitmap[3] = {
 233         FATTR4_WORD0_TYPE
 234         | FATTR4_WORD0_FILEID,
 235 };
 236
 237 const u32 nfs4_statfs_bitmap[3] = {
 238         FATTR4_WORD0_FILES_AVAIL
 239         | FATTR4_WORD0_FILES_FREE
 240         | FATTR4_WORD0_FILES_TOTAL,
 241         FATTR4_WORD1_SPACE_AVAIL
 242         | FATTR4_WORD1_SPACE_FREE
 243         | FATTR4_WORD1_SPACE_TOTAL
 244 };
 245
 246 const u32 nfs4_pathconf_bitmap[3] = {
 247         FATTR4_WORD0_MAXLINK
 248         | FATTR4_WORD0_MAXNAME,
 249         0
 250 };
 251
 252 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 253                         | FATTR4_WORD0_MAXREAD
 254                         | FATTR4_WORD0_MAXWRITE
 255                         | FATTR4_WORD0_LEASE_TIME,
 256                         FATTR4_WORD1_TIME_DELTA
 257                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 258                         FATTR4_WORD2_LAYOUT_BLKSIZE
 259                         | FATTR4_WORD2_CLONE_BLKSIZE
 260 };
 261
 262 const u32 nfs4_fs_locations_bitmap[3] = {
 263         FATTR4_WORD0_CHANGE
 264         | FATTR4_WORD0_SIZE
 265         | FATTR4_WORD0_FSID
 266         | FATTR4_WORD0_FILEID
 267         | FATTR4_WORD0_FS_LOCATIONS,
 268         FATTR4_WORD1_OWNER
 269         | FATTR4_WORD1_OWNER_GROUP
 270         | FATTR4_WORD1_RAWDEV
 271         | FATTR4_WORD1_SPACE_USED
 272         | FATTR4_WORD1_TIME_ACCESS
 273         | FATTR4_WORD1_TIME_METADATA
 274         | FATTR4_WORD1_TIME_MODIFY
 275         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 276 };
 277
 278 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 279                 struct inode *inode)
 280 {
 281         unsigned long cache_validity;
 282
 283         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 284         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 285                 return;
 286
 287         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 288         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 289                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 290                                 | NFS_INO_INVALID_SIZE);
 291
 292         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 293                 dst[0] &= ~FATTR4_WORD0_SIZE;
 294
 295         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 296                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 297 }
 298
 299 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 300                 const __u32 *src, struct inode *inode)
 301 {
 302         nfs4_bitmap_copy_adjust(dst, src, inode);
 303 }
 304
 305 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 306                 struct nfs4_readdir_arg *readdir)
 307 {
 308         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 309         __be32 *start, *p;
 310
 311         if (cookie > 2) {
 312                 readdir->cookie = cookie;
 313                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 314                 return;
 315         }
 316
 317         readdir->cookie = 0;
 318         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 319         if (cookie == 2)
 320                 return;
 321
 322         /*
 323          * NFSv4 servers do not return entries for '.' and '..'
 324          * Therefore, we fake these entries here.  We let '.'
 325          * have cookie 0 and '..' have cookie 1.  Note that
 326          * when talking to the server, we always send cookie 0
 327          * instead of 1 or 2.
 328          */
 329         start = p = kmap_atomic(*readdir->pages);
 330
 331         if (cookie == 0) {
 332                 *p++ = xdr_one;                                  /* next */
 333                 *p++ = xdr_zero;                   /* cookie, first word */
 334                 *p++ = xdr_one;                   /* cookie, second word */
 335                 *p++ = xdr_one;                             /* entry len */
 336                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 337                 p++;
 338                 *p++ = xdr_one;                         /* bitmap length */
 339                 *p++ = htonl(attrs);                           /* bitmap */
 340                 *p++ = htonl(12);             /* attribute buffer length */
 341                 *p++ = htonl(NF4DIR);
 342                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 343         }
 344
 345         *p++ = xdr_one;                                  /* next */
 346         *p++ = xdr_zero;                   /* cookie, first word */
 347         *p++ = xdr_two;                   /* cookie, second word */
 348         *p++ = xdr_two;                             /* entry len */
 349         memcpy(p, "..\0\0", 4);                         /* entry */
 350         p++;
 351         *p++ = xdr_one;                         /* bitmap length */
 352         *p++ = htonl(attrs);                           /* bitmap */
 353         *p++ = htonl(12);             /* attribute buffer length */
 354         *p++ = htonl(NF4DIR);
 355         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 356
 357         readdir->pgbase = (char *)p - (char *)start;
 358         readdir->count -= readdir->pgbase;
 359         kunmap_atomic(start);
 360 }
 361
 362 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 363                 nfs4_stateid *stateid,
 364                 struct rpc_cred *cred)
 365 {
 366         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 367
 368         ops->test_and_free_expired(server, stateid, cred);
 369 }
 370
 371 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 372                 nfs4_stateid *stateid,
 373                 struct rpc_cred *cred)
 374 {
 375         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 376         nfs4_test_and_free_stateid(server, stateid, cred);
 377 }
 378
 379 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 380                 const nfs4_stateid *stateid,
 381                 struct rpc_cred *cred)
 382 {
 383         nfs4_stateid tmp;
 384
 385         nfs4_stateid_copy(&tmp, stateid);
 386         __nfs4_free_revoked_stateid(server, &tmp, cred);
 387 }
 388
 389 static long nfs4_update_delay(long *timeout)
 390 {
 391         long ret;
 392         if (!timeout)
 393                 return NFS4_POLL_RETRY_MAX;
 394         if (*timeout <= 0)
 395                 *timeout = NFS4_POLL_RETRY_MIN;
 396         if (*timeout > NFS4_POLL_RETRY_MAX)
 397                 *timeout = NFS4_POLL_RETRY_MAX;
 398         ret = *timeout;
 399         *timeout <<= 1;
 400         return ret;
 401 }
 402
 403 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 404 {
 405         int res = 0;
 406
 407         might_sleep();
 408
 409         freezable_schedule_timeout_killable_unsafe(
 410                 nfs4_update_delay(timeout));
 411         if (fatal_signal_pending(current))
 412                 res = -ERESTARTSYS;
 413         return res;
 414 }
 415
 416 /* This is the error handling routine for processes that are allowed
 417  * to sleep.
 418  */
 419 static int nfs4_do_handle_exception(struct nfs_server *server,
 420                 int errorcode, struct nfs4_exception *exception)
 421 {
 422         struct nfs_client *clp = server->nfs_client;
 423         struct nfs4_state *state = exception->state;
 424         const nfs4_stateid *stateid = exception->stateid;
 425         struct inode *inode = exception->inode;
 426         int ret = errorcode;
 427
 428         exception->delay = 0;
 429         exception->recovering = 0;
 430         exception->retry = 0;
 431
 432         if (stateid == NULL && state != NULL)
 433                 stateid = &state->stateid;
 434
 435         switch(errorcode) {
 436                 case 0:
 437                         return 0;
 438                 case -NFS4ERR_BADHANDLE:
 439                 case -ESTALE:
 440                         if (inode != NULL && S_ISREG(inode->i_mode))
 441                                 pnfs_destroy_layout(NFS_I(inode));
 442                         break;
 443                 case -NFS4ERR_DELEG_REVOKED:
 444                 case -NFS4ERR_ADMIN_REVOKED:
 445                 case -NFS4ERR_EXPIRED:
 446                 case -NFS4ERR_BAD_STATEID:
 447                         if (inode != NULL && stateid != NULL) {
 448                                 nfs_inode_find_state_and_recover(inode,
 449                                                 stateid);
 450                                 goto wait_on_recovery;
 451                         }
 452                         /* Fall through */
 453                 case -NFS4ERR_OPENMODE:
 454                         if (inode) {
 455                                 int err;
 456
 457                                 err = nfs_async_inode_return_delegation(inode,
 458                                                 stateid);
 459                                 if (err == 0)
 460                                         goto wait_on_recovery;
 461                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 462                                         exception->retry = 1;
 463                                         break;
 464                                 }
 465                         }
 466                         if (state == NULL)
 467                                 break;
 468                         ret = nfs4_schedule_stateid_recovery(server, state);
 469                         if (ret < 0)
 470                                 break;
 471                         goto wait_on_recovery;
 472                 case -NFS4ERR_STALE_STATEID:
 473                 case -NFS4ERR_STALE_CLIENTID:
 474                         nfs4_schedule_lease_recovery(clp);
 475                         goto wait_on_recovery;
 476                 case -NFS4ERR_MOVED:
 477                         ret = nfs4_schedule_migration_recovery(server);
 478                         if (ret < 0)
 479                                 break;
 480                         goto wait_on_recovery;
 481                 case -NFS4ERR_LEASE_MOVED:
 482                         nfs4_schedule_lease_moved_recovery(clp);
 483                         goto wait_on_recovery;
 484 #if defined(CONFIG_NFS_V4_1)
 485                 case -NFS4ERR_BADSESSION:
 486                 case -NFS4ERR_BADSLOT:
 487                 case -NFS4ERR_BAD_HIGH_SLOT:
 488                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 489                 case -NFS4ERR_DEADSESSION:
 490                 case -NFS4ERR_SEQ_FALSE_RETRY:
 491                 case -NFS4ERR_SEQ_MISORDERED:
 492                         dprintk("%s ERROR: %d Reset session\n", __func__,
 493                                 errorcode);
 494                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 495                         goto wait_on_recovery;
 496 #endif /* defined(CONFIG_NFS_V4_1) */
 497                 case -NFS4ERR_FILE_OPEN:
 498                         if (exception->timeout > HZ) {
 499                                 /* We have retried a decent amount, time to
 500                                  * fail
 501                                  */
 502                                 ret = -EBUSY;
 503                                 break;
 504                         }
 505                         /* Fall through */
 506                 case -NFS4ERR_DELAY:
 507                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 508                         /* Fall through */
 509                 case -NFS4ERR_GRACE:
 510                 case -NFS4ERR_LAYOUTTRYLATER:
 511                 case -NFS4ERR_RECALLCONFLICT:
 512                         exception->delay = 1;
 513                         return 0;
 514
 515                 case -NFS4ERR_RETRY_UNCACHED_REP:
 516                 case -NFS4ERR_OLD_STATEID:
 517                         exception->retry = 1;
 518                         break;
 519                 case -NFS4ERR_BADOWNER:
 520                         /* The following works around a Linux server bug! */
 521                 case -NFS4ERR_BADNAME:
 522                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 523                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 524                                 exception->retry = 1;
 525                                 printk(KERN_WARNING "NFS: v4 server %s "
 526                                                 "does not accept raw "
 527                                                 "uid/gids. "
 528                                                 "Reenabling the idmapper.\n",
 529                                                 server->nfs_client->cl_hostname);
 530                         }
 531         }
 532         /* We failed to handle the error */
 533         return nfs4_map_errors(ret);
 534 wait_on_recovery:
 535         exception->recovering = 1;
 536         return 0;
 537 }
 538
 539 /* This is the error handling routine for processes that are allowed
 540  * to sleep.
 541  */
 542 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 543 {
 544         struct nfs_client *clp = server->nfs_client;
 545         int ret;
 546
 547         ret = nfs4_do_handle_exception(server, errorcode, exception);
 548         if (exception->delay) {
 549                 ret = nfs4_delay(server->client, &exception->timeout);
 550                 goto out_retry;
 551         }
 552         if (exception->recovering) {
 553                 ret = nfs4_wait_clnt_recover(clp);
 554                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 555                         return -EIO;
 556                 goto out_retry;
 557         }
 558         return ret;
 559 out_retry:
 560         if (ret == 0)
 561                 exception->retry = 1;
 562         return ret;
 563 }
 564
 565 static int
 566 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 567                 int errorcode, struct nfs4_exception *exception)
 568 {
 569         struct nfs_client *clp = server->nfs_client;
 570         int ret;
 571
 572         ret = nfs4_do_handle_exception(server, errorcode, exception);
 573         if (exception->delay) {
 574                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 575                 goto out_retry;
 576         }
 577         if (exception->recovering) {
 578                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 579                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 580                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 581                 goto out_retry;
 582         }
 583         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 584                 ret = -EIO;
 585         return ret;
 586 out_retry:
 587         if (ret == 0) {
 588                 exception->retry = 1;
 589                 /*
 590                  * For NFS4ERR_MOVED, the client transport will need to
 591                  * be recomputed after migration recovery has completed.
 592                  */
 593                 if (errorcode == -NFS4ERR_MOVED)
 594                         rpc_task_release_transport(task);
 595         }
 596         return ret;
 597 }
 598
 599 int
 600 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 601                         struct nfs4_state *state, long *timeout)
 602 {
 603         struct nfs4_exception exception = {
 604                 .state = state,
 605         };
 606
 607         if (task->tk_status >= 0)
 608                 return 0;
 609         if (timeout)
 610                 exception.timeout = *timeout;
 611         task->tk_status = nfs4_async_handle_exception(task, server,
 612                         task->tk_status,
 613                         &exception);
 614         if (exception.delay && timeout)
 615                 *timeout = exception.timeout;
 616         if (exception.retry)
 617                 return -EAGAIN;
 618         return 0;
 619 }
 620
 621 /*
 622  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 623  * or 'false' otherwise.
 624  */
 625 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 626 {
 627         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 628         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 629 }
 630
 631 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 632 {
 633         spin_lock(&clp->cl_lock);
 634         if (time_before(clp->cl_last_renewal,timestamp))
 635                 clp->cl_last_renewal = timestamp;
 636         spin_unlock(&clp->cl_lock);
 637 }
 638
 639 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 640 {
 641         struct nfs_client *clp = server->nfs_client;
 642
 643         if (!nfs4_has_session(clp))
 644                 do_renew_lease(clp, timestamp);
 645 }
 646
 647 struct nfs4_call_sync_data {
 648         const struct nfs_server *seq_server;
 649         struct nfs4_sequence_args *seq_args;
 650         struct nfs4_sequence_res *seq_res;
 651 };
 652
 653 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 654                         struct nfs4_sequence_res *res, int cache_reply,
 655                         int privileged)
 656 {
 657         args->sa_slot = NULL;
 658         args->sa_cache_this = cache_reply;
 659         args->sa_privileged = privileged;
 660
 661         res->sr_slot = NULL;
 662 }
 663
 664 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 665 {
 666         struct nfs4_slot *slot = res->sr_slot;
 667         struct nfs4_slot_table *tbl;
 668
 669         tbl = slot->table;
 670         spin_lock(&tbl->slot_tbl_lock);
 671         if (!nfs41_wake_and_assign_slot(tbl, slot))
 672                 nfs4_free_slot(tbl, slot);
 673         spin_unlock(&tbl->slot_tbl_lock);
 674
 675         res->sr_slot = NULL;
 676 }
 677
 678 static int nfs40_sequence_done(struct rpc_task *task,
 679                                struct nfs4_sequence_res *res)
 680 {
 681         if (res->sr_slot != NULL)
 682                 nfs40_sequence_free_slot(res);
 683         return 1;
 684 }
 685
 686 #if defined(CONFIG_NFS_V4_1)
 687
 688 static void nfs41_release_slot(struct nfs4_slot *slot)
 689 {
 690         struct nfs4_session *session;
 691         struct nfs4_slot_table *tbl;
 692         bool send_new_highest_used_slotid = false;
 693
 694         if (!slot)
 695                 return;
 696         tbl = slot->table;
 697         session = tbl->session;
 698
 699         /* Bump the slot sequence number */
 700         if (slot->seq_done)
 701                 slot->seq_nr++;
 702         slot->seq_done = 0;
 703
 704         spin_lock(&tbl->slot_tbl_lock);
 705         /* Be nice to the server: try to ensure that the last transmitted
 706          * value for highest_user_slotid <= target_highest_slotid
 707          */
 708         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 709                 send_new_highest_used_slotid = true;
 710
 711         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 712                 send_new_highest_used_slotid = false;
 713                 goto out_unlock;
 714         }
 715         nfs4_free_slot(tbl, slot);
 716
 717         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 718                 send_new_highest_used_slotid = false;
 719 out_unlock:
 720         spin_unlock(&tbl->slot_tbl_lock);
 721         if (send_new_highest_used_slotid)
 722                 nfs41_notify_server(session->clp);
 723         if (waitqueue_active(&tbl->slot_waitq))
 724                 wake_up_all(&tbl->slot_waitq);
 725 }
 726
 727 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 728 {
 729         nfs41_release_slot(res->sr_slot);
 730         res->sr_slot = NULL;
 731 }
 732
 733 static int nfs41_sequence_process(struct rpc_task *task,
 734                 struct nfs4_sequence_res *res)
 735 {
 736         struct nfs4_session *session;
 737         struct nfs4_slot *slot = res->sr_slot;
 738         struct nfs_client *clp;
 739         bool interrupted = false;
 740         int ret = 1;
 741
 742         if (slot == NULL)
 743                 goto out_noaction;
 744         /* don't increment the sequence number if the task wasn't sent */
 745         if (!RPC_WAS_SENT(task))
 746                 goto out;
 747
 748         session = slot->table->session;
 749
 750         if (slot->interrupted) {
 751                 if (res->sr_status != -NFS4ERR_DELAY)
 752                         slot->interrupted = 0;
 753                 interrupted = true;
 754         }
 755
 756         trace_nfs4_sequence_done(session, res);
 757         /* Check the SEQUENCE operation status */
 758         switch (res->sr_status) {
 759         case 0:
 760                 /* Update the slot's sequence and clientid lease timer */
 761                 slot->seq_done = 1;
 762                 clp = session->clp;
 763                 do_renew_lease(clp, res->sr_timestamp);
 764                 /* Check sequence flags */
 765                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 766                                 !!slot->privileged);
 767                 nfs41_update_target_slotid(slot->table, slot, res);
 768                 break;
 769         case 1:
 770                 /*
 771                  * sr_status remains 1 if an RPC level error occurred.
 772                  * The server may or may not have processed the sequence
 773                  * operation..
 774                  * Mark the slot as having hosted an interrupted RPC call.
 775                  */
 776                 slot->interrupted = 1;
 777                 goto out;
 778         case -NFS4ERR_DELAY:
 779                 /* The server detected a resend of the RPC call and
 780                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 781                  * of RFC5661.
 782                  */
 783                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 784                         __func__,
 785                         slot->slot_nr,
 786                         slot->seq_nr);
 787                 goto out_retry;
 788         case -NFS4ERR_RETRY_UNCACHED_REP:
 789         case -NFS4ERR_SEQ_FALSE_RETRY:
 790                 /*
 791                  * The server thinks we tried to replay a request.
 792                  * Retry the call after bumping the sequence ID.
 793                  */
 794                 goto retry_new_seq;
 795         case -NFS4ERR_BADSLOT:
 796                 /*
 797                  * The slot id we used was probably retired. Try again
 798                  * using a different slot id.
 799                  */
 800                 if (slot->slot_nr < slot->table->target_highest_slotid)
 801                         goto session_recover;
 802                 goto retry_nowait;
 803         case -NFS4ERR_SEQ_MISORDERED:
 804                 /*
 805                  * Was the last operation on this sequence interrupted?
 806                  * If so, retry after bumping the sequence number.
 807                  */
 808                 if (interrupted)
 809                         goto retry_new_seq;
 810                 /*
 811                  * Could this slot have been previously retired?
 812                  * If so, then the server may be expecting seq_nr = 1!
 813                  */
 814                 if (slot->seq_nr != 1) {
 815                         slot->seq_nr = 1;
 816                         goto retry_nowait;
 817                 }
 818                 goto session_recover;
 819         default:
 820                 /* Just update the slot sequence no. */
 821                 slot->seq_done = 1;
 822         }
 823 out:
 824         /* The session may be reset by one of the error handlers. */
 825         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 826 out_noaction:
 827         return ret;
 828 session_recover:
 829         nfs4_schedule_session_recovery(session, res->sr_status);
 830         goto retry_nowait;
 831 retry_new_seq:
 832         ++slot->seq_nr;
 833 retry_nowait:
 834         if (rpc_restart_call_prepare(task)) {
 835                 nfs41_sequence_free_slot(res);
 836                 task->tk_status = 0;
 837                 ret = 0;
 838         }
 839         goto out;
 840 out_retry:
 841         if (!rpc_restart_call(task))
 842                 goto out;
 843         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 844         return 0;
 845 }
 846
 847 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 848 {
 849         if (!nfs41_sequence_process(task, res))
 850                 return 0;
 851         if (res->sr_slot != NULL)
 852                 nfs41_sequence_free_slot(res);
 853         return 1;
 854
 855 }
 856 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 857
 858 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 859 {
 860         if (res->sr_slot == NULL)
 861                 return 1;
 862         if (res->sr_slot->table->session != NULL)
 863                 return nfs41_sequence_process(task, res);
 864         return nfs40_sequence_done(task, res);
 865 }
 866
 867 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 868 {
 869         if (res->sr_slot != NULL) {
 870                 if (res->sr_slot->table->session != NULL)
 871                         nfs41_sequence_free_slot(res);
 872                 else
 873                         nfs40_sequence_free_slot(res);
 874         }
 875 }
 876
 877 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 878 {
 879         if (res->sr_slot == NULL)
 880                 return 1;
 881         if (!res->sr_slot->table->session)
 882                 return nfs40_sequence_done(task, res);
 883         return nfs41_sequence_done(task, res);
 884 }
 885 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 886
 887 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 888 {
 889         struct nfs4_call_sync_data *data = calldata;
 890
 891         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 892
 893         nfs4_setup_sequence(data->seq_server->nfs_client,
 894                             data->seq_args, data->seq_res, task);
 895 }
 896
 897 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 898 {
 899         struct nfs4_call_sync_data *data = calldata;
 900
 901         nfs41_sequence_done(task, data->seq_res);
 902 }
 903
 904 static const struct rpc_call_ops nfs41_call_sync_ops = {
 905         .rpc_call_prepare = nfs41_call_sync_prepare,
 906         .rpc_call_done = nfs41_call_sync_done,
 907 };
 908
 909 static void
 910 nfs4_sequence_process_interrupted(struct nfs_client *client,
 911                 struct nfs4_slot *slot, struct rpc_cred *cred)
 912 {
 913         struct rpc_task *task;
 914
 915         task = _nfs41_proc_sequence(client, cred, slot, true);
 916         if (!IS_ERR(task))
 917                 rpc_put_task_async(task);
 918 }
 919
 920 #else   /* !CONFIG_NFS_V4_1 */
 921
 922 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 923 {
 924         return nfs40_sequence_done(task, res);
 925 }
 926
 927 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 928 {
 929         if (res->sr_slot != NULL)
 930                 nfs40_sequence_free_slot(res);
 931 }
 932
 933 int nfs4_sequence_done(struct rpc_task *task,
 934                        struct nfs4_sequence_res *res)
 935 {
 936         return nfs40_sequence_done(task, res);
 937 }
 938 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 939
 940 static void
 941 nfs4_sequence_process_interrupted(struct nfs_client *client,
 942                 struct nfs4_slot *slot, struct rpc_cred *cred)
 943 {
 944         WARN_ON_ONCE(1);
 945         slot->interrupted = 0;
 946 }
 947
 948 #endif  /* !CONFIG_NFS_V4_1 */
 949
 950 static
 951 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
 952                 struct nfs4_sequence_res *res,
 953                 struct nfs4_slot *slot)
 954 {
 955         if (!slot)
 956                 return;
 957         slot->privileged = args->sa_privileged ? 1 : 0;
 958         args->sa_slot = slot;
 959
 960         res->sr_slot = slot;
 961         res->sr_timestamp = jiffies;
 962         res->sr_status_flags = 0;
 963         res->sr_status = 1;
 964
 965 }
 966
 967 int nfs4_setup_sequence(struct nfs_client *client,
 968                         struct nfs4_sequence_args *args,
 969                         struct nfs4_sequence_res *res,
 970                         struct rpc_task *task)
 971 {
 972         struct nfs4_session *session = nfs4_get_session(client);
 973         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
 974         struct nfs4_slot *slot;
 975
 976         /* slot already allocated? */
 977         if (res->sr_slot != NULL)
 978                 goto out_start;
 979
 980         if (session) {
 981                 tbl = &session->fc_slot_table;
 982                 task->tk_timeout = 0;
 983         }
 984
 985         for (;;) {
 986                 spin_lock(&tbl->slot_tbl_lock);
 987                 /* The state manager will wait until the slot table is empty */
 988                 if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 989                         goto out_sleep;
 990
 991                 slot = nfs4_alloc_slot(tbl);
 992                 if (IS_ERR(slot)) {
 993                         /* Try again in 1/4 second */
 994                         if (slot == ERR_PTR(-ENOMEM))
 995                                 task->tk_timeout = HZ >> 2;
 996                         goto out_sleep;
 997                 }
 998                 spin_unlock(&tbl->slot_tbl_lock);
 999
1000                 if (likely(!slot->interrupted))
1001                         break;
1002                 nfs4_sequence_process_interrupted(client,
1003                                 slot, task->tk_msg.rpc_cred);
1004         }
1005
1006         nfs4_sequence_attach_slot(args, res, slot);
1007
1008         trace_nfs4_setup_sequence(session, args);
1009 out_start:
1010         rpc_call_start(task);
1011         return 0;
1012
1013 out_sleep:
1014         if (args->sa_privileged)
1015                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1016                                 NULL, RPC_PRIORITY_PRIVILEGED);
1017         else
1018                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1019         spin_unlock(&tbl->slot_tbl_lock);
1020         return -EAGAIN;
1021 }
1022 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1023
1024 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1025 {
1026         struct nfs4_call_sync_data *data = calldata;
1027         nfs4_setup_sequence(data->seq_server->nfs_client,
1028                                 data->seq_args, data->seq_res, task);
1029 }
1030
1031 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1032 {
1033         struct nfs4_call_sync_data *data = calldata;
1034         nfs4_sequence_done(task, data->seq_res);
1035 }
1036
1037 static const struct rpc_call_ops nfs40_call_sync_ops = {
1038         .rpc_call_prepare = nfs40_call_sync_prepare,
1039         .rpc_call_done = nfs40_call_sync_done,
1040 };
1041
1042 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1043                                    struct nfs_server *server,
1044                                    struct rpc_message *msg,
1045                                    struct nfs4_sequence_args *args,
1046                                    struct nfs4_sequence_res *res)
1047 {
1048         int ret;
1049         struct rpc_task *task;
1050         struct nfs_client *clp = server->nfs_client;
1051         struct nfs4_call_sync_data data = {
1052                 .seq_server = server,
1053                 .seq_args = args,
1054                 .seq_res = res,
1055         };
1056         struct rpc_task_setup task_setup = {
1057                 .rpc_client = clnt,
1058                 .rpc_message = msg,
1059                 .callback_ops = clp->cl_mvops->call_sync_ops,
1060                 .callback_data = &data
1061         };
1062
1063         task = rpc_run_task(&task_setup);
1064         if (IS_ERR(task))
1065                 ret = PTR_ERR(task);
1066         else {
1067                 ret = task->tk_status;
1068                 rpc_put_task(task);
1069         }
1070         return ret;
1071 }
1072
1073 int nfs4_call_sync(struct rpc_clnt *clnt,
1074                    struct nfs_server *server,
1075                    struct rpc_message *msg,
1076                    struct nfs4_sequence_args *args,
1077                    struct nfs4_sequence_res *res,
1078                    int cache_reply)
1079 {
1080         nfs4_init_sequence(args, res, cache_reply, 0);
1081         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1082 }
1083
1084 static void
1085 nfs4_inc_nlink_locked(struct inode *inode)
1086 {
1087         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1088         inc_nlink(inode);
1089 }
1090
1091 static void
1092 nfs4_dec_nlink_locked(struct inode *inode)
1093 {
1094         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1095         drop_nlink(inode);
1096 }
1097
1098 static void
1099 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1100                 unsigned long timestamp, unsigned long cache_validity)
1101 {
1102         struct nfs_inode *nfsi = NFS_I(dir);
1103
1104         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1105                 | NFS_INO_INVALID_MTIME
1106                 | NFS_INO_INVALID_DATA
1107                 | cache_validity;
1108         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1109                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1110                 nfsi->attrtimeo_timestamp = jiffies;
1111         } else {
1112                 nfs_force_lookup_revalidate(dir);
1113                 if (cinfo->before != inode_peek_iversion_raw(dir))
1114                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1115                                 NFS_INO_INVALID_ACL;
1116         }
1117         inode_set_iversion_raw(dir, cinfo->after);
1118         nfsi->read_cache_jiffies = timestamp;
1119         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1120         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1121         nfs_fscache_invalidate(dir);
1122 }
1123
1124 static void
1125 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1126                 unsigned long timestamp, unsigned long cache_validity)
1127 {
1128         spin_lock(&dir->i_lock);
1129         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1130         spin_unlock(&dir->i_lock);
1131 }
1132
1133 struct nfs4_open_createattrs {
1134         struct nfs4_label *label;
1135         struct iattr *sattr;
1136         const __u32 verf[2];
1137 };
1138
1139 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1140                 int err, struct nfs4_exception *exception)
1141 {
1142         if (err != -EINVAL)
1143                 return false;
1144         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1145                 return false;
1146         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1147         exception->retry = 1;
1148         return true;
1149 }
1150
1151 static u32
1152 nfs4_map_atomic_open_share(struct nfs_server *server,
1153                 fmode_t fmode, int openflags)
1154 {
1155         u32 res = 0;
1156
1157         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1158         case FMODE_READ:
1159                 res = NFS4_SHARE_ACCESS_READ;
1160                 break;
1161         case FMODE_WRITE:
1162                 res = NFS4_SHARE_ACCESS_WRITE;
1163                 break;
1164         case FMODE_READ|FMODE_WRITE:
1165                 res = NFS4_SHARE_ACCESS_BOTH;
1166         }
1167         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1168                 goto out;
1169         /* Want no delegation if we're using O_DIRECT */
1170         if (openflags & O_DIRECT)
1171                 res |= NFS4_SHARE_WANT_NO_DELEG;
1172 out:
1173         return res;
1174 }
1175
1176 static enum open_claim_type4
1177 nfs4_map_atomic_open_claim(struct nfs_server *server,
1178                 enum open_claim_type4 claim)
1179 {
1180         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1181                 return claim;
1182         switch (claim) {
1183         default:
1184                 return claim;
1185         case NFS4_OPEN_CLAIM_FH:
1186                 return NFS4_OPEN_CLAIM_NULL;
1187         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1188                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1189         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1190                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1191         }
1192 }
1193
1194 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1195 {
1196         p->o_res.f_attr = &p->f_attr;
1197         p->o_res.f_label = p->f_label;
1198         p->o_res.seqid = p->o_arg.seqid;
1199         p->c_res.seqid = p->c_arg.seqid;
1200         p->o_res.server = p->o_arg.server;
1201         p->o_res.access_request = p->o_arg.access;
1202         nfs_fattr_init(&p->f_attr);
1203         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1204 }
1205
1206 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1207                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1208                 const struct nfs4_open_createattrs *c,
1209                 enum open_claim_type4 claim,
1210                 gfp_t gfp_mask)
1211 {
1212         struct dentry *parent = dget_parent(dentry);
1213         struct inode *dir = d_inode(parent);
1214         struct nfs_server *server = NFS_SERVER(dir);
1215         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1216         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1217         struct nfs4_opendata *p;
1218
1219         p = kzalloc(sizeof(*p), gfp_mask);
1220         if (p == NULL)
1221                 goto err;
1222
1223         p->f_label = nfs4_label_alloc(server, gfp_mask);
1224         if (IS_ERR(p->f_label))
1225                 goto err_free_p;
1226
1227         p->a_label = nfs4_label_alloc(server, gfp_mask);
1228         if (IS_ERR(p->a_label))
1229                 goto err_free_f;
1230
1231         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1232         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1233         if (IS_ERR(p->o_arg.seqid))
1234                 goto err_free_label;
1235         nfs_sb_active(dentry->d_sb);
1236         p->dentry = dget(dentry);
1237         p->dir = parent;
1238         p->owner = sp;
1239         atomic_inc(&sp->so_count);
1240         p->o_arg.open_flags = flags;
1241         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1242         p->o_arg.umask = current_umask();
1243         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1244         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1245                         fmode, flags);
1246         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1247          * will return permission denied for all bits until close */
1248         if (!(flags & O_EXCL)) {
1249                 /* ask server to check for all possible rights as results
1250                  * are cached */
1251                 switch (p->o_arg.claim) {
1252                 default:
1253                         break;
1254                 case NFS4_OPEN_CLAIM_NULL:
1255                 case NFS4_OPEN_CLAIM_FH:
1256                         p->o_arg.access = NFS4_ACCESS_READ |
1257                                 NFS4_ACCESS_MODIFY |
1258                                 NFS4_ACCESS_EXTEND |
1259                                 NFS4_ACCESS_EXECUTE;
1260                 }
1261         }
1262         p->o_arg.clientid = server->nfs_client->cl_clientid;
1263         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1264         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1265         p->o_arg.name = &dentry->d_name;
1266         p->o_arg.server = server;
1267         p->o_arg.bitmask = nfs4_bitmask(server, label);
1268         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1269         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1270         switch (p->o_arg.claim) {
1271         case NFS4_OPEN_CLAIM_NULL:
1272         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1273         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1274                 p->o_arg.fh = NFS_FH(dir);
1275                 break;
1276         case NFS4_OPEN_CLAIM_PREVIOUS:
1277         case NFS4_OPEN_CLAIM_FH:
1278         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1279         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1280                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1281         }
1282         if (c != NULL && c->sattr != NULL && c->sattr->ia_valid != 0) {
1283                 p->o_arg.u.attrs = &p->attrs;
1284                 memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1285
1286                 memcpy(p->o_arg.u.verifier.data, c->verf,
1287                                 sizeof(p->o_arg.u.verifier.data));
1288         }
1289         p->c_arg.fh = &p->o_res.fh;
1290         p->c_arg.stateid = &p->o_res.stateid;
1291         p->c_arg.seqid = p->o_arg.seqid;
1292         nfs4_init_opendata_res(p);
1293         kref_init(&p->kref);
1294         return p;
1295
1296 err_free_label:
1297         nfs4_label_free(p->a_label);
1298 err_free_f:
1299         nfs4_label_free(p->f_label);
1300 err_free_p:
1301         kfree(p);
1302 err:
1303         dput(parent);
1304         return NULL;
1305 }
1306
1307 static void nfs4_opendata_free(struct kref *kref)
1308 {
1309         struct nfs4_opendata *p = container_of(kref,
1310                         struct nfs4_opendata, kref);
1311         struct super_block *sb = p->dentry->d_sb;
1312
1313         nfs4_lgopen_release(p->lgp);
1314         nfs_free_seqid(p->o_arg.seqid);
1315         nfs4_sequence_free_slot(&p->o_res.seq_res);
1316         if (p->state != NULL)
1317                 nfs4_put_open_state(p->state);
1318         nfs4_put_state_owner(p->owner);
1319
1320         nfs4_label_free(p->a_label);
1321         nfs4_label_free(p->f_label);
1322
1323         dput(p->dir);
1324         dput(p->dentry);
1325         nfs_sb_deactive(sb);
1326         nfs_fattr_free_names(&p->f_attr);
1327         kfree(p->f_attr.mdsthreshold);
1328         kfree(p);
1329 }
1330
1331 static void nfs4_opendata_put(struct nfs4_opendata *p)
1332 {
1333         if (p != NULL)
1334                 kref_put(&p->kref, nfs4_opendata_free);
1335 }
1336
1337 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1338                 fmode_t fmode)
1339 {
1340         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1341         case FMODE_READ|FMODE_WRITE:
1342                 return state->n_rdwr != 0;
1343         case FMODE_WRITE:
1344                 return state->n_wronly != 0;
1345         case FMODE_READ:
1346                 return state->n_rdonly != 0;
1347         }
1348         WARN_ON_ONCE(1);
1349         return false;
1350 }
1351
1352 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1353 {
1354         int ret = 0;
1355
1356         if (open_mode & (O_EXCL|O_TRUNC))
1357                 goto out;
1358         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1359                 case FMODE_READ:
1360                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1361                                 && state->n_rdonly != 0;
1362                         break;
1363                 case FMODE_WRITE:
1364                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1365                                 && state->n_wronly != 0;
1366                         break;
1367                 case FMODE_READ|FMODE_WRITE:
1368                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1369                                 && state->n_rdwr != 0;
1370         }
1371 out:
1372         return ret;
1373 }
1374
1375 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1376                 enum open_claim_type4 claim)
1377 {
1378         if (delegation == NULL)
1379                 return 0;
1380         if ((delegation->type & fmode) != fmode)
1381                 return 0;
1382         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1383                 return 0;
1384         switch (claim) {
1385         case NFS4_OPEN_CLAIM_NULL:
1386         case NFS4_OPEN_CLAIM_FH:
1387                 break;
1388         case NFS4_OPEN_CLAIM_PREVIOUS:
1389                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1390                         break;
1391                 /* Fall through */
1392         default:
1393                 return 0;
1394         }
1395         nfs_mark_delegation_referenced(delegation);
1396         return 1;
1397 }
1398
1399 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1400 {
1401         switch (fmode) {
1402                 case FMODE_WRITE:
1403                         state->n_wronly++;
1404                         break;
1405                 case FMODE_READ:
1406                         state->n_rdonly++;
1407                         break;
1408                 case FMODE_READ|FMODE_WRITE:
1409                         state->n_rdwr++;
1410         }
1411         nfs4_state_set_mode_locked(state, state->state | fmode);
1412 }
1413
1414 #ifdef CONFIG_NFS_V4_1
1415 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1416 {
1417         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1418                 return true;
1419         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1420                 return true;
1421         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1422                 return true;
1423         return false;
1424 }
1425 #endif /* CONFIG_NFS_V4_1 */
1426
1427 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1428 {
1429         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1430                 wake_up_all(&state->waitq);
1431 }
1432
1433 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1434                 const nfs4_stateid *stateid)
1435 {
1436         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1437         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1438
1439         if (stateid_seqid == state_seqid + 1U ||
1440             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1441                 nfs_state_log_update_open_stateid(state);
1442         else
1443                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1444 }
1445
1446 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1447 {
1448         struct nfs_client *clp = state->owner->so_server->nfs_client;
1449         bool need_recover = false;
1450
1451         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1452                 need_recover = true;
1453         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1454                 need_recover = true;
1455         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1456                 need_recover = true;
1457         if (need_recover)
1458                 nfs4_state_mark_reclaim_nograce(clp, state);
1459 }
1460
1461 /*
1462  * Check for whether or not the caller may update the open stateid
1463  * to the value passed in by stateid.
1464  *
1465  * Note: This function relies heavily on the server implementing
1466  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1467  * correctly.
1468  * i.e. The stateid seqids have to be initialised to 1, and
1469  * are then incremented on every state transition.
1470  */
1471 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1472                 const nfs4_stateid *stateid)
1473 {
1474         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1475             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1476                 if (stateid->seqid == cpu_to_be32(1))
1477                         nfs_state_log_update_open_stateid(state);
1478                 else
1479                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1480                 return true;
1481         }
1482
1483         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1484                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1485                 return true;
1486         }
1487         return false;
1488 }
1489
1490 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1491 {
1492         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1493                 return;
1494         if (state->n_wronly)
1495                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1496         if (state->n_rdonly)
1497                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1498         if (state->n_rdwr)
1499                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1500         set_bit(NFS_OPEN_STATE, &state->flags);
1501 }
1502
1503 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1504                 nfs4_stateid *stateid, fmode_t fmode)
1505 {
1506         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1507         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1508         case FMODE_WRITE:
1509                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1510                 break;
1511         case FMODE_READ:
1512                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1513                 break;
1514         case 0:
1515                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1516                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1517                 clear_bit(NFS_OPEN_STATE, &state->flags);
1518         }
1519         if (stateid == NULL)
1520                 return;
1521         /* Handle OPEN+OPEN_DOWNGRADE races */
1522         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1523             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1524                 nfs_resync_open_stateid_locked(state);
1525                 goto out;
1526         }
1527         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1528                 nfs4_stateid_copy(&state->stateid, stateid);
1529         nfs4_stateid_copy(&state->open_stateid, stateid);
1530         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1531 out:
1532         nfs_state_log_update_open_stateid(state);
1533 }
1534
1535 static void nfs_clear_open_stateid(struct nfs4_state *state,
1536         nfs4_stateid *arg_stateid,
1537         nfs4_stateid *stateid, fmode_t fmode)
1538 {
1539         write_seqlock(&state->seqlock);
1540         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1541         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1542                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1543         write_sequnlock(&state->seqlock);
1544         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1545                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1546 }
1547
1548 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1549                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1550 {
1551         DEFINE_WAIT(wait);
1552         int status = 0;
1553         for (;;) {
1554
1555                 if (!nfs_need_update_open_stateid(state, stateid))
1556                         return;
1557                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1558                         break;
1559                 if (status)
1560                         break;
1561                 /* Rely on seqids for serialisation with NFSv4.0 */
1562                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1563                         break;
1564
1565                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1566                 /*
1567                  * Ensure we process the state changes in the same order
1568                  * in which the server processed them by delaying the
1569                  * update of the stateid until we are in sequence.
1570                  */
1571                 write_sequnlock(&state->seqlock);
1572                 spin_unlock(&state->owner->so_lock);
1573                 rcu_read_unlock();
1574                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1575                 if (!signal_pending(current)) {
1576                         if (schedule_timeout(5*HZ) == 0)
1577                                 status = -EAGAIN;
1578                         else
1579                                 status = 0;
1580                 } else
1581                         status = -EINTR;
1582                 finish_wait(&state->waitq, &wait);
1583                 rcu_read_lock();
1584                 spin_lock(&state->owner->so_lock);
1585                 write_seqlock(&state->seqlock);
1586         }
1587
1588         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1589             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1590                 nfs4_stateid_copy(freeme, &state->open_stateid);
1591                 nfs_test_and_clear_all_open_stateid(state);
1592         }
1593
1594         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1595                 nfs4_stateid_copy(&state->stateid, stateid);
1596         nfs4_stateid_copy(&state->open_stateid, stateid);
1597         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1598         nfs_state_log_update_open_stateid(state);
1599 }
1600
1601 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1602                 const nfs4_stateid *open_stateid,
1603                 fmode_t fmode,
1604                 nfs4_stateid *freeme)
1605 {
1606         /*
1607          * Protect the call to nfs4_state_set_mode_locked and
1608          * serialise the stateid update
1609          */
1610         write_seqlock(&state->seqlock);
1611         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1612         switch (fmode) {
1613         case FMODE_READ:
1614                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1615                 break;
1616         case FMODE_WRITE:
1617                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1618                 break;
1619         case FMODE_READ|FMODE_WRITE:
1620                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1621         }
1622         set_bit(NFS_OPEN_STATE, &state->flags);
1623         write_sequnlock(&state->seqlock);
1624 }
1625
1626 static void nfs_state_set_delegation(struct nfs4_state *state,
1627                 const nfs4_stateid *deleg_stateid,
1628                 fmode_t fmode)
1629 {
1630         /*
1631          * Protect the call to nfs4_state_set_mode_locked and
1632          * serialise the stateid update
1633          */
1634         write_seqlock(&state->seqlock);
1635         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1636         set_bit(NFS_DELEGATED_STATE, &state->flags);
1637         write_sequnlock(&state->seqlock);
1638 }
1639
1640 static int update_open_stateid(struct nfs4_state *state,
1641                 const nfs4_stateid *open_stateid,
1642                 const nfs4_stateid *delegation,
1643                 fmode_t fmode)
1644 {
1645         struct nfs_server *server = NFS_SERVER(state->inode);
1646         struct nfs_client *clp = server->nfs_client;
1647         struct nfs_inode *nfsi = NFS_I(state->inode);
1648         struct nfs_delegation *deleg_cur;
1649         nfs4_stateid freeme = { };
1650         int ret = 0;
1651
1652         fmode &= (FMODE_READ|FMODE_WRITE);
1653
1654         rcu_read_lock();
1655         spin_lock(&state->owner->so_lock);
1656         if (open_stateid != NULL) {
1657                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1658                 ret = 1;
1659         }
1660
1661         deleg_cur = rcu_dereference(nfsi->delegation);
1662         if (deleg_cur == NULL)
1663                 goto no_delegation;
1664
1665         spin_lock(&deleg_cur->lock);
1666         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1667            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1668             (deleg_cur->type & fmode) != fmode)
1669                 goto no_delegation_unlock;
1670
1671         if (delegation == NULL)
1672                 delegation = &deleg_cur->stateid;
1673         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1674                 goto no_delegation_unlock;
1675
1676         nfs_mark_delegation_referenced(deleg_cur);
1677         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1678         ret = 1;
1679 no_delegation_unlock:
1680         spin_unlock(&deleg_cur->lock);
1681 no_delegation:
1682         if (ret)
1683                 update_open_stateflags(state, fmode);
1684         spin_unlock(&state->owner->so_lock);
1685         rcu_read_unlock();
1686
1687         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1688                 nfs4_schedule_state_manager(clp);
1689         if (freeme.type != 0)
1690                 nfs4_test_and_free_stateid(server, &freeme,
1691                                 state->owner->so_cred);
1692
1693         return ret;
1694 }
1695
1696 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1697                 const nfs4_stateid *stateid)
1698 {
1699         struct nfs4_state *state = lsp->ls_state;
1700         bool ret = false;
1701
1702         spin_lock(&state->state_lock);
1703         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1704                 goto out_noupdate;
1705         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1706                 goto out_noupdate;
1707         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1708         ret = true;
1709 out_noupdate:
1710         spin_unlock(&state->state_lock);
1711         return ret;
1712 }
1713
1714 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1715 {
1716         struct nfs_delegation *delegation;
1717
1718         fmode &= FMODE_READ|FMODE_WRITE;
1719         rcu_read_lock();
1720         delegation = rcu_dereference(NFS_I(inode)->delegation);
1721         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1722                 rcu_read_unlock();
1723                 return;
1724         }
1725         rcu_read_unlock();
1726         nfs4_inode_return_delegation(inode);
1727 }
1728
1729 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1730 {
1731         struct nfs4_state *state = opendata->state;
1732         struct nfs_inode *nfsi = NFS_I(state->inode);
1733         struct nfs_delegation *delegation;
1734         int open_mode = opendata->o_arg.open_flags;
1735         fmode_t fmode = opendata->o_arg.fmode;
1736         enum open_claim_type4 claim = opendata->o_arg.claim;
1737         nfs4_stateid stateid;
1738         int ret = -EAGAIN;
1739
1740         for (;;) {
1741                 spin_lock(&state->owner->so_lock);
1742                 if (can_open_cached(state, fmode, open_mode)) {
1743                         update_open_stateflags(state, fmode);
1744                         spin_unlock(&state->owner->so_lock);
1745                         goto out_return_state;
1746                 }
1747                 spin_unlock(&state->owner->so_lock);
1748                 rcu_read_lock();
1749                 delegation = rcu_dereference(nfsi->delegation);
1750                 if (!can_open_delegated(delegation, fmode, claim)) {
1751                         rcu_read_unlock();
1752                         break;
1753                 }
1754                 /* Save the delegation */
1755                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1756                 rcu_read_unlock();
1757                 nfs_release_seqid(opendata->o_arg.seqid);
1758                 if (!opendata->is_recover) {
1759                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1760                         if (ret != 0)
1761                                 goto out;
1762                 }
1763                 ret = -EAGAIN;
1764
1765                 /* Try to update the stateid using the delegation */
1766                 if (update_open_stateid(state, NULL, &stateid, fmode))
1767                         goto out_return_state;
1768         }
1769 out:
1770         return ERR_PTR(ret);
1771 out_return_state:
1772         atomic_inc(&state->count);
1773         return state;
1774 }
1775
1776 static void
1777 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1778 {
1779         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1780         struct nfs_delegation *delegation;
1781         int delegation_flags = 0;
1782
1783         rcu_read_lock();
1784         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1785         if (delegation)
1786                 delegation_flags = delegation->flags;
1787         rcu_read_unlock();
1788         switch (data->o_arg.claim) {
1789         default:
1790                 break;
1791         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1792         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1793                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1794                                    "returning a delegation for "
1795                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1796                                    clp->cl_hostname);
1797                 return;
1798         }
1799         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1800                 nfs_inode_set_delegation(state->inode,
1801                                 data->owner->so_cred,
1802                                 data->o_res.delegation_type,
1803                                 &data->o_res.delegation,
1804                                 data->o_res.pagemod_limit);
1805         else
1806                 nfs_inode_reclaim_delegation(state->inode,
1807                                 data->owner->so_cred,
1808                                 data->o_res.delegation_type,
1809                                 &data->o_res.delegation,
1810                                 data->o_res.pagemod_limit);
1811
1812         if (data->o_res.do_recall)
1813                 nfs_async_inode_return_delegation(state->inode,
1814                                                   &data->o_res.delegation);
1815 }
1816
1817 /*
1818  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1819  * and update the nfs4_state.
1820  */
1821 static struct nfs4_state *
1822 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1823 {
1824         struct inode *inode = data->state->inode;
1825         struct nfs4_state *state = data->state;
1826         int ret;
1827
1828         if (!data->rpc_done) {
1829                 if (data->rpc_status)
1830                         return ERR_PTR(data->rpc_status);
1831                 /* cached opens have already been processed */
1832                 goto update;
1833         }
1834
1835         ret = nfs_refresh_inode(inode, &data->f_attr);
1836         if (ret)
1837                 return ERR_PTR(ret);
1838
1839         if (data->o_res.delegation_type != 0)
1840                 nfs4_opendata_check_deleg(data, state);
1841 update:
1842         update_open_stateid(state, &data->o_res.stateid, NULL,
1843                             data->o_arg.fmode);
1844         atomic_inc(&state->count);
1845
1846         return state;
1847 }
1848
1849 static struct inode *
1850 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1851 {
1852         struct inode *inode;
1853
1854         switch (data->o_arg.claim) {
1855         case NFS4_OPEN_CLAIM_NULL:
1856         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1857         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1858                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1859                         return ERR_PTR(-EAGAIN);
1860                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1861                                 &data->f_attr, data->f_label);
1862                 break;
1863         default:
1864                 inode = d_inode(data->dentry);
1865                 ihold(inode);
1866                 nfs_refresh_inode(inode, &data->f_attr);
1867         }
1868         return inode;
1869 }
1870
1871 static struct nfs4_state *
1872 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1873 {
1874         struct nfs4_state *state;
1875         struct inode *inode;
1876
1877         inode = nfs4_opendata_get_inode(data);
1878         if (IS_ERR(inode))
1879                 return ERR_CAST(inode);
1880         if (data->state != NULL && data->state->inode == inode) {
1881                 state = data->state;
1882                 atomic_inc(&state->count);
1883         } else
1884                 state = nfs4_get_open_state(inode, data->owner);
1885         iput(inode);
1886         if (state == NULL)
1887                 state = ERR_PTR(-ENOMEM);
1888         return state;
1889 }
1890
1891 static struct nfs4_state *
1892 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1893 {
1894         struct nfs4_state *state;
1895
1896         if (!data->rpc_done) {
1897                 state = nfs4_try_open_cached(data);
1898                 trace_nfs4_cached_open(data->state);
1899                 goto out;
1900         }
1901
1902         state = nfs4_opendata_find_nfs4_state(data);
1903         if (IS_ERR(state))
1904                 goto out;
1905
1906         if (data->o_res.delegation_type != 0)
1907                 nfs4_opendata_check_deleg(data, state);
1908         update_open_stateid(state, &data->o_res.stateid, NULL,
1909                         data->o_arg.fmode);
1910 out:
1911         nfs_release_seqid(data->o_arg.seqid);
1912         return state;
1913 }
1914
1915 static struct nfs4_state *
1916 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1917 {
1918         struct nfs4_state *ret;
1919
1920         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1921                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1922         else
1923                 ret = _nfs4_opendata_to_nfs4_state(data);
1924         nfs4_sequence_free_slot(&data->o_res.seq_res);
1925         return ret;
1926 }
1927
1928 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1929 {
1930         struct nfs_inode *nfsi = NFS_I(state->inode);
1931         struct nfs_open_context *ctx;
1932
1933         spin_lock(&state->inode->i_lock);
1934         list_for_each_entry(ctx, &nfsi->open_files, list) {
1935                 if (ctx->state != state)
1936                         continue;
1937                 get_nfs_open_context(ctx);
1938                 spin_unlock(&state->inode->i_lock);
1939                 return ctx;
1940         }
1941         spin_unlock(&state->inode->i_lock);
1942         return ERR_PTR(-ENOENT);
1943 }
1944
1945 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1946                 struct nfs4_state *state, enum open_claim_type4 claim)
1947 {
1948         struct nfs4_opendata *opendata;
1949
1950         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1951                         NULL, claim, GFP_NOFS);
1952         if (opendata == NULL)
1953                 return ERR_PTR(-ENOMEM);
1954         opendata->state = state;
1955         atomic_inc(&state->count);
1956         return opendata;
1957 }
1958
1959 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1960                 fmode_t fmode)
1961 {
1962         struct nfs4_state *newstate;
1963         int ret;
1964
1965         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1966                 return 0;
1967         opendata->o_arg.open_flags = 0;
1968         opendata->o_arg.fmode = fmode;
1969         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1970                         NFS_SB(opendata->dentry->d_sb),
1971                         fmode, 0);
1972         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1973         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1974         nfs4_init_opendata_res(opendata);
1975         ret = _nfs4_recover_proc_open(opendata);
1976         if (ret != 0)
1977                 return ret;
1978         newstate = nfs4_opendata_to_nfs4_state(opendata);
1979         if (IS_ERR(newstate))
1980                 return PTR_ERR(newstate);
1981         if (newstate != opendata->state)
1982                 ret = -ESTALE;
1983         nfs4_close_state(newstate, fmode);
1984         return ret;
1985 }
1986
1987 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1988 {
1989         int ret;
1990
1991         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1992         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1993         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1994         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1995         /* memory barrier prior to reading state->n_* */
1996         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1997         clear_bit(NFS_OPEN_STATE, &state->flags);
1998         smp_rmb();
1999         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2000         if (ret != 0)
2001                 return ret;
2002         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2003         if (ret != 0)
2004                 return ret;
2005         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2006         if (ret != 0)
2007                 return ret;
2008         /*
2009          * We may have performed cached opens for all three recoveries.
2010          * Check if we need to update the current stateid.
2011          */
2012         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2013             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2014                 write_seqlock(&state->seqlock);
2015                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2016                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2017                 write_sequnlock(&state->seqlock);
2018         }
2019         return 0;
2020 }
2021
2022 /*
2023  * OPEN_RECLAIM:
2024  *      reclaim state on the server after a reboot.
2025  */
2026 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2027 {
2028         struct nfs_delegation *delegation;
2029         struct nfs4_opendata *opendata;
2030         fmode_t delegation_type = 0;
2031         int status;
2032
2033         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2034                         NFS4_OPEN_CLAIM_PREVIOUS);
2035         if (IS_ERR(opendata))
2036                 return PTR_ERR(opendata);
2037         rcu_read_lock();
2038         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2039         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2040                 delegation_type = delegation->type;
2041         rcu_read_unlock();
2042         opendata->o_arg.u.delegation_type = delegation_type;
2043         status = nfs4_open_recover(opendata, state);
2044         nfs4_opendata_put(opendata);
2045         return status;
2046 }
2047
2048 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2049 {
2050         struct nfs_server *server = NFS_SERVER(state->inode);
2051         struct nfs4_exception exception = { };
2052         int err;
2053         do {
2054                 err = _nfs4_do_open_reclaim(ctx, state);
2055                 trace_nfs4_open_reclaim(ctx, 0, err);
2056                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2057                         continue;
2058                 if (err != -NFS4ERR_DELAY)
2059                         break;
2060                 nfs4_handle_exception(server, err, &exception);
2061         } while (exception.retry);
2062         return err;
2063 }
2064
2065 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2066 {
2067         struct nfs_open_context *ctx;
2068         int ret;
2069
2070         ctx = nfs4_state_find_open_context(state);
2071         if (IS_ERR(ctx))
2072                 return -EAGAIN;
2073         ret = nfs4_do_open_reclaim(ctx, state);
2074         put_nfs_open_context(ctx);
2075         return ret;
2076 }
2077
2078 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)
2079 {
2080         switch (err) {
2081                 default:
2082                         printk(KERN_ERR "NFS: %s: unhandled error "
2083                                         "%d.\n", __func__, err);
2084                 case 0:
2085                 case -ENOENT:
2086                 case -EAGAIN:
2087                 case -ESTALE:
2088                         break;
2089                 case -NFS4ERR_BADSESSION:
2090                 case -NFS4ERR_BADSLOT:
2091                 case -NFS4ERR_BAD_HIGH_SLOT:
2092                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2093                 case -NFS4ERR_DEADSESSION:
2094                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2095                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
2096                         return -EAGAIN;
2097                 case -NFS4ERR_STALE_CLIENTID:
2098                 case -NFS4ERR_STALE_STATEID:
2099                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2100                         /* Don't recall a delegation if it was lost */
2101                         nfs4_schedule_lease_recovery(server->nfs_client);
2102                         return -EAGAIN;
2103                 case -NFS4ERR_MOVED:
2104                         nfs4_schedule_migration_recovery(server);
2105                         return -EAGAIN;
2106                 case -NFS4ERR_LEASE_MOVED:
2107                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2108                         return -EAGAIN;
2109                 case -NFS4ERR_DELEG_REVOKED:
2110                 case -NFS4ERR_ADMIN_REVOKED:
2111                 case -NFS4ERR_EXPIRED:
2112                 case -NFS4ERR_BAD_STATEID:
2113                 case -NFS4ERR_OPENMODE:
2114                         nfs_inode_find_state_and_recover(state->inode,
2115                                         stateid);
2116                         nfs4_schedule_stateid_recovery(server, state);
2117                         return -EAGAIN;
2118                 case -NFS4ERR_DELAY:
2119                 case -NFS4ERR_GRACE:
2120                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2121                         ssleep(1);
2122                         return -EAGAIN;
2123                 case -ENOMEM:
2124                 case -NFS4ERR_DENIED:
2125                         if (fl) {
2126                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2127                                 if (lsp)
2128                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2129                         }
2130                         return 0;
2131         }
2132         return err;
2133 }
2134
2135 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2136                 struct nfs4_state *state, const nfs4_stateid *stateid,
2137                 fmode_t type)
2138 {
2139         struct nfs_server *server = NFS_SERVER(state->inode);
2140         struct nfs4_opendata *opendata;
2141         int err = 0;
2142
2143         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2144                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2145         if (IS_ERR(opendata))
2146                 return PTR_ERR(opendata);
2147         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2148         write_seqlock(&state->seqlock);
2149         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2150         write_sequnlock(&state->seqlock);
2151         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2152         switch (type & (FMODE_READ|FMODE_WRITE)) {
2153         case FMODE_READ|FMODE_WRITE:
2154         case FMODE_WRITE:
2155                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2156                 if (err)
2157                         break;
2158                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2159                 if (err)
2160                         break;
2161                 /* Fall through */
2162         case FMODE_READ:
2163                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2164         }
2165         nfs4_opendata_put(opendata);
2166         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2167 }
2168
2169 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2170 {
2171         struct nfs4_opendata *data = calldata;
2172
2173         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2174                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2175 }
2176
2177 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2178 {
2179         struct nfs4_opendata *data = calldata;
2180
2181         nfs40_sequence_done(task, &data->c_res.seq_res);
2182
2183         data->rpc_status = task->tk_status;
2184         if (data->rpc_status == 0) {
2185                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2186                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2187                 renew_lease(data->o_res.server, data->timestamp);
2188                 data->rpc_done = true;
2189         }
2190 }
2191
2192 static void nfs4_open_confirm_release(void *calldata)
2193 {
2194         struct nfs4_opendata *data = calldata;
2195         struct nfs4_state *state = NULL;
2196
2197         /* If this request hasn't been cancelled, do nothing */
2198         if (!data->cancelled)
2199                 goto out_free;
2200         /* In case of error, no cleanup! */
2201         if (!data->rpc_done)
2202                 goto out_free;
2203         state = nfs4_opendata_to_nfs4_state(data);
2204         if (!IS_ERR(state))
2205                 nfs4_close_state(state, data->o_arg.fmode);
2206 out_free:
2207         nfs4_opendata_put(data);
2208 }
2209
2210 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2211         .rpc_call_prepare = nfs4_open_confirm_prepare,
2212         .rpc_call_done = nfs4_open_confirm_done,
2213         .rpc_release = nfs4_open_confirm_release,
2214 };
2215
2216 /*
2217  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2218  */
2219 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2220 {
2221         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2222         struct rpc_task *task;
2223         struct  rpc_message msg = {
2224                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2225                 .rpc_argp = &data->c_arg,
2226                 .rpc_resp = &data->c_res,
2227                 .rpc_cred = data->owner->so_cred,
2228         };
2229         struct rpc_task_setup task_setup_data = {
2230                 .rpc_client = server->client,
2231                 .rpc_message = &msg,
2232                 .callback_ops = &nfs4_open_confirm_ops,
2233                 .callback_data = data,
2234                 .workqueue = nfsiod_workqueue,
2235                 .flags = RPC_TASK_ASYNC,
2236         };
2237         int status;
2238
2239         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2240                                 data->is_recover);
2241         kref_get(&data->kref);
2242         data->rpc_done = false;
2243         data->rpc_status = 0;
2244         data->timestamp = jiffies;
2245         task = rpc_run_task(&task_setup_data);
2246         if (IS_ERR(task))
2247                 return PTR_ERR(task);
2248         status = rpc_wait_for_completion_task(task);
2249         if (status != 0) {
2250                 data->cancelled = true;
2251                 smp_wmb();
2252         } else
2253                 status = data->rpc_status;
2254         rpc_put_task(task);
2255         return status;
2256 }
2257
2258 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2259 {
2260         struct nfs4_opendata *data = calldata;
2261         struct nfs4_state_owner *sp = data->owner;
2262         struct nfs_client *clp = sp->so_server->nfs_client;
2263         enum open_claim_type4 claim = data->o_arg.claim;
2264
2265         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2266                 goto out_wait;
2267         /*
2268          * Check if we still need to send an OPEN call, or if we can use
2269          * a delegation instead.
2270          */
2271         if (data->state != NULL) {
2272                 struct nfs_delegation *delegation;
2273
2274                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2275                         goto out_no_action;
2276                 rcu_read_lock();
2277                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2278                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2279                         goto unlock_no_action;
2280                 rcu_read_unlock();
2281         }
2282         /* Update client id. */
2283         data->o_arg.clientid = clp->cl_clientid;
2284         switch (claim) {
2285         default:
2286                 break;
2287         case NFS4_OPEN_CLAIM_PREVIOUS:
2288         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2289         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2290                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2291                 /* Fall through */
2292         case NFS4_OPEN_CLAIM_FH:
2293                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2294         }
2295         data->timestamp = jiffies;
2296         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2297                                 &data->o_arg.seq_args,
2298                                 &data->o_res.seq_res,
2299                                 task) != 0)
2300                 nfs_release_seqid(data->o_arg.seqid);
2301
2302         /* Set the create mode (note dependency on the session type) */
2303         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2304         if (data->o_arg.open_flags & O_EXCL) {
2305                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2306                 if (nfs4_has_persistent_session(clp))
2307                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2308                 else if (clp->cl_mvops->minor_version > 0)
2309                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2310         }
2311         return;
2312 unlock_no_action:
2313         trace_nfs4_cached_open(data->state);
2314         rcu_read_unlock();
2315 out_no_action:
2316         task->tk_action = NULL;
2317 out_wait:
2318         nfs4_sequence_done(task, &data->o_res.seq_res);
2319 }
2320
2321 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2322 {
2323         struct nfs4_opendata *data = calldata;
2324
2325         data->rpc_status = task->tk_status;
2326
2327         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2328                 return;
2329
2330         if (task->tk_status == 0) {
2331                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2332                         switch (data->o_res.f_attr->mode & S_IFMT) {
2333                         case S_IFREG:
2334                                 break;
2335                         case S_IFLNK:
2336                                 data->rpc_status = -ELOOP;
2337                                 break;
2338                         case S_IFDIR:
2339                                 data->rpc_status = -EISDIR;
2340                                 break;
2341                         default:
2342                                 data->rpc_status = -ENOTDIR;
2343                         }
2344                 }
2345                 renew_lease(data->o_res.server, data->timestamp);
2346                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2347                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2348         }
2349         data->rpc_done = true;
2350 }
2351
2352 static void nfs4_open_release(void *calldata)
2353 {
2354         struct nfs4_opendata *data = calldata;
2355         struct nfs4_state *state = NULL;
2356
2357         /* If this request hasn't been cancelled, do nothing */
2358         if (!data->cancelled)
2359                 goto out_free;
2360         /* In case of error, no cleanup! */
2361         if (data->rpc_status != 0 || !data->rpc_done)
2362                 goto out_free;
2363         /* In case we need an open_confirm, no cleanup! */
2364         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2365                 goto out_free;
2366         state = nfs4_opendata_to_nfs4_state(data);
2367         if (!IS_ERR(state))
2368                 nfs4_close_state(state, data->o_arg.fmode);
2369 out_free:
2370         nfs4_opendata_put(data);
2371 }
2372
2373 static const struct rpc_call_ops nfs4_open_ops = {
2374         .rpc_call_prepare = nfs4_open_prepare,
2375         .rpc_call_done = nfs4_open_done,
2376         .rpc_release = nfs4_open_release,
2377 };
2378
2379 static int nfs4_run_open_task(struct nfs4_opendata *data,
2380                               struct nfs_open_context *ctx)
2381 {
2382         struct inode *dir = d_inode(data->dir);
2383         struct nfs_server *server = NFS_SERVER(dir);
2384         struct nfs_openargs *o_arg = &data->o_arg;
2385         struct nfs_openres *o_res = &data->o_res;
2386         struct rpc_task *task;
2387         struct rpc_message msg = {
2388                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2389                 .rpc_argp = o_arg,
2390                 .rpc_resp = o_res,
2391                 .rpc_cred = data->owner->so_cred,
2392         };
2393         struct rpc_task_setup task_setup_data = {
2394                 .rpc_client = server->client,
2395                 .rpc_message = &msg,
2396                 .callback_ops = &nfs4_open_ops,
2397                 .callback_data = data,
2398                 .workqueue = nfsiod_workqueue,
2399                 .flags = RPC_TASK_ASYNC,
2400         };
2401         int status;
2402
2403         kref_get(&data->kref);
2404         data->rpc_done = false;
2405         data->rpc_status = 0;
2406         data->cancelled = false;
2407         data->is_recover = false;
2408         if (!ctx) {
2409                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2410                 data->is_recover = true;
2411         } else {
2412                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2413                 pnfs_lgopen_prepare(data, ctx);
2414         }
2415         task = rpc_run_task(&task_setup_data);
2416         if (IS_ERR(task))
2417                 return PTR_ERR(task);
2418         status = rpc_wait_for_completion_task(task);
2419         if (status != 0) {
2420                 data->cancelled = true;
2421                 smp_wmb();
2422         } else
2423                 status = data->rpc_status;
2424         rpc_put_task(task);
2425
2426         return status;
2427 }
2428
2429 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2430 {
2431         struct inode *dir = d_inode(data->dir);
2432         struct nfs_openres *o_res = &data->o_res;
2433         int status;
2434
2435         status = nfs4_run_open_task(data, NULL);
2436         if (status != 0 || !data->rpc_done)
2437                 return status;
2438
2439         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2440
2441         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2442                 status = _nfs4_proc_open_confirm(data);
2443
2444         return status;
2445 }
2446
2447 /*
2448  * Additional permission checks in order to distinguish between an
2449  * open for read, and an open for execute. This works around the
2450  * fact that NFSv4 OPEN treats read and execute permissions as being
2451  * the same.
2452  * Note that in the non-execute case, we want to turn off permission
2453  * checking if we just created a new file (POSIX open() semantics).
2454  */
2455 static int nfs4_opendata_access(struct rpc_cred *cred,
2456                                 struct nfs4_opendata *opendata,
2457                                 struct nfs4_state *state, fmode_t fmode,
2458                                 int openflags)
2459 {
2460         struct nfs_access_entry cache;
2461         u32 mask, flags;
2462
2463         /* access call failed or for some reason the server doesn't
2464          * support any access modes -- defer access call until later */
2465         if (opendata->o_res.access_supported == 0)
2466                 return 0;
2467
2468         mask = 0;
2469         /*
2470          * Use openflags to check for exec, because fmode won't
2471          * always have FMODE_EXEC set when file open for exec.
2472          */
2473         if (openflags & __FMODE_EXEC) {
2474                 /* ONLY check for exec rights */
2475                 if (S_ISDIR(state->inode->i_mode))
2476                         mask = NFS4_ACCESS_LOOKUP;
2477                 else
2478                         mask = NFS4_ACCESS_EXECUTE;
2479         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2480                 mask = NFS4_ACCESS_READ;
2481
2482         cache.cred = cred;
2483         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2484         nfs_access_add_cache(state->inode, &cache);
2485
2486         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2487         if ((mask & ~cache.mask & flags) == 0)
2488                 return 0;
2489
2490         return -EACCES;
2491 }
2492
2493 /*
2494  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2495  */
2496 static int _nfs4_proc_open(struct nfs4_opendata *data,
2497                            struct nfs_open_context *ctx)
2498 {
2499         struct inode *dir = d_inode(data->dir);
2500         struct nfs_server *server = NFS_SERVER(dir);
2501         struct nfs_openargs *o_arg = &data->o_arg;
2502         struct nfs_openres *o_res = &data->o_res;
2503         int status;
2504
2505         status = nfs4_run_open_task(data, ctx);
2506         if (!data->rpc_done)
2507                 return status;
2508         if (status != 0) {
2509                 if (status == -NFS4ERR_BADNAME &&
2510                                 !(o_arg->open_flags & O_CREAT))
2511                         return -ENOENT;
2512                 return status;
2513         }
2514
2515         nfs_fattr_map_and_free_names(server, &data->f_attr);
2516
2517         if (o_arg->open_flags & O_CREAT) {
2518                 if (o_arg->open_flags & O_EXCL)
2519                         data->file_created = true;
2520                 else if (o_res->cinfo.before != o_res->cinfo.after)
2521                         data->file_created = true;
2522                 if (data->file_created ||
2523                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2524                         update_changeattr(dir, &o_res->cinfo,
2525                                         o_res->f_attr->time_start, 0);
2526         }
2527         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2528                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2529         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2530                 status = _nfs4_proc_open_confirm(data);
2531                 if (status != 0)
2532                         return status;
2533         }
2534         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2535                 nfs4_sequence_free_slot(&o_res->seq_res);
2536                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2537                                 o_res->f_label, NULL);
2538         }
2539         return 0;
2540 }
2541
2542 /*
2543  * OPEN_EXPIRED:
2544  *      reclaim state on the server after a network partition.
2545  *      Assumes caller holds the appropriate lock
2546  */
2547 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2548 {
2549         struct nfs4_opendata *opendata;
2550         int ret;
2551
2552         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2553                         NFS4_OPEN_CLAIM_FH);
2554         if (IS_ERR(opendata))
2555                 return PTR_ERR(opendata);
2556         ret = nfs4_open_recover(opendata, state);
2557         if (ret == -ESTALE)
2558                 d_drop(ctx->dentry);
2559         nfs4_opendata_put(opendata);
2560         return ret;
2561 }
2562
2563 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2564 {
2565         struct nfs_server *server = NFS_SERVER(state->inode);
2566         struct nfs4_exception exception = { };
2567         int err;
2568
2569         do {
2570                 err = _nfs4_open_expired(ctx, state);
2571                 trace_nfs4_open_expired(ctx, 0, err);
2572                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2573                         continue;
2574                 switch (err) {
2575                 default:
2576                         goto out;
2577                 case -NFS4ERR_GRACE:
2578                 case -NFS4ERR_DELAY:
2579                         nfs4_handle_exception(server, err, &exception);
2580                         err = 0;
2581                 }
2582         } while (exception.retry);
2583 out:
2584         return err;
2585 }
2586
2587 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2588 {
2589         struct nfs_open_context *ctx;
2590         int ret;
2591
2592         ctx = nfs4_state_find_open_context(state);
2593         if (IS_ERR(ctx))
2594                 return -EAGAIN;
2595         ret = nfs4_do_open_expired(ctx, state);
2596         put_nfs_open_context(ctx);
2597         return ret;
2598 }
2599
2600 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2601                 const nfs4_stateid *stateid)
2602 {
2603         nfs_remove_bad_delegation(state->inode, stateid);
2604         write_seqlock(&state->seqlock);
2605         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2606         write_sequnlock(&state->seqlock);
2607         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2608 }
2609
2610 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2611 {
2612         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2613                 nfs_finish_clear_delegation_stateid(state, NULL);
2614 }
2615
2616 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2617 {
2618         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2619         nfs40_clear_delegation_stateid(state);
2620         return nfs4_open_expired(sp, state);
2621 }
2622
2623 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2624                 nfs4_stateid *stateid,
2625                 struct rpc_cred *cred)
2626 {
2627         return -NFS4ERR_BAD_STATEID;
2628 }
2629
2630 #if defined(CONFIG_NFS_V4_1)
2631 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2632                 nfs4_stateid *stateid,
2633                 struct rpc_cred *cred)
2634 {
2635         int status;
2636
2637         switch (stateid->type) {
2638         default:
2639                 break;
2640         case NFS4_INVALID_STATEID_TYPE:
2641         case NFS4_SPECIAL_STATEID_TYPE:
2642                 return -NFS4ERR_BAD_STATEID;
2643         case NFS4_REVOKED_STATEID_TYPE:
2644                 goto out_free;
2645         }
2646
2647         status = nfs41_test_stateid(server, stateid, cred);
2648         switch (status) {
2649         case -NFS4ERR_EXPIRED:
2650         case -NFS4ERR_ADMIN_REVOKED:
2651         case -NFS4ERR_DELEG_REVOKED:
2652                 break;
2653         default:
2654                 return status;
2655         }
2656 out_free:
2657         /* Ack the revoked state to the server */
2658         nfs41_free_stateid(server, stateid, cred, true);
2659         return -NFS4ERR_EXPIRED;
2660 }
2661
2662 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2663 {
2664         struct nfs_server *server = NFS_SERVER(state->inode);
2665         nfs4_stateid stateid;
2666         struct nfs_delegation *delegation;
2667         struct rpc_cred *cred;
2668         int status;
2669
2670         /* Get the delegation credential for use by test/free_stateid */
2671         rcu_read_lock();
2672         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2673         if (delegation == NULL) {
2674                 rcu_read_unlock();
2675                 return;
2676         }
2677
2678         nfs4_stateid_copy(&stateid, &delegation->stateid);
2679         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags) ||
2680                 !test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2681                         &delegation->flags)) {
2682                 rcu_read_unlock();
2683                 nfs_finish_clear_delegation_stateid(state, &stateid);
2684                 return;
2685         }
2686
2687         cred = get_rpccred(delegation->cred);
2688         rcu_read_unlock();
2689         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2690         trace_nfs4_test_delegation_stateid(state, NULL, status);
2691         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2692                 nfs_finish_clear_delegation_stateid(state, &stateid);
2693
2694         put_rpccred(cred);
2695 }
2696
2697 /**
2698  * nfs41_check_expired_locks - possibly free a lock stateid
2699  *
2700  * @state: NFSv4 state for an inode
2701  *
2702  * Returns NFS_OK if recovery for this stateid is now finished.
2703  * Otherwise a negative NFS4ERR value is returned.
2704  */
2705 static int nfs41_check_expired_locks(struct nfs4_state *state)
2706 {
2707         int status, ret = NFS_OK;
2708         struct nfs4_lock_state *lsp, *prev = NULL;
2709         struct nfs_server *server = NFS_SERVER(state->inode);
2710
2711         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2712                 goto out;
2713
2714         spin_lock(&state->state_lock);
2715         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2716                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2717                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2718
2719                         refcount_inc(&lsp->ls_count);
2720                         spin_unlock(&state->state_lock);
2721
2722                         nfs4_put_lock_state(prev);
2723                         prev = lsp;
2724
2725                         status = nfs41_test_and_free_expired_stateid(server,
2726                                         &lsp->ls_stateid,
2727                                         cred);
2728                         trace_nfs4_test_lock_stateid(state, lsp, status);
2729                         if (status == -NFS4ERR_EXPIRED ||
2730                             status == -NFS4ERR_BAD_STATEID) {
2731                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2732                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2733                                 if (!recover_lost_locks)
2734                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2735                         } else if (status != NFS_OK) {
2736                                 ret = status;
2737                                 nfs4_put_lock_state(prev);
2738                                 goto out;
2739                         }
2740                         spin_lock(&state->state_lock);
2741                 }
2742         }
2743         spin_unlock(&state->state_lock);
2744         nfs4_put_lock_state(prev);
2745 out:
2746         return ret;
2747 }
2748
2749 /**
2750  * nfs41_check_open_stateid - possibly free an open stateid
2751  *
2752  * @state: NFSv4 state for an inode
2753  *
2754  * Returns NFS_OK if recovery for this stateid is now finished.
2755  * Otherwise a negative NFS4ERR value is returned.
2756  */
2757 static int nfs41_check_open_stateid(struct nfs4_state *state)
2758 {
2759         struct nfs_server *server = NFS_SERVER(state->inode);
2760         nfs4_stateid *stateid = &state->open_stateid;
2761         struct rpc_cred *cred = state->owner->so_cred;
2762         int status;
2763
2764         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2765                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2766                         if (nfs4_have_delegation(state->inode, state->state))
2767                                 return NFS_OK;
2768                         return -NFS4ERR_OPENMODE;
2769                 }
2770                 return -NFS4ERR_BAD_STATEID;
2771         }
2772         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2773         trace_nfs4_test_open_stateid(state, NULL, status);
2774         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2775                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2776                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2777                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2778                 clear_bit(NFS_OPEN_STATE, &state->flags);
2779                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2780                 return status;
2781         }
2782         if (nfs_open_stateid_recover_openmode(state))
2783                 return -NFS4ERR_OPENMODE;
2784         return NFS_OK;
2785 }
2786
2787 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2788 {
2789         int status;
2790
2791         nfs41_check_delegation_stateid(state);
2792         status = nfs41_check_expired_locks(state);
2793         if (status != NFS_OK)
2794                 return status;
2795         status = nfs41_check_open_stateid(state);
2796         if (status != NFS_OK)
2797                 status = nfs4_open_expired(sp, state);
2798         return status;
2799 }
2800 #endif
2801
2802 /*
2803  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2804  * fields corresponding to attributes that were used to store the verifier.
2805  * Make sure we clobber those fields in the later setattr call
2806  */
2807 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2808                                 struct iattr *sattr, struct nfs4_label **label)
2809 {
2810         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2811         __u32 attrset[3];
2812         unsigned ret;
2813         unsigned i;
2814
2815         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2816                 attrset[i] = opendata->o_res.attrset[i];
2817                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2818                         attrset[i] &= ~bitmask[i];
2819         }
2820
2821         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2822                 sattr->ia_valid : 0;
2823
2824         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2825                 if (sattr->ia_valid & ATTR_ATIME_SET)
2826                         ret |= ATTR_ATIME_SET;
2827                 else
2828                         ret |= ATTR_ATIME;
2829         }
2830
2831         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2832                 if (sattr->ia_valid & ATTR_MTIME_SET)
2833                         ret |= ATTR_MTIME_SET;
2834                 else
2835                         ret |= ATTR_MTIME;
2836         }
2837
2838         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2839                 *label = NULL;
2840         return ret;
2841 }
2842
2843 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2844                 fmode_t fmode,
2845                 int flags,
2846                 struct nfs_open_context *ctx)
2847 {
2848         struct nfs4_state_owner *sp = opendata->owner;
2849         struct nfs_server *server = sp->so_server;
2850         struct dentry *dentry;
2851         struct nfs4_state *state;
2852         unsigned int seq;
2853         int ret;
2854
2855         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2856
2857         ret = _nfs4_proc_open(opendata, ctx);
2858         if (ret != 0)
2859                 goto out;
2860
2861         state = _nfs4_opendata_to_nfs4_state(opendata);
2862         ret = PTR_ERR(state);
2863         if (IS_ERR(state))
2864                 goto out;
2865         ctx->state = state;
2866         if (server->caps & NFS_CAP_POSIX_LOCK)
2867                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2868         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2869                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2870
2871         dentry = opendata->dentry;
2872         if (d_really_is_negative(dentry)) {
2873                 struct dentry *alias;
2874                 d_drop(dentry);
2875                 alias = d_exact_alias(dentry, state->inode);
2876                 if (!alias)
2877                         alias = d_splice_alias(igrab(state->inode), dentry);
2878                 /* d_splice_alias() can't fail here - it's a non-directory */
2879                 if (alias) {
2880                         dput(ctx->dentry);
2881                         ctx->dentry = dentry = alias;
2882                 }
2883                 nfs_set_verifier(dentry,
2884                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2885         }
2886
2887         /* Parse layoutget results before we check for access */
2888         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
2889
2890         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2891         if (ret != 0)
2892                 goto out;
2893
2894         if (d_inode(dentry) == state->inode) {
2895                 nfs_inode_attach_open_context(ctx);
2896                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2897                         nfs4_schedule_stateid_recovery(server, state);
2898         }
2899
2900 out:
2901         nfs4_sequence_free_slot(&opendata->o_res.seq_res);
2902         return ret;
2903 }
2904
2905 /*
2906  * Returns a referenced nfs4_state
2907  */
2908 static int _nfs4_do_open(struct inode *dir,
2909                         struct nfs_open_context *ctx,
2910                         int flags,
2911                         const struct nfs4_open_createattrs *c,
2912                         int *opened)
2913 {
2914         struct nfs4_state_owner  *sp;
2915         struct nfs4_state     *state = NULL;
2916         struct nfs_server       *server = NFS_SERVER(dir);
2917         struct nfs4_opendata *opendata;
2918         struct dentry *dentry = ctx->dentry;
2919         struct rpc_cred *cred = ctx->cred;
2920         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2921         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2922         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2923         struct iattr *sattr = c->sattr;
2924         struct nfs4_label *label = c->label;
2925         struct nfs4_label *olabel = NULL;
2926         int status;
2927
2928         /* Protect against reboot recovery conflicts */
2929         status = -ENOMEM;
2930         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2931         if (sp == NULL) {
2932                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2933                 goto out_err;
2934         }
2935         status = nfs4_client_recover_expired_lease(server->nfs_client);
2936         if (status != 0)
2937                 goto err_put_state_owner;
2938         if (d_really_is_positive(dentry))
2939                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2940         status = -ENOMEM;
2941         if (d_really_is_positive(dentry))
2942                 claim = NFS4_OPEN_CLAIM_FH;
2943         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
2944                         c, claim, GFP_KERNEL);
2945         if (opendata == NULL)
2946                 goto err_put_state_owner;
2947
2948         if (label) {
2949                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2950                 if (IS_ERR(olabel)) {
2951                         status = PTR_ERR(olabel);
2952                         goto err_opendata_put;
2953                 }
2954         }
2955
2956         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2957                 if (!opendata->f_attr.mdsthreshold) {
2958                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2959                         if (!opendata->f_attr.mdsthreshold)
2960                                 goto err_free_label;
2961                 }
2962                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2963         }
2964         if (d_really_is_positive(dentry))
2965                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2966
2967         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2968         if (status != 0)
2969                 goto err_free_label;
2970         state = ctx->state;
2971
2972         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2973             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2974                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
2975                 /*
2976                  * send create attributes which was not set by open
2977                  * with an extra setattr.
2978                  */
2979                 if (attrs || label) {
2980                         unsigned ia_old = sattr->ia_valid;
2981
2982                         sattr->ia_valid = attrs;
2983                         nfs_fattr_init(opendata->o_res.f_attr);
2984                         status = nfs4_do_setattr(state->inode, cred,
2985                                         opendata->o_res.f_attr, sattr,
2986                                         ctx, label, olabel);
2987                         if (status == 0) {
2988                                 nfs_setattr_update_inode(state->inode, sattr,
2989                                                 opendata->o_res.f_attr);
2990                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2991                         }
2992                         sattr->ia_valid = ia_old;
2993                 }
2994         }
2995         if (opened && opendata->file_created)
2996                 *opened = 1;
2997
2998         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2999                 *ctx_th = opendata->f_attr.mdsthreshold;
3000                 opendata->f_attr.mdsthreshold = NULL;
3001         }
3002
3003         nfs4_label_free(olabel);
3004
3005         nfs4_opendata_put(opendata);
3006         nfs4_put_state_owner(sp);
3007         return 0;
3008 err_free_label:
3009         nfs4_label_free(olabel);
3010 err_opendata_put:
3011         nfs4_opendata_put(opendata);
3012 err_put_state_owner:
3013         nfs4_put_state_owner(sp);
3014 out_err:
3015         return status;
3016 }
3017
3018
3019 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3020                                         struct nfs_open_context *ctx,
3021                                         int flags,
3022                                         struct iattr *sattr,
3023                                         struct nfs4_label *label,
3024                                         int *opened)
3025 {
3026         struct nfs_server *server = NFS_SERVER(dir);
3027         struct nfs4_exception exception = { };
3028         struct nfs4_state *res;
3029         struct nfs4_open_createattrs c = {
3030                 .label = label,
3031                 .sattr = sattr,
3032                 .verf = {
3033                         [0] = (__u32)jiffies,
3034                         [1] = (__u32)current->pid,
3035                 },
3036         };
3037         int status;
3038
3039         do {
3040                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3041                 res = ctx->state;
3042                 trace_nfs4_open_file(ctx, flags, status);
3043                 if (status == 0)
3044                         break;
3045                 /* NOTE: BAD_SEQID means the server and client disagree about the
3046                  * book-keeping w.r.t. state-changing operations
3047                  * (OPEN/CLOSE/LOCK/LOCKU...)
3048                  * It is actually a sign of a bug on the client or on the server.
3049                  *
3050                  * If we receive a BAD_SEQID error in the particular case of
3051                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3052                  * have unhashed the old state_owner for us, and that we can
3053                  * therefore safely retry using a new one. We should still warn
3054                  * the user though...
3055                  */
3056                 if (status == -NFS4ERR_BAD_SEQID) {
3057                         pr_warn_ratelimited("NFS: v4 server %s "
3058                                         " returned a bad sequence-id error!\n",
3059                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3060                         exception.retry = 1;
3061                         continue;
3062                 }
3063                 /*
3064                  * BAD_STATEID on OPEN means that the server cancelled our
3065                  * state before it received the OPEN_CONFIRM.
3066                  * Recover by retrying the request as per the discussion
3067                  * on Page 181 of RFC3530.
3068                  */
3069                 if (status == -NFS4ERR_BAD_STATEID) {
3070                         exception.retry = 1;
3071                         continue;
3072                 }
3073                 if (status == -EAGAIN) {
3074                         /* We must have found a delegation */
3075                         exception.retry = 1;
3076                         continue;
3077                 }
3078                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3079                         continue;
3080                 res = ERR_PTR(nfs4_handle_exception(server,
3081                                         status, &exception));
3082         } while (exception.retry);
3083         return res;
3084 }
3085
3086 static int _nfs4_do_setattr(struct inode *inode,
3087                             struct nfs_setattrargs *arg,
3088                             struct nfs_setattrres *res,
3089                             struct rpc_cred *cred,
3090                             struct nfs_open_context *ctx)
3091 {
3092         struct nfs_server *server = NFS_SERVER(inode);
3093         struct rpc_message msg = {
3094                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3095                 .rpc_argp       = arg,
3096                 .rpc_resp       = res,
3097                 .rpc_cred       = cred,
3098         };
3099         struct rpc_cred *delegation_cred = NULL;
3100         unsigned long timestamp = jiffies;
3101         bool truncate;
3102         int status;
3103
3104         nfs_fattr_init(res->fattr);
3105
3106         /* Servers should only apply open mode checks for file size changes */
3107         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3108         if (!truncate)
3109                 goto zero_stateid;
3110
3111         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3112                 /* Use that stateid */
3113         } else if (ctx != NULL) {
3114                 struct nfs_lock_context *l_ctx;
3115                 if (!nfs4_valid_open_stateid(ctx->state))
3116                         return -EBADF;
3117                 l_ctx = nfs_get_lock_context(ctx);
3118                 if (IS_ERR(l_ctx))
3119                         return PTR_ERR(l_ctx);
3120                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3121                                                 &arg->stateid, &delegation_cred);
3122                 nfs_put_lock_context(l_ctx);
3123                 if (status == -EIO)
3124                         return -EBADF;
3125         } else {
3126 zero_stateid:
3127                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3128         }
3129         if (delegation_cred)
3130                 msg.rpc_cred = delegation_cred;
3131
3132         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3133
3134         put_rpccred(delegation_cred);
3135         if (status == 0 && ctx != NULL)
3136                 renew_lease(server, timestamp);
3137         trace_nfs4_setattr(inode, &arg->stateid, status);
3138         return status;
3139 }
3140
3141 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
3142                            struct nfs_fattr *fattr, struct iattr *sattr,
3143                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3144                            struct nfs4_label *olabel)
3145 {
3146         struct nfs_server *server = NFS_SERVER(inode);
3147         __u32 bitmask[NFS4_BITMASK_SZ];
3148         struct nfs4_state *state = ctx ? ctx->state : NULL;
3149         struct nfs_setattrargs  arg = {
3150                 .fh             = NFS_FH(inode),
3151                 .iap            = sattr,
3152                 .server         = server,
3153                 .bitmask = bitmask,
3154                 .label          = ilabel,
3155         };
3156         struct nfs_setattrres  res = {
3157                 .fattr          = fattr,
3158                 .label          = olabel,
3159                 .server         = server,
3160         };
3161         struct nfs4_exception exception = {
3162                 .state = state,
3163                 .inode = inode,
3164                 .stateid = &arg.stateid,
3165         };
3166         int err;
3167
3168         do {
3169                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3170                                 nfs4_bitmask(server, olabel),
3171                                 inode);
3172
3173                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3174                 switch (err) {
3175                 case -NFS4ERR_OPENMODE:
3176                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3177                                 pr_warn_once("NFSv4: server %s is incorrectly "
3178                                                 "applying open mode checks to "
3179                                                 "a SETATTR that is not "
3180                                                 "changing file size.\n",
3181                                                 server->nfs_client->cl_hostname);
3182                         }
3183                         if (state && !(state->state & FMODE_WRITE)) {
3184                                 err = -EBADF;
3185                                 if (sattr->ia_valid & ATTR_OPEN)
3186                                         err = -EACCES;
3187                                 goto out;
3188                         }
3189                 }
3190                 err = nfs4_handle_exception(server, err, &exception);
3191         } while (exception.retry);
3192 out:
3193         return err;
3194 }
3195
3196 static bool
3197 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3198 {
3199         if (inode == NULL || !nfs_have_layout(inode))
3200                 return false;
3201
3202         return pnfs_wait_on_layoutreturn(inode, task);
3203 }
3204
3205 struct nfs4_closedata {
3206         struct inode *inode;
3207         struct nfs4_state *state;
3208         struct nfs_closeargs arg;
3209         struct nfs_closeres res;
3210         struct {
3211                 struct nfs4_layoutreturn_args arg;
3212                 struct nfs4_layoutreturn_res res;
3213                 struct nfs4_xdr_opaque_data ld_private;
3214                 u32 roc_barrier;
3215                 bool roc;
3216         } lr;
3217         struct nfs_fattr fattr;
3218         unsigned long timestamp;
3219 };
3220
3221 static void nfs4_free_closedata(void *data)
3222 {
3223         struct nfs4_closedata *calldata = data;
3224         struct nfs4_state_owner *sp = calldata->state->owner;
3225         struct super_block *sb = calldata->state->inode->i_sb;
3226
3227         if (calldata->lr.roc)
3228                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3229                                 calldata->res.lr_ret);
3230         nfs4_put_open_state(calldata->state);
3231         nfs_free_seqid(calldata->arg.seqid);
3232         nfs4_put_state_owner(sp);
3233         nfs_sb_deactive(sb);
3234         kfree(calldata);
3235 }
3236
3237 static void nfs4_close_done(struct rpc_task *task, void *data)
3238 {
3239         struct nfs4_closedata *calldata = data;
3240         struct nfs4_state *state = calldata->state;
3241         struct nfs_server *server = NFS_SERVER(calldata->inode);
3242         nfs4_stateid *res_stateid = NULL;
3243         struct nfs4_exception exception = {
3244                 .state = state,
3245                 .inode = calldata->inode,
3246                 .stateid = &calldata->arg.stateid,
3247         };
3248
3249         dprintk("%s: begin!\n", __func__);
3250         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3251                 return;
3252         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3253
3254         /* Handle Layoutreturn errors */
3255         if (calldata->arg.lr_args && task->tk_status != 0) {
3256                 switch (calldata->res.lr_ret) {
3257                 default:
3258                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3259                         break;
3260                 case 0:
3261                         calldata->arg.lr_args = NULL;
3262                         calldata->res.lr_res = NULL;
3263                         break;
3264                 case -NFS4ERR_OLD_STATEID:
3265                         if (nfs4_layoutreturn_refresh_stateid(&calldata->arg.lr_args->stateid,
3266                                                 &calldata->arg.lr_args->range,
3267                                                 calldata->inode))
3268                                 goto lr_restart;
3269                         /* Fallthrough */
3270                 case -NFS4ERR_ADMIN_REVOKED:
3271                 case -NFS4ERR_DELEG_REVOKED:
3272                 case -NFS4ERR_EXPIRED:
3273                 case -NFS4ERR_BAD_STATEID:
3274                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3275                 case -NFS4ERR_WRONG_CRED:
3276                         calldata->arg.lr_args = NULL;
3277                         calldata->res.lr_res = NULL;
3278                         goto lr_restart;
3279                 }
3280         }
3281
3282         /* hmm. we are done with the inode, and in the process of freeing
3283          * the state_owner. we keep this around to process errors
3284          */
3285         switch (task->tk_status) {
3286                 case 0:
3287                         res_stateid = &calldata->res.stateid;
3288                         renew_lease(server, calldata->timestamp);
3289                         break;
3290                 case -NFS4ERR_ACCESS:
3291                         if (calldata->arg.bitmask != NULL) {
3292                                 calldata->arg.bitmask = NULL;
3293                                 calldata->res.fattr = NULL;
3294                                 goto out_restart;
3295
3296                         }
3297                         break;
3298                 case -NFS4ERR_OLD_STATEID:
3299                         /* Did we race with OPEN? */
3300                         if (nfs4_refresh_open_stateid(&calldata->arg.stateid,
3301                                                 state))
3302                                 goto out_restart;
3303                         goto out_release;
3304                 case -NFS4ERR_ADMIN_REVOKED:
3305                 case -NFS4ERR_STALE_STATEID:
3306                 case -NFS4ERR_EXPIRED:
3307                         nfs4_free_revoked_stateid(server,
3308                                         &calldata->arg.stateid,
3309                                         task->tk_msg.rpc_cred);
3310                         /* Fallthrough */
3311                 case -NFS4ERR_BAD_STATEID:
3312                         break;
3313                 default:
3314                         task->tk_status = nfs4_async_handle_exception(task,
3315                                         server, task->tk_status, &exception);
3316                         if (exception.retry)
3317                                 goto out_restart;
3318         }
3319         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3320                         res_stateid, calldata->arg.fmode);
3321 out_release:
3322         task->tk_status = 0;
3323         nfs_release_seqid(calldata->arg.seqid);
3324         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3325         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3326         return;
3327 lr_restart:
3328         calldata->res.lr_ret = 0;
3329 out_restart:
3330         task->tk_status = 0;
3331         rpc_restart_call_prepare(task);
3332         goto out_release;
3333 }
3334
3335 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3336 {
3337         struct nfs4_closedata *calldata = data;
3338         struct nfs4_state *state = calldata->state;
3339         struct inode *inode = calldata->inode;
3340         struct pnfs_layout_hdr *lo;
3341         bool is_rdonly, is_wronly, is_rdwr;
3342         int call_close = 0;
3343
3344         dprintk("%s: begin!\n", __func__);
3345         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3346                 goto out_wait;
3347
3348         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3349         spin_lock(&state->owner->so_lock);
3350         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3351         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3352         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3353         /* Calculate the change in open mode */
3354         calldata->arg.fmode = 0;
3355         if (state->n_rdwr == 0) {
3356                 if (state->n_rdonly == 0)
3357                         call_close |= is_rdonly;
3358                 else if (is_rdonly)
3359                         calldata->arg.fmode |= FMODE_READ;
3360                 if (state->n_wronly == 0)
3361                         call_close |= is_wronly;
3362                 else if (is_wronly)
3363                         calldata->arg.fmode |= FMODE_WRITE;
3364                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3365                         call_close |= is_rdwr;
3366         } else if (is_rdwr)
3367                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3368
3369         if (!nfs4_valid_open_stateid(state) ||
3370             !nfs4_refresh_open_stateid(&calldata->arg.stateid, state))
3371                 call_close = 0;
3372         spin_unlock(&state->owner->so_lock);
3373
3374         if (!call_close) {
3375                 /* Note: exit _without_ calling nfs4_close_done */
3376                 goto out_no_action;
3377         }
3378
3379         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3380                 nfs_release_seqid(calldata->arg.seqid);
3381                 goto out_wait;
3382         }
3383
3384         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3385         if (lo && !pnfs_layout_is_valid(lo)) {
3386                 calldata->arg.lr_args = NULL;
3387                 calldata->res.lr_res = NULL;
3388         }
3389
3390         if (calldata->arg.fmode == 0)
3391                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3392
3393         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3394                 /* Close-to-open cache consistency revalidation */
3395                 if (!nfs4_have_delegation(inode, FMODE_READ))
3396                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3397                 else
3398                         calldata->arg.bitmask = NULL;
3399         }
3400
3401         calldata->arg.share_access =
3402                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3403                                 calldata->arg.fmode, 0);
3404
3405         if (calldata->res.fattr == NULL)
3406                 calldata->arg.bitmask = NULL;
3407         else if (calldata->arg.bitmask == NULL)
3408                 calldata->res.fattr = NULL;
3409         calldata->timestamp = jiffies;
3410         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3411                                 &calldata->arg.seq_args,
3412                                 &calldata->res.seq_res,
3413                                 task) != 0)
3414                 nfs_release_seqid(calldata->arg.seqid);
3415         dprintk("%s: done!\n", __func__);
3416         return;
3417 out_no_action:
3418         task->tk_action = NULL;
3419 out_wait:
3420         nfs4_sequence_done(task, &calldata->res.seq_res);
3421 }
3422
3423 static const struct rpc_call_ops nfs4_close_ops = {
3424         .rpc_call_prepare = nfs4_close_prepare,
3425         .rpc_call_done = nfs4_close_done,
3426         .rpc_release = nfs4_free_closedata,
3427 };
3428
3429 /*
3430  * It is possible for data to be read/written from a mem-mapped file
3431  * after the sys_close call (which hits the vfs layer as a flush).
3432  * This means that we can't safely call nfsv4 close on a file until
3433  * the inode is cleared. This in turn means that we are not good
3434  * NFSv4 citizens - we do not indicate to the server to update the file's
3435  * share state even when we are done with one of the three share
3436  * stateid's in the inode.
3437  *
3438  * NOTE: Caller must be holding the sp->so_owner semaphore!
3439  */
3440 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3441 {
3442         struct nfs_server *server = NFS_SERVER(state->inode);
3443         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3444         struct nfs4_closedata *calldata;
3445         struct nfs4_state_owner *sp = state->owner;
3446         struct rpc_task *task;
3447         struct rpc_message msg = {
3448                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3449                 .rpc_cred = state->owner->so_cred,
3450         };
3451         struct rpc_task_setup task_setup_data = {
3452                 .rpc_client = server->client,
3453                 .rpc_message = &msg,
3454                 .callback_ops = &nfs4_close_ops,
3455                 .workqueue = nfsiod_workqueue,
3456                 .flags = RPC_TASK_ASYNC,
3457         };
3458         int status = -ENOMEM;
3459
3460         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3461                 &task_setup_data.rpc_client, &msg);
3462
3463         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3464         if (calldata == NULL)
3465                 goto out;
3466         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3467         calldata->inode = state->inode;
3468         calldata->state = state;
3469         calldata->arg.fh = NFS_FH(state->inode);
3470         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3471                 goto out_free_calldata;
3472         /* Serialization for the sequence id */
3473         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3474         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3475         if (IS_ERR(calldata->arg.seqid))
3476                 goto out_free_calldata;
3477         nfs_fattr_init(&calldata->fattr);
3478         calldata->arg.fmode = 0;
3479         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3480         calldata->res.fattr = &calldata->fattr;
3481         calldata->res.seqid = calldata->arg.seqid;
3482         calldata->res.server = server;
3483         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3484         calldata->lr.roc = pnfs_roc(state->inode,
3485                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3486         if (calldata->lr.roc) {
3487                 calldata->arg.lr_args = &calldata->lr.arg;
3488                 calldata->res.lr_res = &calldata->lr.res;
3489         }
3490         nfs_sb_active(calldata->inode->i_sb);
3491
3492         msg.rpc_argp = &calldata->arg;
3493         msg.rpc_resp = &calldata->res;
3494         task_setup_data.callback_data = calldata;
3495         task = rpc_run_task(&task_setup_data);
3496         if (IS_ERR(task))
3497                 return PTR_ERR(task);
3498         status = 0;
3499         if (wait)
3500                 status = rpc_wait_for_completion_task(task);
3501         rpc_put_task(task);
3502         return status;
3503 out_free_calldata:
3504         kfree(calldata);
3505 out:
3506         nfs4_put_open_state(state);
3507         nfs4_put_state_owner(sp);
3508         return status;
3509 }
3510
3511 static struct inode *
3512 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3513                 int open_flags, struct iattr *attr, int *opened)
3514 {
3515         struct nfs4_state *state;
3516         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3517
3518         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3519
3520         /* Protect against concurrent sillydeletes */
3521         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3522
3523         nfs4_label_release_security(label);
3524
3525         if (IS_ERR(state))
3526                 return ERR_CAST(state);
3527         return state->inode;
3528 }
3529
3530 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3531 {
3532         if (ctx->state == NULL)
3533                 return;
3534         if (is_sync)
3535                 nfs4_close_sync(ctx->state, ctx->mode);
3536         else
3537                 nfs4_close_state(ctx->state, ctx->mode);
3538 }
3539
3540 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3541 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3542 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3543
3544 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3545 {
3546         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3547         struct nfs4_server_caps_arg args = {
3548                 .fhandle = fhandle,
3549                 .bitmask = bitmask,
3550         };
3551         struct nfs4_server_caps_res res = {};
3552         struct rpc_message msg = {
3553                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3554                 .rpc_argp = &args,
3555                 .rpc_resp = &res,
3556         };
3557         int status;
3558         int i;
3559
3560         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3561                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3562                      FATTR4_WORD0_LINK_SUPPORT |
3563                      FATTR4_WORD0_SYMLINK_SUPPORT |
3564                      FATTR4_WORD0_ACLSUPPORT;
3565         if (minorversion)
3566                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3567
3568         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3569         if (status == 0) {
3570                 /* Sanity check the server answers */
3571                 switch (minorversion) {
3572                 case 0:
3573                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3574                         res.attr_bitmask[2] = 0;
3575                         break;
3576                 case 1:
3577                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3578                         break;
3579                 case 2:
3580                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3581                 }
3582                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3583                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3584                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3585                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3586                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3587                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3588                                 NFS_CAP_SECURITY_LABEL);
3589                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3590                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3591                         server->caps |= NFS_CAP_ACLS;
3592                 if (res.has_links != 0)
3593                         server->caps |= NFS_CAP_HARDLINKS;
3594                 if (res.has_symlinks != 0)
3595                         server->caps |= NFS_CAP_SYMLINKS;
3596                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3597                         server->caps |= NFS_CAP_FILEID;
3598                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3599                         server->caps |= NFS_CAP_MODE;
3600                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3601                         server->caps |= NFS_CAP_NLINK;
3602                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3603                         server->caps |= NFS_CAP_OWNER;
3604                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3605                         server->caps |= NFS_CAP_OWNER_GROUP;
3606                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3607                         server->caps |= NFS_CAP_ATIME;
3608                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3609                         server->caps |= NFS_CAP_CTIME;
3610                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3611                         server->caps |= NFS_CAP_MTIME;
3612 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3613                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3614                         server->caps |= NFS_CAP_SECURITY_LABEL;
3615 #endif
3616                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3617                                 sizeof(server->attr_bitmask));
3618                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3619
3620                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3621                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3622                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3623                 server->cache_consistency_bitmask[2] = 0;
3624
3625                 /* Avoid a regression due to buggy server */
3626                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3627                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3628                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3629                         sizeof(server->exclcreat_bitmask));
3630
3631                 server->acl_bitmask = res.acl_bitmask;
3632                 server->fh_expire_type = res.fh_expire_type;
3633         }
3634
3635         return status;
3636 }
3637
3638 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3639 {
3640         struct nfs4_exception exception = { };
3641         int err;
3642         do {
3643                 err = nfs4_handle_exception(server,
3644                                 _nfs4_server_capabilities(server, fhandle),
3645                                 &exception);
3646         } while (exception.retry);
3647         return err;
3648 }
3649
3650 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3651                 struct nfs_fsinfo *info)
3652 {
3653         u32 bitmask[3];
3654         struct nfs4_lookup_root_arg args = {
3655                 .bitmask = bitmask,
3656         };
3657         struct nfs4_lookup_res res = {
3658                 .server = server,
3659                 .fattr = info->fattr,
3660                 .fh = fhandle,
3661         };
3662         struct rpc_message msg = {
3663                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3664                 .rpc_argp = &args,
3665                 .rpc_resp = &res,
3666         };
3667
3668         bitmask[0] = nfs4_fattr_bitmap[0];
3669         bitmask[1] = nfs4_fattr_bitmap[1];
3670         /*
3671          * Process the label in the upcoming getfattr
3672          */
3673         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3674
3675         nfs_fattr_init(info->fattr);
3676         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3677 }
3678
3679 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3680                 struct nfs_fsinfo *info)
3681 {
3682         struct nfs4_exception exception = { };
3683         int err;
3684         do {
3685                 err = _nfs4_lookup_root(server, fhandle, info);
3686                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3687                 switch (err) {
3688                 case 0:
3689                 case -NFS4ERR_WRONGSEC:
3690                         goto out;
3691                 default:
3692                         err = nfs4_handle_exception(server, err, &exception);
3693                 }
3694         } while (exception.retry);
3695 out:
3696         return err;
3697 }
3698
3699 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3700                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3701 {
3702         struct rpc_auth_create_args auth_args = {
3703                 .pseudoflavor = flavor,
3704         };
3705         struct rpc_auth *auth;
3706
3707         auth = rpcauth_create(&auth_args, server->client);
3708         if (IS_ERR(auth))
3709                 return -EACCES;
3710         return nfs4_lookup_root(server, fhandle, info);
3711 }
3712
3713 /*
3714  * Retry pseudoroot lookup with various security flavors.  We do this when:
3715  *
3716  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3717  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3718  *
3719  * Returns zero on success, or a negative NFS4ERR value, or a
3720  * negative errno value.
3721  */
3722 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3723                               struct nfs_fsinfo *info)
3724 {
3725         /* Per 3530bis 15.33.5 */
3726         static const rpc_authflavor_t flav_array[] = {
3727                 RPC_AUTH_GSS_KRB5P,
3728                 RPC_AUTH_GSS_KRB5I,
3729                 RPC_AUTH_GSS_KRB5,
3730                 RPC_AUTH_UNIX,                  /* courtesy */
3731                 RPC_AUTH_NULL,
3732         };
3733         int status = -EPERM;
3734         size_t i;
3735
3736         if (server->auth_info.flavor_len > 0) {
3737                 /* try each flavor specified by user */
3738                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3739                         status = nfs4_lookup_root_sec(server, fhandle, info,
3740                                                 server->auth_info.flavors[i]);
3741                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3742                                 continue;
3743                         break;
3744                 }
3745         } else {
3746                 /* no flavors specified by user, try default list */
3747                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3748                         status = nfs4_lookup_root_sec(server, fhandle, info,
3749                                                       flav_array[i]);
3750                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3751                                 continue;
3752                         break;
3753                 }
3754         }
3755
3756         /*
3757          * -EACCESS could mean that the user doesn't have correct permissions
3758          * to access the mount.  It could also mean that we tried to mount
3759          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3760          * existing mount programs don't handle -EACCES very well so it should
3761          * be mapped to -EPERM instead.
3762          */
3763         if (status == -EACCES)
3764                 status = -EPERM;
3765         return status;
3766 }
3767
3768 /**
3769  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3770  * @server: initialized nfs_server handle
3771  * @fhandle: we fill in the pseudo-fs root file handle
3772  * @info: we fill in an FSINFO struct
3773  * @auth_probe: probe the auth flavours
3774  *
3775  * Returns zero on success, or a negative errno.
3776  */
3777 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3778                          struct nfs_fsinfo *info,
3779                          bool auth_probe)
3780 {
3781         int status = 0;
3782
3783         if (!auth_probe)
3784                 status = nfs4_lookup_root(server, fhandle, info);
3785
3786         if (auth_probe || status == NFS4ERR_WRONGSEC)
3787                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3788                                 fhandle, info);
3789
3790         if (status == 0)
3791                 status = nfs4_server_capabilities(server, fhandle);
3792         if (status == 0)
3793                 status = nfs4_do_fsinfo(server, fhandle, info);
3794
3795         return nfs4_map_errors(status);
3796 }
3797
3798 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3799                               struct nfs_fsinfo *info)
3800 {
3801         int error;
3802         struct nfs_fattr *fattr = info->fattr;
3803         struct nfs4_label *label = NULL;
3804
3805         error = nfs4_server_capabilities(server, mntfh);
3806         if (error < 0) {
3807                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3808                 return error;
3809         }
3810
3811         label = nfs4_label_alloc(server, GFP_KERNEL);
3812         if (IS_ERR(label))
3813                 return PTR_ERR(label);
3814
3815         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
3816         if (error < 0) {
3817                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3818                 goto err_free_label;
3819         }
3820
3821         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3822             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3823                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3824
3825 err_free_label:
3826         nfs4_label_free(label);
3827
3828         return error;
3829 }
3830
3831 /*
3832  * Get locations and (maybe) other attributes of a referral.
3833  * Note that we'll actually follow the referral later when
3834  * we detect fsid mismatch in inode revalidation
3835  */
3836 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3837                              const struct qstr *name, struct nfs_fattr *fattr,
3838                              struct nfs_fh *fhandle)
3839 {
3840         int status = -ENOMEM;
3841         struct page *page = NULL;
3842         struct nfs4_fs_locations *locations = NULL;
3843
3844         page = alloc_page(GFP_KERNEL);
3845         if (page == NULL)
3846                 goto out;
3847         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3848         if (locations == NULL)
3849                 goto out;
3850
3851         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3852         if (status != 0)
3853                 goto out;
3854
3855         /*
3856          * If the fsid didn't change, this is a migration event, not a
3857          * referral.  Cause us to drop into the exception handler, which
3858          * will kick off migration recovery.
3859          */
3860         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3861                 dprintk("%s: server did not return a different fsid for"
3862                         " a referral at %s\n", __func__, name->name);
3863                 status = -NFS4ERR_MOVED;
3864                 goto out;
3865         }
3866         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3867         nfs_fixup_referral_attributes(&locations->fattr);
3868
3869         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3870         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3871         memset(fhandle, 0, sizeof(struct nfs_fh));
3872 out:
3873         if (page)
3874                 __free_page(page);
3875         kfree(locations);
3876         return status;
3877 }
3878
3879 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3880                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3881                                 struct inode *inode)
3882 {
3883         __u32 bitmask[NFS4_BITMASK_SZ];
3884         struct nfs4_getattr_arg args = {
3885                 .fh = fhandle,
3886                 .bitmask = bitmask,
3887         };
3888         struct nfs4_getattr_res res = {
3889                 .fattr = fattr,
3890                 .label = label,
3891                 .server = server,
3892         };
3893         struct rpc_message msg = {
3894                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3895                 .rpc_argp = &args,
3896                 .rpc_resp = &res,
3897         };
3898
3899         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
3900
3901         nfs_fattr_init(fattr);
3902         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3903 }
3904
3905 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3906                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3907                                 struct inode *inode)
3908 {
3909         struct nfs4_exception exception = { };
3910         int err;
3911         do {
3912                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
3913                 trace_nfs4_getattr(server, fhandle, fattr, err);
3914                 err = nfs4_handle_exception(server, err,
3915                                 &exception);
3916         } while (exception.retry);
3917         return err;
3918 }
3919
3920 /*
3921  * The file is not closed if it is opened due to the a request to change
3922  * the size of the file. The open call will not be needed once the
3923  * VFS layer lookup-intents are implemented.
3924  *
3925  * Close is called when the inode is destroyed.
3926  * If we haven't opened the file for O_WRONLY, we
3927  * need to in the size_change case to obtain a stateid.
3928  *
3929  * Got race?
3930  * Because OPEN is always done by name in nfsv4, it is
3931  * possible that we opened a different file by the same
3932  * name.  We can recognize this race condition, but we
3933  * can't do anything about it besides returning an error.
3934  *
3935  * This will be fixed with VFS changes (lookup-intent).
3936  */
3937 static int
3938 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3939                   struct iattr *sattr)
3940 {
3941         struct inode *inode = d_inode(dentry);
3942         struct rpc_cred *cred = NULL;
3943         struct nfs_open_context *ctx = NULL;
3944         struct nfs4_label *label = NULL;
3945         int status;
3946
3947         if (pnfs_ld_layoutret_on_setattr(inode) &&
3948             sattr->ia_valid & ATTR_SIZE &&
3949             sattr->ia_size < i_size_read(inode))
3950                 pnfs_commit_and_return_layout(inode);
3951
3952         nfs_fattr_init(fattr);
3953
3954         /* Deal with open(O_TRUNC) */
3955         if (sattr->ia_valid & ATTR_OPEN)
3956                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3957
3958         /* Optimization: if the end result is no change, don't RPC */
3959         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3960                 return 0;
3961
3962         /* Search for an existing open(O_WRITE) file */
3963         if (sattr->ia_valid & ATTR_FILE) {
3964
3965                 ctx = nfs_file_open_context(sattr->ia_file);
3966                 if (ctx)
3967                         cred = ctx->cred;
3968         }
3969
3970         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3971         if (IS_ERR(label))
3972                 return PTR_ERR(label);
3973
3974         /* Return any delegations if we're going to change ACLs */
3975         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
3976                 nfs4_inode_make_writeable(inode);
3977
3978         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3979         if (status == 0) {
3980                 nfs_setattr_update_inode(inode, sattr, fattr);
3981                 nfs_setsecurity(inode, fattr, label);
3982         }
3983         nfs4_label_free(label);
3984         return status;
3985 }
3986
3987 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3988                 const struct qstr *name, struct nfs_fh *fhandle,
3989                 struct nfs_fattr *fattr, struct nfs4_label *label)
3990 {
3991         struct nfs_server *server = NFS_SERVER(dir);
3992         int                    status;
3993         struct nfs4_lookup_arg args = {
3994                 .bitmask = server->attr_bitmask,
3995                 .dir_fh = NFS_FH(dir),
3996                 .name = name,
3997         };
3998         struct nfs4_lookup_res res = {
3999                 .server = server,
4000                 .fattr = fattr,
4001                 .label = label,
4002                 .fh = fhandle,
4003         };
4004         struct rpc_message msg = {
4005                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4006                 .rpc_argp = &args,
4007                 .rpc_resp = &res,
4008         };
4009
4010         args.bitmask = nfs4_bitmask(server, label);
4011
4012         nfs_fattr_init(fattr);
4013
4014         dprintk("NFS call  lookup %s\n", name->name);
4015         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
4016         dprintk("NFS reply lookup: %d\n", status);
4017         return status;
4018 }
4019
4020 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4021 {
4022         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4023                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4024         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4025         fattr->nlink = 2;
4026 }
4027
4028 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4029                                    const struct qstr *name, struct nfs_fh *fhandle,
4030                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4031 {
4032         struct nfs4_exception exception = { };
4033         struct rpc_clnt *client = *clnt;
4034         int err;
4035         do {
4036                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
4037                 trace_nfs4_lookup(dir, name, err);
4038                 switch (err) {
4039                 case -NFS4ERR_BADNAME:
4040                         err = -ENOENT;
4041                         goto out;
4042                 case -NFS4ERR_MOVED:
4043                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4044                         if (err == -NFS4ERR_MOVED)
4045                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4046                         goto out;
4047                 case -NFS4ERR_WRONGSEC:
4048                         err = -EPERM;
4049                         if (client != *clnt)
4050                                 goto out;
4051                         client = nfs4_negotiate_security(client, dir, name);
4052                         if (IS_ERR(client))
4053                                 return PTR_ERR(client);
4054
4055                         exception.retry = 1;
4056                         break;
4057                 default:
4058                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4059                 }
4060         } while (exception.retry);
4061
4062 out:
4063         if (err == 0)
4064                 *clnt = client;
4065         else if (client != *clnt)
4066                 rpc_shutdown_client(client);
4067
4068         return err;
4069 }
4070
4071 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
4072                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4073                             struct nfs4_label *label)
4074 {
4075         int status;
4076         struct rpc_clnt *client = NFS_CLIENT(dir);
4077
4078         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
4079         if (client != NFS_CLIENT(dir)) {
4080                 rpc_shutdown_client(client);
4081                 nfs_fixup_secinfo_attributes(fattr);
4082         }
4083         return status;
4084 }
4085
4086 struct rpc_clnt *
4087 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
4088                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4089 {
4090         struct rpc_clnt *client = NFS_CLIENT(dir);
4091         int status;
4092
4093         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
4094         if (status < 0)
4095                 return ERR_PTR(status);
4096         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4097 }
4098
4099 static int _nfs4_proc_lookupp(struct inode *inode,
4100                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4101                 struct nfs4_label *label)
4102 {
4103         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4104         struct nfs_server *server = NFS_SERVER(inode);
4105         int                    status;
4106         struct nfs4_lookupp_arg args = {
4107                 .bitmask = server->attr_bitmask,
4108                 .fh = NFS_FH(inode),
4109         };
4110         struct nfs4_lookupp_res res = {
4111                 .server = server,
4112                 .fattr = fattr,
4113                 .label = label,
4114                 .fh = fhandle,
4115         };
4116         struct rpc_message msg = {
4117                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4118                 .rpc_argp = &args,
4119                 .rpc_resp = &res,
4120         };
4121
4122         args.bitmask = nfs4_bitmask(server, label);
4123
4124         nfs_fattr_init(fattr);
4125
4126         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4127         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4128                                 &res.seq_res, 0);
4129         dprintk("NFS reply lookupp: %d\n", status);
4130         return status;
4131 }
4132
4133 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4134                              struct nfs_fattr *fattr, struct nfs4_label *label)
4135 {
4136         struct nfs4_exception exception = { };
4137         int err;
4138         do {
4139                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4140                 trace_nfs4_lookupp(inode, err);
4141                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4142                                 &exception);
4143         } while (exception.retry);
4144         return err;
4145 }
4146
4147 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4148 {
4149         struct nfs_server *server = NFS_SERVER(inode);
4150         struct nfs4_accessargs args = {
4151                 .fh = NFS_FH(inode),
4152                 .access = entry->mask,
4153         };
4154         struct nfs4_accessres res = {
4155                 .server = server,
4156         };
4157         struct rpc_message msg = {
4158                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4159                 .rpc_argp = &args,
4160                 .rpc_resp = &res,
4161                 .rpc_cred = entry->cred,
4162         };
4163         int status = 0;
4164
4165         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4166                 res.fattr = nfs_alloc_fattr();
4167                 if (res.fattr == NULL)
4168                         return -ENOMEM;
4169                 args.bitmask = server->cache_consistency_bitmask;
4170         }
4171
4172         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4173         if (!status) {
4174                 nfs_access_set_mask(entry, res.access);
4175                 if (res.fattr)
4176                         nfs_refresh_inode(inode, res.fattr);
4177         }
4178         nfs_free_fattr(res.fattr);
4179         return status;
4180 }
4181
4182 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4183 {
4184         struct nfs4_exception exception = { };
4185         int err;
4186         do {
4187                 err = _nfs4_proc_access(inode, entry);
4188                 trace_nfs4_access(inode, err);
4189                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4190                                 &exception);
4191         } while (exception.retry);
4192         return err;
4193 }
4194
4195 /*
4196  * TODO: For the time being, we don't try to get any attributes
4197  * along with any of the zero-copy operations READ, READDIR,
4198  * READLINK, WRITE.
4199  *
4200  * In the case of the first three, we want to put the GETATTR
4201  * after the read-type operation -- this is because it is hard
4202  * to predict the length of a GETATTR response in v4, and thus
4203  * align the READ data correctly.  This means that the GETATTR
4204  * may end up partially falling into the page cache, and we should
4205  * shift it into the 'tail' of the xdr_buf before processing.
4206  * To do this efficiently, we need to know the total length
4207  * of data received, which doesn't seem to be available outside
4208  * of the RPC layer.
4209  *
4210  * In the case of WRITE, we also want to put the GETATTR after
4211  * the operation -- in this case because we want to make sure
4212  * we get the post-operation mtime and size.
4213  *
4214  * Both of these changes to the XDR layer would in fact be quite
4215  * minor, but I decided to leave them for a subsequent patch.
4216  */
4217 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4218                 unsigned int pgbase, unsigned int pglen)
4219 {
4220         struct nfs4_readlink args = {
4221                 .fh       = NFS_FH(inode),
4222                 .pgbase   = pgbase,
4223                 .pglen    = pglen,
4224                 .pages    = &page,
4225         };
4226         struct nfs4_readlink_res res;
4227         struct rpc_message msg = {
4228                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4229                 .rpc_argp = &args,
4230                 .rpc_resp = &res,
4231         };
4232
4233         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4234 }
4235
4236 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4237                 unsigned int pgbase, unsigned int pglen)
4238 {
4239         struct nfs4_exception exception = { };
4240         int err;
4241         do {
4242                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4243                 trace_nfs4_readlink(inode, err);
4244                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4245                                 &exception);
4246         } while (exception.retry);
4247         return err;
4248 }
4249
4250 /*
4251  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4252  */
4253 static int
4254 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4255                  int flags)
4256 {
4257         struct nfs_server *server = NFS_SERVER(dir);
4258         struct nfs4_label l, *ilabel = NULL;
4259         struct nfs_open_context *ctx;
4260         struct nfs4_state *state;
4261         int status = 0;
4262
4263         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4264         if (IS_ERR(ctx))
4265                 return PTR_ERR(ctx);
4266
4267         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4268
4269         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4270                 sattr->ia_mode &= ~current_umask();
4271         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4272         if (IS_ERR(state)) {
4273                 status = PTR_ERR(state);
4274                 goto out;
4275         }
4276 out:
4277         nfs4_label_release_security(ilabel);
4278         put_nfs_open_context(ctx);
4279         return status;
4280 }
4281
4282 static int
4283 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4284 {
4285         struct nfs_server *server = NFS_SERVER(dir);
4286         struct nfs_removeargs args = {
4287                 .fh = NFS_FH(dir),
4288                 .name = *name,
4289         };
4290         struct nfs_removeres res = {
4291                 .server = server,
4292         };
4293         struct rpc_message msg = {
4294                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4295                 .rpc_argp = &args,
4296                 .rpc_resp = &res,
4297         };
4298         unsigned long timestamp = jiffies;
4299         int status;
4300
4301         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4302         if (status == 0) {
4303                 spin_lock(&dir->i_lock);
4304                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4305                 /* Removing a directory decrements nlink in the parent */
4306                 if (ftype == NF4DIR && dir->i_nlink > 2)
4307                         nfs4_dec_nlink_locked(dir);
4308                 spin_unlock(&dir->i_lock);
4309         }
4310         return status;
4311 }
4312
4313 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4314 {
4315         struct nfs4_exception exception = { };
4316         struct inode *inode = d_inode(dentry);
4317         int err;
4318
4319         if (inode) {
4320                 if (inode->i_nlink == 1)
4321                         nfs4_inode_return_delegation(inode);
4322                 else
4323                         nfs4_inode_make_writeable(inode);
4324         }
4325         do {
4326                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4327                 trace_nfs4_remove(dir, &dentry->d_name, err);
4328                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4329                                 &exception);
4330         } while (exception.retry);
4331         return err;
4332 }
4333
4334 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4335 {
4336         struct nfs4_exception exception = { };
4337         int err;
4338
4339         do {
4340                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4341                 trace_nfs4_remove(dir, name, err);
4342                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4343                                 &exception);
4344         } while (exception.retry);
4345         return err;
4346 }
4347
4348 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4349                 struct dentry *dentry,
4350                 struct inode *inode)
4351 {
4352         struct nfs_removeargs *args = msg->rpc_argp;
4353         struct nfs_removeres *res = msg->rpc_resp;
4354
4355         res->server = NFS_SB(dentry->d_sb);
4356         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4357         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4358
4359         nfs_fattr_init(res->dir_attr);
4360
4361         if (inode)
4362                 nfs4_inode_return_delegation(inode);
4363 }
4364
4365 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4366 {
4367         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4368                         &data->args.seq_args,
4369                         &data->res.seq_res,
4370                         task);
4371 }
4372
4373 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4374 {
4375         struct nfs_unlinkdata *data = task->tk_calldata;
4376         struct nfs_removeres *res = &data->res;
4377
4378         if (!nfs4_sequence_done(task, &res->seq_res))
4379                 return 0;
4380         if (nfs4_async_handle_error(task, res->server, NULL,
4381                                     &data->timeout) == -EAGAIN)
4382                 return 0;
4383         if (task->tk_status == 0)
4384                 update_changeattr(dir, &res->cinfo,
4385                                 res->dir_attr->time_start, 0);
4386         return 1;
4387 }
4388
4389 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4390                 struct dentry *old_dentry,
4391                 struct dentry *new_dentry)
4392 {
4393         struct nfs_renameargs *arg = msg->rpc_argp;
4394         struct nfs_renameres *res = msg->rpc_resp;
4395         struct inode *old_inode = d_inode(old_dentry);
4396         struct inode *new_inode = d_inode(new_dentry);
4397
4398         if (old_inode)
4399                 nfs4_inode_make_writeable(old_inode);
4400         if (new_inode)
4401                 nfs4_inode_return_delegation(new_inode);
4402         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4403         res->server = NFS_SB(old_dentry->d_sb);
4404         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4405 }
4406
4407 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4408 {
4409         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4410                         &data->args.seq_args,
4411                         &data->res.seq_res,
4412                         task);
4413 }
4414
4415 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4416                                  struct inode *new_dir)
4417 {
4418         struct nfs_renamedata *data = task->tk_calldata;
4419         struct nfs_renameres *res = &data->res;
4420
4421         if (!nfs4_sequence_done(task, &res->seq_res))
4422                 return 0;
4423         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4424                 return 0;
4425
4426         if (task->tk_status == 0) {
4427                 if (new_dir != old_dir) {
4428                         /* Note: If we moved a directory, nlink will change */
4429                         update_changeattr(old_dir, &res->old_cinfo,
4430                                         res->old_fattr->time_start,
4431                                         NFS_INO_INVALID_OTHER);
4432                         update_changeattr(new_dir, &res->new_cinfo,
4433                                         res->new_fattr->time_start,
4434                                         NFS_INO_INVALID_OTHER);
4435                 } else
4436                         update_changeattr(old_dir, &res->old_cinfo,
4437                                         res->old_fattr->time_start,
4438                                         0);
4439         }
4440         return 1;
4441 }
4442
4443 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4444 {
4445         struct nfs_server *server = NFS_SERVER(inode);
4446         __u32 bitmask[NFS4_BITMASK_SZ];
4447         struct nfs4_link_arg arg = {
4448                 .fh     = NFS_FH(inode),
4449                 .dir_fh = NFS_FH(dir),
4450                 .name   = name,
4451                 .bitmask = bitmask,
4452         };
4453         struct nfs4_link_res res = {
4454                 .server = server,
4455                 .label = NULL,
4456         };
4457         struct rpc_message msg = {
4458                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4459                 .rpc_argp = &arg,
4460                 .rpc_resp = &res,
4461         };
4462         int status = -ENOMEM;
4463
4464         res.fattr = nfs_alloc_fattr();
4465         if (res.fattr == NULL)
4466                 goto out;
4467
4468         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4469         if (IS_ERR(res.label)) {
4470                 status = PTR_ERR(res.label);
4471                 goto out;
4472         }
4473
4474         nfs4_inode_make_writeable(inode);
4475         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4476
4477         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4478         if (!status) {
4479                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4480                 status = nfs_post_op_update_inode(inode, res.fattr);
4481                 if (!status)
4482                         nfs_setsecurity(inode, res.fattr, res.label);
4483         }
4484
4485
4486         nfs4_label_free(res.label);
4487
4488 out:
4489         nfs_free_fattr(res.fattr);
4490         return status;
4491 }
4492
4493 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4494 {
4495         struct nfs4_exception exception = { };
4496         int err;
4497         do {
4498                 err = nfs4_handle_exception(NFS_SERVER(inode),
4499                                 _nfs4_proc_link(inode, dir, name),
4500                                 &exception);
4501         } while (exception.retry);
4502         return err;
4503 }
4504
4505 struct nfs4_createdata {
4506         struct rpc_message msg;
4507         struct nfs4_create_arg arg;
4508         struct nfs4_create_res res;
4509         struct nfs_fh fh;
4510         struct nfs_fattr fattr;
4511         struct nfs4_label *label;
4512 };
4513
4514 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4515                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4516 {
4517         struct nfs4_createdata *data;
4518
4519         data = kzalloc(sizeof(*data), GFP_KERNEL);
4520         if (data != NULL) {
4521                 struct nfs_server *server = NFS_SERVER(dir);
4522
4523                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4524                 if (IS_ERR(data->label))
4525                         goto out_free;
4526
4527                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4528                 data->msg.rpc_argp = &data->arg;
4529                 data->msg.rpc_resp = &data->res;
4530                 data->arg.dir_fh = NFS_FH(dir);
4531                 data->arg.server = server;
4532                 data->arg.name = name;
4533                 data->arg.attrs = sattr;
4534                 data->arg.ftype = ftype;
4535                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4536                 data->arg.umask = current_umask();
4537                 data->res.server = server;
4538                 data->res.fh = &data->fh;
4539                 data->res.fattr = &data->fattr;
4540                 data->res.label = data->label;
4541                 nfs_fattr_init(data->res.fattr);
4542         }
4543         return data;
4544 out_free:
4545         kfree(data);
4546         return NULL;
4547 }
4548
4549 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4550 {
4551         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4552                                     &data->arg.seq_args, &data->res.seq_res, 1);
4553         if (status == 0) {
4554                 spin_lock(&dir->i_lock);
4555                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4556                                 data->res.fattr->time_start, 0);
4557                 /* Creating a directory bumps nlink in the parent */
4558                 if (data->arg.ftype == NF4DIR)
4559                         nfs4_inc_nlink_locked(dir);
4560                 spin_unlock(&dir->i_lock);
4561                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4562         }
4563         return status;
4564 }
4565
4566 static void nfs4_free_createdata(struct nfs4_createdata *data)
4567 {
4568         nfs4_label_free(data->label);
4569         kfree(data);
4570 }
4571
4572 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4573                 struct page *page, unsigned int len, struct iattr *sattr,
4574                 struct nfs4_label *label)
4575 {
4576         struct nfs4_createdata *data;
4577         int status = -ENAMETOOLONG;
4578
4579         if (len > NFS4_MAXPATHLEN)
4580                 goto out;
4581
4582         status = -ENOMEM;
4583         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4584         if (data == NULL)
4585                 goto out;
4586
4587         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4588         data->arg.u.symlink.pages = &page;
4589         data->arg.u.symlink.len = len;
4590         data->arg.label = label;
4591
4592         status = nfs4_do_create(dir, dentry, data);
4593
4594         nfs4_free_createdata(data);
4595 out:
4596         return status;
4597 }
4598
4599 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4600                 struct page *page, unsigned int len, struct iattr *sattr)
4601 {
4602         struct nfs4_exception exception = { };
4603         struct nfs4_label l, *label = NULL;
4604         int err;
4605
4606         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4607
4608         do {
4609                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4610                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4611                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4612                                 &exception);
4613         } while (exception.retry);
4614
4615         nfs4_label_release_security(label);
4616         return err;
4617 }
4618
4619 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4620                 struct iattr *sattr, struct nfs4_label *label)
4621 {
4622         struct nfs4_createdata *data;
4623         int status = -ENOMEM;
4624
4625         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4626         if (data == NULL)
4627                 goto out;
4628
4629         data->arg.label = label;
4630         status = nfs4_do_create(dir, dentry, data);
4631
4632         nfs4_free_createdata(data);
4633 out:
4634         return status;
4635 }
4636
4637 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4638                 struct iattr *sattr)
4639 {
4640         struct nfs_server *server = NFS_SERVER(dir);
4641         struct nfs4_exception exception = { };
4642         struct nfs4_label l, *label = NULL;
4643         int err;
4644
4645         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4646
4647         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4648                 sattr->ia_mode &= ~current_umask();
4649         do {
4650                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4651                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4652                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4653                                 &exception);
4654         } while (exception.retry);
4655         nfs4_label_release_security(label);
4656
4657         return err;
4658 }
4659
4660 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4661                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4662 {
4663         struct inode            *dir = d_inode(dentry);
4664         struct nfs4_readdir_arg args = {
4665                 .fh = NFS_FH(dir),
4666                 .pages = pages,
4667                 .pgbase = 0,
4668                 .count = count,
4669                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4670                 .plus = plus,
4671         };
4672         struct nfs4_readdir_res res;
4673         struct rpc_message msg = {
4674                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4675                 .rpc_argp = &args,
4676                 .rpc_resp = &res,
4677                 .rpc_cred = cred,
4678         };
4679         int                     status;
4680
4681         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4682                         dentry,
4683                         (unsigned long long)cookie);
4684         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4685         res.pgbase = args.pgbase;
4686         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4687         if (status >= 0) {
4688                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4689                 status += args.pgbase;
4690         }
4691
4692         nfs_invalidate_atime(dir);
4693
4694         dprintk("%s: returns %d\n", __func__, status);
4695         return status;
4696 }
4697
4698 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4699                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4700 {
4701         struct nfs4_exception exception = { };
4702         int err;
4703         do {
4704                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4705                                 pages, count, plus);
4706                 trace_nfs4_readdir(d_inode(dentry), err);
4707                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4708                                 &exception);
4709         } while (exception.retry);
4710         return err;
4711 }
4712
4713 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4714                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4715 {
4716         struct nfs4_createdata *data;
4717         int mode = sattr->ia_mode;
4718         int status = -ENOMEM;
4719
4720         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4721         if (data == NULL)
4722                 goto out;
4723
4724         if (S_ISFIFO(mode))
4725                 data->arg.ftype = NF4FIFO;
4726         else if (S_ISBLK(mode)) {
4727                 data->arg.ftype = NF4BLK;
4728                 data->arg.u.device.specdata1 = MAJOR(rdev);
4729                 data->arg.u.device.specdata2 = MINOR(rdev);
4730         }
4731         else if (S_ISCHR(mode)) {
4732                 data->arg.ftype = NF4CHR;
4733                 data->arg.u.device.specdata1 = MAJOR(rdev);
4734                 data->arg.u.device.specdata2 = MINOR(rdev);
4735         } else if (!S_ISSOCK(mode)) {
4736                 status = -EINVAL;
4737                 goto out_free;
4738         }
4739
4740         data->arg.label = label;
4741         status = nfs4_do_create(dir, dentry, data);
4742 out_free:
4743         nfs4_free_createdata(data);
4744 out:
4745         return status;
4746 }
4747
4748 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4749                 struct iattr *sattr, dev_t rdev)
4750 {
4751         struct nfs_server *server = NFS_SERVER(dir);
4752         struct nfs4_exception exception = { };
4753         struct nfs4_label l, *label = NULL;
4754         int err;
4755
4756         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4757
4758         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4759                 sattr->ia_mode &= ~current_umask();
4760         do {
4761                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4762                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4763                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4764                                 &exception);
4765         } while (exception.retry);
4766
4767         nfs4_label_release_security(label);
4768
4769         return err;
4770 }
4771
4772 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4773                  struct nfs_fsstat *fsstat)
4774 {
4775         struct nfs4_statfs_arg args = {
4776                 .fh = fhandle,
4777                 .bitmask = server->attr_bitmask,
4778         };
4779         struct nfs4_statfs_res res = {
4780                 .fsstat = fsstat,
4781         };
4782         struct rpc_message msg = {
4783                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4784                 .rpc_argp = &args,
4785                 .rpc_resp = &res,
4786         };
4787
4788         nfs_fattr_init(fsstat->fattr);
4789         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4790 }
4791
4792 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4793 {
4794         struct nfs4_exception exception = { };
4795         int err;
4796         do {
4797                 err = nfs4_handle_exception(server,
4798                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4799                                 &exception);
4800         } while (exception.retry);
4801         return err;
4802 }
4803
4804 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4805                 struct nfs_fsinfo *fsinfo)
4806 {
4807         struct nfs4_fsinfo_arg args = {
4808                 .fh = fhandle,
4809                 .bitmask = server->attr_bitmask,
4810         };
4811         struct nfs4_fsinfo_res res = {
4812                 .fsinfo = fsinfo,
4813         };
4814         struct rpc_message msg = {
4815                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4816                 .rpc_argp = &args,
4817                 .rpc_resp = &res,
4818         };
4819
4820         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4821 }
4822
4823 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4824 {
4825         struct nfs4_exception exception = { };
4826         unsigned long now = jiffies;
4827         int err;
4828
4829         do {
4830                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4831                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4832                 if (err == 0) {
4833                         nfs4_set_lease_period(server->nfs_client,
4834                                         fsinfo->lease_time * HZ,
4835                                         now);
4836                         break;
4837                 }
4838                 err = nfs4_handle_exception(server, err, &exception);
4839         } while (exception.retry);
4840         return err;
4841 }
4842
4843 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4844 {
4845         int error;
4846
4847         nfs_fattr_init(fsinfo->fattr);
4848         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4849         if (error == 0) {
4850                 /* block layout checks this! */
4851                 server->pnfs_blksize = fsinfo->blksize;
4852                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4853         }
4854
4855         return error;
4856 }
4857
4858 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4859                 struct nfs_pathconf *pathconf)
4860 {
4861         struct nfs4_pathconf_arg args = {
4862                 .fh = fhandle,
4863                 .bitmask = server->attr_bitmask,
4864         };
4865         struct nfs4_pathconf_res res = {
4866                 .pathconf = pathconf,
4867         };
4868         struct rpc_message msg = {
4869                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4870                 .rpc_argp = &args,
4871                 .rpc_resp = &res,
4872         };
4873
4874         /* None of the pathconf attributes are mandatory to implement */
4875         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4876                 memset(pathconf, 0, sizeof(*pathconf));
4877                 return 0;
4878         }
4879
4880         nfs_fattr_init(pathconf->fattr);
4881         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4882 }
4883
4884 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4885                 struct nfs_pathconf *pathconf)
4886 {
4887         struct nfs4_exception exception = { };
4888         int err;
4889
4890         do {
4891                 err = nfs4_handle_exception(server,
4892                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4893                                 &exception);
4894         } while (exception.retry);
4895         return err;
4896 }
4897
4898 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4899                 const struct nfs_open_context *ctx,
4900                 const struct nfs_lock_context *l_ctx,
4901                 fmode_t fmode)
4902 {
4903         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4904 }
4905 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4906
4907 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4908                 const struct nfs_open_context *ctx,
4909                 const struct nfs_lock_context *l_ctx,
4910                 fmode_t fmode)
4911 {
4912         nfs4_stateid current_stateid;
4913
4914         /* If the current stateid represents a lost lock, then exit */
4915         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4916                 return true;
4917         return nfs4_stateid_match(stateid, &current_stateid);
4918 }
4919
4920 static bool nfs4_error_stateid_expired(int err)
4921 {
4922         switch (err) {
4923         case -NFS4ERR_DELEG_REVOKED:
4924         case -NFS4ERR_ADMIN_REVOKED:
4925         case -NFS4ERR_BAD_STATEID:
4926         case -NFS4ERR_STALE_STATEID:
4927         case -NFS4ERR_OLD_STATEID:
4928         case -NFS4ERR_OPENMODE:
4929         case -NFS4ERR_EXPIRED:
4930                 return true;
4931         }
4932         return false;
4933 }
4934
4935 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4936 {
4937         struct nfs_server *server = NFS_SERVER(hdr->inode);
4938
4939         trace_nfs4_read(hdr, task->tk_status);
4940         if (task->tk_status < 0) {
4941                 struct nfs4_exception exception = {
4942                         .inode = hdr->inode,
4943                         .state = hdr->args.context->state,
4944                         .stateid = &hdr->args.stateid,
4945                 };
4946                 task->tk_status = nfs4_async_handle_exception(task,
4947                                 server, task->tk_status, &exception);
4948                 if (exception.retry) {
4949                         rpc_restart_call_prepare(task);
4950                         return -EAGAIN;
4951                 }
4952         }
4953
4954         if (task->tk_status > 0)
4955                 renew_lease(server, hdr->timestamp);
4956         return 0;
4957 }
4958
4959 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4960                 struct nfs_pgio_args *args)
4961 {
4962
4963         if (!nfs4_error_stateid_expired(task->tk_status) ||
4964                 nfs4_stateid_is_current(&args->stateid,
4965                                 args->context,
4966                                 args->lock_context,
4967                                 FMODE_READ))
4968                 return false;
4969         rpc_restart_call_prepare(task);
4970         return true;
4971 }
4972
4973 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4974 {
4975
4976         dprintk("--> %s\n", __func__);
4977
4978         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4979                 return -EAGAIN;
4980         if (nfs4_read_stateid_changed(task, &hdr->args))
4981                 return -EAGAIN;
4982         if (task->tk_status > 0)
4983                 nfs_invalidate_atime(hdr->inode);
4984         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4985                                     nfs4_read_done_cb(task, hdr);
4986 }
4987
4988 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4989                                  struct rpc_message *msg)
4990 {
4991         hdr->timestamp   = jiffies;
4992         if (!hdr->pgio_done_cb)
4993                 hdr->pgio_done_cb = nfs4_read_done_cb;
4994         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4995         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
4996 }
4997
4998 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4999                                       struct nfs_pgio_header *hdr)
5000 {
5001         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5002                         &hdr->args.seq_args,
5003                         &hdr->res.seq_res,
5004                         task))
5005                 return 0;
5006         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5007                                 hdr->args.lock_context,
5008                                 hdr->rw_mode) == -EIO)
5009                 return -EIO;
5010         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5011                 return -EIO;
5012         return 0;
5013 }
5014
5015 static int nfs4_write_done_cb(struct rpc_task *task,
5016                               struct nfs_pgio_header *hdr)
5017 {
5018         struct inode *inode = hdr->inode;
5019
5020         trace_nfs4_write(hdr, task->tk_status);
5021         if (task->tk_status < 0) {
5022                 struct nfs4_exception exception = {
5023                         .inode = hdr->inode,
5024                         .state = hdr->args.context->state,
5025                         .stateid = &hdr->args.stateid,
5026                 };
5027                 task->tk_status = nfs4_async_handle_exception(task,
5028                                 NFS_SERVER(inode), task->tk_status,
5029                                 &exception);
5030                 if (exception.retry) {
5031                         rpc_restart_call_prepare(task);
5032                         return -EAGAIN;
5033                 }
5034         }
5035         if (task->tk_status >= 0) {
5036                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5037                 nfs_writeback_update_inode(hdr);
5038         }
5039         return 0;
5040 }
5041
5042 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5043                 struct nfs_pgio_args *args)
5044 {
5045
5046         if (!nfs4_error_stateid_expired(task->tk_status) ||
5047                 nfs4_stateid_is_current(&args->stateid,
5048                                 args->context,
5049                                 args->lock_context,
5050                                 FMODE_WRITE))
5051                 return false;
5052         rpc_restart_call_prepare(task);
5053         return true;
5054 }
5055
5056 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5057 {
5058         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5059                 return -EAGAIN;
5060         if (nfs4_write_stateid_changed(task, &hdr->args))
5061                 return -EAGAIN;
5062         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5063                 nfs4_write_done_cb(task, hdr);
5064 }
5065
5066 static
5067 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5068 {
5069         /* Don't request attributes for pNFS or O_DIRECT writes */
5070         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5071                 return false;
5072         /* Otherwise, request attributes if and only if we don't hold
5073          * a delegation
5074          */
5075         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5076 }
5077
5078 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5079                                   struct rpc_message *msg,
5080                                   struct rpc_clnt **clnt)
5081 {
5082         struct nfs_server *server = NFS_SERVER(hdr->inode);
5083
5084         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5085                 hdr->args.bitmask = NULL;
5086                 hdr->res.fattr = NULL;
5087         } else
5088                 hdr->args.bitmask = server->cache_consistency_bitmask;
5089
5090         if (!hdr->pgio_done_cb)
5091                 hdr->pgio_done_cb = nfs4_write_done_cb;
5092         hdr->res.server = server;
5093         hdr->timestamp   = jiffies;
5094
5095         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5096         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1, 0);
5097         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5098 }
5099
5100 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5101 {
5102         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5103                         &data->args.seq_args,
5104                         &data->res.seq_res,
5105                         task);
5106 }
5107
5108 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5109 {
5110         struct inode *inode = data->inode;
5111
5112         trace_nfs4_commit(data, task->tk_status);
5113         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5114                                     NULL, NULL) == -EAGAIN) {
5115                 rpc_restart_call_prepare(task);
5116                 return -EAGAIN;
5117         }
5118         return 0;
5119 }
5120
5121 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5122 {
5123         if (!nfs4_sequence_done(task, &data->res.seq_res))
5124                 return -EAGAIN;
5125         return data->commit_done_cb(task, data);
5126 }
5127
5128 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5129                                    struct rpc_clnt **clnt)
5130 {
5131         struct nfs_server *server = NFS_SERVER(data->inode);
5132
5133         if (data->commit_done_cb == NULL)
5134                 data->commit_done_cb = nfs4_commit_done_cb;
5135         data->res.server = server;
5136         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5137         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5138         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5139 }
5140
5141 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5142                                 struct nfs_commitres *res)
5143 {
5144         struct inode *dst_inode = file_inode(dst);
5145         struct nfs_server *server = NFS_SERVER(dst_inode);
5146         struct rpc_message msg = {
5147                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5148                 .rpc_argp = args,
5149                 .rpc_resp = res,
5150         };
5151
5152         args->fh = NFS_FH(dst_inode);
5153         return nfs4_call_sync(server->client, server, &msg,
5154                         &args->seq_args, &res->seq_res, 1);
5155 }
5156
5157 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5158 {
5159         struct nfs_commitargs args = {
5160                 .offset = offset,
5161                 .count = count,
5162         };
5163         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5164         struct nfs4_exception exception = { };
5165         int status;
5166
5167         do {
5168                 status = _nfs4_proc_commit(dst, &args, res);
5169                 status = nfs4_handle_exception(dst_server, status, &exception);
5170         } while (exception.retry);
5171
5172         return status;
5173 }
5174
5175 struct nfs4_renewdata {
5176         struct nfs_client       *client;
5177         unsigned long           timestamp;
5178 };
5179
5180 /*
5181  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5182  * standalone procedure for queueing an asynchronous RENEW.
5183  */
5184 static void nfs4_renew_release(void *calldata)
5185 {
5186         struct nfs4_renewdata *data = calldata;
5187         struct nfs_client *clp = data->client;
5188
5189         if (refcount_read(&clp->cl_count) > 1)
5190                 nfs4_schedule_state_renewal(clp);
5191         nfs_put_client(clp);
5192         kfree(data);
5193 }
5194
5195 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5196 {
5197         struct nfs4_renewdata *data = calldata;
5198         struct nfs_client *clp = data->client;
5199         unsigned long timestamp = data->timestamp;
5200
5201         trace_nfs4_renew_async(clp, task->tk_status);
5202         switch (task->tk_status) {
5203         case 0:
5204                 break;
5205         case -NFS4ERR_LEASE_MOVED:
5206                 nfs4_schedule_lease_moved_recovery(clp);
5207                 break;
5208         default:
5209                 /* Unless we're shutting down, schedule state recovery! */
5210                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5211                         return;
5212                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5213                         nfs4_schedule_lease_recovery(clp);
5214                         return;
5215                 }
5216                 nfs4_schedule_path_down_recovery(clp);
5217         }
5218         do_renew_lease(clp, timestamp);
5219 }
5220
5221 static const struct rpc_call_ops nfs4_renew_ops = {
5222         .rpc_call_done = nfs4_renew_done,
5223         .rpc_release = nfs4_renew_release,
5224 };
5225
5226 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5227 {
5228         struct rpc_message msg = {
5229                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5230                 .rpc_argp       = clp,
5231                 .rpc_cred       = cred,
5232         };
5233         struct nfs4_renewdata *data;
5234
5235         if (renew_flags == 0)
5236                 return 0;
5237         if (!refcount_inc_not_zero(&clp->cl_count))
5238                 return -EIO;
5239         data = kmalloc(sizeof(*data), GFP_NOFS);
5240         if (data == NULL) {
5241                 nfs_put_client(clp);
5242                 return -ENOMEM;
5243         }
5244         data->client = clp;
5245         data->timestamp = jiffies;
5246         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5247                         &nfs4_renew_ops, data);
5248 }
5249
5250 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
5251 {
5252         struct rpc_message msg = {
5253                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5254                 .rpc_argp       = clp,
5255                 .rpc_cred       = cred,
5256         };
5257         unsigned long now = jiffies;
5258         int status;
5259
5260         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5261         if (status < 0)
5262                 return status;
5263         do_renew_lease(clp, now);
5264         return 0;
5265 }
5266
5267 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5268 {
5269         return server->caps & NFS_CAP_ACLS;
5270 }
5271
5272 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5273  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5274  * the stack.
5275  */
5276 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5277
5278 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5279                 struct page **pages)
5280 {
5281         struct page *newpage, **spages;
5282         int rc = 0;
5283         size_t len;
5284         spages = pages;
5285
5286         do {
5287                 len = min_t(size_t, PAGE_SIZE, buflen);
5288                 newpage = alloc_page(GFP_KERNEL);
5289
5290                 if (newpage == NULL)
5291                         goto unwind;
5292                 memcpy(page_address(newpage), buf, len);
5293                 buf += len;
5294                 buflen -= len;
5295                 *pages++ = newpage;
5296                 rc++;
5297         } while (buflen != 0);
5298
5299         return rc;
5300
5301 unwind:
5302         for(; rc > 0; rc--)
5303                 __free_page(spages[rc-1]);
5304         return -ENOMEM;
5305 }
5306
5307 struct nfs4_cached_acl {
5308         int cached;
5309         size_t len;
5310         char data[0];
5311 };
5312
5313 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5314 {
5315         struct nfs_inode *nfsi = NFS_I(inode);
5316
5317         spin_lock(&inode->i_lock);
5318         kfree(nfsi->nfs4_acl);
5319         nfsi->nfs4_acl = acl;
5320         spin_unlock(&inode->i_lock);
5321 }
5322
5323 static void nfs4_zap_acl_attr(struct inode *inode)
5324 {
5325         nfs4_set_cached_acl(inode, NULL);
5326 }
5327
5328 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5329 {
5330         struct nfs_inode *nfsi = NFS_I(inode);
5331         struct nfs4_cached_acl *acl;
5332         int ret = -ENOENT;
5333
5334         spin_lock(&inode->i_lock);
5335         acl = nfsi->nfs4_acl;
5336         if (acl == NULL)
5337                 goto out;
5338         if (buf == NULL) /* user is just asking for length */
5339                 goto out_len;
5340         if (acl->cached == 0)
5341                 goto out;
5342         ret = -ERANGE; /* see getxattr(2) man page */
5343         if (acl->len > buflen)
5344                 goto out;
5345         memcpy(buf, acl->data, acl->len);
5346 out_len:
5347         ret = acl->len;
5348 out:
5349         spin_unlock(&inode->i_lock);
5350         return ret;
5351 }
5352
5353 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5354 {
5355         struct nfs4_cached_acl *acl;
5356         size_t buflen = sizeof(*acl) + acl_len;
5357
5358         if (buflen <= PAGE_SIZE) {
5359                 acl = kmalloc(buflen, GFP_KERNEL);
5360                 if (acl == NULL)
5361                         goto out;
5362                 acl->cached = 1;
5363                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5364         } else {
5365                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5366                 if (acl == NULL)
5367                         goto out;
5368                 acl->cached = 0;
5369         }
5370         acl->len = acl_len;
5371 out:
5372         nfs4_set_cached_acl(inode, acl);
5373 }
5374
5375 /*
5376  * The getxattr API returns the required buffer length when called with a
5377  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5378  * the required buf.  On a NULL buf, we send a page of data to the server
5379  * guessing that the ACL request can be serviced by a page. If so, we cache
5380  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5381  * the cache. If not so, we throw away the page, and cache the required
5382  * length. The next getxattr call will then produce another round trip to
5383  * the server, this time with the input buf of the required size.
5384  */
5385 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5386 {
5387         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5388         struct nfs_getaclargs args = {
5389                 .fh = NFS_FH(inode),
5390                 .acl_pages = pages,
5391                 .acl_len = buflen,
5392         };
5393         struct nfs_getaclres res = {
5394                 .acl_len = buflen,
5395         };
5396         struct rpc_message msg = {
5397                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5398                 .rpc_argp = &args,
5399                 .rpc_resp = &res,
5400         };
5401         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5402         int ret = -ENOMEM, i;
5403
5404         if (npages > ARRAY_SIZE(pages))
5405                 return -ERANGE;
5406
5407         for (i = 0; i < npages; i++) {
5408                 pages[i] = alloc_page(GFP_KERNEL);
5409                 if (!pages[i])
5410                         goto out_free;
5411         }
5412
5413         /* for decoding across pages */
5414         res.acl_scratch = alloc_page(GFP_KERNEL);
5415         if (!res.acl_scratch)
5416                 goto out_free;
5417
5418         args.acl_len = npages * PAGE_SIZE;
5419
5420         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5421                 __func__, buf, buflen, npages, args.acl_len);
5422         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5423                              &msg, &args.seq_args, &res.seq_res, 0);
5424         if (ret)
5425                 goto out_free;
5426
5427         /* Handle the case where the passed-in buffer is too short */
5428         if (res.acl_flags & NFS4_ACL_TRUNC) {
5429                 /* Did the user only issue a request for the acl length? */
5430                 if (buf == NULL)
5431                         goto out_ok;
5432                 ret = -ERANGE;
5433                 goto out_free;
5434         }
5435         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5436         if (buf) {
5437                 if (res.acl_len > buflen) {
5438                         ret = -ERANGE;
5439                         goto out_free;
5440                 }
5441                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5442         }
5443 out_ok:
5444         ret = res.acl_len;
5445 out_free:
5446         for (i = 0; i < npages; i++)
5447                 if (pages[i])
5448                         __free_page(pages[i]);
5449         if (res.acl_scratch)
5450                 __free_page(res.acl_scratch);
5451         return ret;
5452 }
5453
5454 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5455 {
5456         struct nfs4_exception exception = { };
5457         ssize_t ret;
5458         do {
5459                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5460                 trace_nfs4_get_acl(inode, ret);
5461                 if (ret >= 0)
5462                         break;
5463                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5464         } while (exception.retry);
5465         return ret;
5466 }
5467
5468 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5469 {
5470         struct nfs_server *server = NFS_SERVER(inode);
5471         int ret;
5472
5473         if (!nfs4_server_supports_acls(server))
5474                 return -EOPNOTSUPP;
5475         ret = nfs_revalidate_inode(server, inode);
5476         if (ret < 0)
5477                 return ret;
5478         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5479                 nfs_zap_acl_cache(inode);
5480         ret = nfs4_read_cached_acl(inode, buf, buflen);
5481         if (ret != -ENOENT)
5482                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5483                  * but no cached acl data, just the acl length */
5484                 return ret;
5485         return nfs4_get_acl_uncached(inode, buf, buflen);
5486 }
5487
5488 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5489 {
5490         struct nfs_server *server = NFS_SERVER(inode);
5491         struct page *pages[NFS4ACL_MAXPAGES];
5492         struct nfs_setaclargs arg = {
5493                 .fh             = NFS_FH(inode),
5494                 .acl_pages      = pages,
5495                 .acl_len        = buflen,
5496         };
5497         struct nfs_setaclres res;
5498         struct rpc_message msg = {
5499                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5500                 .rpc_argp       = &arg,
5501                 .rpc_resp       = &res,
5502         };
5503         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5504         int ret, i;
5505
5506         if (!nfs4_server_supports_acls(server))
5507                 return -EOPNOTSUPP;
5508         if (npages > ARRAY_SIZE(pages))
5509                 return -ERANGE;
5510         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5511         if (i < 0)
5512                 return i;
5513         nfs4_inode_make_writeable(inode);
5514         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5515
5516         /*
5517          * Free each page after tx, so the only ref left is
5518          * held by the network stack
5519          */
5520         for (; i > 0; i--)
5521                 put_page(pages[i-1]);
5522
5523         /*
5524          * Acl update can result in inode attribute update.
5525          * so mark the attribute cache invalid.
5526          */
5527         spin_lock(&inode->i_lock);
5528         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5529                 | NFS_INO_INVALID_CTIME
5530                 | NFS_INO_REVAL_FORCED;
5531         spin_unlock(&inode->i_lock);
5532         nfs_access_zap_cache(inode);
5533         nfs_zap_acl_cache(inode);
5534         return ret;
5535 }
5536
5537 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5538 {
5539         struct nfs4_exception exception = { };
5540         int err;
5541         do {
5542                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5543                 trace_nfs4_set_acl(inode, err);
5544                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5545                                 &exception);
5546         } while (exception.retry);
5547         return err;
5548 }
5549
5550 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5551 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5552                                         size_t buflen)
5553 {
5554         struct nfs_server *server = NFS_SERVER(inode);
5555         struct nfs_fattr fattr;
5556         struct nfs4_label label = {0, 0, buflen, buf};
5557
5558         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5559         struct nfs4_getattr_arg arg = {
5560                 .fh             = NFS_FH(inode),
5561                 .bitmask        = bitmask,
5562         };
5563         struct nfs4_getattr_res res = {
5564                 .fattr          = &fattr,
5565                 .label          = &label,
5566                 .server         = server,
5567         };
5568         struct rpc_message msg = {
5569                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5570                 .rpc_argp       = &arg,
5571                 .rpc_resp       = &res,
5572         };
5573         int ret;
5574
5575         nfs_fattr_init(&fattr);
5576
5577         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5578         if (ret)
5579                 return ret;
5580         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5581                 return -ENOENT;
5582         if (buflen < label.len)
5583                 return -ERANGE;
5584         return 0;
5585 }
5586
5587 static int nfs4_get_security_label(struct inode *inode, void *buf,
5588                                         size_t buflen)
5589 {
5590         struct nfs4_exception exception = { };
5591         int err;
5592
5593         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5594                 return -EOPNOTSUPP;
5595
5596         do {
5597                 err = _nfs4_get_security_label(inode, buf, buflen);
5598                 trace_nfs4_get_security_label(inode, err);
5599                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5600                                 &exception);
5601         } while (exception.retry);
5602         return err;
5603 }
5604
5605 static int _nfs4_do_set_security_label(struct inode *inode,
5606                 struct nfs4_label *ilabel,
5607                 struct nfs_fattr *fattr,
5608                 struct nfs4_label *olabel)
5609 {
5610
5611         struct iattr sattr = {0};
5612         struct nfs_server *server = NFS_SERVER(inode);
5613         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5614         struct nfs_setattrargs arg = {
5615                 .fh             = NFS_FH(inode),
5616                 .iap            = &sattr,
5617                 .server         = server,
5618                 .bitmask        = bitmask,
5619                 .label          = ilabel,
5620         };
5621         struct nfs_setattrres res = {
5622                 .fattr          = fattr,
5623                 .label          = olabel,
5624                 .server         = server,
5625         };
5626         struct rpc_message msg = {
5627                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5628                 .rpc_argp       = &arg,
5629                 .rpc_resp       = &res,
5630         };
5631         int status;
5632
5633         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5634
5635         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5636         if (status)
5637                 dprintk("%s failed: %d\n", __func__, status);
5638
5639         return status;
5640 }
5641
5642 static int nfs4_do_set_security_label(struct inode *inode,
5643                 struct nfs4_label *ilabel,
5644                 struct nfs_fattr *fattr,
5645                 struct nfs4_label *olabel)
5646 {
5647         struct nfs4_exception exception = { };
5648         int err;
5649
5650         do {
5651                 err = _nfs4_do_set_security_label(inode, ilabel,
5652                                 fattr, olabel);
5653                 trace_nfs4_set_security_label(inode, err);
5654                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5655                                 &exception);
5656         } while (exception.retry);
5657         return err;
5658 }
5659
5660 static int
5661 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5662 {
5663         struct nfs4_label ilabel, *olabel = NULL;
5664         struct nfs_fattr fattr;
5665         struct rpc_cred *cred;
5666         int status;
5667
5668         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5669                 return -EOPNOTSUPP;
5670
5671         nfs_fattr_init(&fattr);
5672
5673         ilabel.pi = 0;
5674         ilabel.lfs = 0;
5675         ilabel.label = (char *)buf;
5676         ilabel.len = buflen;
5677
5678         cred = rpc_lookup_cred();
5679         if (IS_ERR(cred))
5680                 return PTR_ERR(cred);
5681
5682         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5683         if (IS_ERR(olabel)) {
5684                 status = -PTR_ERR(olabel);
5685                 goto out;
5686         }
5687
5688         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5689         if (status == 0)
5690                 nfs_setsecurity(inode, &fattr, olabel);
5691
5692         nfs4_label_free(olabel);
5693 out:
5694         put_rpccred(cred);
5695         return status;
5696 }
5697 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5698
5699
5700 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5701                                     nfs4_verifier *bootverf)
5702 {
5703         __be32 verf[2];
5704
5705         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5706                 /* An impossible timestamp guarantees this value
5707                  * will never match a generated boot time. */
5708                 verf[0] = cpu_to_be32(U32_MAX);
5709                 verf[1] = cpu_to_be32(U32_MAX);
5710         } else {
5711                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5712                 u64 ns = ktime_to_ns(nn->boot_time);
5713
5714                 verf[0] = cpu_to_be32(ns >> 32);
5715                 verf[1] = cpu_to_be32(ns);
5716         }
5717         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5718 }
5719
5720 static int
5721 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5722 {
5723         size_t len;
5724         char *str;
5725
5726         if (clp->cl_owner_id != NULL)
5727                 return 0;
5728
5729         rcu_read_lock();
5730         len = 14 +
5731                 strlen(clp->cl_rpcclient->cl_nodename) +
5732                 1 +
5733                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5734                 1;
5735         rcu_read_unlock();
5736         if (nfs4_client_id_uniquifier[0] != '\0')
5737                 len += strlen(nfs4_client_id_uniquifier) + 1;
5738         if (len > NFS4_OPAQUE_LIMIT + 1)
5739                 return -EINVAL;
5740
5741         /*
5742          * Since this string is allocated at mount time, and held until the
5743          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5744          * about a memory-reclaim deadlock.
5745          */
5746         str = kmalloc(len, GFP_KERNEL);
5747         if (!str)
5748                 return -ENOMEM;
5749
5750         rcu_read_lock();
5751         if (nfs4_client_id_uniquifier[0] != '\0')
5752                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5753                           clp->cl_rpcclient->cl_nodename,
5754                           nfs4_client_id_uniquifier,
5755                           rpc_peeraddr2str(clp->cl_rpcclient,
5756                                            RPC_DISPLAY_ADDR));
5757         else
5758                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
5759                           clp->cl_rpcclient->cl_nodename,
5760                           rpc_peeraddr2str(clp->cl_rpcclient,
5761                                            RPC_DISPLAY_ADDR));
5762         rcu_read_unlock();
5763
5764         clp->cl_owner_id = str;
5765         return 0;
5766 }
5767
5768 static int
5769 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5770 {
5771         size_t len;
5772         char *str;
5773
5774         len = 10 + 10 + 1 + 10 + 1 +
5775                 strlen(nfs4_client_id_uniquifier) + 1 +
5776                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5777
5778         if (len > NFS4_OPAQUE_LIMIT + 1)
5779                 return -EINVAL;
5780
5781         /*
5782          * Since this string is allocated at mount time, and held until the
5783          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5784          * about a memory-reclaim deadlock.
5785          */
5786         str = kmalloc(len, GFP_KERNEL);
5787         if (!str)
5788                 return -ENOMEM;
5789
5790         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5791                         clp->rpc_ops->version, clp->cl_minorversion,
5792                         nfs4_client_id_uniquifier,
5793                         clp->cl_rpcclient->cl_nodename);
5794         clp->cl_owner_id = str;
5795         return 0;
5796 }
5797
5798 static int
5799 nfs4_init_uniform_client_string(struct nfs_client *clp)
5800 {
5801         size_t len;
5802         char *str;
5803
5804         if (clp->cl_owner_id != NULL)
5805                 return 0;
5806
5807         if (nfs4_client_id_uniquifier[0] != '\0')
5808                 return nfs4_init_uniquifier_client_string(clp);
5809
5810         len = 10 + 10 + 1 + 10 + 1 +
5811                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5812
5813         if (len > NFS4_OPAQUE_LIMIT + 1)
5814                 return -EINVAL;
5815
5816         /*
5817          * Since this string is allocated at mount time, and held until the
5818          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5819          * about a memory-reclaim deadlock.
5820          */
5821         str = kmalloc(len, GFP_KERNEL);
5822         if (!str)
5823                 return -ENOMEM;
5824
5825         scnprintf(str, len, "Linux NFSv%u.%u %s",
5826                         clp->rpc_ops->version, clp->cl_minorversion,
5827                         clp->cl_rpcclient->cl_nodename);
5828         clp->cl_owner_id = str;
5829         return 0;
5830 }
5831
5832 /*
5833  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5834  * services.  Advertise one based on the address family of the
5835  * clientaddr.
5836  */
5837 static unsigned int
5838 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5839 {
5840         if (strchr(clp->cl_ipaddr, ':') != NULL)
5841                 return scnprintf(buf, len, "tcp6");
5842         else
5843                 return scnprintf(buf, len, "tcp");
5844 }
5845
5846 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5847 {
5848         struct nfs4_setclientid *sc = calldata;
5849
5850         if (task->tk_status == 0)
5851                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5852 }
5853
5854 static const struct rpc_call_ops nfs4_setclientid_ops = {
5855         .rpc_call_done = nfs4_setclientid_done,
5856 };
5857
5858 /**
5859  * nfs4_proc_setclientid - Negotiate client ID
5860  * @clp: state data structure
5861  * @program: RPC program for NFSv4 callback service
5862  * @port: IP port number for NFS4 callback service
5863  * @cred: RPC credential to use for this call
5864  * @res: where to place the result
5865  *
5866  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5867  */
5868 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5869                 unsigned short port, struct rpc_cred *cred,
5870                 struct nfs4_setclientid_res *res)
5871 {
5872         nfs4_verifier sc_verifier;
5873         struct nfs4_setclientid setclientid = {
5874                 .sc_verifier = &sc_verifier,
5875                 .sc_prog = program,
5876                 .sc_clnt = clp,
5877         };
5878         struct rpc_message msg = {
5879                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5880                 .rpc_argp = &setclientid,
5881                 .rpc_resp = res,
5882                 .rpc_cred = cred,
5883         };
5884         struct rpc_task *task;
5885         struct rpc_task_setup task_setup_data = {
5886                 .rpc_client = clp->cl_rpcclient,
5887                 .rpc_message = &msg,
5888                 .callback_ops = &nfs4_setclientid_ops,
5889                 .callback_data = &setclientid,
5890                 .flags = RPC_TASK_TIMEOUT,
5891         };
5892         int status;
5893
5894         /* nfs_client_id4 */
5895         nfs4_init_boot_verifier(clp, &sc_verifier);
5896
5897         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5898                 status = nfs4_init_uniform_client_string(clp);
5899         else
5900                 status = nfs4_init_nonuniform_client_string(clp);
5901
5902         if (status)
5903                 goto out;
5904
5905         /* cb_client4 */
5906         setclientid.sc_netid_len =
5907                                 nfs4_init_callback_netid(clp,
5908                                                 setclientid.sc_netid,
5909                                                 sizeof(setclientid.sc_netid));
5910         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5911                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5912                                 clp->cl_ipaddr, port >> 8, port & 255);
5913
5914         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5915                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5916                 clp->cl_owner_id);
5917         task = rpc_run_task(&task_setup_data);
5918         if (IS_ERR(task)) {
5919                 status = PTR_ERR(task);
5920                 goto out;
5921         }
5922         status = task->tk_status;
5923         if (setclientid.sc_cred) {
5924                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5925                 put_rpccred(setclientid.sc_cred);
5926         }
5927         rpc_put_task(task);
5928 out:
5929         trace_nfs4_setclientid(clp, status);
5930         dprintk("NFS reply setclientid: %d\n", status);
5931         return status;
5932 }
5933
5934 /**
5935  * nfs4_proc_setclientid_confirm - Confirm client ID
5936  * @clp: state data structure
5937  * @res: result of a previous SETCLIENTID
5938  * @cred: RPC credential to use for this call
5939  *
5940  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5941  */
5942 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5943                 struct nfs4_setclientid_res *arg,
5944                 struct rpc_cred *cred)
5945 {
5946         struct rpc_message msg = {
5947                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5948                 .rpc_argp = arg,
5949                 .rpc_cred = cred,
5950         };
5951         int status;
5952
5953         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5954                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5955                 clp->cl_clientid);
5956         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5957         trace_nfs4_setclientid_confirm(clp, status);
5958         dprintk("NFS reply setclientid_confirm: %d\n", status);
5959         return status;
5960 }
5961
5962 struct nfs4_delegreturndata {
5963         struct nfs4_delegreturnargs args;
5964         struct nfs4_delegreturnres res;
5965         struct nfs_fh fh;
5966         nfs4_stateid stateid;
5967         unsigned long timestamp;
5968         struct {
5969                 struct nfs4_layoutreturn_args arg;
5970                 struct nfs4_layoutreturn_res res;
5971                 struct nfs4_xdr_opaque_data ld_private;
5972                 u32 roc_barrier;
5973                 bool roc;
5974         } lr;
5975         struct nfs_fattr fattr;
5976         int rpc_status;
5977         struct inode *inode;
5978 };
5979
5980 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5981 {
5982         struct nfs4_delegreturndata *data = calldata;
5983         struct nfs4_exception exception = {
5984                 .inode = data->inode,
5985                 .stateid = &data->stateid,
5986         };
5987
5988         if (!nfs4_sequence_done(task, &data->res.seq_res))
5989                 return;
5990
5991         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5992
5993         /* Handle Layoutreturn errors */
5994         if (data->args.lr_args && task->tk_status != 0) {
5995                 switch(data->res.lr_ret) {
5996                 default:
5997                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5998                         break;
5999                 case 0:
6000                         data->args.lr_args = NULL;
6001                         data->res.lr_res = NULL;
6002                         break;
6003                 case -NFS4ERR_OLD_STATEID:
6004                         if (nfs4_layoutreturn_refresh_stateid(&data->args.lr_args->stateid,
6005                                                 &data->args.lr_args->range,
6006                                                 data->inode))
6007                                 goto lr_restart;
6008                         /* Fallthrough */
6009                 case -NFS4ERR_ADMIN_REVOKED:
6010                 case -NFS4ERR_DELEG_REVOKED:
6011                 case -NFS4ERR_EXPIRED:
6012                 case -NFS4ERR_BAD_STATEID:
6013                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
6014                 case -NFS4ERR_WRONG_CRED:
6015                         data->args.lr_args = NULL;
6016                         data->res.lr_res = NULL;
6017                         goto lr_restart;
6018                 }
6019         }
6020
6021         switch (task->tk_status) {
6022         case 0:
6023                 renew_lease(data->res.server, data->timestamp);
6024                 break;
6025         case -NFS4ERR_ADMIN_REVOKED:
6026         case -NFS4ERR_DELEG_REVOKED:
6027         case -NFS4ERR_EXPIRED:
6028                 nfs4_free_revoked_stateid(data->res.server,
6029                                 data->args.stateid,
6030                                 task->tk_msg.rpc_cred);
6031                 /* Fallthrough */
6032         case -NFS4ERR_BAD_STATEID:
6033         case -NFS4ERR_STALE_STATEID:
6034                 task->tk_status = 0;
6035                 break;
6036         case -NFS4ERR_OLD_STATEID:
6037                 if (nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6038                         goto out_restart;
6039                 task->tk_status = 0;
6040                 break;
6041         case -NFS4ERR_ACCESS:
6042                 if (data->args.bitmask) {
6043                         data->args.bitmask = NULL;
6044                         data->res.fattr = NULL;
6045                         goto out_restart;
6046                 }
6047                 /* Fallthrough */
6048         default:
6049                 task->tk_status = nfs4_async_handle_exception(task,
6050                                 data->res.server, task->tk_status,
6051                                 &exception);
6052                 if (exception.retry)
6053                         goto out_restart;
6054         }
6055         data->rpc_status = task->tk_status;
6056         return;
6057 lr_restart:
6058         data->res.lr_ret = 0;
6059 out_restart:
6060         task->tk_status = 0;
6061         rpc_restart_call_prepare(task);
6062 }
6063
6064 static void nfs4_delegreturn_release(void *calldata)
6065 {
6066         struct nfs4_delegreturndata *data = calldata;
6067         struct inode *inode = data->inode;
6068
6069         if (inode) {
6070                 if (data->lr.roc)
6071                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6072                                         data->res.lr_ret);
6073                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6074                 nfs_iput_and_deactive(inode);
6075         }
6076         kfree(calldata);
6077 }
6078
6079 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6080 {
6081         struct nfs4_delegreturndata *d_data;
6082         struct pnfs_layout_hdr *lo;
6083
6084         d_data = (struct nfs4_delegreturndata *)data;
6085
6086         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
6087                 return;
6088
6089         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6090         if (lo && !pnfs_layout_is_valid(lo)) {
6091                 d_data->args.lr_args = NULL;
6092                 d_data->res.lr_res = NULL;
6093         }
6094
6095         nfs4_setup_sequence(d_data->res.server->nfs_client,
6096                         &d_data->args.seq_args,
6097                         &d_data->res.seq_res,
6098                         task);
6099 }
6100
6101 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6102         .rpc_call_prepare = nfs4_delegreturn_prepare,
6103         .rpc_call_done = nfs4_delegreturn_done,
6104         .rpc_release = nfs4_delegreturn_release,
6105 };
6106
6107 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
6108 {
6109         struct nfs4_delegreturndata *data;
6110         struct nfs_server *server = NFS_SERVER(inode);
6111         struct rpc_task *task;
6112         struct rpc_message msg = {
6113                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6114                 .rpc_cred = cred,
6115         };
6116         struct rpc_task_setup task_setup_data = {
6117                 .rpc_client = server->client,
6118                 .rpc_message = &msg,
6119                 .callback_ops = &nfs4_delegreturn_ops,
6120                 .flags = RPC_TASK_ASYNC,
6121         };
6122         int status = 0;
6123
6124         data = kzalloc(sizeof(*data), GFP_NOFS);
6125         if (data == NULL)
6126                 return -ENOMEM;
6127         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6128
6129         nfs4_state_protect(server->nfs_client,
6130                         NFS_SP4_MACH_CRED_CLEANUP,
6131                         &task_setup_data.rpc_client, &msg);
6132
6133         data->args.fhandle = &data->fh;
6134         data->args.stateid = &data->stateid;
6135         data->args.bitmask = server->cache_consistency_bitmask;
6136         nfs_copy_fh(&data->fh, NFS_FH(inode));
6137         nfs4_stateid_copy(&data->stateid, stateid);
6138         data->res.fattr = &data->fattr;
6139         data->res.server = server;
6140         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6141         data->lr.arg.ld_private = &data->lr.ld_private;
6142         nfs_fattr_init(data->res.fattr);
6143         data->timestamp = jiffies;
6144         data->rpc_status = 0;
6145         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6146         data->inode = nfs_igrab_and_active(inode);
6147         if (data->inode) {
6148                 if (data->lr.roc) {
6149                         data->args.lr_args = &data->lr.arg;
6150                         data->res.lr_res = &data->lr.res;
6151                 }
6152         } else if (data->lr.roc) {
6153                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6154                 data->lr.roc = false;
6155         }
6156
6157         task_setup_data.callback_data = data;
6158         msg.rpc_argp = &data->args;
6159         msg.rpc_resp = &data->res;
6160         task = rpc_run_task(&task_setup_data);
6161         if (IS_ERR(task))
6162                 return PTR_ERR(task);
6163         if (!issync)
6164                 goto out;
6165         status = rpc_wait_for_completion_task(task);
6166         if (status != 0)
6167                 goto out;
6168         status = data->rpc_status;
6169 out:
6170         rpc_put_task(task);
6171         return status;
6172 }
6173
6174 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
6175 {
6176         struct nfs_server *server = NFS_SERVER(inode);
6177         struct nfs4_exception exception = { };
6178         int err;
6179         do {
6180                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6181                 trace_nfs4_delegreturn(inode, stateid, err);
6182                 switch (err) {
6183                         case -NFS4ERR_STALE_STATEID:
6184                         case -NFS4ERR_EXPIRED:
6185                         case 0:
6186                                 return 0;
6187                 }
6188                 err = nfs4_handle_exception(server, err, &exception);
6189         } while (exception.retry);
6190         return err;
6191 }
6192
6193 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6194 {
6195         struct inode *inode = state->inode;
6196         struct nfs_server *server = NFS_SERVER(inode);
6197         struct nfs_client *clp = server->nfs_client;
6198         struct nfs_lockt_args arg = {
6199                 .fh = NFS_FH(inode),
6200                 .fl = request,
6201         };
6202         struct nfs_lockt_res res = {
6203                 .denied = request,
6204         };
6205         struct rpc_message msg = {
6206                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6207                 .rpc_argp       = &arg,
6208                 .rpc_resp       = &res,
6209                 .rpc_cred       = state->owner->so_cred,
6210         };
6211         struct nfs4_lock_state *lsp;
6212         int status;
6213
6214         arg.lock_owner.clientid = clp->cl_clientid;
6215         status = nfs4_set_lock_state(state, request);
6216         if (status != 0)
6217                 goto out;
6218         lsp = request->fl_u.nfs4_fl.owner;
6219         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6220         arg.lock_owner.s_dev = server->s_dev;
6221         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6222         switch (status) {
6223                 case 0:
6224                         request->fl_type = F_UNLCK;
6225                         break;
6226                 case -NFS4ERR_DENIED:
6227                         status = 0;
6228         }
6229         request->fl_ops->fl_release_private(request);
6230         request->fl_ops = NULL;
6231 out:
6232         return status;
6233 }
6234
6235 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6236 {
6237         struct nfs4_exception exception = { };
6238         int err;
6239
6240         do {
6241                 err = _nfs4_proc_getlk(state, cmd, request);
6242                 trace_nfs4_get_lock(request, state, cmd, err);
6243                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6244                                 &exception);
6245         } while (exception.retry);
6246         return err;
6247 }
6248
6249 struct nfs4_unlockdata {
6250         struct nfs_locku_args arg;
6251         struct nfs_locku_res res;
6252         struct nfs4_lock_state *lsp;
6253         struct nfs_open_context *ctx;
6254         struct nfs_lock_context *l_ctx;
6255         struct file_lock fl;
6256         struct nfs_server *server;
6257         unsigned long timestamp;
6258 };
6259
6260 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6261                 struct nfs_open_context *ctx,
6262                 struct nfs4_lock_state *lsp,
6263                 struct nfs_seqid *seqid)
6264 {
6265         struct nfs4_unlockdata *p;
6266         struct inode *inode = lsp->ls_state->inode;
6267
6268         p = kzalloc(sizeof(*p), GFP_NOFS);
6269         if (p == NULL)
6270                 return NULL;
6271         p->arg.fh = NFS_FH(inode);
6272         p->arg.fl = &p->fl;
6273         p->arg.seqid = seqid;
6274         p->res.seqid = seqid;
6275         p->lsp = lsp;
6276         refcount_inc(&lsp->ls_count);
6277         /* Ensure we don't close file until we're done freeing locks! */
6278         p->ctx = get_nfs_open_context(ctx);
6279         p->l_ctx = nfs_get_lock_context(ctx);
6280         memcpy(&p->fl, fl, sizeof(p->fl));
6281         p->server = NFS_SERVER(inode);
6282         return p;
6283 }
6284
6285 static void nfs4_locku_release_calldata(void *data)
6286 {
6287         struct nfs4_unlockdata *calldata = data;
6288         nfs_free_seqid(calldata->arg.seqid);
6289         nfs4_put_lock_state(calldata->lsp);
6290         nfs_put_lock_context(calldata->l_ctx);
6291         put_nfs_open_context(calldata->ctx);
6292         kfree(calldata);
6293 }
6294
6295 static void nfs4_locku_done(struct rpc_task *task, void *data)
6296 {
6297         struct nfs4_unlockdata *calldata = data;
6298         struct nfs4_exception exception = {
6299                 .inode = calldata->lsp->ls_state->inode,
6300                 .stateid = &calldata->arg.stateid,
6301         };
6302
6303         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6304                 return;
6305         switch (task->tk_status) {
6306                 case 0:
6307                         renew_lease(calldata->server, calldata->timestamp);
6308                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6309                         if (nfs4_update_lock_stateid(calldata->lsp,
6310                                         &calldata->res.stateid))
6311                                 break;
6312                         /* Fall through */
6313                 case -NFS4ERR_ADMIN_REVOKED:
6314                 case -NFS4ERR_EXPIRED:
6315                         nfs4_free_revoked_stateid(calldata->server,
6316                                         &calldata->arg.stateid,
6317                                         task->tk_msg.rpc_cred);
6318                         /* Fall through */
6319                 case -NFS4ERR_BAD_STATEID:
6320                 case -NFS4ERR_OLD_STATEID:
6321                 case -NFS4ERR_STALE_STATEID:
6322                         if (!nfs4_stateid_match(&calldata->arg.stateid,
6323                                                 &calldata->lsp->ls_stateid))
6324                                 rpc_restart_call_prepare(task);
6325                         break;
6326                 default:
6327                         task->tk_status = nfs4_async_handle_exception(task,
6328                                         calldata->server, task->tk_status,
6329                                         &exception);
6330                         if (exception.retry)
6331                                 rpc_restart_call_prepare(task);
6332         }
6333         nfs_release_seqid(calldata->arg.seqid);
6334 }
6335
6336 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6337 {
6338         struct nfs4_unlockdata *calldata = data;
6339
6340         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6341                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6342                 return;
6343
6344         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6345                 goto out_wait;
6346         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
6347         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6348                 /* Note: exit _without_ running nfs4_locku_done */
6349                 goto out_no_action;
6350         }
6351         calldata->timestamp = jiffies;
6352         if (nfs4_setup_sequence(calldata->server->nfs_client,
6353                                 &calldata->arg.seq_args,
6354                                 &calldata->res.seq_res,
6355                                 task) != 0)
6356                 nfs_release_seqid(calldata->arg.seqid);
6357         return;
6358 out_no_action:
6359         task->tk_action = NULL;
6360 out_wait:
6361         nfs4_sequence_done(task, &calldata->res.seq_res);
6362 }
6363
6364 static const struct rpc_call_ops nfs4_locku_ops = {
6365         .rpc_call_prepare = nfs4_locku_prepare,
6366         .rpc_call_done = nfs4_locku_done,
6367         .rpc_release = nfs4_locku_release_calldata,
6368 };
6369
6370 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6371                 struct nfs_open_context *ctx,
6372                 struct nfs4_lock_state *lsp,
6373                 struct nfs_seqid *seqid)
6374 {
6375         struct nfs4_unlockdata *data;
6376         struct rpc_message msg = {
6377                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6378                 .rpc_cred = ctx->cred,
6379         };
6380         struct rpc_task_setup task_setup_data = {
6381                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6382                 .rpc_message = &msg,
6383                 .callback_ops = &nfs4_locku_ops,
6384                 .workqueue = nfsiod_workqueue,
6385                 .flags = RPC_TASK_ASYNC,
6386         };
6387
6388         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6389                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6390
6391         /* Ensure this is an unlock - when canceling a lock, the
6392          * canceled lock is passed in, and it won't be an unlock.
6393          */
6394         fl->fl_type = F_UNLCK;
6395         if (fl->fl_flags & FL_CLOSE)
6396                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6397
6398         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6399         if (data == NULL) {
6400                 nfs_free_seqid(seqid);
6401                 return ERR_PTR(-ENOMEM);
6402         }
6403
6404         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6405         msg.rpc_argp = &data->arg;
6406         msg.rpc_resp = &data->res;
6407         task_setup_data.callback_data = data;
6408         return rpc_run_task(&task_setup_data);
6409 }
6410
6411 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6412 {
6413         struct inode *inode = state->inode;
6414         struct nfs4_state_owner *sp = state->owner;
6415         struct nfs_inode *nfsi = NFS_I(inode);
6416         struct nfs_seqid *seqid;
6417         struct nfs4_lock_state *lsp;
6418         struct rpc_task *task;
6419         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6420         int status = 0;
6421         unsigned char fl_flags = request->fl_flags;
6422
6423         status = nfs4_set_lock_state(state, request);
6424         /* Unlock _before_ we do the RPC call */
6425         request->fl_flags |= FL_EXISTS;
6426         /* Exclude nfs_delegation_claim_locks() */
6427         mutex_lock(&sp->so_delegreturn_mutex);
6428         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6429         down_read(&nfsi->rwsem);
6430         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6431                 up_read(&nfsi->rwsem);
6432                 mutex_unlock(&sp->so_delegreturn_mutex);
6433                 goto out;
6434         }
6435         up_read(&nfsi->rwsem);
6436         mutex_unlock(&sp->so_delegreturn_mutex);
6437         if (status != 0)
6438                 goto out;
6439         /* Is this a delegated lock? */
6440         lsp = request->fl_u.nfs4_fl.owner;
6441         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6442                 goto out;
6443         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6444         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6445         status = -ENOMEM;
6446         if (IS_ERR(seqid))
6447                 goto out;
6448         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6449         status = PTR_ERR(task);
6450         if (IS_ERR(task))
6451                 goto out;
6452         status = rpc_wait_for_completion_task(task);
6453         rpc_put_task(task);
6454 out:
6455         request->fl_flags = fl_flags;
6456         trace_nfs4_unlock(request, state, F_SETLK, status);
6457         return status;
6458 }
6459
6460 struct nfs4_lockdata {
6461         struct nfs_lock_args arg;
6462         struct nfs_lock_res res;
6463         struct nfs4_lock_state *lsp;
6464         struct nfs_open_context *ctx;
6465         struct file_lock fl;
6466         unsigned long timestamp;
6467         int rpc_status;
6468         int cancelled;
6469         struct nfs_server *server;
6470 };
6471
6472 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6473                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6474                 gfp_t gfp_mask)
6475 {
6476         struct nfs4_lockdata *p;
6477         struct inode *inode = lsp->ls_state->inode;
6478         struct nfs_server *server = NFS_SERVER(inode);
6479         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6480
6481         p = kzalloc(sizeof(*p), gfp_mask);
6482         if (p == NULL)
6483                 return NULL;
6484
6485         p->arg.fh = NFS_FH(inode);
6486         p->arg.fl = &p->fl;
6487         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6488         if (IS_ERR(p->arg.open_seqid))
6489                 goto out_free;
6490         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6491         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6492         if (IS_ERR(p->arg.lock_seqid))
6493                 goto out_free_seqid;
6494         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6495         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6496         p->arg.lock_owner.s_dev = server->s_dev;
6497         p->res.lock_seqid = p->arg.lock_seqid;
6498         p->lsp = lsp;
6499         p->server = server;
6500         refcount_inc(&lsp->ls_count);
6501         p->ctx = get_nfs_open_context(ctx);
6502         memcpy(&p->fl, fl, sizeof(p->fl));
6503         return p;
6504 out_free_seqid:
6505         nfs_free_seqid(p->arg.open_seqid);
6506 out_free:
6507         kfree(p);
6508         return NULL;
6509 }
6510
6511 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6512 {
6513         struct nfs4_lockdata *data = calldata;
6514         struct nfs4_state *state = data->lsp->ls_state;
6515
6516         dprintk("%s: begin!\n", __func__);
6517         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6518                 goto out_wait;
6519         /* Do we need to do an open_to_lock_owner? */
6520         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6521                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6522                         goto out_release_lock_seqid;
6523                 }
6524                 nfs4_stateid_copy(&data->arg.open_stateid,
6525                                 &state->open_stateid);
6526                 data->arg.new_lock_owner = 1;
6527                 data->res.open_seqid = data->arg.open_seqid;
6528         } else {
6529                 data->arg.new_lock_owner = 0;
6530                 nfs4_stateid_copy(&data->arg.lock_stateid,
6531                                 &data->lsp->ls_stateid);
6532         }
6533         if (!nfs4_valid_open_stateid(state)) {
6534                 data->rpc_status = -EBADF;
6535                 task->tk_action = NULL;
6536                 goto out_release_open_seqid;
6537         }
6538         data->timestamp = jiffies;
6539         if (nfs4_setup_sequence(data->server->nfs_client,
6540                                 &data->arg.seq_args,
6541                                 &data->res.seq_res,
6542                                 task) == 0)
6543                 return;
6544 out_release_open_seqid:
6545         nfs_release_seqid(data->arg.open_seqid);
6546 out_release_lock_seqid:
6547         nfs_release_seqid(data->arg.lock_seqid);
6548 out_wait:
6549         nfs4_sequence_done(task, &data->res.seq_res);
6550         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6551 }
6552
6553 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6554 {
6555         struct nfs4_lockdata *data = calldata;
6556         struct nfs4_lock_state *lsp = data->lsp;
6557
6558         dprintk("%s: begin!\n", __func__);
6559
6560         if (!nfs4_sequence_done(task, &data->res.seq_res))
6561                 return;
6562
6563         data->rpc_status = task->tk_status;
6564         switch (task->tk_status) {
6565         case 0:
6566                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6567                                 data->timestamp);
6568                 if (data->arg.new_lock && !data->cancelled) {
6569                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6570                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6571                                 goto out_restart;
6572                 }
6573                 if (data->arg.new_lock_owner != 0) {
6574                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6575                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6576                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6577                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6578                         goto out_restart;
6579                 break;
6580         case -NFS4ERR_BAD_STATEID:
6581         case -NFS4ERR_OLD_STATEID:
6582         case -NFS4ERR_STALE_STATEID:
6583         case -NFS4ERR_EXPIRED:
6584                 if (data->arg.new_lock_owner != 0) {
6585                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6586                                                 &lsp->ls_state->open_stateid))
6587                                 goto out_restart;
6588                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6589                                                 &lsp->ls_stateid))
6590                                 goto out_restart;
6591         }
6592 out_done:
6593         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6594         return;
6595 out_restart:
6596         if (!data->cancelled)
6597                 rpc_restart_call_prepare(task);
6598         goto out_done;
6599 }
6600
6601 static void nfs4_lock_release(void *calldata)
6602 {
6603         struct nfs4_lockdata *data = calldata;
6604
6605         dprintk("%s: begin!\n", __func__);
6606         nfs_free_seqid(data->arg.open_seqid);
6607         if (data->cancelled && data->rpc_status == 0) {
6608                 struct rpc_task *task;
6609                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6610                                 data->arg.lock_seqid);
6611                 if (!IS_ERR(task))
6612                         rpc_put_task_async(task);
6613                 dprintk("%s: cancelling lock!\n", __func__);
6614         } else
6615                 nfs_free_seqid(data->arg.lock_seqid);
6616         nfs4_put_lock_state(data->lsp);
6617         put_nfs_open_context(data->ctx);
6618         kfree(data);
6619         dprintk("%s: done!\n", __func__);
6620 }
6621
6622 static const struct rpc_call_ops nfs4_lock_ops = {
6623         .rpc_call_prepare = nfs4_lock_prepare,
6624         .rpc_call_done = nfs4_lock_done,
6625         .rpc_release = nfs4_lock_release,
6626 };
6627
6628 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6629 {
6630         switch (error) {
6631         case -NFS4ERR_ADMIN_REVOKED:
6632         case -NFS4ERR_EXPIRED:
6633         case -NFS4ERR_BAD_STATEID:
6634                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6635                 if (new_lock_owner != 0 ||
6636                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6637                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6638                 break;
6639         case -NFS4ERR_STALE_STATEID:
6640                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6641                 nfs4_schedule_lease_recovery(server->nfs_client);
6642         };
6643 }
6644
6645 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6646 {
6647         struct nfs4_lockdata *data;
6648         struct rpc_task *task;
6649         struct rpc_message msg = {
6650                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6651                 .rpc_cred = state->owner->so_cred,
6652         };
6653         struct rpc_task_setup task_setup_data = {
6654                 .rpc_client = NFS_CLIENT(state->inode),
6655                 .rpc_message = &msg,
6656                 .callback_ops = &nfs4_lock_ops,
6657                 .workqueue = nfsiod_workqueue,
6658                 .flags = RPC_TASK_ASYNC,
6659         };
6660         int ret;
6661
6662         dprintk("%s: begin!\n", __func__);
6663         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6664                         fl->fl_u.nfs4_fl.owner,
6665                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6666         if (data == NULL)
6667                 return -ENOMEM;
6668         if (IS_SETLKW(cmd))
6669                 data->arg.block = 1;
6670         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6671                                 recovery_type > NFS_LOCK_NEW);
6672         msg.rpc_argp = &data->arg;
6673         msg.rpc_resp = &data->res;
6674         task_setup_data.callback_data = data;
6675         if (recovery_type > NFS_LOCK_NEW) {
6676                 if (recovery_type == NFS_LOCK_RECLAIM)
6677                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6678         } else
6679                 data->arg.new_lock = 1;
6680         task = rpc_run_task(&task_setup_data);
6681         if (IS_ERR(task))
6682                 return PTR_ERR(task);
6683         ret = rpc_wait_for_completion_task(task);
6684         if (ret == 0) {
6685                 ret = data->rpc_status;
6686                 if (ret)
6687                         nfs4_handle_setlk_error(data->server, data->lsp,
6688                                         data->arg.new_lock_owner, ret);
6689         } else
6690                 data->cancelled = true;
6691         rpc_put_task(task);
6692         dprintk("%s: done, ret = %d!\n", __func__, ret);
6693         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6694         return ret;
6695 }
6696
6697 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6698 {
6699         struct nfs_server *server = NFS_SERVER(state->inode);
6700         struct nfs4_exception exception = {
6701                 .inode = state->inode,
6702         };
6703         int err;
6704
6705         do {
6706                 /* Cache the lock if possible... */
6707                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6708                         return 0;
6709                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6710                 if (err != -NFS4ERR_DELAY)
6711                         break;
6712                 nfs4_handle_exception(server, err, &exception);
6713         } while (exception.retry);
6714         return err;
6715 }
6716
6717 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6718 {
6719         struct nfs_server *server = NFS_SERVER(state->inode);
6720         struct nfs4_exception exception = {
6721                 .inode = state->inode,
6722         };
6723         int err;
6724
6725         err = nfs4_set_lock_state(state, request);
6726         if (err != 0)
6727                 return err;
6728         if (!recover_lost_locks) {
6729                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6730                 return 0;
6731         }
6732         do {
6733                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6734                         return 0;
6735                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6736                 switch (err) {
6737                 default:
6738                         goto out;
6739                 case -NFS4ERR_GRACE:
6740                 case -NFS4ERR_DELAY:
6741                         nfs4_handle_exception(server, err, &exception);
6742                         err = 0;
6743                 }
6744         } while (exception.retry);
6745 out:
6746         return err;
6747 }
6748
6749 #if defined(CONFIG_NFS_V4_1)
6750 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6751 {
6752         struct nfs4_lock_state *lsp;
6753         int status;
6754
6755         status = nfs4_set_lock_state(state, request);
6756         if (status != 0)
6757                 return status;
6758         lsp = request->fl_u.nfs4_fl.owner;
6759         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6760             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6761                 return 0;
6762         return nfs4_lock_expired(state, request);
6763 }
6764 #endif
6765
6766 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6767 {
6768         struct nfs_inode *nfsi = NFS_I(state->inode);
6769         struct nfs4_state_owner *sp = state->owner;
6770         unsigned char fl_flags = request->fl_flags;
6771         int status;
6772
6773         request->fl_flags |= FL_ACCESS;
6774         status = locks_lock_inode_wait(state->inode, request);
6775         if (status < 0)
6776                 goto out;
6777         mutex_lock(&sp->so_delegreturn_mutex);
6778         down_read(&nfsi->rwsem);
6779         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6780                 /* Yes: cache locks! */
6781                 /* ...but avoid races with delegation recall... */
6782                 request->fl_flags = fl_flags & ~FL_SLEEP;
6783                 status = locks_lock_inode_wait(state->inode, request);
6784                 up_read(&nfsi->rwsem);
6785                 mutex_unlock(&sp->so_delegreturn_mutex);
6786                 goto out;
6787         }
6788         up_read(&nfsi->rwsem);
6789         mutex_unlock(&sp->so_delegreturn_mutex);
6790         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6791 out:
6792         request->fl_flags = fl_flags;
6793         return status;
6794 }
6795
6796 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6797 {
6798         struct nfs4_exception exception = {
6799                 .state = state,
6800                 .inode = state->inode,
6801         };
6802         int err;
6803
6804         do {
6805                 err = _nfs4_proc_setlk(state, cmd, request);
6806                 if (err == -NFS4ERR_DENIED)
6807                         err = -EAGAIN;
6808                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6809                                 err, &exception);
6810         } while (exception.retry);
6811         return err;
6812 }
6813
6814 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6815 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6816
6817 static int
6818 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6819                         struct file_lock *request)
6820 {
6821         int             status = -ERESTARTSYS;
6822         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6823
6824         while(!signalled()) {
6825                 status = nfs4_proc_setlk(state, cmd, request);
6826                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6827                         break;
6828                 freezable_schedule_timeout_interruptible(timeout);
6829                 timeout *= 2;
6830                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6831                 status = -ERESTARTSYS;
6832         }
6833         return status;
6834 }
6835
6836 #ifdef CONFIG_NFS_V4_1
6837 struct nfs4_lock_waiter {
6838         struct task_struct      *task;
6839         struct inode            *inode;
6840         struct nfs_lowner       *owner;
6841         bool                    notified;
6842 };
6843
6844 static int
6845 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6846 {
6847         int ret;
6848         struct nfs4_lock_waiter *waiter = wait->private;
6849
6850         /* NULL key means to wake up everyone */
6851         if (key) {
6852                 struct cb_notify_lock_args      *cbnl = key;
6853                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
6854                                                 *wowner = waiter->owner;
6855
6856                 /* Only wake if the callback was for the same owner. */
6857                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
6858                         return 0;
6859
6860                 /* Make sure it's for the right inode */
6861                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6862                         return 0;
6863
6864                 waiter->notified = true;
6865         }
6866
6867         /* override "private" so we can use default_wake_function */
6868         wait->private = waiter->task;
6869         ret = autoremove_wake_function(wait, mode, flags, key);
6870         wait->private = waiter;
6871         return ret;
6872 }
6873
6874 static int
6875 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6876 {
6877         int status = -ERESTARTSYS;
6878         unsigned long flags;
6879         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6880         struct nfs_server *server = NFS_SERVER(state->inode);
6881         struct nfs_client *clp = server->nfs_client;
6882         wait_queue_head_t *q = &clp->cl_lock_waitq;
6883         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6884                                     .id = lsp->ls_seqid.owner_id,
6885                                     .s_dev = server->s_dev };
6886         struct nfs4_lock_waiter waiter = { .task  = current,
6887                                            .inode = state->inode,
6888                                            .owner = &owner,
6889                                            .notified = false };
6890         wait_queue_entry_t wait;
6891
6892         /* Don't bother with waitqueue if we don't expect a callback */
6893         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6894                 return nfs4_retry_setlk_simple(state, cmd, request);
6895
6896         init_wait(&wait);
6897         wait.private = &waiter;
6898         wait.func = nfs4_wake_lock_waiter;
6899         add_wait_queue(q, &wait);
6900
6901         while(!signalled()) {
6902                 waiter.notified = false;
6903                 status = nfs4_proc_setlk(state, cmd, request);
6904                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6905                         break;
6906
6907                 status = -ERESTARTSYS;
6908                 spin_lock_irqsave(&q->lock, flags);
6909                 if (waiter.notified) {
6910                         spin_unlock_irqrestore(&q->lock, flags);
6911                         continue;
6912                 }
6913                 set_current_state(TASK_INTERRUPTIBLE);
6914                 spin_unlock_irqrestore(&q->lock, flags);
6915
6916                 freezable_schedule_timeout(NFS4_LOCK_MAXTIMEOUT);
6917         }
6918
6919         finish_wait(q, &wait);
6920         return status;
6921 }
6922 #else /* !CONFIG_NFS_V4_1 */
6923 static inline int
6924 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6925 {
6926         return nfs4_retry_setlk_simple(state, cmd, request);
6927 }
6928 #endif
6929
6930 static int
6931 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6932 {
6933         struct nfs_open_context *ctx;
6934         struct nfs4_state *state;
6935         int status;
6936
6937         /* verify open state */
6938         ctx = nfs_file_open_context(filp);
6939         state = ctx->state;
6940
6941         if (IS_GETLK(cmd)) {
6942                 if (state != NULL)
6943                         return nfs4_proc_getlk(state, F_GETLK, request);
6944                 return 0;
6945         }
6946
6947         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6948                 return -EINVAL;
6949
6950         if (request->fl_type == F_UNLCK) {
6951                 if (state != NULL)
6952                         return nfs4_proc_unlck(state, cmd, request);
6953                 return 0;
6954         }
6955
6956         if (state == NULL)
6957                 return -ENOLCK;
6958
6959         if ((request->fl_flags & FL_POSIX) &&
6960             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6961                 return -ENOLCK;
6962
6963         /*
6964          * Don't rely on the VFS having checked the file open mode,
6965          * since it won't do this for flock() locks.
6966          */
6967         switch (request->fl_type) {
6968         case F_RDLCK:
6969                 if (!(filp->f_mode & FMODE_READ))
6970                         return -EBADF;
6971                 break;
6972         case F_WRLCK:
6973                 if (!(filp->f_mode & FMODE_WRITE))
6974                         return -EBADF;
6975         }
6976
6977         status = nfs4_set_lock_state(state, request);
6978         if (status != 0)
6979                 return status;
6980
6981         return nfs4_retry_setlk(state, cmd, request);
6982 }
6983
6984 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6985 {
6986         struct nfs_server *server = NFS_SERVER(state->inode);
6987         int err;
6988
6989         err = nfs4_set_lock_state(state, fl);
6990         if (err != 0)
6991                 return err;
6992         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6993         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
6994 }
6995
6996 struct nfs_release_lockowner_data {
6997         struct nfs4_lock_state *lsp;
6998         struct nfs_server *server;
6999         struct nfs_release_lockowner_args args;
7000         struct nfs_release_lockowner_res res;
7001         unsigned long timestamp;
7002 };
7003
7004 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7005 {
7006         struct nfs_release_lockowner_data *data = calldata;
7007         struct nfs_server *server = data->server;
7008         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7009                            &data->res.seq_res, task);
7010         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7011         data->timestamp = jiffies;
7012 }
7013
7014 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7015 {
7016         struct nfs_release_lockowner_data *data = calldata;
7017         struct nfs_server *server = data->server;
7018
7019         nfs40_sequence_done(task, &data->res.seq_res);
7020
7021         switch (task->tk_status) {
7022         case 0:
7023                 renew_lease(server, data->timestamp);
7024                 break;
7025         case -NFS4ERR_STALE_CLIENTID:
7026         case -NFS4ERR_EXPIRED:
7027                 nfs4_schedule_lease_recovery(server->nfs_client);
7028                 break;
7029         case -NFS4ERR_LEASE_MOVED:
7030         case -NFS4ERR_DELAY:
7031                 if (nfs4_async_handle_error(task, server,
7032                                             NULL, NULL) == -EAGAIN)
7033                         rpc_restart_call_prepare(task);
7034         }
7035 }
7036
7037 static void nfs4_release_lockowner_release(void *calldata)
7038 {
7039         struct nfs_release_lockowner_data *data = calldata;
7040         nfs4_free_lock_state(data->server, data->lsp);
7041         kfree(calldata);
7042 }
7043
7044 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7045         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7046         .rpc_call_done = nfs4_release_lockowner_done,
7047         .rpc_release = nfs4_release_lockowner_release,
7048 };
7049
7050 static void
7051 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7052 {
7053         struct nfs_release_lockowner_data *data;
7054         struct rpc_message msg = {
7055                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7056         };
7057
7058         if (server->nfs_client->cl_mvops->minor_version != 0)
7059                 return;
7060
7061         data = kmalloc(sizeof(*data), GFP_NOFS);
7062         if (!data)
7063                 return;
7064         data->lsp = lsp;
7065         data->server = server;
7066         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7067         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7068         data->args.lock_owner.s_dev = server->s_dev;
7069
7070         msg.rpc_argp = &data->args;
7071         msg.rpc_resp = &data->res;
7072         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7073         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7074 }
7075
7076 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7077
7078 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7079                                    struct dentry *unused, struct inode *inode,
7080                                    const char *key, const void *buf,
7081                                    size_t buflen, int flags)
7082 {
7083         return nfs4_proc_set_acl(inode, buf, buflen);
7084 }
7085
7086 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7087                                    struct dentry *unused, struct inode *inode,
7088                                    const char *key, void *buf, size_t buflen)
7089 {
7090         return nfs4_proc_get_acl(inode, buf, buflen);
7091 }
7092
7093 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7094 {
7095         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7096 }
7097
7098 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7099
7100 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7101                                      struct dentry *unused, struct inode *inode,
7102                                      const char *key, const void *buf,
7103                                      size_t buflen, int flags)
7104 {
7105         if (security_ismaclabel(key))
7106                 return nfs4_set_security_label(inode, buf, buflen);
7107
7108         return -EOPNOTSUPP;
7109 }
7110
7111 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7112                                      struct dentry *unused, struct inode *inode,
7113                                      const char *key, void *buf, size_t buflen)
7114 {
7115         if (security_ismaclabel(key))
7116                 return nfs4_get_security_label(inode, buf, buflen);
7117         return -EOPNOTSUPP;
7118 }
7119
7120 static ssize_t
7121 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7122 {
7123         int len = 0;
7124
7125         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7126                 len = security_inode_listsecurity(inode, list, list_len);
7127                 if (list_len && len > list_len)
7128                         return -ERANGE;
7129         }
7130         return len;
7131 }
7132
7133 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7134         .prefix = XATTR_SECURITY_PREFIX,
7135         .get    = nfs4_xattr_get_nfs4_label,
7136         .set    = nfs4_xattr_set_nfs4_label,
7137 };
7138
7139 #else
7140
7141 static ssize_t
7142 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7143 {
7144         return 0;
7145 }
7146
7147 #endif
7148
7149 /*
7150  * nfs_fhget will use either the mounted_on_fileid or the fileid
7151  */
7152 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7153 {
7154         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7155                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7156               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7157               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7158                 return;
7159
7160         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7161                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7162         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7163         fattr->nlink = 2;
7164 }
7165
7166 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7167                                    const struct qstr *name,
7168                                    struct nfs4_fs_locations *fs_locations,
7169                                    struct page *page)
7170 {
7171         struct nfs_server *server = NFS_SERVER(dir);
7172         u32 bitmask[3];
7173         struct nfs4_fs_locations_arg args = {
7174                 .dir_fh = NFS_FH(dir),
7175                 .name = name,
7176                 .page = page,
7177                 .bitmask = bitmask,
7178         };
7179         struct nfs4_fs_locations_res res = {
7180                 .fs_locations = fs_locations,
7181         };
7182         struct rpc_message msg = {
7183                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7184                 .rpc_argp = &args,
7185                 .rpc_resp = &res,
7186         };
7187         int status;
7188
7189         dprintk("%s: start\n", __func__);
7190
7191         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7192         bitmask[1] = nfs4_fattr_bitmap[1];
7193
7194         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7195          * is not supported */
7196         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7197                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7198         else
7199                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7200
7201         nfs_fattr_init(&fs_locations->fattr);
7202         fs_locations->server = server;
7203         fs_locations->nlocations = 0;
7204         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7205         dprintk("%s: returned status = %d\n", __func__, status);
7206         return status;
7207 }
7208
7209 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7210                            const struct qstr *name,
7211                            struct nfs4_fs_locations *fs_locations,
7212                            struct page *page)
7213 {
7214         struct nfs4_exception exception = { };
7215         int err;
7216         do {
7217                 err = _nfs4_proc_fs_locations(client, dir, name,
7218                                 fs_locations, page);
7219                 trace_nfs4_get_fs_locations(dir, name, err);
7220                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7221                                 &exception);
7222         } while (exception.retry);
7223         return err;
7224 }
7225
7226 /*
7227  * This operation also signals the server that this client is
7228  * performing migration recovery.  The server can stop returning
7229  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7230  * appended to this compound to identify the client ID which is
7231  * performing recovery.
7232  */
7233 static int _nfs40_proc_get_locations(struct inode *inode,
7234                                      struct nfs4_fs_locations *locations,
7235                                      struct page *page, struct rpc_cred *cred)
7236 {
7237         struct nfs_server *server = NFS_SERVER(inode);
7238         struct rpc_clnt *clnt = server->client;
7239         u32 bitmask[2] = {
7240                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7241         };
7242         struct nfs4_fs_locations_arg args = {
7243                 .clientid       = server->nfs_client->cl_clientid,
7244                 .fh             = NFS_FH(inode),
7245                 .page           = page,
7246                 .bitmask        = bitmask,
7247                 .migration      = 1,            /* skip LOOKUP */
7248                 .renew          = 1,            /* append RENEW */
7249         };
7250         struct nfs4_fs_locations_res res = {
7251                 .fs_locations   = locations,
7252                 .migration      = 1,
7253                 .renew          = 1,
7254         };
7255         struct rpc_message msg = {
7256                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7257                 .rpc_argp       = &args,
7258                 .rpc_resp       = &res,
7259                 .rpc_cred       = cred,
7260         };
7261         unsigned long now = jiffies;
7262         int status;
7263
7264         nfs_fattr_init(&locations->fattr);
7265         locations->server = server;
7266         locations->nlocations = 0;
7267
7268         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7269         status = nfs4_call_sync_sequence(clnt, server, &msg,
7270                                         &args.seq_args, &res.seq_res);
7271         if (status)
7272                 return status;
7273
7274         renew_lease(server, now);
7275         return 0;
7276 }
7277
7278 #ifdef CONFIG_NFS_V4_1
7279
7280 /*
7281  * This operation also signals the server that this client is
7282  * performing migration recovery.  The server can stop asserting
7283  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7284  * performing this operation is identified in the SEQUENCE
7285  * operation in this compound.
7286  *
7287  * When the client supports GETATTR(fs_locations_info), it can
7288  * be plumbed in here.
7289  */
7290 static int _nfs41_proc_get_locations(struct inode *inode,
7291                                      struct nfs4_fs_locations *locations,
7292                                      struct page *page, struct rpc_cred *cred)
7293 {
7294         struct nfs_server *server = NFS_SERVER(inode);
7295         struct rpc_clnt *clnt = server->client;
7296         u32 bitmask[2] = {
7297                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7298         };
7299         struct nfs4_fs_locations_arg args = {
7300                 .fh             = NFS_FH(inode),
7301                 .page           = page,
7302                 .bitmask        = bitmask,
7303                 .migration      = 1,            /* skip LOOKUP */
7304         };
7305         struct nfs4_fs_locations_res res = {
7306                 .fs_locations   = locations,
7307                 .migration      = 1,
7308         };
7309         struct rpc_message msg = {
7310                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7311                 .rpc_argp       = &args,
7312                 .rpc_resp       = &res,
7313                 .rpc_cred       = cred,
7314         };
7315         int status;
7316
7317         nfs_fattr_init(&locations->fattr);
7318         locations->server = server;
7319         locations->nlocations = 0;
7320
7321         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7322         status = nfs4_call_sync_sequence(clnt, server, &msg,
7323                                         &args.seq_args, &res.seq_res);
7324         if (status == NFS4_OK &&
7325             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7326                 status = -NFS4ERR_LEASE_MOVED;
7327         return status;
7328 }
7329
7330 #endif  /* CONFIG_NFS_V4_1 */
7331
7332 /**
7333  * nfs4_proc_get_locations - discover locations for a migrated FSID
7334  * @inode: inode on FSID that is migrating
7335  * @locations: result of query
7336  * @page: buffer
7337  * @cred: credential to use for this operation
7338  *
7339  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7340  * operation failed, or a negative errno if a local error occurred.
7341  *
7342  * On success, "locations" is filled in, but if the server has
7343  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7344  * asserted.
7345  *
7346  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7347  * from this client that require migration recovery.
7348  */
7349 int nfs4_proc_get_locations(struct inode *inode,
7350                             struct nfs4_fs_locations *locations,
7351                             struct page *page, struct rpc_cred *cred)
7352 {
7353         struct nfs_server *server = NFS_SERVER(inode);
7354         struct nfs_client *clp = server->nfs_client;
7355         const struct nfs4_mig_recovery_ops *ops =
7356                                         clp->cl_mvops->mig_recovery_ops;
7357         struct nfs4_exception exception = { };
7358         int status;
7359
7360         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7361                 (unsigned long long)server->fsid.major,
7362                 (unsigned long long)server->fsid.minor,
7363                 clp->cl_hostname);
7364         nfs_display_fhandle(NFS_FH(inode), __func__);
7365
7366         do {
7367                 status = ops->get_locations(inode, locations, page, cred);
7368                 if (status != -NFS4ERR_DELAY)
7369                         break;
7370                 nfs4_handle_exception(server, status, &exception);
7371         } while (exception.retry);
7372         return status;
7373 }
7374
7375 /*
7376  * This operation also signals the server that this client is
7377  * performing "lease moved" recovery.  The server can stop
7378  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7379  * is appended to this compound to identify the client ID which is
7380  * performing recovery.
7381  */
7382 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7383 {
7384         struct nfs_server *server = NFS_SERVER(inode);
7385         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7386         struct rpc_clnt *clnt = server->client;
7387         struct nfs4_fsid_present_arg args = {
7388                 .fh             = NFS_FH(inode),
7389                 .clientid       = clp->cl_clientid,
7390                 .renew          = 1,            /* append RENEW */
7391         };
7392         struct nfs4_fsid_present_res res = {
7393                 .renew          = 1,
7394         };
7395         struct rpc_message msg = {
7396                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7397                 .rpc_argp       = &args,
7398                 .rpc_resp       = &res,
7399                 .rpc_cred       = cred,
7400         };
7401         unsigned long now = jiffies;
7402         int status;
7403
7404         res.fh = nfs_alloc_fhandle();
7405         if (res.fh == NULL)
7406                 return -ENOMEM;
7407
7408         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7409         status = nfs4_call_sync_sequence(clnt, server, &msg,
7410                                                 &args.seq_args, &res.seq_res);
7411         nfs_free_fhandle(res.fh);
7412         if (status)
7413                 return status;
7414
7415         do_renew_lease(clp, now);
7416         return 0;
7417 }
7418
7419 #ifdef CONFIG_NFS_V4_1
7420
7421 /*
7422  * This operation also signals the server that this client is
7423  * performing "lease moved" recovery.  The server can stop asserting
7424  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7425  * this operation is identified in the SEQUENCE operation in this
7426  * compound.
7427  */
7428 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7429 {
7430         struct nfs_server *server = NFS_SERVER(inode);
7431         struct rpc_clnt *clnt = server->client;
7432         struct nfs4_fsid_present_arg args = {
7433                 .fh             = NFS_FH(inode),
7434         };
7435         struct nfs4_fsid_present_res res = {
7436         };
7437         struct rpc_message msg = {
7438                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7439                 .rpc_argp       = &args,
7440                 .rpc_resp       = &res,
7441                 .rpc_cred       = cred,
7442         };
7443         int status;
7444
7445         res.fh = nfs_alloc_fhandle();
7446         if (res.fh == NULL)
7447                 return -ENOMEM;
7448
7449         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7450         status = nfs4_call_sync_sequence(clnt, server, &msg,
7451                                                 &args.seq_args, &res.seq_res);
7452         nfs_free_fhandle(res.fh);
7453         if (status == NFS4_OK &&
7454             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7455                 status = -NFS4ERR_LEASE_MOVED;
7456         return status;
7457 }
7458
7459 #endif  /* CONFIG_NFS_V4_1 */
7460
7461 /**
7462  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7463  * @inode: inode on FSID to check
7464  * @cred: credential to use for this operation
7465  *
7466  * Server indicates whether the FSID is present, moved, or not
7467  * recognized.  This operation is necessary to clear a LEASE_MOVED
7468  * condition for this client ID.
7469  *
7470  * Returns NFS4_OK if the FSID is present on this server,
7471  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7472  *  NFS4ERR code if some error occurred on the server, or a
7473  *  negative errno if a local failure occurred.
7474  */
7475 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7476 {
7477         struct nfs_server *server = NFS_SERVER(inode);
7478         struct nfs_client *clp = server->nfs_client;
7479         const struct nfs4_mig_recovery_ops *ops =
7480                                         clp->cl_mvops->mig_recovery_ops;
7481         struct nfs4_exception exception = { };
7482         int status;
7483
7484         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7485                 (unsigned long long)server->fsid.major,
7486                 (unsigned long long)server->fsid.minor,
7487                 clp->cl_hostname);
7488         nfs_display_fhandle(NFS_FH(inode), __func__);
7489
7490         do {
7491                 status = ops->fsid_present(inode, cred);
7492                 if (status != -NFS4ERR_DELAY)
7493                         break;
7494                 nfs4_handle_exception(server, status, &exception);
7495         } while (exception.retry);
7496         return status;
7497 }
7498
7499 /**
7500  * If 'use_integrity' is true and the state managment nfs_client
7501  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7502  * and the machine credential as per RFC3530bis and RFC5661 Security
7503  * Considerations sections. Otherwise, just use the user cred with the
7504  * filesystem's rpc_client.
7505  */
7506 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7507 {
7508         int status;
7509         struct nfs4_secinfo_arg args = {
7510                 .dir_fh = NFS_FH(dir),
7511                 .name   = name,
7512         };
7513         struct nfs4_secinfo_res res = {
7514                 .flavors     = flavors,
7515         };
7516         struct rpc_message msg = {
7517                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7518                 .rpc_argp = &args,
7519                 .rpc_resp = &res,
7520         };
7521         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7522         struct rpc_cred *cred = NULL;
7523
7524         if (use_integrity) {
7525                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7526                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7527                 msg.rpc_cred = cred;
7528         }
7529
7530         dprintk("NFS call  secinfo %s\n", name->name);
7531
7532         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7533                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7534
7535         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7536                                 &res.seq_res, 0);
7537         dprintk("NFS reply  secinfo: %d\n", status);
7538
7539         if (cred)
7540                 put_rpccred(cred);
7541
7542         return status;
7543 }
7544
7545 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7546                       struct nfs4_secinfo_flavors *flavors)
7547 {
7548         struct nfs4_exception exception = { };
7549         int err;
7550         do {
7551                 err = -NFS4ERR_WRONGSEC;
7552
7553                 /* try to use integrity protection with machine cred */
7554                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7555                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7556
7557                 /*
7558                  * if unable to use integrity protection, or SECINFO with
7559                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7560                  * disallowed by spec, but exists in deployed servers) use
7561                  * the current filesystem's rpc_client and the user cred.
7562                  */
7563                 if (err == -NFS4ERR_WRONGSEC)
7564                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7565
7566                 trace_nfs4_secinfo(dir, name, err);
7567                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7568                                 &exception);
7569         } while (exception.retry);
7570         return err;
7571 }
7572
7573 #ifdef CONFIG_NFS_V4_1
7574 /*
7575  * Check the exchange flags returned by the server for invalid flags, having
7576  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7577  * DS flags set.
7578  */
7579 static int nfs4_check_cl_exchange_flags(u32 flags)
7580 {
7581         if (flags & ~EXCHGID4_FLAG_MASK_R)
7582                 goto out_inval;
7583         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7584             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7585                 goto out_inval;
7586         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7587                 goto out_inval;
7588         return NFS_OK;
7589 out_inval:
7590         return -NFS4ERR_INVAL;
7591 }
7592
7593 static bool
7594 nfs41_same_server_scope(struct nfs41_server_scope *a,
7595                         struct nfs41_server_scope *b)
7596 {
7597         if (a->server_scope_sz != b->server_scope_sz)
7598                 return false;
7599         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7600 }
7601
7602 static void
7603 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7604 {
7605 }
7606
7607 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7608         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7609 };
7610
7611 /*
7612  * nfs4_proc_bind_one_conn_to_session()
7613  *
7614  * The 4.1 client currently uses the same TCP connection for the
7615  * fore and backchannel.
7616  */
7617 static
7618 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7619                 struct rpc_xprt *xprt,
7620                 struct nfs_client *clp,
7621                 struct rpc_cred *cred)
7622 {
7623         int status;
7624         struct nfs41_bind_conn_to_session_args args = {
7625                 .client = clp,
7626                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7627         };
7628         struct nfs41_bind_conn_to_session_res res;
7629         struct rpc_message msg = {
7630                 .rpc_proc =
7631                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7632                 .rpc_argp = &args,
7633                 .rpc_resp = &res,
7634                 .rpc_cred = cred,
7635         };
7636         struct rpc_task_setup task_setup_data = {
7637                 .rpc_client = clnt,
7638                 .rpc_xprt = xprt,
7639                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7640                 .rpc_message = &msg,
7641                 .flags = RPC_TASK_TIMEOUT,
7642         };
7643         struct rpc_task *task;
7644
7645         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7646         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7647                 args.dir = NFS4_CDFC4_FORE;
7648
7649         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7650         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7651                 args.dir = NFS4_CDFC4_FORE;
7652
7653         task = rpc_run_task(&task_setup_data);
7654         if (!IS_ERR(task)) {
7655                 status = task->tk_status;
7656                 rpc_put_task(task);
7657         } else
7658                 status = PTR_ERR(task);
7659         trace_nfs4_bind_conn_to_session(clp, status);
7660         if (status == 0) {
7661                 if (memcmp(res.sessionid.data,
7662                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7663                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7664                         return -EIO;
7665                 }
7666                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7667                         dprintk("NFS: %s: Unexpected direction from server\n",
7668                                 __func__);
7669                         return -EIO;
7670                 }
7671                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7672                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7673                                 __func__);
7674                         return -EIO;
7675                 }
7676         }
7677
7678         return status;
7679 }
7680
7681 struct rpc_bind_conn_calldata {
7682         struct nfs_client *clp;
7683         struct rpc_cred *cred;
7684 };
7685
7686 static int
7687 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7688                 struct rpc_xprt *xprt,
7689                 void *calldata)
7690 {
7691         struct rpc_bind_conn_calldata *p = calldata;
7692
7693         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7694 }
7695
7696 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7697 {
7698         struct rpc_bind_conn_calldata data = {
7699                 .clp = clp,
7700                 .cred = cred,
7701         };
7702         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7703                         nfs4_proc_bind_conn_to_session_callback, &data);
7704 }
7705
7706 /*
7707  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7708  * and operations we'd like to see to enable certain features in the allow map
7709  */
7710 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7711         .how = SP4_MACH_CRED,
7712         .enforce.u.words = {
7713                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7714                       1 << (OP_EXCHANGE_ID - 32) |
7715                       1 << (OP_CREATE_SESSION - 32) |
7716                       1 << (OP_DESTROY_SESSION - 32) |
7717                       1 << (OP_DESTROY_CLIENTID - 32)
7718         },
7719         .allow.u.words = {
7720                 [0] = 1 << (OP_CLOSE) |
7721                       1 << (OP_OPEN_DOWNGRADE) |
7722                       1 << (OP_LOCKU) |
7723                       1 << (OP_DELEGRETURN) |
7724                       1 << (OP_COMMIT),
7725                 [1] = 1 << (OP_SECINFO - 32) |
7726                       1 << (OP_SECINFO_NO_NAME - 32) |
7727                       1 << (OP_LAYOUTRETURN - 32) |
7728                       1 << (OP_TEST_STATEID - 32) |
7729                       1 << (OP_FREE_STATEID - 32) |
7730                       1 << (OP_WRITE - 32)
7731         }
7732 };
7733
7734 /*
7735  * Select the state protection mode for client `clp' given the server results
7736  * from exchange_id in `sp'.
7737  *
7738  * Returns 0 on success, negative errno otherwise.
7739  */
7740 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7741                                  struct nfs41_state_protection *sp)
7742 {
7743         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7744                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7745                       1 << (OP_EXCHANGE_ID - 32) |
7746                       1 << (OP_CREATE_SESSION - 32) |
7747                       1 << (OP_DESTROY_SESSION - 32) |
7748                       1 << (OP_DESTROY_CLIENTID - 32)
7749         };
7750         unsigned long flags = 0;
7751         unsigned int i;
7752         int ret = 0;
7753
7754         if (sp->how == SP4_MACH_CRED) {
7755                 /* Print state protect result */
7756                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7757                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7758                         if (test_bit(i, sp->enforce.u.longs))
7759                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7760                         if (test_bit(i, sp->allow.u.longs))
7761                                 dfprintk(MOUNT, "  allow op %d\n", i);
7762                 }
7763
7764                 /* make sure nothing is on enforce list that isn't supported */
7765                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7766                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7767                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7768                                 ret = -EINVAL;
7769                                 goto out;
7770                         }
7771                 }
7772
7773                 /*
7774                  * Minimal mode - state operations are allowed to use machine
7775                  * credential.  Note this already happens by default, so the
7776                  * client doesn't have to do anything more than the negotiation.
7777                  *
7778                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7779                  *       we're already using the machine cred for exchange_id
7780                  *       and will never use a different cred.
7781                  */
7782                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7783                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7784                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7785                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7786                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7787                         dfprintk(MOUNT, "  minimal mode enabled\n");
7788                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
7789                 } else {
7790                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7791                         ret = -EINVAL;
7792                         goto out;
7793                 }
7794
7795                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7796                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7797                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7798                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7799                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7800                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
7801                 }
7802
7803                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7804                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7805                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
7806                 }
7807
7808                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7809                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7810                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7811                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
7812                 }
7813
7814                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7815                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7816                         dfprintk(MOUNT, "  stateid mode enabled\n");
7817                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
7818                 }
7819
7820                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7821                         dfprintk(MOUNT, "  write mode enabled\n");
7822                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
7823                 }
7824
7825                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7826                         dfprintk(MOUNT, "  commit mode enabled\n");
7827                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
7828                 }
7829         }
7830 out:
7831         clp->cl_sp4_flags = flags;
7832         return ret;
7833 }
7834
7835 struct nfs41_exchange_id_data {
7836         struct nfs41_exchange_id_res res;
7837         struct nfs41_exchange_id_args args;
7838 };
7839
7840 static void nfs4_exchange_id_release(void *data)
7841 {
7842         struct nfs41_exchange_id_data *cdata =
7843                                         (struct nfs41_exchange_id_data *)data;
7844
7845         nfs_put_client(cdata->args.client);
7846         kfree(cdata->res.impl_id);
7847         kfree(cdata->res.server_scope);
7848         kfree(cdata->res.server_owner);
7849         kfree(cdata);
7850 }
7851
7852 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7853         .rpc_release = nfs4_exchange_id_release,
7854 };
7855
7856 /*
7857  * _nfs4_proc_exchange_id()
7858  *
7859  * Wrapper for EXCHANGE_ID operation.
7860  */
7861 static struct rpc_task *
7862 nfs4_run_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7863                         u32 sp4_how, struct rpc_xprt *xprt)
7864 {
7865         struct rpc_message msg = {
7866                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7867                 .rpc_cred = cred,
7868         };
7869         struct rpc_task_setup task_setup_data = {
7870                 .rpc_client = clp->cl_rpcclient,
7871                 .callback_ops = &nfs4_exchange_id_call_ops,
7872                 .rpc_message = &msg,
7873                 .flags = RPC_TASK_TIMEOUT,
7874         };
7875         struct nfs41_exchange_id_data *calldata;
7876         int status;
7877
7878         if (!refcount_inc_not_zero(&clp->cl_count))
7879                 return ERR_PTR(-EIO);
7880
7881         status = -ENOMEM;
7882         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7883         if (!calldata)
7884                 goto out;
7885
7886         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
7887
7888         status = nfs4_init_uniform_client_string(clp);
7889         if (status)
7890                 goto out_calldata;
7891
7892         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7893                                                 GFP_NOFS);
7894         status = -ENOMEM;
7895         if (unlikely(calldata->res.server_owner == NULL))
7896                 goto out_calldata;
7897
7898         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7899                                         GFP_NOFS);
7900         if (unlikely(calldata->res.server_scope == NULL))
7901                 goto out_server_owner;
7902
7903         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7904         if (unlikely(calldata->res.impl_id == NULL))
7905                 goto out_server_scope;
7906
7907         switch (sp4_how) {
7908         case SP4_NONE:
7909                 calldata->args.state_protect.how = SP4_NONE;
7910                 break;
7911
7912         case SP4_MACH_CRED:
7913                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7914                 break;
7915
7916         default:
7917                 /* unsupported! */
7918                 WARN_ON_ONCE(1);
7919                 status = -EINVAL;
7920                 goto out_impl_id;
7921         }
7922         if (xprt) {
7923                 task_setup_data.rpc_xprt = xprt;
7924                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
7925                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
7926                                 sizeof(calldata->args.verifier.data));
7927         }
7928         calldata->args.client = clp;
7929         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7930         EXCHGID4_FLAG_BIND_PRINC_STATEID;
7931 #ifdef CONFIG_NFS_V4_1_MIGRATION
7932         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
7933 #endif
7934         msg.rpc_argp = &calldata->args;
7935         msg.rpc_resp = &calldata->res;
7936         task_setup_data.callback_data = calldata;
7937
7938         return rpc_run_task(&task_setup_data);
7939
7940 out_impl_id:
7941         kfree(calldata->res.impl_id);
7942 out_server_scope:
7943         kfree(calldata->res.server_scope);
7944 out_server_owner:
7945         kfree(calldata->res.server_owner);
7946 out_calldata:
7947         kfree(calldata);
7948 out:
7949         nfs_put_client(clp);
7950         return ERR_PTR(status);
7951 }
7952
7953 /*
7954  * _nfs4_proc_exchange_id()
7955  *
7956  * Wrapper for EXCHANGE_ID operation.
7957  */
7958 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7959                         u32 sp4_how)
7960 {
7961         struct rpc_task *task;
7962         struct nfs41_exchange_id_args *argp;
7963         struct nfs41_exchange_id_res *resp;
7964         int status;
7965
7966         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
7967         if (IS_ERR(task))
7968                 return PTR_ERR(task);
7969
7970         argp = task->tk_msg.rpc_argp;
7971         resp = task->tk_msg.rpc_resp;
7972         status = task->tk_status;
7973         if (status  != 0)
7974                 goto out;
7975
7976         status = nfs4_check_cl_exchange_flags(resp->flags);
7977         if (status  != 0)
7978                 goto out;
7979
7980         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
7981         if (status != 0)
7982                 goto out;
7983
7984         clp->cl_clientid = resp->clientid;
7985         clp->cl_exchange_flags = resp->flags;
7986         clp->cl_seqid = resp->seqid;
7987         /* Client ID is not confirmed */
7988         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
7989                 clear_bit(NFS4_SESSION_ESTABLISHED,
7990                           &clp->cl_session->session_state);
7991
7992         if (clp->cl_serverscope != NULL &&
7993             !nfs41_same_server_scope(clp->cl_serverscope,
7994                                 resp->server_scope)) {
7995                 dprintk("%s: server_scope mismatch detected\n",
7996                         __func__);
7997                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7998         }
7999
8000         swap(clp->cl_serverowner, resp->server_owner);
8001         swap(clp->cl_serverscope, resp->server_scope);
8002         swap(clp->cl_implid, resp->impl_id);
8003
8004         /* Save the EXCHANGE_ID verifier session trunk tests */
8005         memcpy(clp->cl_confirm.data, argp->verifier.data,
8006                sizeof(clp->cl_confirm.data));
8007 out:
8008         trace_nfs4_exchange_id(clp, status);
8009         rpc_put_task(task);
8010         return status;
8011 }
8012
8013 /*
8014  * nfs4_proc_exchange_id()
8015  *
8016  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8017  *
8018  * Since the clientid has expired, all compounds using sessions
8019  * associated with the stale clientid will be returning
8020  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8021  * be in some phase of session reset.
8022  *
8023  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8024  */
8025 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
8026 {
8027         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8028         int status;
8029
8030         /* try SP4_MACH_CRED if krb5i/p */
8031         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8032             authflavor == RPC_AUTH_GSS_KRB5P) {
8033                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8034                 if (!status)
8035                         return 0;
8036         }
8037
8038         /* try SP4_NONE */
8039         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8040 }
8041
8042 /**
8043  * nfs4_test_session_trunk
8044  *
8045  * This is an add_xprt_test() test function called from
8046  * rpc_clnt_setup_test_and_add_xprt.
8047  *
8048  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8049  * and is dereferrenced in nfs4_exchange_id_release
8050  *
8051  * Upon success, add the new transport to the rpc_clnt
8052  *
8053  * @clnt: struct rpc_clnt to get new transport
8054  * @xprt: the rpc_xprt to test
8055  * @data: call data for _nfs4_proc_exchange_id.
8056  */
8057 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8058                             void *data)
8059 {
8060         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8061         struct rpc_task *task;
8062         int status;
8063
8064         u32 sp4_how;
8065
8066         dprintk("--> %s try %s\n", __func__,
8067                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8068
8069         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8070
8071         /* Test connection for session trunking. Async exchange_id call */
8072         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8073         if (IS_ERR(task))
8074                 return PTR_ERR(task);
8075
8076         status = task->tk_status;
8077         if (status == 0)
8078                 status = nfs4_detect_session_trunking(adata->clp,
8079                                 task->tk_msg.rpc_resp, xprt);
8080
8081         rpc_put_task(task);
8082         return status;
8083 }
8084 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8085
8086 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8087                 struct rpc_cred *cred)
8088 {
8089         struct rpc_message msg = {
8090                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8091                 .rpc_argp = clp,
8092                 .rpc_cred = cred,
8093         };
8094         int status;
8095
8096         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8097         trace_nfs4_destroy_clientid(clp, status);
8098         if (status)
8099                 dprintk("NFS: Got error %d from the server %s on "
8100                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8101         return status;
8102 }
8103
8104 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8105                 struct rpc_cred *cred)
8106 {
8107         unsigned int loop;
8108         int ret;
8109
8110         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8111                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8112                 switch (ret) {
8113                 case -NFS4ERR_DELAY:
8114                 case -NFS4ERR_CLIENTID_BUSY:
8115                         ssleep(1);
8116                         break;
8117                 default:
8118                         return ret;
8119                 }
8120         }
8121         return 0;
8122 }
8123
8124 int nfs4_destroy_clientid(struct nfs_client *clp)
8125 {
8126         struct rpc_cred *cred;
8127         int ret = 0;
8128
8129         if (clp->cl_mvops->minor_version < 1)
8130                 goto out;
8131         if (clp->cl_exchange_flags == 0)
8132                 goto out;
8133         if (clp->cl_preserve_clid)
8134                 goto out;
8135         cred = nfs4_get_clid_cred(clp);
8136         ret = nfs4_proc_destroy_clientid(clp, cred);
8137         if (cred)
8138                 put_rpccred(cred);
8139         switch (ret) {
8140         case 0:
8141         case -NFS4ERR_STALE_CLIENTID:
8142                 clp->cl_exchange_flags = 0;
8143         }
8144 out:
8145         return ret;
8146 }
8147
8148 struct nfs4_get_lease_time_data {
8149         struct nfs4_get_lease_time_args *args;
8150         struct nfs4_get_lease_time_res *res;
8151         struct nfs_client *clp;
8152 };
8153
8154 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8155                                         void *calldata)
8156 {
8157         struct nfs4_get_lease_time_data *data =
8158                         (struct nfs4_get_lease_time_data *)calldata;
8159
8160         dprintk("--> %s\n", __func__);
8161         /* just setup sequence, do not trigger session recovery
8162            since we're invoked within one */
8163         nfs4_setup_sequence(data->clp,
8164                         &data->args->la_seq_args,
8165                         &data->res->lr_seq_res,
8166                         task);
8167         dprintk("<-- %s\n", __func__);
8168 }
8169
8170 /*
8171  * Called from nfs4_state_manager thread for session setup, so don't recover
8172  * from sequence operation or clientid errors.
8173  */
8174 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8175 {
8176         struct nfs4_get_lease_time_data *data =
8177                         (struct nfs4_get_lease_time_data *)calldata;
8178
8179         dprintk("--> %s\n", __func__);
8180         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
8181                 return;
8182         switch (task->tk_status) {
8183         case -NFS4ERR_DELAY:
8184         case -NFS4ERR_GRACE:
8185                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8186                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8187                 task->tk_status = 0;
8188                 /* fall through */
8189         case -NFS4ERR_RETRY_UNCACHED_REP:
8190                 rpc_restart_call_prepare(task);
8191                 return;
8192         }
8193         dprintk("<-- %s\n", __func__);
8194 }
8195
8196 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8197         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8198         .rpc_call_done = nfs4_get_lease_time_done,
8199 };
8200
8201 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8202 {
8203         struct rpc_task *task;
8204         struct nfs4_get_lease_time_args args;
8205         struct nfs4_get_lease_time_res res = {
8206                 .lr_fsinfo = fsinfo,
8207         };
8208         struct nfs4_get_lease_time_data data = {
8209                 .args = &args,
8210                 .res = &res,
8211                 .clp = clp,
8212         };
8213         struct rpc_message msg = {
8214                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8215                 .rpc_argp = &args,
8216                 .rpc_resp = &res,
8217         };
8218         struct rpc_task_setup task_setup = {
8219                 .rpc_client = clp->cl_rpcclient,
8220                 .rpc_message = &msg,
8221                 .callback_ops = &nfs4_get_lease_time_ops,
8222                 .callback_data = &data,
8223                 .flags = RPC_TASK_TIMEOUT,
8224         };
8225         int status;
8226
8227         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8228         task = rpc_run_task(&task_setup);
8229
8230         if (IS_ERR(task))
8231                 return PTR_ERR(task);
8232
8233         status = task->tk_status;
8234         rpc_put_task(task);
8235         return status;
8236 }
8237
8238 /*
8239  * Initialize the values to be used by the client in CREATE_SESSION
8240  * If nfs4_init_session set the fore channel request and response sizes,
8241  * use them.
8242  *
8243  * Set the back channel max_resp_sz_cached to zero to force the client to
8244  * always set csa_cachethis to FALSE because the current implementation
8245  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8246  */
8247 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8248                                     struct rpc_clnt *clnt)
8249 {
8250         unsigned int max_rqst_sz, max_resp_sz;
8251         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8252
8253         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8254         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8255
8256         /* Fore channel attributes */
8257         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8258         args->fc_attrs.max_resp_sz = max_resp_sz;
8259         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8260         args->fc_attrs.max_reqs = max_session_slots;
8261
8262         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8263                 "max_ops=%u max_reqs=%u\n",
8264                 __func__,
8265                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8266                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8267
8268         /* Back channel attributes */
8269         args->bc_attrs.max_rqst_sz = max_bc_payload;
8270         args->bc_attrs.max_resp_sz = max_bc_payload;
8271         args->bc_attrs.max_resp_sz_cached = 0;
8272         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8273         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8274
8275         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8276                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8277                 __func__,
8278                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8279                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8280                 args->bc_attrs.max_reqs);
8281 }
8282
8283 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8284                 struct nfs41_create_session_res *res)
8285 {
8286         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8287         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8288
8289         if (rcvd->max_resp_sz > sent->max_resp_sz)
8290                 return -EINVAL;
8291         /*
8292          * Our requested max_ops is the minimum we need; we're not
8293          * prepared to break up compounds into smaller pieces than that.
8294          * So, no point even trying to continue if the server won't
8295          * cooperate:
8296          */
8297         if (rcvd->max_ops < sent->max_ops)
8298                 return -EINVAL;
8299         if (rcvd->max_reqs == 0)
8300                 return -EINVAL;
8301         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8302                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8303         return 0;
8304 }
8305
8306 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8307                 struct nfs41_create_session_res *res)
8308 {
8309         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8310         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8311
8312         if (!(res->flags & SESSION4_BACK_CHAN))
8313                 goto out;
8314         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8315                 return -EINVAL;
8316         if (rcvd->max_resp_sz < sent->max_resp_sz)
8317                 return -EINVAL;
8318         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8319                 return -EINVAL;
8320         if (rcvd->max_ops > sent->max_ops)
8321                 return -EINVAL;
8322         if (rcvd->max_reqs > sent->max_reqs)
8323                 return -EINVAL;
8324 out:
8325         return 0;
8326 }
8327
8328 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8329                                      struct nfs41_create_session_res *res)
8330 {
8331         int ret;
8332
8333         ret = nfs4_verify_fore_channel_attrs(args, res);
8334         if (ret)
8335                 return ret;
8336         return nfs4_verify_back_channel_attrs(args, res);
8337 }
8338
8339 static void nfs4_update_session(struct nfs4_session *session,
8340                 struct nfs41_create_session_res *res)
8341 {
8342         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8343         /* Mark client id and session as being confirmed */
8344         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8345         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8346         session->flags = res->flags;
8347         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8348         if (res->flags & SESSION4_BACK_CHAN)
8349                 memcpy(&session->bc_attrs, &res->bc_attrs,
8350                                 sizeof(session->bc_attrs));
8351 }
8352
8353 static int _nfs4_proc_create_session(struct nfs_client *clp,
8354                 struct rpc_cred *cred)
8355 {
8356         struct nfs4_session *session = clp->cl_session;
8357         struct nfs41_create_session_args args = {
8358                 .client = clp,
8359                 .clientid = clp->cl_clientid,
8360                 .seqid = clp->cl_seqid,
8361                 .cb_program = NFS4_CALLBACK,
8362         };
8363         struct nfs41_create_session_res res;
8364
8365         struct rpc_message msg = {
8366                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8367                 .rpc_argp = &args,
8368                 .rpc_resp = &res,
8369                 .rpc_cred = cred,
8370         };
8371         int status;
8372
8373         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8374         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8375
8376         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8377         trace_nfs4_create_session(clp, status);
8378
8379         switch (status) {
8380         case -NFS4ERR_STALE_CLIENTID:
8381         case -NFS4ERR_DELAY:
8382         case -ETIMEDOUT:
8383         case -EACCES:
8384         case -EAGAIN:
8385                 goto out;
8386         };
8387
8388         clp->cl_seqid++;
8389         if (!status) {
8390                 /* Verify the session's negotiated channel_attrs values */
8391                 status = nfs4_verify_channel_attrs(&args, &res);
8392                 /* Increment the clientid slot sequence id */
8393                 if (status)
8394                         goto out;
8395                 nfs4_update_session(session, &res);
8396         }
8397 out:
8398         return status;
8399 }
8400
8401 /*
8402  * Issues a CREATE_SESSION operation to the server.
8403  * It is the responsibility of the caller to verify the session is
8404  * expired before calling this routine.
8405  */
8406 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
8407 {
8408         int status;
8409         unsigned *ptr;
8410         struct nfs4_session *session = clp->cl_session;
8411
8412         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8413
8414         status = _nfs4_proc_create_session(clp, cred);
8415         if (status)
8416                 goto out;
8417
8418         /* Init or reset the session slot tables */
8419         status = nfs4_setup_session_slot_tables(session);
8420         dprintk("slot table setup returned %d\n", status);
8421         if (status)
8422                 goto out;
8423
8424         ptr = (unsigned *)&session->sess_id.data[0];
8425         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8426                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8427 out:
8428         dprintk("<-- %s\n", __func__);
8429         return status;
8430 }
8431
8432 /*
8433  * Issue the over-the-wire RPC DESTROY_SESSION.
8434  * The caller must serialize access to this routine.
8435  */
8436 int nfs4_proc_destroy_session(struct nfs4_session *session,
8437                 struct rpc_cred *cred)
8438 {
8439         struct rpc_message msg = {
8440                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8441                 .rpc_argp = session,
8442                 .rpc_cred = cred,
8443         };
8444         int status = 0;
8445
8446         dprintk("--> nfs4_proc_destroy_session\n");
8447
8448         /* session is still being setup */
8449         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8450                 return 0;
8451
8452         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8453         trace_nfs4_destroy_session(session->clp, status);
8454
8455         if (status)
8456                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8457                         "Session has been destroyed regardless...\n", status);
8458
8459         dprintk("<-- nfs4_proc_destroy_session\n");
8460         return status;
8461 }
8462
8463 /*
8464  * Renew the cl_session lease.
8465  */
8466 struct nfs4_sequence_data {
8467         struct nfs_client *clp;
8468         struct nfs4_sequence_args args;
8469         struct nfs4_sequence_res res;
8470 };
8471
8472 static void nfs41_sequence_release(void *data)
8473 {
8474         struct nfs4_sequence_data *calldata = data;
8475         struct nfs_client *clp = calldata->clp;
8476
8477         if (refcount_read(&clp->cl_count) > 1)
8478                 nfs4_schedule_state_renewal(clp);
8479         nfs_put_client(clp);
8480         kfree(calldata);
8481 }
8482
8483 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8484 {
8485         switch(task->tk_status) {
8486         case -NFS4ERR_DELAY:
8487                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8488                 return -EAGAIN;
8489         default:
8490                 nfs4_schedule_lease_recovery(clp);
8491         }
8492         return 0;
8493 }
8494
8495 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8496 {
8497         struct nfs4_sequence_data *calldata = data;
8498         struct nfs_client *clp = calldata->clp;
8499
8500         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8501                 return;
8502
8503         trace_nfs4_sequence(clp, task->tk_status);
8504         if (task->tk_status < 0) {
8505                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8506                 if (refcount_read(&clp->cl_count) == 1)
8507                         goto out;
8508
8509                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8510                         rpc_restart_call_prepare(task);
8511                         return;
8512                 }
8513         }
8514         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8515 out:
8516         dprintk("<-- %s\n", __func__);
8517 }
8518
8519 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8520 {
8521         struct nfs4_sequence_data *calldata = data;
8522         struct nfs_client *clp = calldata->clp;
8523         struct nfs4_sequence_args *args;
8524         struct nfs4_sequence_res *res;
8525
8526         args = task->tk_msg.rpc_argp;
8527         res = task->tk_msg.rpc_resp;
8528
8529         nfs4_setup_sequence(clp, args, res, task);
8530 }
8531
8532 static const struct rpc_call_ops nfs41_sequence_ops = {
8533         .rpc_call_done = nfs41_sequence_call_done,
8534         .rpc_call_prepare = nfs41_sequence_prepare,
8535         .rpc_release = nfs41_sequence_release,
8536 };
8537
8538 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8539                 struct rpc_cred *cred,
8540                 struct nfs4_slot *slot,
8541                 bool is_privileged)
8542 {
8543         struct nfs4_sequence_data *calldata;
8544         struct rpc_message msg = {
8545                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8546                 .rpc_cred = cred,
8547         };
8548         struct rpc_task_setup task_setup_data = {
8549                 .rpc_client = clp->cl_rpcclient,
8550                 .rpc_message = &msg,
8551                 .callback_ops = &nfs41_sequence_ops,
8552                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8553         };
8554         struct rpc_task *ret;
8555
8556         ret = ERR_PTR(-EIO);
8557         if (!refcount_inc_not_zero(&clp->cl_count))
8558                 goto out_err;
8559
8560         ret = ERR_PTR(-ENOMEM);
8561         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8562         if (calldata == NULL)
8563                 goto out_put_clp;
8564         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8565         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8566         msg.rpc_argp = &calldata->args;
8567         msg.rpc_resp = &calldata->res;
8568         calldata->clp = clp;
8569         task_setup_data.callback_data = calldata;
8570
8571         ret = rpc_run_task(&task_setup_data);
8572         if (IS_ERR(ret))
8573                 goto out_err;
8574         return ret;
8575 out_put_clp:
8576         nfs_put_client(clp);
8577 out_err:
8578         nfs41_release_slot(slot);
8579         return ret;
8580 }
8581
8582 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8583 {
8584         struct rpc_task *task;
8585         int ret = 0;
8586
8587         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8588                 return -EAGAIN;
8589         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8590         if (IS_ERR(task))
8591                 ret = PTR_ERR(task);
8592         else
8593                 rpc_put_task_async(task);
8594         dprintk("<-- %s status=%d\n", __func__, ret);
8595         return ret;
8596 }
8597
8598 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8599 {
8600         struct rpc_task *task;
8601         int ret;
8602
8603         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8604         if (IS_ERR(task)) {
8605                 ret = PTR_ERR(task);
8606                 goto out;
8607         }
8608         ret = rpc_wait_for_completion_task(task);
8609         if (!ret)
8610                 ret = task->tk_status;
8611         rpc_put_task(task);
8612 out:
8613         dprintk("<-- %s status=%d\n", __func__, ret);
8614         return ret;
8615 }
8616
8617 struct nfs4_reclaim_complete_data {
8618         struct nfs_client *clp;
8619         struct nfs41_reclaim_complete_args arg;
8620         struct nfs41_reclaim_complete_res res;
8621 };
8622
8623 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8624 {
8625         struct nfs4_reclaim_complete_data *calldata = data;
8626
8627         nfs4_setup_sequence(calldata->clp,
8628                         &calldata->arg.seq_args,
8629                         &calldata->res.seq_res,
8630                         task);
8631 }
8632
8633 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8634 {
8635         switch(task->tk_status) {
8636         case 0:
8637                 wake_up_all(&clp->cl_lock_waitq);
8638                 /* Fallthrough */
8639         case -NFS4ERR_COMPLETE_ALREADY:
8640         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8641                 break;
8642         case -NFS4ERR_DELAY:
8643                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8644                 /* fall through */
8645         case -NFS4ERR_RETRY_UNCACHED_REP:
8646                 return -EAGAIN;
8647         case -NFS4ERR_BADSESSION:
8648         case -NFS4ERR_DEADSESSION:
8649         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8650                 nfs4_schedule_session_recovery(clp->cl_session,
8651                                 task->tk_status);
8652                 break;
8653         default:
8654                 nfs4_schedule_lease_recovery(clp);
8655         }
8656         return 0;
8657 }
8658
8659 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8660 {
8661         struct nfs4_reclaim_complete_data *calldata = data;
8662         struct nfs_client *clp = calldata->clp;
8663         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8664
8665         dprintk("--> %s\n", __func__);
8666         if (!nfs41_sequence_done(task, res))
8667                 return;
8668
8669         trace_nfs4_reclaim_complete(clp, task->tk_status);
8670         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8671                 rpc_restart_call_prepare(task);
8672                 return;
8673         }
8674         dprintk("<-- %s\n", __func__);
8675 }
8676
8677 static void nfs4_free_reclaim_complete_data(void *data)
8678 {
8679         struct nfs4_reclaim_complete_data *calldata = data;
8680
8681         kfree(calldata);
8682 }
8683
8684 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8685         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8686         .rpc_call_done = nfs4_reclaim_complete_done,
8687         .rpc_release = nfs4_free_reclaim_complete_data,
8688 };
8689
8690 /*
8691  * Issue a global reclaim complete.
8692  */
8693 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8694                 struct rpc_cred *cred)
8695 {
8696         struct nfs4_reclaim_complete_data *calldata;
8697         struct rpc_task *task;
8698         struct rpc_message msg = {
8699                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8700                 .rpc_cred = cred,
8701         };
8702         struct rpc_task_setup task_setup_data = {
8703                 .rpc_client = clp->cl_rpcclient,
8704                 .rpc_message = &msg,
8705                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8706                 .flags = RPC_TASK_ASYNC,
8707         };
8708         int status = -ENOMEM;
8709
8710         dprintk("--> %s\n", __func__);
8711         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8712         if (calldata == NULL)
8713                 goto out;
8714         calldata->clp = clp;
8715         calldata->arg.one_fs = 0;
8716
8717         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
8718         msg.rpc_argp = &calldata->arg;
8719         msg.rpc_resp = &calldata->res;
8720         task_setup_data.callback_data = calldata;
8721         task = rpc_run_task(&task_setup_data);
8722         if (IS_ERR(task)) {
8723                 status = PTR_ERR(task);
8724                 goto out;
8725         }
8726         status = rpc_wait_for_completion_task(task);
8727         if (status == 0)
8728                 status = task->tk_status;
8729         rpc_put_task(task);
8730 out:
8731         dprintk("<-- %s status=%d\n", __func__, status);
8732         return status;
8733 }
8734
8735 static void
8736 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8737 {
8738         struct nfs4_layoutget *lgp = calldata;
8739         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8740
8741         dprintk("--> %s\n", __func__);
8742         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8743                                 &lgp->res.seq_res, task);
8744         dprintk("<-- %s\n", __func__);
8745 }
8746
8747 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8748 {
8749         struct nfs4_layoutget *lgp = calldata;
8750
8751         dprintk("--> %s\n", __func__);
8752         nfs41_sequence_process(task, &lgp->res.seq_res);
8753         dprintk("<-- %s\n", __func__);
8754 }
8755
8756 static int
8757 nfs4_layoutget_handle_exception(struct rpc_task *task,
8758                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8759 {
8760         struct inode *inode = lgp->args.inode;
8761         struct nfs_server *server = NFS_SERVER(inode);
8762         struct pnfs_layout_hdr *lo;
8763         int nfs4err = task->tk_status;
8764         int err, status = 0;
8765         LIST_HEAD(head);
8766
8767         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8768
8769         nfs4_sequence_free_slot(&lgp->res.seq_res);
8770
8771         switch (nfs4err) {
8772         case 0:
8773                 goto out;
8774
8775         /*
8776          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8777          * on the file. set tk_status to -ENODATA to tell upper layer to
8778          * retry go inband.
8779          */
8780         case -NFS4ERR_LAYOUTUNAVAILABLE:
8781                 status = -ENODATA;
8782                 goto out;
8783         /*
8784          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8785          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8786          */
8787         case -NFS4ERR_BADLAYOUT:
8788                 status = -EOVERFLOW;
8789                 goto out;
8790         /*
8791          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8792          * (or clients) writing to the same RAID stripe except when
8793          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8794          *
8795          * Treat it like we would RECALLCONFLICT -- we retry for a little
8796          * while, and then eventually give up.
8797          */
8798         case -NFS4ERR_LAYOUTTRYLATER:
8799                 if (lgp->args.minlength == 0) {
8800                         status = -EOVERFLOW;
8801                         goto out;
8802                 }
8803                 status = -EBUSY;
8804                 break;
8805         case -NFS4ERR_RECALLCONFLICT:
8806                 status = -ERECALLCONFLICT;
8807                 break;
8808         case -NFS4ERR_DELEG_REVOKED:
8809         case -NFS4ERR_ADMIN_REVOKED:
8810         case -NFS4ERR_EXPIRED:
8811         case -NFS4ERR_BAD_STATEID:
8812                 exception->timeout = 0;
8813                 spin_lock(&inode->i_lock);
8814                 lo = NFS_I(inode)->layout;
8815                 /* If the open stateid was bad, then recover it. */
8816                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8817                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
8818                         spin_unlock(&inode->i_lock);
8819                         exception->state = lgp->args.ctx->state;
8820                         exception->stateid = &lgp->args.stateid;
8821                         break;
8822                 }
8823
8824                 /*
8825                  * Mark the bad layout state as invalid, then retry
8826                  */
8827                 pnfs_mark_layout_stateid_invalid(lo, &head);
8828                 spin_unlock(&inode->i_lock);
8829                 nfs_commit_inode(inode, 0);
8830                 pnfs_free_lseg_list(&head);
8831                 status = -EAGAIN;
8832                 goto out;
8833         }
8834
8835         err = nfs4_handle_exception(server, nfs4err, exception);
8836         if (!status) {
8837                 if (exception->retry)
8838                         status = -EAGAIN;
8839                 else
8840                         status = err;
8841         }
8842 out:
8843         dprintk("<-- %s\n", __func__);
8844         return status;
8845 }
8846
8847 size_t max_response_pages(struct nfs_server *server)
8848 {
8849         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8850         return nfs_page_array_len(0, max_resp_sz);
8851 }
8852
8853 static void nfs4_layoutget_release(void *calldata)
8854 {
8855         struct nfs4_layoutget *lgp = calldata;
8856
8857         dprintk("--> %s\n", __func__);
8858         nfs4_sequence_free_slot(&lgp->res.seq_res);
8859         pnfs_layoutget_free(lgp);
8860         dprintk("<-- %s\n", __func__);
8861 }
8862
8863 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8864         .rpc_call_prepare = nfs4_layoutget_prepare,
8865         .rpc_call_done = nfs4_layoutget_done,
8866         .rpc_release = nfs4_layoutget_release,
8867 };
8868
8869 struct pnfs_layout_segment *
8870 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
8871 {
8872         struct inode *inode = lgp->args.inode;
8873         struct nfs_server *server = NFS_SERVER(inode);
8874         struct rpc_task *task;
8875         struct rpc_message msg = {
8876                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8877                 .rpc_argp = &lgp->args,
8878                 .rpc_resp = &lgp->res,
8879                 .rpc_cred = lgp->cred,
8880         };
8881         struct rpc_task_setup task_setup_data = {
8882                 .rpc_client = server->client,
8883                 .rpc_message = &msg,
8884                 .callback_ops = &nfs4_layoutget_call_ops,
8885                 .callback_data = lgp,
8886                 .flags = RPC_TASK_ASYNC,
8887         };
8888         struct pnfs_layout_segment *lseg = NULL;
8889         struct nfs4_exception exception = {
8890                 .inode = inode,
8891                 .timeout = *timeout,
8892         };
8893         int status = 0;
8894
8895         dprintk("--> %s\n", __func__);
8896
8897         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8898         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8899
8900         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
8901
8902         task = rpc_run_task(&task_setup_data);
8903         if (IS_ERR(task))
8904                 return ERR_CAST(task);
8905         status = rpc_wait_for_completion_task(task);
8906         if (status != 0)
8907                 goto out;
8908
8909         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8910         if (task->tk_status < 0 || lgp->res.layoutp->len == 0) {
8911                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8912                 *timeout = exception.timeout;
8913         } else
8914                 lseg = pnfs_layout_process(lgp);
8915 out:
8916         trace_nfs4_layoutget(lgp->args.ctx,
8917                         &lgp->args.range,
8918                         &lgp->res.range,
8919                         &lgp->res.stateid,
8920                         status);
8921
8922         rpc_put_task(task);
8923         dprintk("<-- %s status=%d\n", __func__, status);
8924         if (status)
8925                 return ERR_PTR(status);
8926         return lseg;
8927 }
8928
8929 static void
8930 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8931 {
8932         struct nfs4_layoutreturn *lrp = calldata;
8933
8934         dprintk("--> %s\n", __func__);
8935         nfs4_setup_sequence(lrp->clp,
8936                         &lrp->args.seq_args,
8937                         &lrp->res.seq_res,
8938                         task);
8939         if (!pnfs_layout_is_valid(lrp->args.layout))
8940                 rpc_exit(task, 0);
8941 }
8942
8943 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8944 {
8945         struct nfs4_layoutreturn *lrp = calldata;
8946         struct nfs_server *server;
8947
8948         dprintk("--> %s\n", __func__);
8949
8950         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8951                 return;
8952
8953         server = NFS_SERVER(lrp->args.inode);
8954         switch (task->tk_status) {
8955         case -NFS4ERR_OLD_STATEID:
8956                 if (nfs4_layoutreturn_refresh_stateid(&lrp->args.stateid,
8957                                         &lrp->args.range,
8958                                         lrp->args.inode))
8959                         goto out_restart;
8960                 /* Fallthrough */
8961         default:
8962                 task->tk_status = 0;
8963                 /* Fallthrough */
8964         case 0:
8965                 break;
8966         case -NFS4ERR_DELAY:
8967                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8968                         break;
8969                 goto out_restart;
8970         }
8971         dprintk("<-- %s\n", __func__);
8972         return;
8973 out_restart:
8974         task->tk_status = 0;
8975         nfs4_sequence_free_slot(&lrp->res.seq_res);
8976         rpc_restart_call_prepare(task);
8977 }
8978
8979 static void nfs4_layoutreturn_release(void *calldata)
8980 {
8981         struct nfs4_layoutreturn *lrp = calldata;
8982         struct pnfs_layout_hdr *lo = lrp->args.layout;
8983
8984         dprintk("--> %s\n", __func__);
8985         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8986                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8987         nfs4_sequence_free_slot(&lrp->res.seq_res);
8988         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8989                 lrp->ld_private.ops->free(&lrp->ld_private);
8990         pnfs_put_layout_hdr(lrp->args.layout);
8991         nfs_iput_and_deactive(lrp->inode);
8992         kfree(calldata);
8993         dprintk("<-- %s\n", __func__);
8994 }
8995
8996 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8997         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8998         .rpc_call_done = nfs4_layoutreturn_done,
8999         .rpc_release = nfs4_layoutreturn_release,
9000 };
9001
9002 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9003 {
9004         struct rpc_task *task;
9005         struct rpc_message msg = {
9006                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9007                 .rpc_argp = &lrp->args,
9008                 .rpc_resp = &lrp->res,
9009                 .rpc_cred = lrp->cred,
9010         };
9011         struct rpc_task_setup task_setup_data = {
9012                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9013                 .rpc_message = &msg,
9014                 .callback_ops = &nfs4_layoutreturn_call_ops,
9015                 .callback_data = lrp,
9016         };
9017         int status = 0;
9018
9019         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9020                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9021                         &task_setup_data.rpc_client, &msg);
9022
9023         dprintk("--> %s\n", __func__);
9024         if (!sync) {
9025                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9026                 if (!lrp->inode) {
9027                         nfs4_layoutreturn_release(lrp);
9028                         return -EAGAIN;
9029                 }
9030                 task_setup_data.flags |= RPC_TASK_ASYNC;
9031         }
9032         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9033         task = rpc_run_task(&task_setup_data);
9034         if (IS_ERR(task))
9035                 return PTR_ERR(task);
9036         if (sync)
9037                 status = task->tk_status;
9038         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9039         dprintk("<-- %s status=%d\n", __func__, status);
9040         rpc_put_task(task);
9041         return status;
9042 }
9043
9044 static int
9045 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9046                 struct pnfs_device *pdev,
9047                 struct rpc_cred *cred)
9048 {
9049         struct nfs4_getdeviceinfo_args args = {
9050                 .pdev = pdev,
9051                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9052                         NOTIFY_DEVICEID4_DELETE,
9053         };
9054         struct nfs4_getdeviceinfo_res res = {
9055                 .pdev = pdev,
9056         };
9057         struct rpc_message msg = {
9058                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9059                 .rpc_argp = &args,
9060                 .rpc_resp = &res,
9061                 .rpc_cred = cred,
9062         };
9063         int status;
9064
9065         dprintk("--> %s\n", __func__);
9066         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9067         if (res.notification & ~args.notify_types)
9068                 dprintk("%s: unsupported notification\n", __func__);
9069         if (res.notification != args.notify_types)
9070                 pdev->nocache = 1;
9071
9072         dprintk("<-- %s status=%d\n", __func__, status);
9073
9074         return status;
9075 }
9076
9077 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9078                 struct pnfs_device *pdev,
9079                 struct rpc_cred *cred)
9080 {
9081         struct nfs4_exception exception = { };
9082         int err;
9083
9084         do {
9085                 err = nfs4_handle_exception(server,
9086                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9087                                         &exception);
9088         } while (exception.retry);
9089         return err;
9090 }
9091 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9092
9093 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9094 {
9095         struct nfs4_layoutcommit_data *data = calldata;
9096         struct nfs_server *server = NFS_SERVER(data->args.inode);
9097
9098         nfs4_setup_sequence(server->nfs_client,
9099                         &data->args.seq_args,
9100                         &data->res.seq_res,
9101                         task);
9102 }
9103
9104 static void
9105 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9106 {
9107         struct nfs4_layoutcommit_data *data = calldata;
9108         struct nfs_server *server = NFS_SERVER(data->args.inode);
9109
9110         if (!nfs41_sequence_done(task, &data->res.seq_res))
9111                 return;
9112
9113         switch (task->tk_status) { /* Just ignore these failures */
9114         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9115         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9116         case -NFS4ERR_BADLAYOUT:     /* no layout */
9117         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9118                 task->tk_status = 0;
9119         case 0:
9120                 break;
9121         default:
9122                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9123                         rpc_restart_call_prepare(task);
9124                         return;
9125                 }
9126         }
9127 }
9128
9129 static void nfs4_layoutcommit_release(void *calldata)
9130 {
9131         struct nfs4_layoutcommit_data *data = calldata;
9132
9133         pnfs_cleanup_layoutcommit(data);
9134         nfs_post_op_update_inode_force_wcc(data->args.inode,
9135                                            data->res.fattr);
9136         put_rpccred(data->cred);
9137         nfs_iput_and_deactive(data->inode);
9138         kfree(data);
9139 }
9140
9141 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9142         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9143         .rpc_call_done = nfs4_layoutcommit_done,
9144         .rpc_release = nfs4_layoutcommit_release,
9145 };
9146
9147 int
9148 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9149 {
9150         struct rpc_message msg = {
9151                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9152                 .rpc_argp = &data->args,
9153                 .rpc_resp = &data->res,
9154                 .rpc_cred = data->cred,
9155         };
9156         struct rpc_task_setup task_setup_data = {
9157                 .task = &data->task,
9158                 .rpc_client = NFS_CLIENT(data->args.inode),
9159                 .rpc_message = &msg,
9160                 .callback_ops = &nfs4_layoutcommit_ops,
9161                 .callback_data = data,
9162         };
9163         struct rpc_task *task;
9164         int status = 0;
9165
9166         dprintk("NFS: initiating layoutcommit call. sync %d "
9167                 "lbw: %llu inode %lu\n", sync,
9168                 data->args.lastbytewritten,
9169                 data->args.inode->i_ino);
9170
9171         if (!sync) {
9172                 data->inode = nfs_igrab_and_active(data->args.inode);
9173                 if (data->inode == NULL) {
9174                         nfs4_layoutcommit_release(data);
9175                         return -EAGAIN;
9176                 }
9177                 task_setup_data.flags = RPC_TASK_ASYNC;
9178         }
9179         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9180         task = rpc_run_task(&task_setup_data);
9181         if (IS_ERR(task))
9182                 return PTR_ERR(task);
9183         if (sync)
9184                 status = task->tk_status;
9185         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9186         dprintk("%s: status %d\n", __func__, status);
9187         rpc_put_task(task);
9188         return status;
9189 }
9190
9191 /**
9192  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9193  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9194  */
9195 static int
9196 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9197                     struct nfs_fsinfo *info,
9198                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9199 {
9200         struct nfs41_secinfo_no_name_args args = {
9201                 .style = SECINFO_STYLE_CURRENT_FH,
9202         };
9203         struct nfs4_secinfo_res res = {
9204                 .flavors = flavors,
9205         };
9206         struct rpc_message msg = {
9207                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9208                 .rpc_argp = &args,
9209                 .rpc_resp = &res,
9210         };
9211         struct rpc_clnt *clnt = server->client;
9212         struct rpc_cred *cred = NULL;
9213         int status;
9214
9215         if (use_integrity) {
9216                 clnt = server->nfs_client->cl_rpcclient;
9217                 cred = nfs4_get_clid_cred(server->nfs_client);
9218                 msg.rpc_cred = cred;
9219         }
9220
9221         dprintk("--> %s\n", __func__);
9222         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
9223                                 &res.seq_res, 0);
9224         dprintk("<-- %s status=%d\n", __func__, status);
9225
9226         if (cred)
9227                 put_rpccred(cred);
9228
9229         return status;
9230 }
9231
9232 static int
9233 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9234                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9235 {
9236         struct nfs4_exception exception = { };
9237         int err;
9238         do {
9239                 /* first try using integrity protection */
9240                 err = -NFS4ERR_WRONGSEC;
9241
9242                 /* try to use integrity protection with machine cred */
9243                 if (_nfs4_is_integrity_protected(server->nfs_client))
9244                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9245                                                           flavors, true);
9246
9247                 /*
9248                  * if unable to use integrity protection, or SECINFO with
9249                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9250                  * disallowed by spec, but exists in deployed servers) use
9251                  * the current filesystem's rpc_client and the user cred.
9252                  */
9253                 if (err == -NFS4ERR_WRONGSEC)
9254                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9255                                                           flavors, false);
9256
9257                 switch (err) {
9258                 case 0:
9259                 case -NFS4ERR_WRONGSEC:
9260                 case -ENOTSUPP:
9261                         goto out;
9262                 default:
9263                         err = nfs4_handle_exception(server, err, &exception);
9264                 }
9265         } while (exception.retry);
9266 out:
9267         return err;
9268 }
9269
9270 static int
9271 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9272                     struct nfs_fsinfo *info)
9273 {
9274         int err;
9275         struct page *page;
9276         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9277         struct nfs4_secinfo_flavors *flavors;
9278         struct nfs4_secinfo4 *secinfo;
9279         int i;
9280
9281         page = alloc_page(GFP_KERNEL);
9282         if (!page) {
9283                 err = -ENOMEM;
9284                 goto out;
9285         }
9286
9287         flavors = page_address(page);
9288         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9289
9290         /*
9291          * Fall back on "guess and check" method if
9292          * the server doesn't support SECINFO_NO_NAME
9293          */
9294         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9295                 err = nfs4_find_root_sec(server, fhandle, info);
9296                 goto out_freepage;
9297         }
9298         if (err)
9299                 goto out_freepage;
9300
9301         for (i = 0; i < flavors->num_flavors; i++) {
9302                 secinfo = &flavors->flavors[i];
9303
9304                 switch (secinfo->flavor) {
9305                 case RPC_AUTH_NULL:
9306                 case RPC_AUTH_UNIX:
9307                 case RPC_AUTH_GSS:
9308                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9309                                         &secinfo->flavor_info);
9310                         break;
9311                 default:
9312                         flavor = RPC_AUTH_MAXFLAVOR;
9313                         break;
9314                 }
9315
9316                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9317                         flavor = RPC_AUTH_MAXFLAVOR;
9318
9319                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9320                         err = nfs4_lookup_root_sec(server, fhandle,
9321                                                    info, flavor);
9322                         if (!err)
9323                                 break;
9324                 }
9325         }
9326
9327         if (flavor == RPC_AUTH_MAXFLAVOR)
9328                 err = -EPERM;
9329
9330 out_freepage:
9331         put_page(page);
9332         if (err == -EACCES)
9333                 return -EPERM;
9334 out:
9335         return err;
9336 }
9337
9338 static int _nfs41_test_stateid(struct nfs_server *server,
9339                 nfs4_stateid *stateid,
9340                 struct rpc_cred *cred)
9341 {
9342         int status;
9343         struct nfs41_test_stateid_args args = {
9344                 .stateid = stateid,
9345         };
9346         struct nfs41_test_stateid_res res;
9347         struct rpc_message msg = {
9348                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9349                 .rpc_argp = &args,
9350                 .rpc_resp = &res,
9351                 .rpc_cred = cred,
9352         };
9353         struct rpc_clnt *rpc_client = server->client;
9354
9355         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9356                 &rpc_client, &msg);
9357
9358         dprintk("NFS call  test_stateid %p\n", stateid);
9359         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9360         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9361                         &args.seq_args, &res.seq_res);
9362         if (status != NFS_OK) {
9363                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9364                 return status;
9365         }
9366         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9367         return -res.status;
9368 }
9369
9370 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9371                 int err, struct nfs4_exception *exception)
9372 {
9373         exception->retry = 0;
9374         switch(err) {
9375         case -NFS4ERR_DELAY:
9376         case -NFS4ERR_RETRY_UNCACHED_REP:
9377                 nfs4_handle_exception(server, err, exception);
9378                 break;
9379         case -NFS4ERR_BADSESSION:
9380         case -NFS4ERR_BADSLOT:
9381         case -NFS4ERR_BAD_HIGH_SLOT:
9382         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9383         case -NFS4ERR_DEADSESSION:
9384                 nfs4_do_handle_exception(server, err, exception);
9385         }
9386 }
9387
9388 /**
9389  * nfs41_test_stateid - perform a TEST_STATEID operation
9390  *
9391  * @server: server / transport on which to perform the operation
9392  * @stateid: state ID to test
9393  * @cred: credential
9394  *
9395  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9396  * Otherwise a negative NFS4ERR value is returned if the operation
9397  * failed or the state ID is not currently valid.
9398  */
9399 static int nfs41_test_stateid(struct nfs_server *server,
9400                 nfs4_stateid *stateid,
9401                 struct rpc_cred *cred)
9402 {
9403         struct nfs4_exception exception = { };
9404         int err;
9405         do {
9406                 err = _nfs41_test_stateid(server, stateid, cred);
9407                 nfs4_handle_delay_or_session_error(server, err, &exception);
9408         } while (exception.retry);
9409         return err;
9410 }
9411
9412 struct nfs_free_stateid_data {
9413         struct nfs_server *server;
9414         struct nfs41_free_stateid_args args;
9415         struct nfs41_free_stateid_res res;
9416 };
9417
9418 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9419 {
9420         struct nfs_free_stateid_data *data = calldata;
9421         nfs4_setup_sequence(data->server->nfs_client,
9422                         &data->args.seq_args,
9423                         &data->res.seq_res,
9424                         task);
9425 }
9426
9427 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9428 {
9429         struct nfs_free_stateid_data *data = calldata;
9430
9431         nfs41_sequence_done(task, &data->res.seq_res);
9432
9433         switch (task->tk_status) {
9434         case -NFS4ERR_DELAY:
9435                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9436                         rpc_restart_call_prepare(task);
9437         }
9438 }
9439
9440 static void nfs41_free_stateid_release(void *calldata)
9441 {
9442         kfree(calldata);
9443 }
9444
9445 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9446         .rpc_call_prepare = nfs41_free_stateid_prepare,
9447         .rpc_call_done = nfs41_free_stateid_done,
9448         .rpc_release = nfs41_free_stateid_release,
9449 };
9450
9451 /**
9452  * nfs41_free_stateid - perform a FREE_STATEID operation
9453  *
9454  * @server: server / transport on which to perform the operation
9455  * @stateid: state ID to release
9456  * @cred: credential
9457  * @is_recovery: set to true if this call needs to be privileged
9458  *
9459  * Note: this function is always asynchronous.
9460  */
9461 static int nfs41_free_stateid(struct nfs_server *server,
9462                 const nfs4_stateid *stateid,
9463                 struct rpc_cred *cred,
9464                 bool privileged)
9465 {
9466         struct rpc_message msg = {
9467                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9468                 .rpc_cred = cred,
9469         };
9470         struct rpc_task_setup task_setup = {
9471                 .rpc_client = server->client,
9472                 .rpc_message = &msg,
9473                 .callback_ops = &nfs41_free_stateid_ops,
9474                 .flags = RPC_TASK_ASYNC,
9475         };
9476         struct nfs_free_stateid_data *data;
9477         struct rpc_task *task;
9478
9479         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9480                 &task_setup.rpc_client, &msg);
9481
9482         dprintk("NFS call  free_stateid %p\n", stateid);
9483         data = kmalloc(sizeof(*data), GFP_NOFS);
9484         if (!data)
9485                 return -ENOMEM;
9486         data->server = server;
9487         nfs4_stateid_copy(&data->args.stateid, stateid);
9488
9489         task_setup.callback_data = data;
9490
9491         msg.rpc_argp = &data->args;
9492         msg.rpc_resp = &data->res;
9493         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9494         task = rpc_run_task(&task_setup);
9495         if (IS_ERR(task))
9496                 return PTR_ERR(task);
9497         rpc_put_task(task);
9498         return 0;
9499 }
9500
9501 static void
9502 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9503 {
9504         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9505
9506         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9507         nfs4_free_lock_state(server, lsp);
9508 }
9509
9510 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9511                 const nfs4_stateid *s2)
9512 {
9513         if (s1->type != s2->type)
9514                 return false;
9515
9516         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9517                 return false;
9518
9519         if (s1->seqid == s2->seqid)
9520                 return true;
9521
9522         return s1->seqid == 0 || s2->seqid == 0;
9523 }
9524
9525 #endif /* CONFIG_NFS_V4_1 */
9526
9527 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9528                 const nfs4_stateid *s2)
9529 {
9530         return nfs4_stateid_match(s1, s2);
9531 }
9532
9533
9534 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9535         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9536         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9537         .recover_open   = nfs4_open_reclaim,
9538         .recover_lock   = nfs4_lock_reclaim,
9539         .establish_clid = nfs4_init_clientid,
9540         .detect_trunking = nfs40_discover_server_trunking,
9541 };
9542
9543 #if defined(CONFIG_NFS_V4_1)
9544 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9545         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9546         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9547         .recover_open   = nfs4_open_reclaim,
9548         .recover_lock   = nfs4_lock_reclaim,
9549         .establish_clid = nfs41_init_clientid,
9550         .reclaim_complete = nfs41_proc_reclaim_complete,
9551         .detect_trunking = nfs41_discover_server_trunking,
9552 };
9553 #endif /* CONFIG_NFS_V4_1 */
9554
9555 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9556         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9557         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9558         .recover_open   = nfs40_open_expired,
9559         .recover_lock   = nfs4_lock_expired,
9560         .establish_clid = nfs4_init_clientid,
9561 };
9562
9563 #if defined(CONFIG_NFS_V4_1)
9564 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9565         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9566         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9567         .recover_open   = nfs41_open_expired,
9568         .recover_lock   = nfs41_lock_expired,
9569         .establish_clid = nfs41_init_clientid,
9570 };
9571 #endif /* CONFIG_NFS_V4_1 */
9572
9573 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9574         .sched_state_renewal = nfs4_proc_async_renew,
9575         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9576         .renew_lease = nfs4_proc_renew,
9577 };
9578
9579 #if defined(CONFIG_NFS_V4_1)
9580 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9581         .sched_state_renewal = nfs41_proc_async_sequence,
9582         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9583         .renew_lease = nfs4_proc_sequence,
9584 };
9585 #endif
9586
9587 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9588         .get_locations = _nfs40_proc_get_locations,
9589         .fsid_present = _nfs40_proc_fsid_present,
9590 };
9591
9592 #if defined(CONFIG_NFS_V4_1)
9593 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9594         .get_locations = _nfs41_proc_get_locations,
9595         .fsid_present = _nfs41_proc_fsid_present,
9596 };
9597 #endif  /* CONFIG_NFS_V4_1 */
9598
9599 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9600         .minor_version = 0,
9601         .init_caps = NFS_CAP_READDIRPLUS
9602                 | NFS_CAP_ATOMIC_OPEN
9603                 | NFS_CAP_POSIX_LOCK,
9604         .init_client = nfs40_init_client,
9605         .shutdown_client = nfs40_shutdown_client,
9606         .match_stateid = nfs4_match_stateid,
9607         .find_root_sec = nfs4_find_root_sec,
9608         .free_lock_state = nfs4_release_lockowner,
9609         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9610         .alloc_seqid = nfs_alloc_seqid,
9611         .call_sync_ops = &nfs40_call_sync_ops,
9612         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9613         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9614         .state_renewal_ops = &nfs40_state_renewal_ops,
9615         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9616 };
9617
9618 #if defined(CONFIG_NFS_V4_1)
9619 static struct nfs_seqid *
9620 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9621 {
9622         return NULL;
9623 }
9624
9625 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9626         .minor_version = 1,
9627         .init_caps = NFS_CAP_READDIRPLUS
9628                 | NFS_CAP_ATOMIC_OPEN
9629                 | NFS_CAP_POSIX_LOCK
9630                 | NFS_CAP_STATEID_NFSV41
9631                 | NFS_CAP_ATOMIC_OPEN_V1
9632                 | NFS_CAP_LGOPEN,
9633         .init_client = nfs41_init_client,
9634         .shutdown_client = nfs41_shutdown_client,
9635         .match_stateid = nfs41_match_stateid,
9636         .find_root_sec = nfs41_find_root_sec,
9637         .free_lock_state = nfs41_free_lock_state,
9638         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9639         .alloc_seqid = nfs_alloc_no_seqid,
9640         .session_trunk = nfs4_test_session_trunk,
9641         .call_sync_ops = &nfs41_call_sync_ops,
9642         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9643         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9644         .state_renewal_ops = &nfs41_state_renewal_ops,
9645         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9646 };
9647 #endif
9648
9649 #if defined(CONFIG_NFS_V4_2)
9650 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9651         .minor_version = 2,
9652         .init_caps = NFS_CAP_READDIRPLUS
9653                 | NFS_CAP_ATOMIC_OPEN
9654                 | NFS_CAP_POSIX_LOCK
9655                 | NFS_CAP_STATEID_NFSV41
9656                 | NFS_CAP_ATOMIC_OPEN_V1
9657                 | NFS_CAP_LGOPEN
9658                 | NFS_CAP_ALLOCATE
9659                 | NFS_CAP_COPY
9660                 | NFS_CAP_OFFLOAD_CANCEL
9661                 | NFS_CAP_DEALLOCATE
9662                 | NFS_CAP_SEEK
9663                 | NFS_CAP_LAYOUTSTATS
9664                 | NFS_CAP_CLONE,
9665         .init_client = nfs41_init_client,
9666         .shutdown_client = nfs41_shutdown_client,
9667         .match_stateid = nfs41_match_stateid,
9668         .find_root_sec = nfs41_find_root_sec,
9669         .free_lock_state = nfs41_free_lock_state,
9670         .call_sync_ops = &nfs41_call_sync_ops,
9671         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9672         .alloc_seqid = nfs_alloc_no_seqid,
9673         .session_trunk = nfs4_test_session_trunk,
9674         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9675         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9676         .state_renewal_ops = &nfs41_state_renewal_ops,
9677         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9678 };
9679 #endif
9680
9681 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9682         [0] = &nfs_v4_0_minor_ops,
9683 #if defined(CONFIG_NFS_V4_1)
9684         [1] = &nfs_v4_1_minor_ops,
9685 #endif
9686 #if defined(CONFIG_NFS_V4_2)
9687         [2] = &nfs_v4_2_minor_ops,
9688 #endif
9689 };
9690
9691 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9692 {
9693         ssize_t error, error2;
9694
9695         error = generic_listxattr(dentry, list, size);
9696         if (error < 0)
9697                 return error;
9698         if (list) {
9699                 list += error;
9700                 size -= error;
9701         }
9702
9703         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9704         if (error2 < 0)
9705                 return error2;
9706         return error + error2;
9707 }
9708
9709 static const struct inode_operations nfs4_dir_inode_operations = {
9710         .create         = nfs_create,
9711         .lookup         = nfs_lookup,
9712         .atomic_open    = nfs_atomic_open,
9713         .link           = nfs_link,
9714         .unlink         = nfs_unlink,
9715         .symlink        = nfs_symlink,
9716         .mkdir          = nfs_mkdir,
9717         .rmdir          = nfs_rmdir,
9718         .mknod          = nfs_mknod,
9719         .rename         = nfs_rename,
9720         .permission     = nfs_permission,
9721         .getattr        = nfs_getattr,
9722         .setattr        = nfs_setattr,
9723         .listxattr      = nfs4_listxattr,
9724 };
9725
9726 static const struct inode_operations nfs4_file_inode_operations = {
9727         .permission     = nfs_permission,
9728         .getattr        = nfs_getattr,
9729         .setattr        = nfs_setattr,
9730         .listxattr      = nfs4_listxattr,
9731 };
9732
9733 const struct nfs_rpc_ops nfs_v4_clientops = {
9734         .version        = 4,                    /* protocol version */
9735         .dentry_ops     = &nfs4_dentry_operations,
9736         .dir_inode_ops  = &nfs4_dir_inode_operations,
9737         .file_inode_ops = &nfs4_file_inode_operations,
9738         .file_ops       = &nfs4_file_operations,
9739         .getroot        = nfs4_proc_get_root,
9740         .submount       = nfs4_submount,
9741         .try_mount      = nfs4_try_mount,
9742         .getattr        = nfs4_proc_getattr,
9743         .setattr        = nfs4_proc_setattr,
9744         .lookup         = nfs4_proc_lookup,
9745         .lookupp        = nfs4_proc_lookupp,
9746         .access         = nfs4_proc_access,
9747         .readlink       = nfs4_proc_readlink,
9748         .create         = nfs4_proc_create,
9749         .remove         = nfs4_proc_remove,
9750         .unlink_setup   = nfs4_proc_unlink_setup,
9751         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9752         .unlink_done    = nfs4_proc_unlink_done,
9753         .rename_setup   = nfs4_proc_rename_setup,
9754         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9755         .rename_done    = nfs4_proc_rename_done,
9756         .link           = nfs4_proc_link,
9757         .symlink        = nfs4_proc_symlink,
9758         .mkdir          = nfs4_proc_mkdir,
9759         .rmdir          = nfs4_proc_rmdir,
9760         .readdir        = nfs4_proc_readdir,
9761         .mknod          = nfs4_proc_mknod,
9762         .statfs         = nfs4_proc_statfs,
9763         .fsinfo         = nfs4_proc_fsinfo,
9764         .pathconf       = nfs4_proc_pathconf,
9765         .set_capabilities = nfs4_server_capabilities,
9766         .decode_dirent  = nfs4_decode_dirent,
9767         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9768         .read_setup     = nfs4_proc_read_setup,
9769         .read_done      = nfs4_read_done,
9770         .write_setup    = nfs4_proc_write_setup,
9771         .write_done     = nfs4_write_done,
9772         .commit_setup   = nfs4_proc_commit_setup,
9773         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9774         .commit_done    = nfs4_commit_done,
9775         .lock           = nfs4_proc_lock,
9776         .clear_acl_cache = nfs4_zap_acl_attr,
9777         .close_context  = nfs4_close_context,
9778         .open_context   = nfs4_atomic_open,
9779         .have_delegation = nfs4_have_delegation,
9780         .alloc_client   = nfs4_alloc_client,
9781         .init_client    = nfs4_init_client,
9782         .free_client    = nfs4_free_client,
9783         .create_server  = nfs4_create_server,
9784         .clone_server   = nfs_clone_server,
9785 };
9786
9787 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9788         .name   = XATTR_NAME_NFSV4_ACL,
9789         .list   = nfs4_xattr_list_nfs4_acl,
9790         .get    = nfs4_xattr_get_nfs4_acl,
9791         .set    = nfs4_xattr_set_nfs4_acl,
9792 };
9793
9794 const struct xattr_handler *nfs4_xattr_handlers[] = {
9795         &nfs4_xattr_nfs4_acl_handler,
9796 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9797         &nfs4_xattr_nfs4_label_handler,
9798 #endif
9799         NULL
9800 };
9801
9802 /*
9803  * Local variables:
9804  *  c-basic-offset: 8
9805  * End:
9806  */