fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57 #include <linux/iversion.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_BITMASK_SZ         3
  75
  76 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  77 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  78
  79 /* file attributes which can be mapped to nfs attributes */
  80 #define NFS4_VALID_ATTRS (ATTR_MODE \
  81         | ATTR_UID \
  82         | ATTR_GID \
  83         | ATTR_SIZE \
  84         | ATTR_ATIME \
  85         | ATTR_MTIME \
  86         | ATTR_CTIME \
  87         | ATTR_ATIME_SET \
  88         | ATTR_MTIME_SET)
  89
  90 struct nfs4_opendata;
  91 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  92 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  93 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  94 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label, struct inode *inode);
  95 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  96 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  97                             struct nfs_fattr *fattr, struct iattr *sattr,
  98                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  99                             struct nfs4_label *olabel);
 100 #ifdef CONFIG_NFS_V4_1
 101 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 102                 const struct cred *cred,
 103                 struct nfs4_slot *slot,
 104                 bool is_privileged);
 105 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 106                 const struct cred *);
 107 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 108                 const struct cred *, bool);
 109 #endif
 110
 111 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 112 static inline struct nfs4_label *
 113 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 114         struct iattr *sattr, struct nfs4_label *label)
 115 {
 116         int err;
 117
 118         if (label == NULL)
 119                 return NULL;
 120
 121         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 122                 return NULL;
 123
 124         err = security_dentry_init_security(dentry, sattr->ia_mode,
 125                                 &dentry->d_name, (void **)&label->label, &label->len);
 126         if (err == 0)
 127                 return label;
 128
 129         return NULL;
 130 }
 131 static inline void
 132 nfs4_label_release_security(struct nfs4_label *label)
 133 {
 134         if (label)
 135                 security_release_secctx(label->label, label->len);
 136 }
 137 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 138 {
 139         if (label)
 140                 return server->attr_bitmask;
 141
 142         return server->attr_bitmask_nl;
 143 }
 144 #else
 145 static inline struct nfs4_label *
 146 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 147         struct iattr *sattr, struct nfs4_label *l)
 148 { return NULL; }
 149 static inline void
 150 nfs4_label_release_security(struct nfs4_label *label)
 151 { return; }
 152 static inline u32 *
 153 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 154 { return server->attr_bitmask; }
 155 #endif
 156
 157 /* Prevent leaks of NFSv4 errors into userland */
 158 static int nfs4_map_errors(int err)
 159 {
 160         if (err >= -1000)
 161                 return err;
 162         switch (err) {
 163         case -NFS4ERR_RESOURCE:
 164         case -NFS4ERR_LAYOUTTRYLATER:
 165         case -NFS4ERR_RECALLCONFLICT:
 166                 return -EREMOTEIO;
 167         case -NFS4ERR_WRONGSEC:
 168         case -NFS4ERR_WRONG_CRED:
 169                 return -EPERM;
 170         case -NFS4ERR_BADOWNER:
 171         case -NFS4ERR_BADNAME:
 172                 return -EINVAL;
 173         case -NFS4ERR_SHARE_DENIED:
 174                 return -EACCES;
 175         case -NFS4ERR_MINOR_VERS_MISMATCH:
 176                 return -EPROTONOSUPPORT;
 177         case -NFS4ERR_FILE_OPEN:
 178                 return -EBUSY;
 179         default:
 180                 dprintk("%s could not handle NFSv4 error %d\n",
 181                                 __func__, -err);
 182                 break;
 183         }
 184         return -EIO;
 185 }
 186
 187 /*
 188  * This is our standard bitmap for GETATTR requests.
 189  */
 190 const u32 nfs4_fattr_bitmap[3] = {
 191         FATTR4_WORD0_TYPE
 192         | FATTR4_WORD0_CHANGE
 193         | FATTR4_WORD0_SIZE
 194         | FATTR4_WORD0_FSID
 195         | FATTR4_WORD0_FILEID,
 196         FATTR4_WORD1_MODE
 197         | FATTR4_WORD1_NUMLINKS
 198         | FATTR4_WORD1_OWNER
 199         | FATTR4_WORD1_OWNER_GROUP
 200         | FATTR4_WORD1_RAWDEV
 201         | FATTR4_WORD1_SPACE_USED
 202         | FATTR4_WORD1_TIME_ACCESS
 203         | FATTR4_WORD1_TIME_METADATA
 204         | FATTR4_WORD1_TIME_MODIFY
 205         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 206 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 207         FATTR4_WORD2_SECURITY_LABEL
 208 #endif
 209 };
 210
 211 static const u32 nfs4_pnfs_open_bitmap[3] = {
 212         FATTR4_WORD0_TYPE
 213         | FATTR4_WORD0_CHANGE
 214         | FATTR4_WORD0_SIZE
 215         | FATTR4_WORD0_FSID
 216         | FATTR4_WORD0_FILEID,
 217         FATTR4_WORD1_MODE
 218         | FATTR4_WORD1_NUMLINKS
 219         | FATTR4_WORD1_OWNER
 220         | FATTR4_WORD1_OWNER_GROUP
 221         | FATTR4_WORD1_RAWDEV
 222         | FATTR4_WORD1_SPACE_USED
 223         | FATTR4_WORD1_TIME_ACCESS
 224         | FATTR4_WORD1_TIME_METADATA
 225         | FATTR4_WORD1_TIME_MODIFY,
 226         FATTR4_WORD2_MDSTHRESHOLD
 227 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 228         | FATTR4_WORD2_SECURITY_LABEL
 229 #endif
 230 };
 231
 232 static const u32 nfs4_open_noattr_bitmap[3] = {
 233         FATTR4_WORD0_TYPE
 234         | FATTR4_WORD0_FILEID,
 235 };
 236
 237 const u32 nfs4_statfs_bitmap[3] = {
 238         FATTR4_WORD0_FILES_AVAIL
 239         | FATTR4_WORD0_FILES_FREE
 240         | FATTR4_WORD0_FILES_TOTAL,
 241         FATTR4_WORD1_SPACE_AVAIL
 242         | FATTR4_WORD1_SPACE_FREE
 243         | FATTR4_WORD1_SPACE_TOTAL
 244 };
 245
 246 const u32 nfs4_pathconf_bitmap[3] = {
 247         FATTR4_WORD0_MAXLINK
 248         | FATTR4_WORD0_MAXNAME,
 249         0
 250 };
 251
 252 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 253                         | FATTR4_WORD0_MAXREAD
 254                         | FATTR4_WORD0_MAXWRITE
 255                         | FATTR4_WORD0_LEASE_TIME,
 256                         FATTR4_WORD1_TIME_DELTA
 257                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 258                         FATTR4_WORD2_LAYOUT_BLKSIZE
 259                         | FATTR4_WORD2_CLONE_BLKSIZE
 260 };
 261
 262 const u32 nfs4_fs_locations_bitmap[3] = {
 263         FATTR4_WORD0_CHANGE
 264         | FATTR4_WORD0_SIZE
 265         | FATTR4_WORD0_FSID
 266         | FATTR4_WORD0_FILEID
 267         | FATTR4_WORD0_FS_LOCATIONS,
 268         FATTR4_WORD1_OWNER
 269         | FATTR4_WORD1_OWNER_GROUP
 270         | FATTR4_WORD1_RAWDEV
 271         | FATTR4_WORD1_SPACE_USED
 272         | FATTR4_WORD1_TIME_ACCESS
 273         | FATTR4_WORD1_TIME_METADATA
 274         | FATTR4_WORD1_TIME_MODIFY
 275         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 276 };
 277
 278 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 279                 struct inode *inode)
 280 {
 281         unsigned long cache_validity;
 282
 283         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 284         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 285                 return;
 286
 287         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 288         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 289                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 290                                 | NFS_INO_INVALID_SIZE);
 291
 292         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 293                 dst[0] &= ~FATTR4_WORD0_SIZE;
 294
 295         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 296                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 297 }
 298
 299 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 300                 const __u32 *src, struct inode *inode)
 301 {
 302         nfs4_bitmap_copy_adjust(dst, src, inode);
 303 }
 304
 305 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 306                 struct nfs4_readdir_arg *readdir)
 307 {
 308         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 309         __be32 *start, *p;
 310
 311         if (cookie > 2) {
 312                 readdir->cookie = cookie;
 313                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 314                 return;
 315         }
 316
 317         readdir->cookie = 0;
 318         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 319         if (cookie == 2)
 320                 return;
 321
 322         /*
 323          * NFSv4 servers do not return entries for '.' and '..'
 324          * Therefore, we fake these entries here.  We let '.'
 325          * have cookie 0 and '..' have cookie 1.  Note that
 326          * when talking to the server, we always send cookie 0
 327          * instead of 1 or 2.
 328          */
 329         start = p = kmap_atomic(*readdir->pages);
 330
 331         if (cookie == 0) {
 332                 *p++ = xdr_one;                                  /* next */
 333                 *p++ = xdr_zero;                   /* cookie, first word */
 334                 *p++ = xdr_one;                   /* cookie, second word */
 335                 *p++ = xdr_one;                             /* entry len */
 336                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 337                 p++;
 338                 *p++ = xdr_one;                         /* bitmap length */
 339                 *p++ = htonl(attrs);                           /* bitmap */
 340                 *p++ = htonl(12);             /* attribute buffer length */
 341                 *p++ = htonl(NF4DIR);
 342                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 343         }
 344
 345         *p++ = xdr_one;                                  /* next */
 346         *p++ = xdr_zero;                   /* cookie, first word */
 347         *p++ = xdr_two;                   /* cookie, second word */
 348         *p++ = xdr_two;                             /* entry len */
 349         memcpy(p, "..\0\0", 4);                         /* entry */
 350         p++;
 351         *p++ = xdr_one;                         /* bitmap length */
 352         *p++ = htonl(attrs);                           /* bitmap */
 353         *p++ = htonl(12);             /* attribute buffer length */
 354         *p++ = htonl(NF4DIR);
 355         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 356
 357         readdir->pgbase = (char *)p - (char *)start;
 358         readdir->count -= readdir->pgbase;
 359         kunmap_atomic(start);
 360 }
 361
 362 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 363                 nfs4_stateid *stateid,
 364                 const struct cred *cred)
 365 {
 366         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 367
 368         ops->test_and_free_expired(server, stateid, cred);
 369 }
 370
 371 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 372                 nfs4_stateid *stateid,
 373                 const struct cred *cred)
 374 {
 375         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 376         nfs4_test_and_free_stateid(server, stateid, cred);
 377 }
 378
 379 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 380                 const nfs4_stateid *stateid,
 381                 const struct cred *cred)
 382 {
 383         nfs4_stateid tmp;
 384
 385         nfs4_stateid_copy(&tmp, stateid);
 386         __nfs4_free_revoked_stateid(server, &tmp, cred);
 387 }
 388
 389 static long nfs4_update_delay(long *timeout)
 390 {
 391         long ret;
 392         if (!timeout)
 393                 return NFS4_POLL_RETRY_MAX;
 394         if (*timeout <= 0)
 395                 *timeout = NFS4_POLL_RETRY_MIN;
 396         if (*timeout > NFS4_POLL_RETRY_MAX)
 397                 *timeout = NFS4_POLL_RETRY_MAX;
 398         ret = *timeout;
 399         *timeout <<= 1;
 400         return ret;
 401 }
 402
 403 static int nfs4_delay(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, const struct 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, const struct 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,
1353                 int open_mode, enum open_claim_type4 claim)
1354 {
1355         int ret = 0;
1356
1357         if (open_mode & (O_EXCL|O_TRUNC))
1358                 goto out;
1359         switch (claim) {
1360         case NFS4_OPEN_CLAIM_NULL:
1361         case NFS4_OPEN_CLAIM_FH:
1362                 goto out;
1363         default:
1364                 break;
1365         }
1366         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1367                 case FMODE_READ:
1368                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1369                                 && state->n_rdonly != 0;
1370                         break;
1371                 case FMODE_WRITE:
1372                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1373                                 && state->n_wronly != 0;
1374                         break;
1375                 case FMODE_READ|FMODE_WRITE:
1376                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1377                                 && state->n_rdwr != 0;
1378         }
1379 out:
1380         return ret;
1381 }
1382
1383 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1384                 enum open_claim_type4 claim)
1385 {
1386         if (delegation == NULL)
1387                 return 0;
1388         if ((delegation->type & fmode) != fmode)
1389                 return 0;
1390         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1391                 return 0;
1392         switch (claim) {
1393         case NFS4_OPEN_CLAIM_NULL:
1394         case NFS4_OPEN_CLAIM_FH:
1395                 break;
1396         case NFS4_OPEN_CLAIM_PREVIOUS:
1397                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1398                         break;
1399                 /* Fall through */
1400         default:
1401                 return 0;
1402         }
1403         nfs_mark_delegation_referenced(delegation);
1404         return 1;
1405 }
1406
1407 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1408 {
1409         switch (fmode) {
1410                 case FMODE_WRITE:
1411                         state->n_wronly++;
1412                         break;
1413                 case FMODE_READ:
1414                         state->n_rdonly++;
1415                         break;
1416                 case FMODE_READ|FMODE_WRITE:
1417                         state->n_rdwr++;
1418         }
1419         nfs4_state_set_mode_locked(state, state->state | fmode);
1420 }
1421
1422 #ifdef CONFIG_NFS_V4_1
1423 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1424 {
1425         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1426                 return true;
1427         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1428                 return true;
1429         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1430                 return true;
1431         return false;
1432 }
1433 #endif /* CONFIG_NFS_V4_1 */
1434
1435 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1436 {
1437         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1438                 wake_up_all(&state->waitq);
1439 }
1440
1441 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1442                 const nfs4_stateid *stateid)
1443 {
1444         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1445         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1446
1447         if (stateid_seqid == state_seqid + 1U ||
1448             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1449                 nfs_state_log_update_open_stateid(state);
1450         else
1451                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1452 }
1453
1454 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1455 {
1456         struct nfs_client *clp = state->owner->so_server->nfs_client;
1457         bool need_recover = false;
1458
1459         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1460                 need_recover = true;
1461         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1462                 need_recover = true;
1463         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1464                 need_recover = true;
1465         if (need_recover)
1466                 nfs4_state_mark_reclaim_nograce(clp, state);
1467 }
1468
1469 /*
1470  * Check for whether or not the caller may update the open stateid
1471  * to the value passed in by stateid.
1472  *
1473  * Note: This function relies heavily on the server implementing
1474  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1475  * correctly.
1476  * i.e. The stateid seqids have to be initialised to 1, and
1477  * are then incremented on every state transition.
1478  */
1479 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1480                 const nfs4_stateid *stateid)
1481 {
1482         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1483             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1484                 if (stateid->seqid == cpu_to_be32(1))
1485                         nfs_state_log_update_open_stateid(state);
1486                 else
1487                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1488                 return true;
1489         }
1490
1491         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1492                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1493                 return true;
1494         }
1495         return false;
1496 }
1497
1498 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1499 {
1500         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1501                 return;
1502         if (state->n_wronly)
1503                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1504         if (state->n_rdonly)
1505                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1506         if (state->n_rdwr)
1507                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1508         set_bit(NFS_OPEN_STATE, &state->flags);
1509 }
1510
1511 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1512                 nfs4_stateid *stateid, fmode_t fmode)
1513 {
1514         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1515         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1516         case FMODE_WRITE:
1517                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1518                 break;
1519         case FMODE_READ:
1520                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1521                 break;
1522         case 0:
1523                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1524                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1525                 clear_bit(NFS_OPEN_STATE, &state->flags);
1526         }
1527         if (stateid == NULL)
1528                 return;
1529         /* Handle OPEN+OPEN_DOWNGRADE races */
1530         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1531             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1532                 nfs_resync_open_stateid_locked(state);
1533                 goto out;
1534         }
1535         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1536                 nfs4_stateid_copy(&state->stateid, stateid);
1537         nfs4_stateid_copy(&state->open_stateid, stateid);
1538         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1539 out:
1540         nfs_state_log_update_open_stateid(state);
1541 }
1542
1543 static void nfs_clear_open_stateid(struct nfs4_state *state,
1544         nfs4_stateid *arg_stateid,
1545         nfs4_stateid *stateid, fmode_t fmode)
1546 {
1547         write_seqlock(&state->seqlock);
1548         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1549         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1550                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1551         write_sequnlock(&state->seqlock);
1552         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1553                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1554 }
1555
1556 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1557                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1558 {
1559         DEFINE_WAIT(wait);
1560         int status = 0;
1561         for (;;) {
1562
1563                 if (!nfs_need_update_open_stateid(state, stateid))
1564                         return;
1565                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1566                         break;
1567                 if (status)
1568                         break;
1569                 /* Rely on seqids for serialisation with NFSv4.0 */
1570                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1571                         break;
1572
1573                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1574                 /*
1575                  * Ensure we process the state changes in the same order
1576                  * in which the server processed them by delaying the
1577                  * update of the stateid until we are in sequence.
1578                  */
1579                 write_sequnlock(&state->seqlock);
1580                 spin_unlock(&state->owner->so_lock);
1581                 rcu_read_unlock();
1582                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1583                 if (!signal_pending(current)) {
1584                         if (schedule_timeout(5*HZ) == 0)
1585                                 status = -EAGAIN;
1586                         else
1587                                 status = 0;
1588                 } else
1589                         status = -EINTR;
1590                 finish_wait(&state->waitq, &wait);
1591                 rcu_read_lock();
1592                 spin_lock(&state->owner->so_lock);
1593                 write_seqlock(&state->seqlock);
1594         }
1595
1596         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1597             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1598                 nfs4_stateid_copy(freeme, &state->open_stateid);
1599                 nfs_test_and_clear_all_open_stateid(state);
1600         }
1601
1602         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1603                 nfs4_stateid_copy(&state->stateid, stateid);
1604         nfs4_stateid_copy(&state->open_stateid, stateid);
1605         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1606         nfs_state_log_update_open_stateid(state);
1607 }
1608
1609 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1610                 const nfs4_stateid *open_stateid,
1611                 fmode_t fmode,
1612                 nfs4_stateid *freeme)
1613 {
1614         /*
1615          * Protect the call to nfs4_state_set_mode_locked and
1616          * serialise the stateid update
1617          */
1618         write_seqlock(&state->seqlock);
1619         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1620         switch (fmode) {
1621         case FMODE_READ:
1622                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1623                 break;
1624         case FMODE_WRITE:
1625                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1626                 break;
1627         case FMODE_READ|FMODE_WRITE:
1628                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1629         }
1630         set_bit(NFS_OPEN_STATE, &state->flags);
1631         write_sequnlock(&state->seqlock);
1632 }
1633
1634 static void nfs_state_set_delegation(struct nfs4_state *state,
1635                 const nfs4_stateid *deleg_stateid,
1636                 fmode_t fmode)
1637 {
1638         /*
1639          * Protect the call to nfs4_state_set_mode_locked and
1640          * serialise the stateid update
1641          */
1642         write_seqlock(&state->seqlock);
1643         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1644         set_bit(NFS_DELEGATED_STATE, &state->flags);
1645         write_sequnlock(&state->seqlock);
1646 }
1647
1648 static void nfs_state_clear_delegation(struct nfs4_state *state)
1649 {
1650         write_seqlock(&state->seqlock);
1651         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1652         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1653         write_sequnlock(&state->seqlock);
1654 }
1655
1656 static int update_open_stateid(struct nfs4_state *state,
1657                 const nfs4_stateid *open_stateid,
1658                 const nfs4_stateid *delegation,
1659                 fmode_t fmode)
1660 {
1661         struct nfs_server *server = NFS_SERVER(state->inode);
1662         struct nfs_client *clp = server->nfs_client;
1663         struct nfs_inode *nfsi = NFS_I(state->inode);
1664         struct nfs_delegation *deleg_cur;
1665         nfs4_stateid freeme = { };
1666         int ret = 0;
1667
1668         fmode &= (FMODE_READ|FMODE_WRITE);
1669
1670         rcu_read_lock();
1671         spin_lock(&state->owner->so_lock);
1672         if (open_stateid != NULL) {
1673                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1674                 ret = 1;
1675         }
1676
1677         deleg_cur = rcu_dereference(nfsi->delegation);
1678         if (deleg_cur == NULL)
1679                 goto no_delegation;
1680
1681         spin_lock(&deleg_cur->lock);
1682         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1683            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1684             (deleg_cur->type & fmode) != fmode)
1685                 goto no_delegation_unlock;
1686
1687         if (delegation == NULL)
1688                 delegation = &deleg_cur->stateid;
1689         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1690                 goto no_delegation_unlock;
1691
1692         nfs_mark_delegation_referenced(deleg_cur);
1693         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1694         ret = 1;
1695 no_delegation_unlock:
1696         spin_unlock(&deleg_cur->lock);
1697 no_delegation:
1698         if (ret)
1699                 update_open_stateflags(state, fmode);
1700         spin_unlock(&state->owner->so_lock);
1701         rcu_read_unlock();
1702
1703         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1704                 nfs4_schedule_state_manager(clp);
1705         if (freeme.type != 0)
1706                 nfs4_test_and_free_stateid(server, &freeme,
1707                                 state->owner->so_cred);
1708
1709         return ret;
1710 }
1711
1712 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1713                 const nfs4_stateid *stateid)
1714 {
1715         struct nfs4_state *state = lsp->ls_state;
1716         bool ret = false;
1717
1718         spin_lock(&state->state_lock);
1719         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1720                 goto out_noupdate;
1721         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1722                 goto out_noupdate;
1723         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1724         ret = true;
1725 out_noupdate:
1726         spin_unlock(&state->state_lock);
1727         return ret;
1728 }
1729
1730 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1731 {
1732         struct nfs_delegation *delegation;
1733
1734         fmode &= FMODE_READ|FMODE_WRITE;
1735         rcu_read_lock();
1736         delegation = rcu_dereference(NFS_I(inode)->delegation);
1737         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1738                 rcu_read_unlock();
1739                 return;
1740         }
1741         rcu_read_unlock();
1742         nfs4_inode_return_delegation(inode);
1743 }
1744
1745 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1746 {
1747         struct nfs4_state *state = opendata->state;
1748         struct nfs_inode *nfsi = NFS_I(state->inode);
1749         struct nfs_delegation *delegation;
1750         int open_mode = opendata->o_arg.open_flags;
1751         fmode_t fmode = opendata->o_arg.fmode;
1752         enum open_claim_type4 claim = opendata->o_arg.claim;
1753         nfs4_stateid stateid;
1754         int ret = -EAGAIN;
1755
1756         for (;;) {
1757                 spin_lock(&state->owner->so_lock);
1758                 if (can_open_cached(state, fmode, open_mode, claim)) {
1759                         update_open_stateflags(state, fmode);
1760                         spin_unlock(&state->owner->so_lock);
1761                         goto out_return_state;
1762                 }
1763                 spin_unlock(&state->owner->so_lock);
1764                 rcu_read_lock();
1765                 delegation = rcu_dereference(nfsi->delegation);
1766                 if (!can_open_delegated(delegation, fmode, claim)) {
1767                         rcu_read_unlock();
1768                         break;
1769                 }
1770                 /* Save the delegation */
1771                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1772                 rcu_read_unlock();
1773                 nfs_release_seqid(opendata->o_arg.seqid);
1774                 if (!opendata->is_recover) {
1775                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1776                         if (ret != 0)
1777                                 goto out;
1778                 }
1779                 ret = -EAGAIN;
1780
1781                 /* Try to update the stateid using the delegation */
1782                 if (update_open_stateid(state, NULL, &stateid, fmode))
1783                         goto out_return_state;
1784         }
1785 out:
1786         return ERR_PTR(ret);
1787 out_return_state:
1788         refcount_inc(&state->count);
1789         return state;
1790 }
1791
1792 static void
1793 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1794 {
1795         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1796         struct nfs_delegation *delegation;
1797         int delegation_flags = 0;
1798
1799         rcu_read_lock();
1800         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1801         if (delegation)
1802                 delegation_flags = delegation->flags;
1803         rcu_read_unlock();
1804         switch (data->o_arg.claim) {
1805         default:
1806                 break;
1807         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1808         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1809                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1810                                    "returning a delegation for "
1811                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1812                                    clp->cl_hostname);
1813                 return;
1814         }
1815         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1816                 nfs_inode_set_delegation(state->inode,
1817                                 data->owner->so_cred,
1818                                 data->o_res.delegation_type,
1819                                 &data->o_res.delegation,
1820                                 data->o_res.pagemod_limit);
1821         else
1822                 nfs_inode_reclaim_delegation(state->inode,
1823                                 data->owner->so_cred,
1824                                 data->o_res.delegation_type,
1825                                 &data->o_res.delegation,
1826                                 data->o_res.pagemod_limit);
1827
1828         if (data->o_res.do_recall)
1829                 nfs_async_inode_return_delegation(state->inode,
1830                                                   &data->o_res.delegation);
1831 }
1832
1833 /*
1834  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1835  * and update the nfs4_state.
1836  */
1837 static struct nfs4_state *
1838 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1839 {
1840         struct inode *inode = data->state->inode;
1841         struct nfs4_state *state = data->state;
1842         int ret;
1843
1844         if (!data->rpc_done) {
1845                 if (data->rpc_status)
1846                         return ERR_PTR(data->rpc_status);
1847                 /* cached opens have already been processed */
1848                 goto update;
1849         }
1850
1851         ret = nfs_refresh_inode(inode, &data->f_attr);
1852         if (ret)
1853                 return ERR_PTR(ret);
1854
1855         if (data->o_res.delegation_type != 0)
1856                 nfs4_opendata_check_deleg(data, state);
1857 update:
1858         update_open_stateid(state, &data->o_res.stateid, NULL,
1859                             data->o_arg.fmode);
1860         refcount_inc(&state->count);
1861
1862         return state;
1863 }
1864
1865 static struct inode *
1866 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1867 {
1868         struct inode *inode;
1869
1870         switch (data->o_arg.claim) {
1871         case NFS4_OPEN_CLAIM_NULL:
1872         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1873         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1874                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1875                         return ERR_PTR(-EAGAIN);
1876                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1877                                 &data->f_attr, data->f_label);
1878                 break;
1879         default:
1880                 inode = d_inode(data->dentry);
1881                 ihold(inode);
1882                 nfs_refresh_inode(inode, &data->f_attr);
1883         }
1884         return inode;
1885 }
1886
1887 static struct nfs4_state *
1888 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1889 {
1890         struct nfs4_state *state;
1891         struct inode *inode;
1892
1893         inode = nfs4_opendata_get_inode(data);
1894         if (IS_ERR(inode))
1895                 return ERR_CAST(inode);
1896         if (data->state != NULL && data->state->inode == inode) {
1897                 state = data->state;
1898                 refcount_inc(&state->count);
1899         } else
1900                 state = nfs4_get_open_state(inode, data->owner);
1901         iput(inode);
1902         if (state == NULL)
1903                 state = ERR_PTR(-ENOMEM);
1904         return state;
1905 }
1906
1907 static struct nfs4_state *
1908 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1909 {
1910         struct nfs4_state *state;
1911
1912         if (!data->rpc_done) {
1913                 state = nfs4_try_open_cached(data);
1914                 trace_nfs4_cached_open(data->state);
1915                 goto out;
1916         }
1917
1918         state = nfs4_opendata_find_nfs4_state(data);
1919         if (IS_ERR(state))
1920                 goto out;
1921
1922         if (data->o_res.delegation_type != 0)
1923                 nfs4_opendata_check_deleg(data, state);
1924         update_open_stateid(state, &data->o_res.stateid, NULL,
1925                         data->o_arg.fmode);
1926 out:
1927         nfs_release_seqid(data->o_arg.seqid);
1928         return state;
1929 }
1930
1931 static struct nfs4_state *
1932 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1933 {
1934         struct nfs4_state *ret;
1935
1936         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1937                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1938         else
1939                 ret = _nfs4_opendata_to_nfs4_state(data);
1940         nfs4_sequence_free_slot(&data->o_res.seq_res);
1941         return ret;
1942 }
1943
1944 static struct nfs_open_context *
1945 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
1946 {
1947         struct nfs_inode *nfsi = NFS_I(state->inode);
1948         struct nfs_open_context *ctx;
1949
1950         rcu_read_lock();
1951         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
1952                 if (ctx->state != state)
1953                         continue;
1954                 if ((ctx->mode & mode) != mode)
1955                         continue;
1956                 if (!get_nfs_open_context(ctx))
1957                         continue;
1958                 rcu_read_unlock();
1959                 return ctx;
1960         }
1961         rcu_read_unlock();
1962         return ERR_PTR(-ENOENT);
1963 }
1964
1965 static struct nfs_open_context *
1966 nfs4_state_find_open_context(struct nfs4_state *state)
1967 {
1968         struct nfs_open_context *ctx;
1969
1970         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
1971         if (!IS_ERR(ctx))
1972                 return ctx;
1973         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
1974         if (!IS_ERR(ctx))
1975                 return ctx;
1976         return nfs4_state_find_open_context_mode(state, FMODE_READ);
1977 }
1978
1979 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1980                 struct nfs4_state *state, enum open_claim_type4 claim)
1981 {
1982         struct nfs4_opendata *opendata;
1983
1984         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1985                         NULL, claim, GFP_NOFS);
1986         if (opendata == NULL)
1987                 return ERR_PTR(-ENOMEM);
1988         opendata->state = state;
1989         refcount_inc(&state->count);
1990         return opendata;
1991 }
1992
1993 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1994                 fmode_t fmode)
1995 {
1996         struct nfs4_state *newstate;
1997         int ret;
1998
1999         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2000                 return 0;
2001         opendata->o_arg.open_flags = 0;
2002         opendata->o_arg.fmode = fmode;
2003         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2004                         NFS_SB(opendata->dentry->d_sb),
2005                         fmode, 0);
2006         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2007         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2008         nfs4_init_opendata_res(opendata);
2009         ret = _nfs4_recover_proc_open(opendata);
2010         if (ret != 0)
2011                 return ret;
2012         newstate = nfs4_opendata_to_nfs4_state(opendata);
2013         if (IS_ERR(newstate))
2014                 return PTR_ERR(newstate);
2015         if (newstate != opendata->state)
2016                 ret = -ESTALE;
2017         nfs4_close_state(newstate, fmode);
2018         return ret;
2019 }
2020
2021 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2022 {
2023         int ret;
2024
2025         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
2026         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2027         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2028         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2029         /* memory barrier prior to reading state->n_* */
2030         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2031         clear_bit(NFS_OPEN_STATE, &state->flags);
2032         smp_rmb();
2033         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2034         if (ret != 0)
2035                 return ret;
2036         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2037         if (ret != 0)
2038                 return ret;
2039         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2040         if (ret != 0)
2041                 return ret;
2042         /*
2043          * We may have performed cached opens for all three recoveries.
2044          * Check if we need to update the current stateid.
2045          */
2046         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2047             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2048                 write_seqlock(&state->seqlock);
2049                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2050                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2051                 write_sequnlock(&state->seqlock);
2052         }
2053         return 0;
2054 }
2055
2056 /*
2057  * OPEN_RECLAIM:
2058  *      reclaim state on the server after a reboot.
2059  */
2060 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2061 {
2062         struct nfs_delegation *delegation;
2063         struct nfs4_opendata *opendata;
2064         fmode_t delegation_type = 0;
2065         int status;
2066
2067         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2068                         NFS4_OPEN_CLAIM_PREVIOUS);
2069         if (IS_ERR(opendata))
2070                 return PTR_ERR(opendata);
2071         rcu_read_lock();
2072         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2073         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2074                 delegation_type = delegation->type;
2075         rcu_read_unlock();
2076         opendata->o_arg.u.delegation_type = delegation_type;
2077         status = nfs4_open_recover(opendata, state);
2078         nfs4_opendata_put(opendata);
2079         return status;
2080 }
2081
2082 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2083 {
2084         struct nfs_server *server = NFS_SERVER(state->inode);
2085         struct nfs4_exception exception = { };
2086         int err;
2087         do {
2088                 err = _nfs4_do_open_reclaim(ctx, state);
2089                 trace_nfs4_open_reclaim(ctx, 0, err);
2090                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2091                         continue;
2092                 if (err != -NFS4ERR_DELAY)
2093                         break;
2094                 nfs4_handle_exception(server, err, &exception);
2095         } while (exception.retry);
2096         return err;
2097 }
2098
2099 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2100 {
2101         struct nfs_open_context *ctx;
2102         int ret;
2103
2104         ctx = nfs4_state_find_open_context(state);
2105         if (IS_ERR(ctx))
2106                 return -EAGAIN;
2107         ret = nfs4_do_open_reclaim(ctx, state);
2108         put_nfs_open_context(ctx);
2109         return ret;
2110 }
2111
2112 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)
2113 {
2114         switch (err) {
2115                 default:
2116                         printk(KERN_ERR "NFS: %s: unhandled error "
2117                                         "%d.\n", __func__, err);
2118                 case 0:
2119                 case -ENOENT:
2120                 case -EAGAIN:
2121                 case -ESTALE:
2122                         break;
2123                 case -NFS4ERR_BADSESSION:
2124                 case -NFS4ERR_BADSLOT:
2125                 case -NFS4ERR_BAD_HIGH_SLOT:
2126                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2127                 case -NFS4ERR_DEADSESSION:
2128                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2129                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
2130                         return -EAGAIN;
2131                 case -NFS4ERR_STALE_CLIENTID:
2132                 case -NFS4ERR_STALE_STATEID:
2133                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2134                         /* Don't recall a delegation if it was lost */
2135                         nfs4_schedule_lease_recovery(server->nfs_client);
2136                         return -EAGAIN;
2137                 case -NFS4ERR_MOVED:
2138                         nfs4_schedule_migration_recovery(server);
2139                         return -EAGAIN;
2140                 case -NFS4ERR_LEASE_MOVED:
2141                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2142                         return -EAGAIN;
2143                 case -NFS4ERR_DELEG_REVOKED:
2144                 case -NFS4ERR_ADMIN_REVOKED:
2145                 case -NFS4ERR_EXPIRED:
2146                 case -NFS4ERR_BAD_STATEID:
2147                 case -NFS4ERR_OPENMODE:
2148                         nfs_inode_find_state_and_recover(state->inode,
2149                                         stateid);
2150                         nfs4_schedule_stateid_recovery(server, state);
2151                         return -EAGAIN;
2152                 case -NFS4ERR_DELAY:
2153                 case -NFS4ERR_GRACE:
2154                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2155                         ssleep(1);
2156                         return -EAGAIN;
2157                 case -ENOMEM:
2158                 case -NFS4ERR_DENIED:
2159                         if (fl) {
2160                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2161                                 if (lsp)
2162                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2163                         }
2164                         return 0;
2165         }
2166         return err;
2167 }
2168
2169 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2170                 struct nfs4_state *state, const nfs4_stateid *stateid,
2171                 fmode_t type)
2172 {
2173         struct nfs_server *server = NFS_SERVER(state->inode);
2174         struct nfs4_opendata *opendata;
2175         int err = 0;
2176
2177         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2178                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2179         if (IS_ERR(opendata))
2180                 return PTR_ERR(opendata);
2181         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2182         nfs_state_clear_delegation(state);
2183         switch (type & (FMODE_READ|FMODE_WRITE)) {
2184         case FMODE_READ|FMODE_WRITE:
2185         case FMODE_WRITE:
2186                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2187                 if (err)
2188                         break;
2189                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2190                 if (err)
2191                         break;
2192                 /* Fall through */
2193         case FMODE_READ:
2194                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2195         }
2196         nfs4_opendata_put(opendata);
2197         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2198 }
2199
2200 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2201 {
2202         struct nfs4_opendata *data = calldata;
2203
2204         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2205                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2206 }
2207
2208 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2209 {
2210         struct nfs4_opendata *data = calldata;
2211
2212         nfs40_sequence_done(task, &data->c_res.seq_res);
2213
2214         data->rpc_status = task->tk_status;
2215         if (data->rpc_status == 0) {
2216                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2217                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2218                 renew_lease(data->o_res.server, data->timestamp);
2219                 data->rpc_done = true;
2220         }
2221 }
2222
2223 static void nfs4_open_confirm_release(void *calldata)
2224 {
2225         struct nfs4_opendata *data = calldata;
2226         struct nfs4_state *state = NULL;
2227
2228         /* If this request hasn't been cancelled, do nothing */
2229         if (!data->cancelled)
2230                 goto out_free;
2231         /* In case of error, no cleanup! */
2232         if (!data->rpc_done)
2233                 goto out_free;
2234         state = nfs4_opendata_to_nfs4_state(data);
2235         if (!IS_ERR(state))
2236                 nfs4_close_state(state, data->o_arg.fmode);
2237 out_free:
2238         nfs4_opendata_put(data);
2239 }
2240
2241 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2242         .rpc_call_prepare = nfs4_open_confirm_prepare,
2243         .rpc_call_done = nfs4_open_confirm_done,
2244         .rpc_release = nfs4_open_confirm_release,
2245 };
2246
2247 /*
2248  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2249  */
2250 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2251 {
2252         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2253         struct rpc_task *task;
2254         struct  rpc_message msg = {
2255                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2256                 .rpc_argp = &data->c_arg,
2257                 .rpc_resp = &data->c_res,
2258                 .rpc_cred = data->owner->so_cred,
2259         };
2260         struct rpc_task_setup task_setup_data = {
2261                 .rpc_client = server->client,
2262                 .rpc_message = &msg,
2263                 .callback_ops = &nfs4_open_confirm_ops,
2264                 .callback_data = data,
2265                 .workqueue = nfsiod_workqueue,
2266                 .flags = RPC_TASK_ASYNC,
2267         };
2268         int status;
2269
2270         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2271                                 data->is_recover);
2272         kref_get(&data->kref);
2273         data->rpc_done = false;
2274         data->rpc_status = 0;
2275         data->timestamp = jiffies;
2276         task = rpc_run_task(&task_setup_data);
2277         if (IS_ERR(task))
2278                 return PTR_ERR(task);
2279         status = rpc_wait_for_completion_task(task);
2280         if (status != 0) {
2281                 data->cancelled = true;
2282                 smp_wmb();
2283         } else
2284                 status = data->rpc_status;
2285         rpc_put_task(task);
2286         return status;
2287 }
2288
2289 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2290 {
2291         struct nfs4_opendata *data = calldata;
2292         struct nfs4_state_owner *sp = data->owner;
2293         struct nfs_client *clp = sp->so_server->nfs_client;
2294         enum open_claim_type4 claim = data->o_arg.claim;
2295
2296         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2297                 goto out_wait;
2298         /*
2299          * Check if we still need to send an OPEN call, or if we can use
2300          * a delegation instead.
2301          */
2302         if (data->state != NULL) {
2303                 struct nfs_delegation *delegation;
2304
2305                 if (can_open_cached(data->state, data->o_arg.fmode,
2306                                         data->o_arg.open_flags, claim))
2307                         goto out_no_action;
2308                 rcu_read_lock();
2309                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2310                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2311                         goto unlock_no_action;
2312                 rcu_read_unlock();
2313         }
2314         /* Update client id. */
2315         data->o_arg.clientid = clp->cl_clientid;
2316         switch (claim) {
2317         default:
2318                 break;
2319         case NFS4_OPEN_CLAIM_PREVIOUS:
2320         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2321         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2322                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2323                 /* Fall through */
2324         case NFS4_OPEN_CLAIM_FH:
2325                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2326         }
2327         data->timestamp = jiffies;
2328         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2329                                 &data->o_arg.seq_args,
2330                                 &data->o_res.seq_res,
2331                                 task) != 0)
2332                 nfs_release_seqid(data->o_arg.seqid);
2333
2334         /* Set the create mode (note dependency on the session type) */
2335         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2336         if (data->o_arg.open_flags & O_EXCL) {
2337                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2338                 if (nfs4_has_persistent_session(clp))
2339                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2340                 else if (clp->cl_mvops->minor_version > 0)
2341                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2342         }
2343         return;
2344 unlock_no_action:
2345         trace_nfs4_cached_open(data->state);
2346         rcu_read_unlock();
2347 out_no_action:
2348         task->tk_action = NULL;
2349 out_wait:
2350         nfs4_sequence_done(task, &data->o_res.seq_res);
2351 }
2352
2353 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2354 {
2355         struct nfs4_opendata *data = calldata;
2356
2357         data->rpc_status = task->tk_status;
2358
2359         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2360                 return;
2361
2362         if (task->tk_status == 0) {
2363                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2364                         switch (data->o_res.f_attr->mode & S_IFMT) {
2365                         case S_IFREG:
2366                                 break;
2367                         case S_IFLNK:
2368                                 data->rpc_status = -ELOOP;
2369                                 break;
2370                         case S_IFDIR:
2371                                 data->rpc_status = -EISDIR;
2372                                 break;
2373                         default:
2374                                 data->rpc_status = -ENOTDIR;
2375                         }
2376                 }
2377                 renew_lease(data->o_res.server, data->timestamp);
2378                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2379                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2380         }
2381         data->rpc_done = true;
2382 }
2383
2384 static void nfs4_open_release(void *calldata)
2385 {
2386         struct nfs4_opendata *data = calldata;
2387         struct nfs4_state *state = NULL;
2388
2389         /* If this request hasn't been cancelled, do nothing */
2390         if (!data->cancelled)
2391                 goto out_free;
2392         /* In case of error, no cleanup! */
2393         if (data->rpc_status != 0 || !data->rpc_done)
2394                 goto out_free;
2395         /* In case we need an open_confirm, no cleanup! */
2396         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2397                 goto out_free;
2398         state = nfs4_opendata_to_nfs4_state(data);
2399         if (!IS_ERR(state))
2400                 nfs4_close_state(state, data->o_arg.fmode);
2401 out_free:
2402         nfs4_opendata_put(data);
2403 }
2404
2405 static const struct rpc_call_ops nfs4_open_ops = {
2406         .rpc_call_prepare = nfs4_open_prepare,
2407         .rpc_call_done = nfs4_open_done,
2408         .rpc_release = nfs4_open_release,
2409 };
2410
2411 static int nfs4_run_open_task(struct nfs4_opendata *data,
2412                               struct nfs_open_context *ctx)
2413 {
2414         struct inode *dir = d_inode(data->dir);
2415         struct nfs_server *server = NFS_SERVER(dir);
2416         struct nfs_openargs *o_arg = &data->o_arg;
2417         struct nfs_openres *o_res = &data->o_res;
2418         struct rpc_task *task;
2419         struct rpc_message msg = {
2420                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2421                 .rpc_argp = o_arg,
2422                 .rpc_resp = o_res,
2423                 .rpc_cred = data->owner->so_cred,
2424         };
2425         struct rpc_task_setup task_setup_data = {
2426                 .rpc_client = server->client,
2427                 .rpc_message = &msg,
2428                 .callback_ops = &nfs4_open_ops,
2429                 .callback_data = data,
2430                 .workqueue = nfsiod_workqueue,
2431                 .flags = RPC_TASK_ASYNC,
2432         };
2433         int status;
2434
2435         kref_get(&data->kref);
2436         data->rpc_done = false;
2437         data->rpc_status = 0;
2438         data->cancelled = false;
2439         data->is_recover = false;
2440         if (!ctx) {
2441                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2442                 data->is_recover = true;
2443         } else {
2444                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2445                 pnfs_lgopen_prepare(data, ctx);
2446         }
2447         task = rpc_run_task(&task_setup_data);
2448         if (IS_ERR(task))
2449                 return PTR_ERR(task);
2450         status = rpc_wait_for_completion_task(task);
2451         if (status != 0) {
2452                 data->cancelled = true;
2453                 smp_wmb();
2454         } else
2455                 status = data->rpc_status;
2456         rpc_put_task(task);
2457
2458         return status;
2459 }
2460
2461 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2462 {
2463         struct inode *dir = d_inode(data->dir);
2464         struct nfs_openres *o_res = &data->o_res;
2465         int status;
2466
2467         status = nfs4_run_open_task(data, NULL);
2468         if (status != 0 || !data->rpc_done)
2469                 return status;
2470
2471         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2472
2473         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2474                 status = _nfs4_proc_open_confirm(data);
2475
2476         return status;
2477 }
2478
2479 /*
2480  * Additional permission checks in order to distinguish between an
2481  * open for read, and an open for execute. This works around the
2482  * fact that NFSv4 OPEN treats read and execute permissions as being
2483  * the same.
2484  * Note that in the non-execute case, we want to turn off permission
2485  * checking if we just created a new file (POSIX open() semantics).
2486  */
2487 static int nfs4_opendata_access(const struct cred *cred,
2488                                 struct nfs4_opendata *opendata,
2489                                 struct nfs4_state *state, fmode_t fmode,
2490                                 int openflags)
2491 {
2492         struct nfs_access_entry cache;
2493         u32 mask, flags;
2494
2495         /* access call failed or for some reason the server doesn't
2496          * support any access modes -- defer access call until later */
2497         if (opendata->o_res.access_supported == 0)
2498                 return 0;
2499
2500         mask = 0;
2501         /*
2502          * Use openflags to check for exec, because fmode won't
2503          * always have FMODE_EXEC set when file open for exec.
2504          */
2505         if (openflags & __FMODE_EXEC) {
2506                 /* ONLY check for exec rights */
2507                 if (S_ISDIR(state->inode->i_mode))
2508                         mask = NFS4_ACCESS_LOOKUP;
2509                 else
2510                         mask = NFS4_ACCESS_EXECUTE;
2511         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2512                 mask = NFS4_ACCESS_READ;
2513
2514         cache.cred = cred;
2515         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2516         nfs_access_add_cache(state->inode, &cache);
2517
2518         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2519         if ((mask & ~cache.mask & flags) == 0)
2520                 return 0;
2521
2522         return -EACCES;
2523 }
2524
2525 /*
2526  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2527  */
2528 static int _nfs4_proc_open(struct nfs4_opendata *data,
2529                            struct nfs_open_context *ctx)
2530 {
2531         struct inode *dir = d_inode(data->dir);
2532         struct nfs_server *server = NFS_SERVER(dir);
2533         struct nfs_openargs *o_arg = &data->o_arg;
2534         struct nfs_openres *o_res = &data->o_res;
2535         int status;
2536
2537         status = nfs4_run_open_task(data, ctx);
2538         if (!data->rpc_done)
2539                 return status;
2540         if (status != 0) {
2541                 if (status == -NFS4ERR_BADNAME &&
2542                                 !(o_arg->open_flags & O_CREAT))
2543                         return -ENOENT;
2544                 return status;
2545         }
2546
2547         nfs_fattr_map_and_free_names(server, &data->f_attr);
2548
2549         if (o_arg->open_flags & O_CREAT) {
2550                 if (o_arg->open_flags & O_EXCL)
2551                         data->file_created = true;
2552                 else if (o_res->cinfo.before != o_res->cinfo.after)
2553                         data->file_created = true;
2554                 if (data->file_created ||
2555                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2556                         update_changeattr(dir, &o_res->cinfo,
2557                                         o_res->f_attr->time_start, 0);
2558         }
2559         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2560                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2561         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2562                 status = _nfs4_proc_open_confirm(data);
2563                 if (status != 0)
2564                         return status;
2565         }
2566         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2567                 nfs4_sequence_free_slot(&o_res->seq_res);
2568                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2569                                 o_res->f_label, NULL);
2570         }
2571         return 0;
2572 }
2573
2574 /*
2575  * OPEN_EXPIRED:
2576  *      reclaim state on the server after a network partition.
2577  *      Assumes caller holds the appropriate lock
2578  */
2579 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2580 {
2581         struct nfs4_opendata *opendata;
2582         int ret;
2583
2584         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2585                         NFS4_OPEN_CLAIM_FH);
2586         if (IS_ERR(opendata))
2587                 return PTR_ERR(opendata);
2588         ret = nfs4_open_recover(opendata, state);
2589         if (ret == -ESTALE)
2590                 d_drop(ctx->dentry);
2591         nfs4_opendata_put(opendata);
2592         return ret;
2593 }
2594
2595 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2596 {
2597         struct nfs_server *server = NFS_SERVER(state->inode);
2598         struct nfs4_exception exception = { };
2599         int err;
2600
2601         do {
2602                 err = _nfs4_open_expired(ctx, state);
2603                 trace_nfs4_open_expired(ctx, 0, err);
2604                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2605                         continue;
2606                 switch (err) {
2607                 default:
2608                         goto out;
2609                 case -NFS4ERR_GRACE:
2610                 case -NFS4ERR_DELAY:
2611                         nfs4_handle_exception(server, err, &exception);
2612                         err = 0;
2613                 }
2614         } while (exception.retry);
2615 out:
2616         return err;
2617 }
2618
2619 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2620 {
2621         struct nfs_open_context *ctx;
2622         int ret;
2623
2624         ctx = nfs4_state_find_open_context(state);
2625         if (IS_ERR(ctx))
2626                 return -EAGAIN;
2627         ret = nfs4_do_open_expired(ctx, state);
2628         put_nfs_open_context(ctx);
2629         return ret;
2630 }
2631
2632 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2633                 const nfs4_stateid *stateid)
2634 {
2635         nfs_remove_bad_delegation(state->inode, stateid);
2636         nfs_state_clear_delegation(state);
2637 }
2638
2639 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2640 {
2641         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2642                 nfs_finish_clear_delegation_stateid(state, NULL);
2643 }
2644
2645 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2646 {
2647         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2648         nfs40_clear_delegation_stateid(state);
2649         return nfs4_open_expired(sp, state);
2650 }
2651
2652 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2653                 nfs4_stateid *stateid,
2654                 const struct cred *cred)
2655 {
2656         return -NFS4ERR_BAD_STATEID;
2657 }
2658
2659 #if defined(CONFIG_NFS_V4_1)
2660 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2661                 nfs4_stateid *stateid,
2662                 const struct cred *cred)
2663 {
2664         int status;
2665
2666         switch (stateid->type) {
2667         default:
2668                 break;
2669         case NFS4_INVALID_STATEID_TYPE:
2670         case NFS4_SPECIAL_STATEID_TYPE:
2671                 return -NFS4ERR_BAD_STATEID;
2672         case NFS4_REVOKED_STATEID_TYPE:
2673                 goto out_free;
2674         }
2675
2676         status = nfs41_test_stateid(server, stateid, cred);
2677         switch (status) {
2678         case -NFS4ERR_EXPIRED:
2679         case -NFS4ERR_ADMIN_REVOKED:
2680         case -NFS4ERR_DELEG_REVOKED:
2681                 break;
2682         default:
2683                 return status;
2684         }
2685 out_free:
2686         /* Ack the revoked state to the server */
2687         nfs41_free_stateid(server, stateid, cred, true);
2688         return -NFS4ERR_EXPIRED;
2689 }
2690
2691 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2692 {
2693         struct nfs_server *server = NFS_SERVER(state->inode);
2694         nfs4_stateid stateid;
2695         struct nfs_delegation *delegation;
2696         const struct cred *cred = NULL;
2697         int status;
2698
2699         /* Get the delegation credential for use by test/free_stateid */
2700         rcu_read_lock();
2701         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2702         if (delegation == NULL) {
2703                 rcu_read_unlock();
2704                 nfs_state_clear_delegation(state);
2705                 return;
2706         }
2707
2708         nfs4_stateid_copy(&stateid, &delegation->stateid);
2709         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2710                 rcu_read_unlock();
2711                 nfs_state_clear_delegation(state);
2712                 return;
2713         }
2714
2715         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2716                                 &delegation->flags)) {
2717                 rcu_read_unlock();
2718                 return;
2719         }
2720
2721         if (delegation->cred)
2722                 cred = get_cred(delegation->cred);
2723         rcu_read_unlock();
2724         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2725         trace_nfs4_test_delegation_stateid(state, NULL, status);
2726         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2727                 nfs_finish_clear_delegation_stateid(state, &stateid);
2728
2729         if (delegation->cred)
2730                 put_cred(cred);
2731 }
2732
2733 /**
2734  * nfs41_check_expired_locks - possibly free a lock stateid
2735  *
2736  * @state: NFSv4 state for an inode
2737  *
2738  * Returns NFS_OK if recovery for this stateid is now finished.
2739  * Otherwise a negative NFS4ERR value is returned.
2740  */
2741 static int nfs41_check_expired_locks(struct nfs4_state *state)
2742 {
2743         int status, ret = NFS_OK;
2744         struct nfs4_lock_state *lsp, *prev = NULL;
2745         struct nfs_server *server = NFS_SERVER(state->inode);
2746
2747         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2748                 goto out;
2749
2750         spin_lock(&state->state_lock);
2751         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2752                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2753                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2754
2755                         refcount_inc(&lsp->ls_count);
2756                         spin_unlock(&state->state_lock);
2757
2758                         nfs4_put_lock_state(prev);
2759                         prev = lsp;
2760
2761                         status = nfs41_test_and_free_expired_stateid(server,
2762                                         &lsp->ls_stateid,
2763                                         cred);
2764                         trace_nfs4_test_lock_stateid(state, lsp, status);
2765                         if (status == -NFS4ERR_EXPIRED ||
2766                             status == -NFS4ERR_BAD_STATEID) {
2767                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2768                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2769                                 if (!recover_lost_locks)
2770                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2771                         } else if (status != NFS_OK) {
2772                                 ret = status;
2773                                 nfs4_put_lock_state(prev);
2774                                 goto out;
2775                         }
2776                         spin_lock(&state->state_lock);
2777                 }
2778         }
2779         spin_unlock(&state->state_lock);
2780         nfs4_put_lock_state(prev);
2781 out:
2782         return ret;
2783 }
2784
2785 /**
2786  * nfs41_check_open_stateid - possibly free an open stateid
2787  *
2788  * @state: NFSv4 state for an inode
2789  *
2790  * Returns NFS_OK if recovery for this stateid is now finished.
2791  * Otherwise a negative NFS4ERR value is returned.
2792  */
2793 static int nfs41_check_open_stateid(struct nfs4_state *state)
2794 {
2795         struct nfs_server *server = NFS_SERVER(state->inode);
2796         nfs4_stateid *stateid = &state->open_stateid;
2797         const struct cred *cred = state->owner->so_cred;
2798         int status;
2799
2800         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2801                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2802                         if (nfs4_have_delegation(state->inode, state->state))
2803                                 return NFS_OK;
2804                         return -NFS4ERR_OPENMODE;
2805                 }
2806                 return -NFS4ERR_BAD_STATEID;
2807         }
2808         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2809         trace_nfs4_test_open_stateid(state, NULL, status);
2810         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2811                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2812                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2813                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2814                 clear_bit(NFS_OPEN_STATE, &state->flags);
2815                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2816                 return status;
2817         }
2818         if (nfs_open_stateid_recover_openmode(state))
2819                 return -NFS4ERR_OPENMODE;
2820         return NFS_OK;
2821 }
2822
2823 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2824 {
2825         int status;
2826
2827         nfs41_check_delegation_stateid(state);
2828         status = nfs41_check_expired_locks(state);
2829         if (status != NFS_OK)
2830                 return status;
2831         status = nfs41_check_open_stateid(state);
2832         if (status != NFS_OK)
2833                 status = nfs4_open_expired(sp, state);
2834         return status;
2835 }
2836 #endif
2837
2838 /*
2839  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2840  * fields corresponding to attributes that were used to store the verifier.
2841  * Make sure we clobber those fields in the later setattr call
2842  */
2843 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2844                                 struct iattr *sattr, struct nfs4_label **label)
2845 {
2846         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2847         __u32 attrset[3];
2848         unsigned ret;
2849         unsigned i;
2850
2851         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2852                 attrset[i] = opendata->o_res.attrset[i];
2853                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2854                         attrset[i] &= ~bitmask[i];
2855         }
2856
2857         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2858                 sattr->ia_valid : 0;
2859
2860         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2861                 if (sattr->ia_valid & ATTR_ATIME_SET)
2862                         ret |= ATTR_ATIME_SET;
2863                 else
2864                         ret |= ATTR_ATIME;
2865         }
2866
2867         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2868                 if (sattr->ia_valid & ATTR_MTIME_SET)
2869                         ret |= ATTR_MTIME_SET;
2870                 else
2871                         ret |= ATTR_MTIME;
2872         }
2873
2874         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2875                 *label = NULL;
2876         return ret;
2877 }
2878
2879 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2880                 fmode_t fmode,
2881                 int flags,
2882                 struct nfs_open_context *ctx)
2883 {
2884         struct nfs4_state_owner *sp = opendata->owner;
2885         struct nfs_server *server = sp->so_server;
2886         struct dentry *dentry;
2887         struct nfs4_state *state;
2888         unsigned int seq;
2889         int ret;
2890
2891         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2892
2893         ret = _nfs4_proc_open(opendata, ctx);
2894         if (ret != 0)
2895                 goto out;
2896
2897         state = _nfs4_opendata_to_nfs4_state(opendata);
2898         ret = PTR_ERR(state);
2899         if (IS_ERR(state))
2900                 goto out;
2901         ctx->state = state;
2902         if (server->caps & NFS_CAP_POSIX_LOCK)
2903                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2904         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2905                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2906
2907         dentry = opendata->dentry;
2908         if (d_really_is_negative(dentry)) {
2909                 struct dentry *alias;
2910                 d_drop(dentry);
2911                 alias = d_exact_alias(dentry, state->inode);
2912                 if (!alias)
2913                         alias = d_splice_alias(igrab(state->inode), dentry);
2914                 /* d_splice_alias() can't fail here - it's a non-directory */
2915                 if (alias) {
2916                         dput(ctx->dentry);
2917                         ctx->dentry = dentry = alias;
2918                 }
2919                 nfs_set_verifier(dentry,
2920                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2921         }
2922
2923         /* Parse layoutget results before we check for access */
2924         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
2925
2926         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2927         if (ret != 0)
2928                 goto out;
2929
2930         if (d_inode(dentry) == state->inode) {
2931                 nfs_inode_attach_open_context(ctx);
2932                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2933                         nfs4_schedule_stateid_recovery(server, state);
2934         }
2935
2936 out:
2937         nfs4_sequence_free_slot(&opendata->o_res.seq_res);
2938         return ret;
2939 }
2940
2941 /*
2942  * Returns a referenced nfs4_state
2943  */
2944 static int _nfs4_do_open(struct inode *dir,
2945                         struct nfs_open_context *ctx,
2946                         int flags,
2947                         const struct nfs4_open_createattrs *c,
2948                         int *opened)
2949 {
2950         struct nfs4_state_owner  *sp;
2951         struct nfs4_state     *state = NULL;
2952         struct nfs_server       *server = NFS_SERVER(dir);
2953         struct nfs4_opendata *opendata;
2954         struct dentry *dentry = ctx->dentry;
2955         const struct cred *cred = ctx->cred;
2956         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2957         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2958         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2959         struct iattr *sattr = c->sattr;
2960         struct nfs4_label *label = c->label;
2961         struct nfs4_label *olabel = NULL;
2962         int status;
2963
2964         /* Protect against reboot recovery conflicts */
2965         status = -ENOMEM;
2966         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2967         if (sp == NULL) {
2968                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2969                 goto out_err;
2970         }
2971         status = nfs4_client_recover_expired_lease(server->nfs_client);
2972         if (status != 0)
2973                 goto err_put_state_owner;
2974         if (d_really_is_positive(dentry))
2975                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2976         status = -ENOMEM;
2977         if (d_really_is_positive(dentry))
2978                 claim = NFS4_OPEN_CLAIM_FH;
2979         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
2980                         c, claim, GFP_KERNEL);
2981         if (opendata == NULL)
2982                 goto err_put_state_owner;
2983
2984         if (label) {
2985                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2986                 if (IS_ERR(olabel)) {
2987                         status = PTR_ERR(olabel);
2988                         goto err_opendata_put;
2989                 }
2990         }
2991
2992         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2993                 if (!opendata->f_attr.mdsthreshold) {
2994                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2995                         if (!opendata->f_attr.mdsthreshold)
2996                                 goto err_free_label;
2997                 }
2998                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2999         }
3000         if (d_really_is_positive(dentry))
3001                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3002
3003         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
3004         if (status != 0)
3005                 goto err_free_label;
3006         state = ctx->state;
3007
3008         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3009             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3010                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3011                 /*
3012                  * send create attributes which was not set by open
3013                  * with an extra setattr.
3014                  */
3015                 if (attrs || label) {
3016                         unsigned ia_old = sattr->ia_valid;
3017
3018                         sattr->ia_valid = attrs;
3019                         nfs_fattr_init(opendata->o_res.f_attr);
3020                         status = nfs4_do_setattr(state->inode, cred,
3021                                         opendata->o_res.f_attr, sattr,
3022                                         ctx, label, olabel);
3023                         if (status == 0) {
3024                                 nfs_setattr_update_inode(state->inode, sattr,
3025                                                 opendata->o_res.f_attr);
3026                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3027                         }
3028                         sattr->ia_valid = ia_old;
3029                 }
3030         }
3031         if (opened && opendata->file_created)
3032                 *opened = 1;
3033
3034         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3035                 *ctx_th = opendata->f_attr.mdsthreshold;
3036                 opendata->f_attr.mdsthreshold = NULL;
3037         }
3038
3039         nfs4_label_free(olabel);
3040
3041         nfs4_opendata_put(opendata);
3042         nfs4_put_state_owner(sp);
3043         return 0;
3044 err_free_label:
3045         nfs4_label_free(olabel);
3046 err_opendata_put:
3047         nfs4_opendata_put(opendata);
3048 err_put_state_owner:
3049         nfs4_put_state_owner(sp);
3050 out_err:
3051         return status;
3052 }
3053
3054
3055 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3056                                         struct nfs_open_context *ctx,
3057                                         int flags,
3058                                         struct iattr *sattr,
3059                                         struct nfs4_label *label,
3060                                         int *opened)
3061 {
3062         struct nfs_server *server = NFS_SERVER(dir);
3063         struct nfs4_exception exception = { };
3064         struct nfs4_state *res;
3065         struct nfs4_open_createattrs c = {
3066                 .label = label,
3067                 .sattr = sattr,
3068                 .verf = {
3069                         [0] = (__u32)jiffies,
3070                         [1] = (__u32)current->pid,
3071                 },
3072         };
3073         int status;
3074
3075         do {
3076                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3077                 res = ctx->state;
3078                 trace_nfs4_open_file(ctx, flags, status);
3079                 if (status == 0)
3080                         break;
3081                 /* NOTE: BAD_SEQID means the server and client disagree about the
3082                  * book-keeping w.r.t. state-changing operations
3083                  * (OPEN/CLOSE/LOCK/LOCKU...)
3084                  * It is actually a sign of a bug on the client or on the server.
3085                  *
3086                  * If we receive a BAD_SEQID error in the particular case of
3087                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3088                  * have unhashed the old state_owner for us, and that we can
3089                  * therefore safely retry using a new one. We should still warn
3090                  * the user though...
3091                  */
3092                 if (status == -NFS4ERR_BAD_SEQID) {
3093                         pr_warn_ratelimited("NFS: v4 server %s "
3094                                         " returned a bad sequence-id error!\n",
3095                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3096                         exception.retry = 1;
3097                         continue;
3098                 }
3099                 /*
3100                  * BAD_STATEID on OPEN means that the server cancelled our
3101                  * state before it received the OPEN_CONFIRM.
3102                  * Recover by retrying the request as per the discussion
3103                  * on Page 181 of RFC3530.
3104                  */
3105                 if (status == -NFS4ERR_BAD_STATEID) {
3106                         exception.retry = 1;
3107                         continue;
3108                 }
3109                 if (status == -EAGAIN) {
3110                         /* We must have found a delegation */
3111                         exception.retry = 1;
3112                         continue;
3113                 }
3114                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3115                         continue;
3116                 res = ERR_PTR(nfs4_handle_exception(server,
3117                                         status, &exception));
3118         } while (exception.retry);
3119         return res;
3120 }
3121
3122 static int _nfs4_do_setattr(struct inode *inode,
3123                             struct nfs_setattrargs *arg,
3124                             struct nfs_setattrres *res,
3125                             const struct cred *cred,
3126                             struct nfs_open_context *ctx)
3127 {
3128         struct nfs_server *server = NFS_SERVER(inode);
3129         struct rpc_message msg = {
3130                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3131                 .rpc_argp       = arg,
3132                 .rpc_resp       = res,
3133                 .rpc_cred       = cred,
3134         };
3135         const struct cred *delegation_cred = NULL;
3136         unsigned long timestamp = jiffies;
3137         bool truncate;
3138         int status;
3139
3140         nfs_fattr_init(res->fattr);
3141
3142         /* Servers should only apply open mode checks for file size changes */
3143         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3144         if (!truncate)
3145                 goto zero_stateid;
3146
3147         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3148                 /* Use that stateid */
3149         } else if (ctx != NULL) {
3150                 struct nfs_lock_context *l_ctx;
3151                 if (!nfs4_valid_open_stateid(ctx->state))
3152                         return -EBADF;
3153                 l_ctx = nfs_get_lock_context(ctx);
3154                 if (IS_ERR(l_ctx))
3155                         return PTR_ERR(l_ctx);
3156                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3157                                                 &arg->stateid, &delegation_cred);
3158                 nfs_put_lock_context(l_ctx);
3159                 if (status == -EIO)
3160                         return -EBADF;
3161         } else {
3162 zero_stateid:
3163                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3164         }
3165         if (delegation_cred)
3166                 msg.rpc_cred = delegation_cred;
3167
3168         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3169
3170         put_cred(delegation_cred);
3171         if (status == 0 && ctx != NULL)
3172                 renew_lease(server, timestamp);
3173         trace_nfs4_setattr(inode, &arg->stateid, status);
3174         return status;
3175 }
3176
3177 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3178                            struct nfs_fattr *fattr, struct iattr *sattr,
3179                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3180                            struct nfs4_label *olabel)
3181 {
3182         struct nfs_server *server = NFS_SERVER(inode);
3183         __u32 bitmask[NFS4_BITMASK_SZ];
3184         struct nfs4_state *state = ctx ? ctx->state : NULL;
3185         struct nfs_setattrargs  arg = {
3186                 .fh             = NFS_FH(inode),
3187                 .iap            = sattr,
3188                 .server         = server,
3189                 .bitmask = bitmask,
3190                 .label          = ilabel,
3191         };
3192         struct nfs_setattrres  res = {
3193                 .fattr          = fattr,
3194                 .label          = olabel,
3195                 .server         = server,
3196         };
3197         struct nfs4_exception exception = {
3198                 .state = state,
3199                 .inode = inode,
3200                 .stateid = &arg.stateid,
3201         };
3202         int err;
3203
3204         do {
3205                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3206                                 nfs4_bitmask(server, olabel),
3207                                 inode);
3208
3209                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3210                 switch (err) {
3211                 case -NFS4ERR_OPENMODE:
3212                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3213                                 pr_warn_once("NFSv4: server %s is incorrectly "
3214                                                 "applying open mode checks to "
3215                                                 "a SETATTR that is not "
3216                                                 "changing file size.\n",
3217                                                 server->nfs_client->cl_hostname);
3218                         }
3219                         if (state && !(state->state & FMODE_WRITE)) {
3220                                 err = -EBADF;
3221                                 if (sattr->ia_valid & ATTR_OPEN)
3222                                         err = -EACCES;
3223                                 goto out;
3224                         }
3225                 }
3226                 err = nfs4_handle_exception(server, err, &exception);
3227         } while (exception.retry);
3228 out:
3229         return err;
3230 }
3231
3232 static bool
3233 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3234 {
3235         if (inode == NULL || !nfs_have_layout(inode))
3236                 return false;
3237
3238         return pnfs_wait_on_layoutreturn(inode, task);
3239 }
3240
3241 struct nfs4_closedata {
3242         struct inode *inode;
3243         struct nfs4_state *state;
3244         struct nfs_closeargs arg;
3245         struct nfs_closeres res;
3246         struct {
3247                 struct nfs4_layoutreturn_args arg;
3248                 struct nfs4_layoutreturn_res res;
3249                 struct nfs4_xdr_opaque_data ld_private;
3250                 u32 roc_barrier;
3251                 bool roc;
3252         } lr;
3253         struct nfs_fattr fattr;
3254         unsigned long timestamp;
3255 };
3256
3257 static void nfs4_free_closedata(void *data)
3258 {
3259         struct nfs4_closedata *calldata = data;
3260         struct nfs4_state_owner *sp = calldata->state->owner;
3261         struct super_block *sb = calldata->state->inode->i_sb;
3262
3263         if (calldata->lr.roc)
3264                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3265                                 calldata->res.lr_ret);
3266         nfs4_put_open_state(calldata->state);
3267         nfs_free_seqid(calldata->arg.seqid);
3268         nfs4_put_state_owner(sp);
3269         nfs_sb_deactive(sb);
3270         kfree(calldata);
3271 }
3272
3273 static void nfs4_close_done(struct rpc_task *task, void *data)
3274 {
3275         struct nfs4_closedata *calldata = data;
3276         struct nfs4_state *state = calldata->state;
3277         struct nfs_server *server = NFS_SERVER(calldata->inode);
3278         nfs4_stateid *res_stateid = NULL;
3279         struct nfs4_exception exception = {
3280                 .state = state,
3281                 .inode = calldata->inode,
3282                 .stateid = &calldata->arg.stateid,
3283         };
3284
3285         dprintk("%s: begin!\n", __func__);
3286         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3287                 return;
3288         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3289
3290         /* Handle Layoutreturn errors */
3291         if (calldata->arg.lr_args && task->tk_status != 0) {
3292                 switch (calldata->res.lr_ret) {
3293                 default:
3294                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3295                         break;
3296                 case 0:
3297                         calldata->arg.lr_args = NULL;
3298                         calldata->res.lr_res = NULL;
3299                         break;
3300                 case -NFS4ERR_OLD_STATEID:
3301                         if (nfs4_layoutreturn_refresh_stateid(&calldata->arg.lr_args->stateid,
3302                                                 &calldata->arg.lr_args->range,
3303                                                 calldata->inode))
3304                                 goto lr_restart;
3305                         /* Fallthrough */
3306                 case -NFS4ERR_ADMIN_REVOKED:
3307                 case -NFS4ERR_DELEG_REVOKED:
3308                 case -NFS4ERR_EXPIRED:
3309                 case -NFS4ERR_BAD_STATEID:
3310                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3311                 case -NFS4ERR_WRONG_CRED:
3312                         calldata->arg.lr_args = NULL;
3313                         calldata->res.lr_res = NULL;
3314                         goto lr_restart;
3315                 }
3316         }
3317
3318         /* hmm. we are done with the inode, and in the process of freeing
3319          * the state_owner. we keep this around to process errors
3320          */
3321         switch (task->tk_status) {
3322                 case 0:
3323                         res_stateid = &calldata->res.stateid;
3324                         renew_lease(server, calldata->timestamp);
3325                         break;
3326                 case -NFS4ERR_ACCESS:
3327                         if (calldata->arg.bitmask != NULL) {
3328                                 calldata->arg.bitmask = NULL;
3329                                 calldata->res.fattr = NULL;
3330                                 goto out_restart;
3331
3332                         }
3333                         break;
3334                 case -NFS4ERR_OLD_STATEID:
3335                         /* Did we race with OPEN? */
3336                         if (nfs4_refresh_open_stateid(&calldata->arg.stateid,
3337                                                 state))
3338                                 goto out_restart;
3339                         goto out_release;
3340                 case -NFS4ERR_ADMIN_REVOKED:
3341                 case -NFS4ERR_STALE_STATEID:
3342                 case -NFS4ERR_EXPIRED:
3343                         nfs4_free_revoked_stateid(server,
3344                                         &calldata->arg.stateid,
3345                                         task->tk_msg.rpc_cred);
3346                         /* Fallthrough */
3347                 case -NFS4ERR_BAD_STATEID:
3348                         break;
3349                 default:
3350                         task->tk_status = nfs4_async_handle_exception(task,
3351                                         server, task->tk_status, &exception);
3352                         if (exception.retry)
3353                                 goto out_restart;
3354         }
3355         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3356                         res_stateid, calldata->arg.fmode);
3357 out_release:
3358         task->tk_status = 0;
3359         nfs_release_seqid(calldata->arg.seqid);
3360         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3361         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3362         return;
3363 lr_restart:
3364         calldata->res.lr_ret = 0;
3365 out_restart:
3366         task->tk_status = 0;
3367         rpc_restart_call_prepare(task);
3368         goto out_release;
3369 }
3370
3371 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3372 {
3373         struct nfs4_closedata *calldata = data;
3374         struct nfs4_state *state = calldata->state;
3375         struct inode *inode = calldata->inode;
3376         struct pnfs_layout_hdr *lo;
3377         bool is_rdonly, is_wronly, is_rdwr;
3378         int call_close = 0;
3379
3380         dprintk("%s: begin!\n", __func__);
3381         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3382                 goto out_wait;
3383
3384         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3385         spin_lock(&state->owner->so_lock);
3386         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3387         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3388         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3389         /* Calculate the change in open mode */
3390         calldata->arg.fmode = 0;
3391         if (state->n_rdwr == 0) {
3392                 if (state->n_rdonly == 0)
3393                         call_close |= is_rdonly;
3394                 else if (is_rdonly)
3395                         calldata->arg.fmode |= FMODE_READ;
3396                 if (state->n_wronly == 0)
3397                         call_close |= is_wronly;
3398                 else if (is_wronly)
3399                         calldata->arg.fmode |= FMODE_WRITE;
3400                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3401                         call_close |= is_rdwr;
3402         } else if (is_rdwr)
3403                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3404
3405         if (!nfs4_valid_open_stateid(state) ||
3406             !nfs4_refresh_open_stateid(&calldata->arg.stateid, state))
3407                 call_close = 0;
3408         spin_unlock(&state->owner->so_lock);
3409
3410         if (!call_close) {
3411                 /* Note: exit _without_ calling nfs4_close_done */
3412                 goto out_no_action;
3413         }
3414
3415         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3416                 nfs_release_seqid(calldata->arg.seqid);
3417                 goto out_wait;
3418         }
3419
3420         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3421         if (lo && !pnfs_layout_is_valid(lo)) {
3422                 calldata->arg.lr_args = NULL;
3423                 calldata->res.lr_res = NULL;
3424         }
3425
3426         if (calldata->arg.fmode == 0)
3427                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3428
3429         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3430                 /* Close-to-open cache consistency revalidation */
3431                 if (!nfs4_have_delegation(inode, FMODE_READ))
3432                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3433                 else
3434                         calldata->arg.bitmask = NULL;
3435         }
3436
3437         calldata->arg.share_access =
3438                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3439                                 calldata->arg.fmode, 0);
3440
3441         if (calldata->res.fattr == NULL)
3442                 calldata->arg.bitmask = NULL;
3443         else if (calldata->arg.bitmask == NULL)
3444                 calldata->res.fattr = NULL;
3445         calldata->timestamp = jiffies;
3446         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3447                                 &calldata->arg.seq_args,
3448                                 &calldata->res.seq_res,
3449                                 task) != 0)
3450                 nfs_release_seqid(calldata->arg.seqid);
3451         dprintk("%s: done!\n", __func__);
3452         return;
3453 out_no_action:
3454         task->tk_action = NULL;
3455 out_wait:
3456         nfs4_sequence_done(task, &calldata->res.seq_res);
3457 }
3458
3459 static const struct rpc_call_ops nfs4_close_ops = {
3460         .rpc_call_prepare = nfs4_close_prepare,
3461         .rpc_call_done = nfs4_close_done,
3462         .rpc_release = nfs4_free_closedata,
3463 };
3464
3465 /*
3466  * It is possible for data to be read/written from a mem-mapped file
3467  * after the sys_close call (which hits the vfs layer as a flush).
3468  * This means that we can't safely call nfsv4 close on a file until
3469  * the inode is cleared. This in turn means that we are not good
3470  * NFSv4 citizens - we do not indicate to the server to update the file's
3471  * share state even when we are done with one of the three share
3472  * stateid's in the inode.
3473  *
3474  * NOTE: Caller must be holding the sp->so_owner semaphore!
3475  */
3476 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3477 {
3478         struct nfs_server *server = NFS_SERVER(state->inode);
3479         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3480         struct nfs4_closedata *calldata;
3481         struct nfs4_state_owner *sp = state->owner;
3482         struct rpc_task *task;
3483         struct rpc_message msg = {
3484                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3485                 .rpc_cred = state->owner->so_cred,
3486         };
3487         struct rpc_task_setup task_setup_data = {
3488                 .rpc_client = server->client,
3489                 .rpc_message = &msg,
3490                 .callback_ops = &nfs4_close_ops,
3491                 .workqueue = nfsiod_workqueue,
3492                 .flags = RPC_TASK_ASYNC,
3493         };
3494         int status = -ENOMEM;
3495
3496         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3497                 &task_setup_data.rpc_client, &msg);
3498
3499         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3500         if (calldata == NULL)
3501                 goto out;
3502         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3503         calldata->inode = state->inode;
3504         calldata->state = state;
3505         calldata->arg.fh = NFS_FH(state->inode);
3506         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3507                 goto out_free_calldata;
3508         /* Serialization for the sequence id */
3509         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3510         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3511         if (IS_ERR(calldata->arg.seqid))
3512                 goto out_free_calldata;
3513         nfs_fattr_init(&calldata->fattr);
3514         calldata->arg.fmode = 0;
3515         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3516         calldata->res.fattr = &calldata->fattr;
3517         calldata->res.seqid = calldata->arg.seqid;
3518         calldata->res.server = server;
3519         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3520         calldata->lr.roc = pnfs_roc(state->inode,
3521                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3522         if (calldata->lr.roc) {
3523                 calldata->arg.lr_args = &calldata->lr.arg;
3524                 calldata->res.lr_res = &calldata->lr.res;
3525         }
3526         nfs_sb_active(calldata->inode->i_sb);
3527
3528         msg.rpc_argp = &calldata->arg;
3529         msg.rpc_resp = &calldata->res;
3530         task_setup_data.callback_data = calldata;
3531         task = rpc_run_task(&task_setup_data);
3532         if (IS_ERR(task))
3533                 return PTR_ERR(task);
3534         status = 0;
3535         if (wait)
3536                 status = rpc_wait_for_completion_task(task);
3537         rpc_put_task(task);
3538         return status;
3539 out_free_calldata:
3540         kfree(calldata);
3541 out:
3542         nfs4_put_open_state(state);
3543         nfs4_put_state_owner(sp);
3544         return status;
3545 }
3546
3547 static struct inode *
3548 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3549                 int open_flags, struct iattr *attr, int *opened)
3550 {
3551         struct nfs4_state *state;
3552         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3553
3554         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3555
3556         /* Protect against concurrent sillydeletes */
3557         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3558
3559         nfs4_label_release_security(label);
3560
3561         if (IS_ERR(state))
3562                 return ERR_CAST(state);
3563         return state->inode;
3564 }
3565
3566 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3567 {
3568         if (ctx->state == NULL)
3569                 return;
3570         if (is_sync)
3571                 nfs4_close_sync(ctx->state, ctx->mode);
3572         else
3573                 nfs4_close_state(ctx->state, ctx->mode);
3574 }
3575
3576 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3577 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3578 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3579
3580 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3581 {
3582         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3583         struct nfs4_server_caps_arg args = {
3584                 .fhandle = fhandle,
3585                 .bitmask = bitmask,
3586         };
3587         struct nfs4_server_caps_res res = {};
3588         struct rpc_message msg = {
3589                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3590                 .rpc_argp = &args,
3591                 .rpc_resp = &res,
3592         };
3593         int status;
3594         int i;
3595
3596         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3597                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3598                      FATTR4_WORD0_LINK_SUPPORT |
3599                      FATTR4_WORD0_SYMLINK_SUPPORT |
3600                      FATTR4_WORD0_ACLSUPPORT;
3601         if (minorversion)
3602                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3603
3604         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3605         if (status == 0) {
3606                 /* Sanity check the server answers */
3607                 switch (minorversion) {
3608                 case 0:
3609                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3610                         res.attr_bitmask[2] = 0;
3611                         break;
3612                 case 1:
3613                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3614                         break;
3615                 case 2:
3616                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3617                 }
3618                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3619                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3620                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3621                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3622                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3623                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3624                                 NFS_CAP_SECURITY_LABEL);
3625                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3626                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3627                         server->caps |= NFS_CAP_ACLS;
3628                 if (res.has_links != 0)
3629                         server->caps |= NFS_CAP_HARDLINKS;
3630                 if (res.has_symlinks != 0)
3631                         server->caps |= NFS_CAP_SYMLINKS;
3632                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3633                         server->caps |= NFS_CAP_FILEID;
3634                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3635                         server->caps |= NFS_CAP_MODE;
3636                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3637                         server->caps |= NFS_CAP_NLINK;
3638                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3639                         server->caps |= NFS_CAP_OWNER;
3640                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3641                         server->caps |= NFS_CAP_OWNER_GROUP;
3642                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3643                         server->caps |= NFS_CAP_ATIME;
3644                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3645                         server->caps |= NFS_CAP_CTIME;
3646                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3647                         server->caps |= NFS_CAP_MTIME;
3648 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3649                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3650                         server->caps |= NFS_CAP_SECURITY_LABEL;
3651 #endif
3652                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3653                                 sizeof(server->attr_bitmask));
3654                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3655
3656                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3657                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3658                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3659                 server->cache_consistency_bitmask[2] = 0;
3660
3661                 /* Avoid a regression due to buggy server */
3662                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3663                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3664                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3665                         sizeof(server->exclcreat_bitmask));
3666
3667                 server->acl_bitmask = res.acl_bitmask;
3668                 server->fh_expire_type = res.fh_expire_type;
3669         }
3670
3671         return status;
3672 }
3673
3674 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3675 {
3676         struct nfs4_exception exception = { };
3677         int err;
3678         do {
3679                 err = nfs4_handle_exception(server,
3680                                 _nfs4_server_capabilities(server, fhandle),
3681                                 &exception);
3682         } while (exception.retry);
3683         return err;
3684 }
3685
3686 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3687                 struct nfs_fsinfo *info)
3688 {
3689         u32 bitmask[3];
3690         struct nfs4_lookup_root_arg args = {
3691                 .bitmask = bitmask,
3692         };
3693         struct nfs4_lookup_res res = {
3694                 .server = server,
3695                 .fattr = info->fattr,
3696                 .fh = fhandle,
3697         };
3698         struct rpc_message msg = {
3699                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3700                 .rpc_argp = &args,
3701                 .rpc_resp = &res,
3702         };
3703
3704         bitmask[0] = nfs4_fattr_bitmap[0];
3705         bitmask[1] = nfs4_fattr_bitmap[1];
3706         /*
3707          * Process the label in the upcoming getfattr
3708          */
3709         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3710
3711         nfs_fattr_init(info->fattr);
3712         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3713 }
3714
3715 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3716                 struct nfs_fsinfo *info)
3717 {
3718         struct nfs4_exception exception = { };
3719         int err;
3720         do {
3721                 err = _nfs4_lookup_root(server, fhandle, info);
3722                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3723                 switch (err) {
3724                 case 0:
3725                 case -NFS4ERR_WRONGSEC:
3726                         goto out;
3727                 default:
3728                         err = nfs4_handle_exception(server, err, &exception);
3729                 }
3730         } while (exception.retry);
3731 out:
3732         return err;
3733 }
3734
3735 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3736                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3737 {
3738         struct rpc_auth_create_args auth_args = {
3739                 .pseudoflavor = flavor,
3740         };
3741         struct rpc_auth *auth;
3742
3743         auth = rpcauth_create(&auth_args, server->client);
3744         if (IS_ERR(auth))
3745                 return -EACCES;
3746         return nfs4_lookup_root(server, fhandle, info);
3747 }
3748
3749 /*
3750  * Retry pseudoroot lookup with various security flavors.  We do this when:
3751  *
3752  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3753  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3754  *
3755  * Returns zero on success, or a negative NFS4ERR value, or a
3756  * negative errno value.
3757  */
3758 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3759                               struct nfs_fsinfo *info)
3760 {
3761         /* Per 3530bis 15.33.5 */
3762         static const rpc_authflavor_t flav_array[] = {
3763                 RPC_AUTH_GSS_KRB5P,
3764                 RPC_AUTH_GSS_KRB5I,
3765                 RPC_AUTH_GSS_KRB5,
3766                 RPC_AUTH_UNIX,                  /* courtesy */
3767                 RPC_AUTH_NULL,
3768         };
3769         int status = -EPERM;
3770         size_t i;
3771
3772         if (server->auth_info.flavor_len > 0) {
3773                 /* try each flavor specified by user */
3774                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3775                         status = nfs4_lookup_root_sec(server, fhandle, info,
3776                                                 server->auth_info.flavors[i]);
3777                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3778                                 continue;
3779                         break;
3780                 }
3781         } else {
3782                 /* no flavors specified by user, try default list */
3783                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3784                         status = nfs4_lookup_root_sec(server, fhandle, info,
3785                                                       flav_array[i]);
3786                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3787                                 continue;
3788                         break;
3789                 }
3790         }
3791
3792         /*
3793          * -EACCES could mean that the user doesn't have correct permissions
3794          * to access the mount.  It could also mean that we tried to mount
3795          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3796          * existing mount programs don't handle -EACCES very well so it should
3797          * be mapped to -EPERM instead.
3798          */
3799         if (status == -EACCES)
3800                 status = -EPERM;
3801         return status;
3802 }
3803
3804 /**
3805  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3806  * @server: initialized nfs_server handle
3807  * @fhandle: we fill in the pseudo-fs root file handle
3808  * @info: we fill in an FSINFO struct
3809  * @auth_probe: probe the auth flavours
3810  *
3811  * Returns zero on success, or a negative errno.
3812  */
3813 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3814                          struct nfs_fsinfo *info,
3815                          bool auth_probe)
3816 {
3817         int status = 0;
3818
3819         if (!auth_probe)
3820                 status = nfs4_lookup_root(server, fhandle, info);
3821
3822         if (auth_probe || status == NFS4ERR_WRONGSEC)
3823                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3824                                 fhandle, info);
3825
3826         if (status == 0)
3827                 status = nfs4_server_capabilities(server, fhandle);
3828         if (status == 0)
3829                 status = nfs4_do_fsinfo(server, fhandle, info);
3830
3831         return nfs4_map_errors(status);
3832 }
3833
3834 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3835                               struct nfs_fsinfo *info)
3836 {
3837         int error;
3838         struct nfs_fattr *fattr = info->fattr;
3839         struct nfs4_label *label = NULL;
3840
3841         error = nfs4_server_capabilities(server, mntfh);
3842         if (error < 0) {
3843                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3844                 return error;
3845         }
3846
3847         label = nfs4_label_alloc(server, GFP_KERNEL);
3848         if (IS_ERR(label))
3849                 return PTR_ERR(label);
3850
3851         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
3852         if (error < 0) {
3853                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3854                 goto err_free_label;
3855         }
3856
3857         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3858             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3859                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3860
3861 err_free_label:
3862         nfs4_label_free(label);
3863
3864         return error;
3865 }
3866
3867 /*
3868  * Get locations and (maybe) other attributes of a referral.
3869  * Note that we'll actually follow the referral later when
3870  * we detect fsid mismatch in inode revalidation
3871  */
3872 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3873                              const struct qstr *name, struct nfs_fattr *fattr,
3874                              struct nfs_fh *fhandle)
3875 {
3876         int status = -ENOMEM;
3877         struct page *page = NULL;
3878         struct nfs4_fs_locations *locations = NULL;
3879
3880         page = alloc_page(GFP_KERNEL);
3881         if (page == NULL)
3882                 goto out;
3883         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3884         if (locations == NULL)
3885                 goto out;
3886
3887         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3888         if (status != 0)
3889                 goto out;
3890
3891         /*
3892          * If the fsid didn't change, this is a migration event, not a
3893          * referral.  Cause us to drop into the exception handler, which
3894          * will kick off migration recovery.
3895          */
3896         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3897                 dprintk("%s: server did not return a different fsid for"
3898                         " a referral at %s\n", __func__, name->name);
3899                 status = -NFS4ERR_MOVED;
3900                 goto out;
3901         }
3902         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3903         nfs_fixup_referral_attributes(&locations->fattr);
3904
3905         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3906         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3907         memset(fhandle, 0, sizeof(struct nfs_fh));
3908 out:
3909         if (page)
3910                 __free_page(page);
3911         kfree(locations);
3912         return status;
3913 }
3914
3915 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3916                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3917                                 struct inode *inode)
3918 {
3919         __u32 bitmask[NFS4_BITMASK_SZ];
3920         struct nfs4_getattr_arg args = {
3921                 .fh = fhandle,
3922                 .bitmask = bitmask,
3923         };
3924         struct nfs4_getattr_res res = {
3925                 .fattr = fattr,
3926                 .label = label,
3927                 .server = server,
3928         };
3929         struct rpc_message msg = {
3930                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3931                 .rpc_argp = &args,
3932                 .rpc_resp = &res,
3933         };
3934
3935         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
3936
3937         nfs_fattr_init(fattr);
3938         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3939 }
3940
3941 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3942                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3943                                 struct inode *inode)
3944 {
3945         struct nfs4_exception exception = { };
3946         int err;
3947         do {
3948                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
3949                 trace_nfs4_getattr(server, fhandle, fattr, err);
3950                 err = nfs4_handle_exception(server, err,
3951                                 &exception);
3952         } while (exception.retry);
3953         return err;
3954 }
3955
3956 /*
3957  * The file is not closed if it is opened due to the a request to change
3958  * the size of the file. The open call will not be needed once the
3959  * VFS layer lookup-intents are implemented.
3960  *
3961  * Close is called when the inode is destroyed.
3962  * If we haven't opened the file for O_WRONLY, we
3963  * need to in the size_change case to obtain a stateid.
3964  *
3965  * Got race?
3966  * Because OPEN is always done by name in nfsv4, it is
3967  * possible that we opened a different file by the same
3968  * name.  We can recognize this race condition, but we
3969  * can't do anything about it besides returning an error.
3970  *
3971  * This will be fixed with VFS changes (lookup-intent).
3972  */
3973 static int
3974 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3975                   struct iattr *sattr)
3976 {
3977         struct inode *inode = d_inode(dentry);
3978         const struct cred *cred = NULL;
3979         struct nfs_open_context *ctx = NULL;
3980         struct nfs4_label *label = NULL;
3981         int status;
3982
3983         if (pnfs_ld_layoutret_on_setattr(inode) &&
3984             sattr->ia_valid & ATTR_SIZE &&
3985             sattr->ia_size < i_size_read(inode))
3986                 pnfs_commit_and_return_layout(inode);
3987
3988         nfs_fattr_init(fattr);
3989
3990         /* Deal with open(O_TRUNC) */
3991         if (sattr->ia_valid & ATTR_OPEN)
3992                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3993
3994         /* Optimization: if the end result is no change, don't RPC */
3995         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3996                 return 0;
3997
3998         /* Search for an existing open(O_WRITE) file */
3999         if (sattr->ia_valid & ATTR_FILE) {
4000
4001                 ctx = nfs_file_open_context(sattr->ia_file);
4002                 if (ctx)
4003                         cred = ctx->cred;
4004         }
4005
4006         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4007         if (IS_ERR(label))
4008                 return PTR_ERR(label);
4009
4010         /* Return any delegations if we're going to change ACLs */
4011         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4012                 nfs4_inode_make_writeable(inode);
4013
4014         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4015         if (status == 0) {
4016                 nfs_setattr_update_inode(inode, sattr, fattr);
4017                 nfs_setsecurity(inode, fattr, label);
4018         }
4019         nfs4_label_free(label);
4020         return status;
4021 }
4022
4023 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4024                 const struct qstr *name, struct nfs_fh *fhandle,
4025                 struct nfs_fattr *fattr, struct nfs4_label *label)
4026 {
4027         struct nfs_server *server = NFS_SERVER(dir);
4028         int                    status;
4029         struct nfs4_lookup_arg args = {
4030                 .bitmask = server->attr_bitmask,
4031                 .dir_fh = NFS_FH(dir),
4032                 .name = name,
4033         };
4034         struct nfs4_lookup_res res = {
4035                 .server = server,
4036                 .fattr = fattr,
4037                 .label = label,
4038                 .fh = fhandle,
4039         };
4040         struct rpc_message msg = {
4041                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4042                 .rpc_argp = &args,
4043                 .rpc_resp = &res,
4044         };
4045
4046         args.bitmask = nfs4_bitmask(server, label);
4047
4048         nfs_fattr_init(fattr);
4049
4050         dprintk("NFS call  lookup %s\n", name->name);
4051         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
4052         dprintk("NFS reply lookup: %d\n", status);
4053         return status;
4054 }
4055
4056 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4057 {
4058         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4059                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4060         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4061         fattr->nlink = 2;
4062 }
4063
4064 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4065                                    const struct qstr *name, struct nfs_fh *fhandle,
4066                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4067 {
4068         struct nfs4_exception exception = { };
4069         struct rpc_clnt *client = *clnt;
4070         int err;
4071         do {
4072                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
4073                 trace_nfs4_lookup(dir, name, err);
4074                 switch (err) {
4075                 case -NFS4ERR_BADNAME:
4076                         err = -ENOENT;
4077                         goto out;
4078                 case -NFS4ERR_MOVED:
4079                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4080                         if (err == -NFS4ERR_MOVED)
4081                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4082                         goto out;
4083                 case -NFS4ERR_WRONGSEC:
4084                         err = -EPERM;
4085                         if (client != *clnt)
4086                                 goto out;
4087                         client = nfs4_negotiate_security(client, dir, name);
4088                         if (IS_ERR(client))
4089                                 return PTR_ERR(client);
4090
4091                         exception.retry = 1;
4092                         break;
4093                 default:
4094                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4095                 }
4096         } while (exception.retry);
4097
4098 out:
4099         if (err == 0)
4100                 *clnt = client;
4101         else if (client != *clnt)
4102                 rpc_shutdown_client(client);
4103
4104         return err;
4105 }
4106
4107 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
4108                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4109                             struct nfs4_label *label)
4110 {
4111         int status;
4112         struct rpc_clnt *client = NFS_CLIENT(dir);
4113
4114         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
4115         if (client != NFS_CLIENT(dir)) {
4116                 rpc_shutdown_client(client);
4117                 nfs_fixup_secinfo_attributes(fattr);
4118         }
4119         return status;
4120 }
4121
4122 struct rpc_clnt *
4123 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
4124                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4125 {
4126         struct rpc_clnt *client = NFS_CLIENT(dir);
4127         int status;
4128
4129         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
4130         if (status < 0)
4131                 return ERR_PTR(status);
4132         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4133 }
4134
4135 static int _nfs4_proc_lookupp(struct inode *inode,
4136                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4137                 struct nfs4_label *label)
4138 {
4139         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4140         struct nfs_server *server = NFS_SERVER(inode);
4141         int                    status;
4142         struct nfs4_lookupp_arg args = {
4143                 .bitmask = server->attr_bitmask,
4144                 .fh = NFS_FH(inode),
4145         };
4146         struct nfs4_lookupp_res res = {
4147                 .server = server,
4148                 .fattr = fattr,
4149                 .label = label,
4150                 .fh = fhandle,
4151         };
4152         struct rpc_message msg = {
4153                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4154                 .rpc_argp = &args,
4155                 .rpc_resp = &res,
4156         };
4157
4158         args.bitmask = nfs4_bitmask(server, label);
4159
4160         nfs_fattr_init(fattr);
4161
4162         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4163         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4164                                 &res.seq_res, 0);
4165         dprintk("NFS reply lookupp: %d\n", status);
4166         return status;
4167 }
4168
4169 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4170                              struct nfs_fattr *fattr, struct nfs4_label *label)
4171 {
4172         struct nfs4_exception exception = { };
4173         int err;
4174         do {
4175                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4176                 trace_nfs4_lookupp(inode, err);
4177                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4178                                 &exception);
4179         } while (exception.retry);
4180         return err;
4181 }
4182
4183 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4184 {
4185         struct nfs_server *server = NFS_SERVER(inode);
4186         struct nfs4_accessargs args = {
4187                 .fh = NFS_FH(inode),
4188                 .access = entry->mask,
4189         };
4190         struct nfs4_accessres res = {
4191                 .server = server,
4192         };
4193         struct rpc_message msg = {
4194                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4195                 .rpc_argp = &args,
4196                 .rpc_resp = &res,
4197                 .rpc_cred = entry->cred,
4198         };
4199         int status = 0;
4200
4201         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4202                 res.fattr = nfs_alloc_fattr();
4203                 if (res.fattr == NULL)
4204                         return -ENOMEM;
4205                 args.bitmask = server->cache_consistency_bitmask;
4206         }
4207         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4208         if (!status) {
4209                 nfs_access_set_mask(entry, res.access);
4210                 if (res.fattr)
4211                         nfs_refresh_inode(inode, res.fattr);
4212         }
4213         nfs_free_fattr(res.fattr);
4214         return status;
4215 }
4216
4217 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4218 {
4219         struct nfs4_exception exception = { };
4220         int err;
4221         do {
4222                 err = _nfs4_proc_access(inode, entry);
4223                 trace_nfs4_access(inode, err);
4224                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4225                                 &exception);
4226         } while (exception.retry);
4227         return err;
4228 }
4229
4230 /*
4231  * TODO: For the time being, we don't try to get any attributes
4232  * along with any of the zero-copy operations READ, READDIR,
4233  * READLINK, WRITE.
4234  *
4235  * In the case of the first three, we want to put the GETATTR
4236  * after the read-type operation -- this is because it is hard
4237  * to predict the length of a GETATTR response in v4, and thus
4238  * align the READ data correctly.  This means that the GETATTR
4239  * may end up partially falling into the page cache, and we should
4240  * shift it into the 'tail' of the xdr_buf before processing.
4241  * To do this efficiently, we need to know the total length
4242  * of data received, which doesn't seem to be available outside
4243  * of the RPC layer.
4244  *
4245  * In the case of WRITE, we also want to put the GETATTR after
4246  * the operation -- in this case because we want to make sure
4247  * we get the post-operation mtime and size.
4248  *
4249  * Both of these changes to the XDR layer would in fact be quite
4250  * minor, but I decided to leave them for a subsequent patch.
4251  */
4252 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4253                 unsigned int pgbase, unsigned int pglen)
4254 {
4255         struct nfs4_readlink args = {
4256                 .fh       = NFS_FH(inode),
4257                 .pgbase   = pgbase,
4258                 .pglen    = pglen,
4259                 .pages    = &page,
4260         };
4261         struct nfs4_readlink_res res;
4262         struct rpc_message msg = {
4263                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4264                 .rpc_argp = &args,
4265                 .rpc_resp = &res,
4266         };
4267
4268         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4269 }
4270
4271 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4272                 unsigned int pgbase, unsigned int pglen)
4273 {
4274         struct nfs4_exception exception = { };
4275         int err;
4276         do {
4277                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4278                 trace_nfs4_readlink(inode, err);
4279                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4280                                 &exception);
4281         } while (exception.retry);
4282         return err;
4283 }
4284
4285 /*
4286  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4287  */
4288 static int
4289 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4290                  int flags)
4291 {
4292         struct nfs_server *server = NFS_SERVER(dir);
4293         struct nfs4_label l, *ilabel = NULL;
4294         struct nfs_open_context *ctx;
4295         struct nfs4_state *state;
4296         int status = 0;
4297
4298         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4299         if (IS_ERR(ctx))
4300                 return PTR_ERR(ctx);
4301
4302         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4303
4304         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4305                 sattr->ia_mode &= ~current_umask();
4306         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4307         if (IS_ERR(state)) {
4308                 status = PTR_ERR(state);
4309                 goto out;
4310         }
4311 out:
4312         nfs4_label_release_security(ilabel);
4313         put_nfs_open_context(ctx);
4314         return status;
4315 }
4316
4317 static int
4318 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4319 {
4320         struct nfs_server *server = NFS_SERVER(dir);
4321         struct nfs_removeargs args = {
4322                 .fh = NFS_FH(dir),
4323                 .name = *name,
4324         };
4325         struct nfs_removeres res = {
4326                 .server = server,
4327         };
4328         struct rpc_message msg = {
4329                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4330                 .rpc_argp = &args,
4331                 .rpc_resp = &res,
4332         };
4333         unsigned long timestamp = jiffies;
4334         int status;
4335
4336         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4337         if (status == 0) {
4338                 spin_lock(&dir->i_lock);
4339                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4340                 /* Removing a directory decrements nlink in the parent */
4341                 if (ftype == NF4DIR && dir->i_nlink > 2)
4342                         nfs4_dec_nlink_locked(dir);
4343                 spin_unlock(&dir->i_lock);
4344         }
4345         return status;
4346 }
4347
4348 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4349 {
4350         struct nfs4_exception exception = { };
4351         struct inode *inode = d_inode(dentry);
4352         int err;
4353
4354         if (inode) {
4355                 if (inode->i_nlink == 1)
4356                         nfs4_inode_return_delegation(inode);
4357                 else
4358                         nfs4_inode_make_writeable(inode);
4359         }
4360         do {
4361                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4362                 trace_nfs4_remove(dir, &dentry->d_name, err);
4363                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4364                                 &exception);
4365         } while (exception.retry);
4366         return err;
4367 }
4368
4369 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4370 {
4371         struct nfs4_exception exception = { };
4372         int err;
4373
4374         do {
4375                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4376                 trace_nfs4_remove(dir, name, err);
4377                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4378                                 &exception);
4379         } while (exception.retry);
4380         return err;
4381 }
4382
4383 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4384                 struct dentry *dentry,
4385                 struct inode *inode)
4386 {
4387         struct nfs_removeargs *args = msg->rpc_argp;
4388         struct nfs_removeres *res = msg->rpc_resp;
4389
4390         res->server = NFS_SB(dentry->d_sb);
4391         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4392         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4393
4394         nfs_fattr_init(res->dir_attr);
4395
4396         if (inode)
4397                 nfs4_inode_return_delegation(inode);
4398 }
4399
4400 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4401 {
4402         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4403                         &data->args.seq_args,
4404                         &data->res.seq_res,
4405                         task);
4406 }
4407
4408 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4409 {
4410         struct nfs_unlinkdata *data = task->tk_calldata;
4411         struct nfs_removeres *res = &data->res;
4412
4413         if (!nfs4_sequence_done(task, &res->seq_res))
4414                 return 0;
4415         if (nfs4_async_handle_error(task, res->server, NULL,
4416                                     &data->timeout) == -EAGAIN)
4417                 return 0;
4418         if (task->tk_status == 0)
4419                 update_changeattr(dir, &res->cinfo,
4420                                 res->dir_attr->time_start, 0);
4421         return 1;
4422 }
4423
4424 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4425                 struct dentry *old_dentry,
4426                 struct dentry *new_dentry)
4427 {
4428         struct nfs_renameargs *arg = msg->rpc_argp;
4429         struct nfs_renameres *res = msg->rpc_resp;
4430         struct inode *old_inode = d_inode(old_dentry);
4431         struct inode *new_inode = d_inode(new_dentry);
4432
4433         if (old_inode)
4434                 nfs4_inode_make_writeable(old_inode);
4435         if (new_inode)
4436                 nfs4_inode_return_delegation(new_inode);
4437         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4438         res->server = NFS_SB(old_dentry->d_sb);
4439         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4440 }
4441
4442 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4443 {
4444         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4445                         &data->args.seq_args,
4446                         &data->res.seq_res,
4447                         task);
4448 }
4449
4450 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4451                                  struct inode *new_dir)
4452 {
4453         struct nfs_renamedata *data = task->tk_calldata;
4454         struct nfs_renameres *res = &data->res;
4455
4456         if (!nfs4_sequence_done(task, &res->seq_res))
4457                 return 0;
4458         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4459                 return 0;
4460
4461         if (task->tk_status == 0) {
4462                 if (new_dir != old_dir) {
4463                         /* Note: If we moved a directory, nlink will change */
4464                         update_changeattr(old_dir, &res->old_cinfo,
4465                                         res->old_fattr->time_start,
4466                                         NFS_INO_INVALID_OTHER);
4467                         update_changeattr(new_dir, &res->new_cinfo,
4468                                         res->new_fattr->time_start,
4469                                         NFS_INO_INVALID_OTHER);
4470                 } else
4471                         update_changeattr(old_dir, &res->old_cinfo,
4472                                         res->old_fattr->time_start,
4473                                         0);
4474         }
4475         return 1;
4476 }
4477
4478 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4479 {
4480         struct nfs_server *server = NFS_SERVER(inode);
4481         __u32 bitmask[NFS4_BITMASK_SZ];
4482         struct nfs4_link_arg arg = {
4483                 .fh     = NFS_FH(inode),
4484                 .dir_fh = NFS_FH(dir),
4485                 .name   = name,
4486                 .bitmask = bitmask,
4487         };
4488         struct nfs4_link_res res = {
4489                 .server = server,
4490                 .label = NULL,
4491         };
4492         struct rpc_message msg = {
4493                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4494                 .rpc_argp = &arg,
4495                 .rpc_resp = &res,
4496         };
4497         int status = -ENOMEM;
4498
4499         res.fattr = nfs_alloc_fattr();
4500         if (res.fattr == NULL)
4501                 goto out;
4502
4503         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4504         if (IS_ERR(res.label)) {
4505                 status = PTR_ERR(res.label);
4506                 goto out;
4507         }
4508
4509         nfs4_inode_make_writeable(inode);
4510         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4511
4512         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4513         if (!status) {
4514                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4515                 status = nfs_post_op_update_inode(inode, res.fattr);
4516                 if (!status)
4517                         nfs_setsecurity(inode, res.fattr, res.label);
4518         }
4519
4520
4521         nfs4_label_free(res.label);
4522
4523 out:
4524         nfs_free_fattr(res.fattr);
4525         return status;
4526 }
4527
4528 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4529 {
4530         struct nfs4_exception exception = { };
4531         int err;
4532         do {
4533                 err = nfs4_handle_exception(NFS_SERVER(inode),
4534                                 _nfs4_proc_link(inode, dir, name),
4535                                 &exception);
4536         } while (exception.retry);
4537         return err;
4538 }
4539
4540 struct nfs4_createdata {
4541         struct rpc_message msg;
4542         struct nfs4_create_arg arg;
4543         struct nfs4_create_res res;
4544         struct nfs_fh fh;
4545         struct nfs_fattr fattr;
4546         struct nfs4_label *label;
4547 };
4548
4549 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4550                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4551 {
4552         struct nfs4_createdata *data;
4553
4554         data = kzalloc(sizeof(*data), GFP_KERNEL);
4555         if (data != NULL) {
4556                 struct nfs_server *server = NFS_SERVER(dir);
4557
4558                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4559                 if (IS_ERR(data->label))
4560                         goto out_free;
4561
4562                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4563                 data->msg.rpc_argp = &data->arg;
4564                 data->msg.rpc_resp = &data->res;
4565                 data->arg.dir_fh = NFS_FH(dir);
4566                 data->arg.server = server;
4567                 data->arg.name = name;
4568                 data->arg.attrs = sattr;
4569                 data->arg.ftype = ftype;
4570                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4571                 data->arg.umask = current_umask();
4572                 data->res.server = server;
4573                 data->res.fh = &data->fh;
4574                 data->res.fattr = &data->fattr;
4575                 data->res.label = data->label;
4576                 nfs_fattr_init(data->res.fattr);
4577         }
4578         return data;
4579 out_free:
4580         kfree(data);
4581         return NULL;
4582 }
4583
4584 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4585 {
4586         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4587                                     &data->arg.seq_args, &data->res.seq_res, 1);
4588         if (status == 0) {
4589                 spin_lock(&dir->i_lock);
4590                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4591                                 data->res.fattr->time_start, 0);
4592                 /* Creating a directory bumps nlink in the parent */
4593                 if (data->arg.ftype == NF4DIR)
4594                         nfs4_inc_nlink_locked(dir);
4595                 spin_unlock(&dir->i_lock);
4596                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4597         }
4598         return status;
4599 }
4600
4601 static void nfs4_free_createdata(struct nfs4_createdata *data)
4602 {
4603         nfs4_label_free(data->label);
4604         kfree(data);
4605 }
4606
4607 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4608                 struct page *page, unsigned int len, struct iattr *sattr,
4609                 struct nfs4_label *label)
4610 {
4611         struct nfs4_createdata *data;
4612         int status = -ENAMETOOLONG;
4613
4614         if (len > NFS4_MAXPATHLEN)
4615                 goto out;
4616
4617         status = -ENOMEM;
4618         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4619         if (data == NULL)
4620                 goto out;
4621
4622         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4623         data->arg.u.symlink.pages = &page;
4624         data->arg.u.symlink.len = len;
4625         data->arg.label = label;
4626
4627         status = nfs4_do_create(dir, dentry, data);
4628
4629         nfs4_free_createdata(data);
4630 out:
4631         return status;
4632 }
4633
4634 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4635                 struct page *page, unsigned int len, struct iattr *sattr)
4636 {
4637         struct nfs4_exception exception = { };
4638         struct nfs4_label l, *label = NULL;
4639         int err;
4640
4641         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4642
4643         do {
4644                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4645                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4646                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4647                                 &exception);
4648         } while (exception.retry);
4649
4650         nfs4_label_release_security(label);
4651         return err;
4652 }
4653
4654 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4655                 struct iattr *sattr, struct nfs4_label *label)
4656 {
4657         struct nfs4_createdata *data;
4658         int status = -ENOMEM;
4659
4660         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4661         if (data == NULL)
4662                 goto out;
4663
4664         data->arg.label = label;
4665         status = nfs4_do_create(dir, dentry, data);
4666
4667         nfs4_free_createdata(data);
4668 out:
4669         return status;
4670 }
4671
4672 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4673                 struct iattr *sattr)
4674 {
4675         struct nfs_server *server = NFS_SERVER(dir);
4676         struct nfs4_exception exception = { };
4677         struct nfs4_label l, *label = NULL;
4678         int err;
4679
4680         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4681
4682         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4683                 sattr->ia_mode &= ~current_umask();
4684         do {
4685                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4686                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4687                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4688                                 &exception);
4689         } while (exception.retry);
4690         nfs4_label_release_security(label);
4691
4692         return err;
4693 }
4694
4695 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4696                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4697 {
4698         struct inode            *dir = d_inode(dentry);
4699         struct nfs4_readdir_arg args = {
4700                 .fh = NFS_FH(dir),
4701                 .pages = pages,
4702                 .pgbase = 0,
4703                 .count = count,
4704                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4705                 .plus = plus,
4706         };
4707         struct nfs4_readdir_res res;
4708         struct rpc_message msg = {
4709                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4710                 .rpc_argp = &args,
4711                 .rpc_resp = &res,
4712                 .rpc_cred = cred,
4713         };
4714         int                     status;
4715
4716         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4717                         dentry,
4718                         (unsigned long long)cookie);
4719         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4720         res.pgbase = args.pgbase;
4721         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4722         if (status >= 0) {
4723                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4724                 status += args.pgbase;
4725         }
4726
4727         nfs_invalidate_atime(dir);
4728
4729         dprintk("%s: returns %d\n", __func__, status);
4730         return status;
4731 }
4732
4733 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4734                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4735 {
4736         struct nfs4_exception exception = { };
4737         int err;
4738         do {
4739                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4740                                 pages, count, plus);
4741                 trace_nfs4_readdir(d_inode(dentry), err);
4742                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4743                                 &exception);
4744         } while (exception.retry);
4745         return err;
4746 }
4747
4748 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4749                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4750 {
4751         struct nfs4_createdata *data;
4752         int mode = sattr->ia_mode;
4753         int status = -ENOMEM;
4754
4755         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4756         if (data == NULL)
4757                 goto out;
4758
4759         if (S_ISFIFO(mode))
4760                 data->arg.ftype = NF4FIFO;
4761         else if (S_ISBLK(mode)) {
4762                 data->arg.ftype = NF4BLK;
4763                 data->arg.u.device.specdata1 = MAJOR(rdev);
4764                 data->arg.u.device.specdata2 = MINOR(rdev);
4765         }
4766         else if (S_ISCHR(mode)) {
4767                 data->arg.ftype = NF4CHR;
4768                 data->arg.u.device.specdata1 = MAJOR(rdev);
4769                 data->arg.u.device.specdata2 = MINOR(rdev);
4770         } else if (!S_ISSOCK(mode)) {
4771                 status = -EINVAL;
4772                 goto out_free;
4773         }
4774
4775         data->arg.label = label;
4776         status = nfs4_do_create(dir, dentry, data);
4777 out_free:
4778         nfs4_free_createdata(data);
4779 out:
4780         return status;
4781 }
4782
4783 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4784                 struct iattr *sattr, dev_t rdev)
4785 {
4786         struct nfs_server *server = NFS_SERVER(dir);
4787         struct nfs4_exception exception = { };
4788         struct nfs4_label l, *label = NULL;
4789         int err;
4790
4791         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4792
4793         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4794                 sattr->ia_mode &= ~current_umask();
4795         do {
4796                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4797                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4798                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4799                                 &exception);
4800         } while (exception.retry);
4801
4802         nfs4_label_release_security(label);
4803
4804         return err;
4805 }
4806
4807 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4808                  struct nfs_fsstat *fsstat)
4809 {
4810         struct nfs4_statfs_arg args = {
4811                 .fh = fhandle,
4812                 .bitmask = server->attr_bitmask,
4813         };
4814         struct nfs4_statfs_res res = {
4815                 .fsstat = fsstat,
4816         };
4817         struct rpc_message msg = {
4818                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4819                 .rpc_argp = &args,
4820                 .rpc_resp = &res,
4821         };
4822
4823         nfs_fattr_init(fsstat->fattr);
4824         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4825 }
4826
4827 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4828 {
4829         struct nfs4_exception exception = { };
4830         int err;
4831         do {
4832                 err = nfs4_handle_exception(server,
4833                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4834                                 &exception);
4835         } while (exception.retry);
4836         return err;
4837 }
4838
4839 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4840                 struct nfs_fsinfo *fsinfo)
4841 {
4842         struct nfs4_fsinfo_arg args = {
4843                 .fh = fhandle,
4844                 .bitmask = server->attr_bitmask,
4845         };
4846         struct nfs4_fsinfo_res res = {
4847                 .fsinfo = fsinfo,
4848         };
4849         struct rpc_message msg = {
4850                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4851                 .rpc_argp = &args,
4852                 .rpc_resp = &res,
4853         };
4854
4855         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4856 }
4857
4858 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4859 {
4860         struct nfs4_exception exception = { };
4861         unsigned long now = jiffies;
4862         int err;
4863
4864         do {
4865                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4866                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4867                 if (err == 0) {
4868                         nfs4_set_lease_period(server->nfs_client,
4869                                         fsinfo->lease_time * HZ,
4870                                         now);
4871                         break;
4872                 }
4873                 err = nfs4_handle_exception(server, err, &exception);
4874         } while (exception.retry);
4875         return err;
4876 }
4877
4878 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4879 {
4880         int error;
4881
4882         nfs_fattr_init(fsinfo->fattr);
4883         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4884         if (error == 0) {
4885                 /* block layout checks this! */
4886                 server->pnfs_blksize = fsinfo->blksize;
4887                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4888         }
4889
4890         return error;
4891 }
4892
4893 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4894                 struct nfs_pathconf *pathconf)
4895 {
4896         struct nfs4_pathconf_arg args = {
4897                 .fh = fhandle,
4898                 .bitmask = server->attr_bitmask,
4899         };
4900         struct nfs4_pathconf_res res = {
4901                 .pathconf = pathconf,
4902         };
4903         struct rpc_message msg = {
4904                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4905                 .rpc_argp = &args,
4906                 .rpc_resp = &res,
4907         };
4908
4909         /* None of the pathconf attributes are mandatory to implement */
4910         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4911                 memset(pathconf, 0, sizeof(*pathconf));
4912                 return 0;
4913         }
4914
4915         nfs_fattr_init(pathconf->fattr);
4916         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4917 }
4918
4919 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4920                 struct nfs_pathconf *pathconf)
4921 {
4922         struct nfs4_exception exception = { };
4923         int err;
4924
4925         do {
4926                 err = nfs4_handle_exception(server,
4927                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4928                                 &exception);
4929         } while (exception.retry);
4930         return err;
4931 }
4932
4933 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4934                 const struct nfs_open_context *ctx,
4935                 const struct nfs_lock_context *l_ctx,
4936                 fmode_t fmode)
4937 {
4938         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4939 }
4940 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4941
4942 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4943                 const struct nfs_open_context *ctx,
4944                 const struct nfs_lock_context *l_ctx,
4945                 fmode_t fmode)
4946 {
4947         nfs4_stateid current_stateid;
4948
4949         /* If the current stateid represents a lost lock, then exit */
4950         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4951                 return true;
4952         return nfs4_stateid_match(stateid, &current_stateid);
4953 }
4954
4955 static bool nfs4_error_stateid_expired(int err)
4956 {
4957         switch (err) {
4958         case -NFS4ERR_DELEG_REVOKED:
4959         case -NFS4ERR_ADMIN_REVOKED:
4960         case -NFS4ERR_BAD_STATEID:
4961         case -NFS4ERR_STALE_STATEID:
4962         case -NFS4ERR_OLD_STATEID:
4963         case -NFS4ERR_OPENMODE:
4964         case -NFS4ERR_EXPIRED:
4965                 return true;
4966         }
4967         return false;
4968 }
4969
4970 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4971 {
4972         struct nfs_server *server = NFS_SERVER(hdr->inode);
4973
4974         trace_nfs4_read(hdr, task->tk_status);
4975         if (task->tk_status < 0) {
4976                 struct nfs4_exception exception = {
4977                         .inode = hdr->inode,
4978                         .state = hdr->args.context->state,
4979                         .stateid = &hdr->args.stateid,
4980                 };
4981                 task->tk_status = nfs4_async_handle_exception(task,
4982                                 server, task->tk_status, &exception);
4983                 if (exception.retry) {
4984                         rpc_restart_call_prepare(task);
4985                         return -EAGAIN;
4986                 }
4987         }
4988
4989         if (task->tk_status > 0)
4990                 renew_lease(server, hdr->timestamp);
4991         return 0;
4992 }
4993
4994 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4995                 struct nfs_pgio_args *args)
4996 {
4997
4998         if (!nfs4_error_stateid_expired(task->tk_status) ||
4999                 nfs4_stateid_is_current(&args->stateid,
5000                                 args->context,
5001                                 args->lock_context,
5002                                 FMODE_READ))
5003                 return false;
5004         rpc_restart_call_prepare(task);
5005         return true;
5006 }
5007
5008 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5009 {
5010
5011         dprintk("--> %s\n", __func__);
5012
5013         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5014                 return -EAGAIN;
5015         if (nfs4_read_stateid_changed(task, &hdr->args))
5016                 return -EAGAIN;
5017         if (task->tk_status > 0)
5018                 nfs_invalidate_atime(hdr->inode);
5019         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5020                                     nfs4_read_done_cb(task, hdr);
5021 }
5022
5023 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5024                                  struct rpc_message *msg)
5025 {
5026         hdr->timestamp   = jiffies;
5027         if (!hdr->pgio_done_cb)
5028                 hdr->pgio_done_cb = nfs4_read_done_cb;
5029         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5030         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5031 }
5032
5033 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5034                                       struct nfs_pgio_header *hdr)
5035 {
5036         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5037                         &hdr->args.seq_args,
5038                         &hdr->res.seq_res,
5039                         task))
5040                 return 0;
5041         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5042                                 hdr->args.lock_context,
5043                                 hdr->rw_mode) == -EIO)
5044                 return -EIO;
5045         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5046                 return -EIO;
5047         return 0;
5048 }
5049
5050 static int nfs4_write_done_cb(struct rpc_task *task,
5051                               struct nfs_pgio_header *hdr)
5052 {
5053         struct inode *inode = hdr->inode;
5054
5055         trace_nfs4_write(hdr, task->tk_status);
5056         if (task->tk_status < 0) {
5057                 struct nfs4_exception exception = {
5058                         .inode = hdr->inode,
5059                         .state = hdr->args.context->state,
5060                         .stateid = &hdr->args.stateid,
5061                 };
5062                 task->tk_status = nfs4_async_handle_exception(task,
5063                                 NFS_SERVER(inode), task->tk_status,
5064                                 &exception);
5065                 if (exception.retry) {
5066                         rpc_restart_call_prepare(task);
5067                         return -EAGAIN;
5068                 }
5069         }
5070         if (task->tk_status >= 0) {
5071                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5072                 nfs_writeback_update_inode(hdr);
5073         }
5074         return 0;
5075 }
5076
5077 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5078                 struct nfs_pgio_args *args)
5079 {
5080
5081         if (!nfs4_error_stateid_expired(task->tk_status) ||
5082                 nfs4_stateid_is_current(&args->stateid,
5083                                 args->context,
5084                                 args->lock_context,
5085                                 FMODE_WRITE))
5086                 return false;
5087         rpc_restart_call_prepare(task);
5088         return true;
5089 }
5090
5091 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5092 {
5093         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5094                 return -EAGAIN;
5095         if (nfs4_write_stateid_changed(task, &hdr->args))
5096                 return -EAGAIN;
5097         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5098                 nfs4_write_done_cb(task, hdr);
5099 }
5100
5101 static
5102 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5103 {
5104         /* Don't request attributes for pNFS or O_DIRECT writes */
5105         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5106                 return false;
5107         /* Otherwise, request attributes if and only if we don't hold
5108          * a delegation
5109          */
5110         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5111 }
5112
5113 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5114                                   struct rpc_message *msg,
5115                                   struct rpc_clnt **clnt)
5116 {
5117         struct nfs_server *server = NFS_SERVER(hdr->inode);
5118
5119         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5120                 hdr->args.bitmask = NULL;
5121                 hdr->res.fattr = NULL;
5122         } else
5123                 hdr->args.bitmask = server->cache_consistency_bitmask;
5124
5125         if (!hdr->pgio_done_cb)
5126                 hdr->pgio_done_cb = nfs4_write_done_cb;
5127         hdr->res.server = server;
5128         hdr->timestamp   = jiffies;
5129
5130         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5131         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1, 0);
5132         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5133 }
5134
5135 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5136 {
5137         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5138                         &data->args.seq_args,
5139                         &data->res.seq_res,
5140                         task);
5141 }
5142
5143 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5144 {
5145         struct inode *inode = data->inode;
5146
5147         trace_nfs4_commit(data, task->tk_status);
5148         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5149                                     NULL, NULL) == -EAGAIN) {
5150                 rpc_restart_call_prepare(task);
5151                 return -EAGAIN;
5152         }
5153         return 0;
5154 }
5155
5156 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5157 {
5158         if (!nfs4_sequence_done(task, &data->res.seq_res))
5159                 return -EAGAIN;
5160         return data->commit_done_cb(task, data);
5161 }
5162
5163 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5164                                    struct rpc_clnt **clnt)
5165 {
5166         struct nfs_server *server = NFS_SERVER(data->inode);
5167
5168         if (data->commit_done_cb == NULL)
5169                 data->commit_done_cb = nfs4_commit_done_cb;
5170         data->res.server = server;
5171         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5172         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5173         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5174 }
5175
5176 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5177                                 struct nfs_commitres *res)
5178 {
5179         struct inode *dst_inode = file_inode(dst);
5180         struct nfs_server *server = NFS_SERVER(dst_inode);
5181         struct rpc_message msg = {
5182                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5183                 .rpc_argp = args,
5184                 .rpc_resp = res,
5185         };
5186
5187         args->fh = NFS_FH(dst_inode);
5188         return nfs4_call_sync(server->client, server, &msg,
5189                         &args->seq_args, &res->seq_res, 1);
5190 }
5191
5192 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5193 {
5194         struct nfs_commitargs args = {
5195                 .offset = offset,
5196                 .count = count,
5197         };
5198         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5199         struct nfs4_exception exception = { };
5200         int status;
5201
5202         do {
5203                 status = _nfs4_proc_commit(dst, &args, res);
5204                 status = nfs4_handle_exception(dst_server, status, &exception);
5205         } while (exception.retry);
5206
5207         return status;
5208 }
5209
5210 struct nfs4_renewdata {
5211         struct nfs_client       *client;
5212         unsigned long           timestamp;
5213 };
5214
5215 /*
5216  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5217  * standalone procedure for queueing an asynchronous RENEW.
5218  */
5219 static void nfs4_renew_release(void *calldata)
5220 {
5221         struct nfs4_renewdata *data = calldata;
5222         struct nfs_client *clp = data->client;
5223
5224         if (refcount_read(&clp->cl_count) > 1)
5225                 nfs4_schedule_state_renewal(clp);
5226         nfs_put_client(clp);
5227         kfree(data);
5228 }
5229
5230 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5231 {
5232         struct nfs4_renewdata *data = calldata;
5233         struct nfs_client *clp = data->client;
5234         unsigned long timestamp = data->timestamp;
5235
5236         trace_nfs4_renew_async(clp, task->tk_status);
5237         switch (task->tk_status) {
5238         case 0:
5239                 break;
5240         case -NFS4ERR_LEASE_MOVED:
5241                 nfs4_schedule_lease_moved_recovery(clp);
5242                 break;
5243         default:
5244                 /* Unless we're shutting down, schedule state recovery! */
5245                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5246                         return;
5247                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5248                         nfs4_schedule_lease_recovery(clp);
5249                         return;
5250                 }
5251                 nfs4_schedule_path_down_recovery(clp);
5252         }
5253         do_renew_lease(clp, timestamp);
5254 }
5255
5256 static const struct rpc_call_ops nfs4_renew_ops = {
5257         .rpc_call_done = nfs4_renew_done,
5258         .rpc_release = nfs4_renew_release,
5259 };
5260
5261 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5262 {
5263         struct rpc_message msg = {
5264                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5265                 .rpc_argp       = clp,
5266                 .rpc_cred       = cred,
5267         };
5268         struct nfs4_renewdata *data;
5269
5270         if (renew_flags == 0)
5271                 return 0;
5272         if (!refcount_inc_not_zero(&clp->cl_count))
5273                 return -EIO;
5274         data = kmalloc(sizeof(*data), GFP_NOFS);
5275         if (data == NULL) {
5276                 nfs_put_client(clp);
5277                 return -ENOMEM;
5278         }
5279         data->client = clp;
5280         data->timestamp = jiffies;
5281         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5282                         &nfs4_renew_ops, data);
5283 }
5284
5285 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5286 {
5287         struct rpc_message msg = {
5288                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5289                 .rpc_argp       = clp,
5290                 .rpc_cred       = cred,
5291         };
5292         unsigned long now = jiffies;
5293         int status;
5294
5295         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5296         if (status < 0)
5297                 return status;
5298         do_renew_lease(clp, now);
5299         return 0;
5300 }
5301
5302 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5303 {
5304         return server->caps & NFS_CAP_ACLS;
5305 }
5306
5307 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5308  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5309  * the stack.
5310  */
5311 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5312
5313 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5314                 struct page **pages)
5315 {
5316         struct page *newpage, **spages;
5317         int rc = 0;
5318         size_t len;
5319         spages = pages;
5320
5321         do {
5322                 len = min_t(size_t, PAGE_SIZE, buflen);
5323                 newpage = alloc_page(GFP_KERNEL);
5324
5325                 if (newpage == NULL)
5326                         goto unwind;
5327                 memcpy(page_address(newpage), buf, len);
5328                 buf += len;
5329                 buflen -= len;
5330                 *pages++ = newpage;
5331                 rc++;
5332         } while (buflen != 0);
5333
5334         return rc;
5335
5336 unwind:
5337         for(; rc > 0; rc--)
5338                 __free_page(spages[rc-1]);
5339         return -ENOMEM;
5340 }
5341
5342 struct nfs4_cached_acl {
5343         int cached;
5344         size_t len;
5345         char data[0];
5346 };
5347
5348 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5349 {
5350         struct nfs_inode *nfsi = NFS_I(inode);
5351
5352         spin_lock(&inode->i_lock);
5353         kfree(nfsi->nfs4_acl);
5354         nfsi->nfs4_acl = acl;
5355         spin_unlock(&inode->i_lock);
5356 }
5357
5358 static void nfs4_zap_acl_attr(struct inode *inode)
5359 {
5360         nfs4_set_cached_acl(inode, NULL);
5361 }
5362
5363 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5364 {
5365         struct nfs_inode *nfsi = NFS_I(inode);
5366         struct nfs4_cached_acl *acl;
5367         int ret = -ENOENT;
5368
5369         spin_lock(&inode->i_lock);
5370         acl = nfsi->nfs4_acl;
5371         if (acl == NULL)
5372                 goto out;
5373         if (buf == NULL) /* user is just asking for length */
5374                 goto out_len;
5375         if (acl->cached == 0)
5376                 goto out;
5377         ret = -ERANGE; /* see getxattr(2) man page */
5378         if (acl->len > buflen)
5379                 goto out;
5380         memcpy(buf, acl->data, acl->len);
5381 out_len:
5382         ret = acl->len;
5383 out:
5384         spin_unlock(&inode->i_lock);
5385         return ret;
5386 }
5387
5388 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5389 {
5390         struct nfs4_cached_acl *acl;
5391         size_t buflen = sizeof(*acl) + acl_len;
5392
5393         if (buflen <= PAGE_SIZE) {
5394                 acl = kmalloc(buflen, GFP_KERNEL);
5395                 if (acl == NULL)
5396                         goto out;
5397                 acl->cached = 1;
5398                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5399         } else {
5400                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5401                 if (acl == NULL)
5402                         goto out;
5403                 acl->cached = 0;
5404         }
5405         acl->len = acl_len;
5406 out:
5407         nfs4_set_cached_acl(inode, acl);
5408 }
5409
5410 /*
5411  * The getxattr API returns the required buffer length when called with a
5412  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5413  * the required buf.  On a NULL buf, we send a page of data to the server
5414  * guessing that the ACL request can be serviced by a page. If so, we cache
5415  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5416  * the cache. If not so, we throw away the page, and cache the required
5417  * length. The next getxattr call will then produce another round trip to
5418  * the server, this time with the input buf of the required size.
5419  */
5420 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5421 {
5422         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5423         struct nfs_getaclargs args = {
5424                 .fh = NFS_FH(inode),
5425                 .acl_pages = pages,
5426                 .acl_len = buflen,
5427         };
5428         struct nfs_getaclres res = {
5429                 .acl_len = buflen,
5430         };
5431         struct rpc_message msg = {
5432                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5433                 .rpc_argp = &args,
5434                 .rpc_resp = &res,
5435         };
5436         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5437         int ret = -ENOMEM, i;
5438
5439         if (npages > ARRAY_SIZE(pages))
5440                 return -ERANGE;
5441
5442         for (i = 0; i < npages; i++) {
5443                 pages[i] = alloc_page(GFP_KERNEL);
5444                 if (!pages[i])
5445                         goto out_free;
5446         }
5447
5448         /* for decoding across pages */
5449         res.acl_scratch = alloc_page(GFP_KERNEL);
5450         if (!res.acl_scratch)
5451                 goto out_free;
5452
5453         args.acl_len = npages * PAGE_SIZE;
5454
5455         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5456                 __func__, buf, buflen, npages, args.acl_len);
5457         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5458                              &msg, &args.seq_args, &res.seq_res, 0);
5459         if (ret)
5460                 goto out_free;
5461
5462         /* Handle the case where the passed-in buffer is too short */
5463         if (res.acl_flags & NFS4_ACL_TRUNC) {
5464                 /* Did the user only issue a request for the acl length? */
5465                 if (buf == NULL)
5466                         goto out_ok;
5467                 ret = -ERANGE;
5468                 goto out_free;
5469         }
5470         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5471         if (buf) {
5472                 if (res.acl_len > buflen) {
5473                         ret = -ERANGE;
5474                         goto out_free;
5475                 }
5476                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5477         }
5478 out_ok:
5479         ret = res.acl_len;
5480 out_free:
5481         for (i = 0; i < npages; i++)
5482                 if (pages[i])
5483                         __free_page(pages[i]);
5484         if (res.acl_scratch)
5485                 __free_page(res.acl_scratch);
5486         return ret;
5487 }
5488
5489 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5490 {
5491         struct nfs4_exception exception = { };
5492         ssize_t ret;
5493         do {
5494                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5495                 trace_nfs4_get_acl(inode, ret);
5496                 if (ret >= 0)
5497                         break;
5498                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5499         } while (exception.retry);
5500         return ret;
5501 }
5502
5503 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5504 {
5505         struct nfs_server *server = NFS_SERVER(inode);
5506         int ret;
5507
5508         if (!nfs4_server_supports_acls(server))
5509                 return -EOPNOTSUPP;
5510         ret = nfs_revalidate_inode(server, inode);
5511         if (ret < 0)
5512                 return ret;
5513         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5514                 nfs_zap_acl_cache(inode);
5515         ret = nfs4_read_cached_acl(inode, buf, buflen);
5516         if (ret != -ENOENT)
5517                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5518                  * but no cached acl data, just the acl length */
5519                 return ret;
5520         return nfs4_get_acl_uncached(inode, buf, buflen);
5521 }
5522
5523 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5524 {
5525         struct nfs_server *server = NFS_SERVER(inode);
5526         struct page *pages[NFS4ACL_MAXPAGES];
5527         struct nfs_setaclargs arg = {
5528                 .fh             = NFS_FH(inode),
5529                 .acl_pages      = pages,
5530                 .acl_len        = buflen,
5531         };
5532         struct nfs_setaclres res;
5533         struct rpc_message msg = {
5534                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5535                 .rpc_argp       = &arg,
5536                 .rpc_resp       = &res,
5537         };
5538         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5539         int ret, i;
5540
5541         if (!nfs4_server_supports_acls(server))
5542                 return -EOPNOTSUPP;
5543         if (npages > ARRAY_SIZE(pages))
5544                 return -ERANGE;
5545         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5546         if (i < 0)
5547                 return i;
5548         nfs4_inode_make_writeable(inode);
5549         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5550
5551         /*
5552          * Free each page after tx, so the only ref left is
5553          * held by the network stack
5554          */
5555         for (; i > 0; i--)
5556                 put_page(pages[i-1]);
5557
5558         /*
5559          * Acl update can result in inode attribute update.
5560          * so mark the attribute cache invalid.
5561          */
5562         spin_lock(&inode->i_lock);
5563         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5564                 | NFS_INO_INVALID_CTIME
5565                 | NFS_INO_REVAL_FORCED;
5566         spin_unlock(&inode->i_lock);
5567         nfs_access_zap_cache(inode);
5568         nfs_zap_acl_cache(inode);
5569         return ret;
5570 }
5571
5572 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5573 {
5574         struct nfs4_exception exception = { };
5575         int err;
5576         do {
5577                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5578                 trace_nfs4_set_acl(inode, err);
5579                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5580                                 &exception);
5581         } while (exception.retry);
5582         return err;
5583 }
5584
5585 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5586 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5587                                         size_t buflen)
5588 {
5589         struct nfs_server *server = NFS_SERVER(inode);
5590         struct nfs_fattr fattr;
5591         struct nfs4_label label = {0, 0, buflen, buf};
5592
5593         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5594         struct nfs4_getattr_arg arg = {
5595                 .fh             = NFS_FH(inode),
5596                 .bitmask        = bitmask,
5597         };
5598         struct nfs4_getattr_res res = {
5599                 .fattr          = &fattr,
5600                 .label          = &label,
5601                 .server         = server,
5602         };
5603         struct rpc_message msg = {
5604                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5605                 .rpc_argp       = &arg,
5606                 .rpc_resp       = &res,
5607         };
5608         int ret;
5609
5610         nfs_fattr_init(&fattr);
5611
5612         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5613         if (ret)
5614                 return ret;
5615         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5616                 return -ENOENT;
5617         if (buflen < label.len)
5618                 return -ERANGE;
5619         return 0;
5620 }
5621
5622 static int nfs4_get_security_label(struct inode *inode, void *buf,
5623                                         size_t buflen)
5624 {
5625         struct nfs4_exception exception = { };
5626         int err;
5627
5628         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5629                 return -EOPNOTSUPP;
5630
5631         do {
5632                 err = _nfs4_get_security_label(inode, buf, buflen);
5633                 trace_nfs4_get_security_label(inode, err);
5634                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5635                                 &exception);
5636         } while (exception.retry);
5637         return err;
5638 }
5639
5640 static int _nfs4_do_set_security_label(struct inode *inode,
5641                 struct nfs4_label *ilabel,
5642                 struct nfs_fattr *fattr,
5643                 struct nfs4_label *olabel)
5644 {
5645
5646         struct iattr sattr = {0};
5647         struct nfs_server *server = NFS_SERVER(inode);
5648         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5649         struct nfs_setattrargs arg = {
5650                 .fh             = NFS_FH(inode),
5651                 .iap            = &sattr,
5652                 .server         = server,
5653                 .bitmask        = bitmask,
5654                 .label          = ilabel,
5655         };
5656         struct nfs_setattrres res = {
5657                 .fattr          = fattr,
5658                 .label          = olabel,
5659                 .server         = server,
5660         };
5661         struct rpc_message msg = {
5662                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5663                 .rpc_argp       = &arg,
5664                 .rpc_resp       = &res,
5665         };
5666         int status;
5667
5668         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5669
5670         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5671         if (status)
5672                 dprintk("%s failed: %d\n", __func__, status);
5673
5674         return status;
5675 }
5676
5677 static int nfs4_do_set_security_label(struct inode *inode,
5678                 struct nfs4_label *ilabel,
5679                 struct nfs_fattr *fattr,
5680                 struct nfs4_label *olabel)
5681 {
5682         struct nfs4_exception exception = { };
5683         int err;
5684
5685         do {
5686                 err = _nfs4_do_set_security_label(inode, ilabel,
5687                                 fattr, olabel);
5688                 trace_nfs4_set_security_label(inode, err);
5689                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5690                                 &exception);
5691         } while (exception.retry);
5692         return err;
5693 }
5694
5695 static int
5696 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5697 {
5698         struct nfs4_label ilabel, *olabel = NULL;
5699         struct nfs_fattr fattr;
5700         int status;
5701
5702         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5703                 return -EOPNOTSUPP;
5704
5705         nfs_fattr_init(&fattr);
5706
5707         ilabel.pi = 0;
5708         ilabel.lfs = 0;
5709         ilabel.label = (char *)buf;
5710         ilabel.len = buflen;
5711
5712         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5713         if (IS_ERR(olabel)) {
5714                 status = -PTR_ERR(olabel);
5715                 goto out;
5716         }
5717
5718         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5719         if (status == 0)
5720                 nfs_setsecurity(inode, &fattr, olabel);
5721
5722         nfs4_label_free(olabel);
5723 out:
5724         return status;
5725 }
5726 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5727
5728
5729 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5730                                     nfs4_verifier *bootverf)
5731 {
5732         __be32 verf[2];
5733
5734         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5735                 /* An impossible timestamp guarantees this value
5736                  * will never match a generated boot time. */
5737                 verf[0] = cpu_to_be32(U32_MAX);
5738                 verf[1] = cpu_to_be32(U32_MAX);
5739         } else {
5740                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5741                 u64 ns = ktime_to_ns(nn->boot_time);
5742
5743                 verf[0] = cpu_to_be32(ns >> 32);
5744                 verf[1] = cpu_to_be32(ns);
5745         }
5746         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5747 }
5748
5749 static int
5750 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5751 {
5752         size_t len;
5753         char *str;
5754
5755         if (clp->cl_owner_id != NULL)
5756                 return 0;
5757
5758         rcu_read_lock();
5759         len = 14 +
5760                 strlen(clp->cl_rpcclient->cl_nodename) +
5761                 1 +
5762                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5763                 1;
5764         rcu_read_unlock();
5765         if (nfs4_client_id_uniquifier[0] != '\0')
5766                 len += strlen(nfs4_client_id_uniquifier) + 1;
5767         if (len > NFS4_OPAQUE_LIMIT + 1)
5768                 return -EINVAL;
5769
5770         /*
5771          * Since this string is allocated at mount time, and held until the
5772          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5773          * about a memory-reclaim deadlock.
5774          */
5775         str = kmalloc(len, GFP_KERNEL);
5776         if (!str)
5777                 return -ENOMEM;
5778
5779         rcu_read_lock();
5780         if (nfs4_client_id_uniquifier[0] != '\0')
5781                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5782                           clp->cl_rpcclient->cl_nodename,
5783                           nfs4_client_id_uniquifier,
5784                           rpc_peeraddr2str(clp->cl_rpcclient,
5785                                            RPC_DISPLAY_ADDR));
5786         else
5787                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
5788                           clp->cl_rpcclient->cl_nodename,
5789                           rpc_peeraddr2str(clp->cl_rpcclient,
5790                                            RPC_DISPLAY_ADDR));
5791         rcu_read_unlock();
5792
5793         clp->cl_owner_id = str;
5794         return 0;
5795 }
5796
5797 static int
5798 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5799 {
5800         size_t len;
5801         char *str;
5802
5803         len = 10 + 10 + 1 + 10 + 1 +
5804                 strlen(nfs4_client_id_uniquifier) + 1 +
5805                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5806
5807         if (len > NFS4_OPAQUE_LIMIT + 1)
5808                 return -EINVAL;
5809
5810         /*
5811          * Since this string is allocated at mount time, and held until the
5812          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5813          * about a memory-reclaim deadlock.
5814          */
5815         str = kmalloc(len, GFP_KERNEL);
5816         if (!str)
5817                 return -ENOMEM;
5818
5819         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5820                         clp->rpc_ops->version, clp->cl_minorversion,
5821                         nfs4_client_id_uniquifier,
5822                         clp->cl_rpcclient->cl_nodename);
5823         clp->cl_owner_id = str;
5824         return 0;
5825 }
5826
5827 static int
5828 nfs4_init_uniform_client_string(struct nfs_client *clp)
5829 {
5830         size_t len;
5831         char *str;
5832
5833         if (clp->cl_owner_id != NULL)
5834                 return 0;
5835
5836         if (nfs4_client_id_uniquifier[0] != '\0')
5837                 return nfs4_init_uniquifier_client_string(clp);
5838
5839         len = 10 + 10 + 1 + 10 + 1 +
5840                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5841
5842         if (len > NFS4_OPAQUE_LIMIT + 1)
5843                 return -EINVAL;
5844
5845         /*
5846          * Since this string is allocated at mount time, and held until the
5847          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5848          * about a memory-reclaim deadlock.
5849          */
5850         str = kmalloc(len, GFP_KERNEL);
5851         if (!str)
5852                 return -ENOMEM;
5853
5854         scnprintf(str, len, "Linux NFSv%u.%u %s",
5855                         clp->rpc_ops->version, clp->cl_minorversion,
5856                         clp->cl_rpcclient->cl_nodename);
5857         clp->cl_owner_id = str;
5858         return 0;
5859 }
5860
5861 /*
5862  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5863  * services.  Advertise one based on the address family of the
5864  * clientaddr.
5865  */
5866 static unsigned int
5867 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5868 {
5869         if (strchr(clp->cl_ipaddr, ':') != NULL)
5870                 return scnprintf(buf, len, "tcp6");
5871         else
5872                 return scnprintf(buf, len, "tcp");
5873 }
5874
5875 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5876 {
5877         struct nfs4_setclientid *sc = calldata;
5878
5879         if (task->tk_status == 0)
5880                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5881 }
5882
5883 static const struct rpc_call_ops nfs4_setclientid_ops = {
5884         .rpc_call_done = nfs4_setclientid_done,
5885 };
5886
5887 /**
5888  * nfs4_proc_setclientid - Negotiate client ID
5889  * @clp: state data structure
5890  * @program: RPC program for NFSv4 callback service
5891  * @port: IP port number for NFS4 callback service
5892  * @cred: credential to use for this call
5893  * @res: where to place the result
5894  *
5895  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5896  */
5897 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5898                 unsigned short port, const struct cred *cred,
5899                 struct nfs4_setclientid_res *res)
5900 {
5901         nfs4_verifier sc_verifier;
5902         struct nfs4_setclientid setclientid = {
5903                 .sc_verifier = &sc_verifier,
5904                 .sc_prog = program,
5905                 .sc_clnt = clp,
5906         };
5907         struct rpc_message msg = {
5908                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5909                 .rpc_argp = &setclientid,
5910                 .rpc_resp = res,
5911                 .rpc_cred = cred,
5912         };
5913         struct rpc_task *task;
5914         struct rpc_task_setup task_setup_data = {
5915                 .rpc_client = clp->cl_rpcclient,
5916                 .rpc_message = &msg,
5917                 .callback_ops = &nfs4_setclientid_ops,
5918                 .callback_data = &setclientid,
5919                 .flags = RPC_TASK_TIMEOUT,
5920         };
5921         int status;
5922
5923         /* nfs_client_id4 */
5924         nfs4_init_boot_verifier(clp, &sc_verifier);
5925
5926         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5927                 status = nfs4_init_uniform_client_string(clp);
5928         else
5929                 status = nfs4_init_nonuniform_client_string(clp);
5930
5931         if (status)
5932                 goto out;
5933
5934         /* cb_client4 */
5935         setclientid.sc_netid_len =
5936                                 nfs4_init_callback_netid(clp,
5937                                                 setclientid.sc_netid,
5938                                                 sizeof(setclientid.sc_netid));
5939         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5940                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5941                                 clp->cl_ipaddr, port >> 8, port & 255);
5942
5943         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5944                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5945                 clp->cl_owner_id);
5946         task = rpc_run_task(&task_setup_data);
5947         if (IS_ERR(task)) {
5948                 status = PTR_ERR(task);
5949                 goto out;
5950         }
5951         status = task->tk_status;
5952         if (setclientid.sc_cred) {
5953                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5954                 put_rpccred(setclientid.sc_cred);
5955         }
5956         rpc_put_task(task);
5957 out:
5958         trace_nfs4_setclientid(clp, status);
5959         dprintk("NFS reply setclientid: %d\n", status);
5960         return status;
5961 }
5962
5963 /**
5964  * nfs4_proc_setclientid_confirm - Confirm client ID
5965  * @clp: state data structure
5966  * @res: result of a previous SETCLIENTID
5967  * @cred: credential to use for this call
5968  *
5969  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5970  */
5971 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5972                 struct nfs4_setclientid_res *arg,
5973                 const struct cred *cred)
5974 {
5975         struct rpc_message msg = {
5976                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5977                 .rpc_argp = arg,
5978                 .rpc_cred = cred,
5979         };
5980         int status;
5981
5982         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5983                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5984                 clp->cl_clientid);
5985         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5986         trace_nfs4_setclientid_confirm(clp, status);
5987         dprintk("NFS reply setclientid_confirm: %d\n", status);
5988         return status;
5989 }
5990
5991 struct nfs4_delegreturndata {
5992         struct nfs4_delegreturnargs args;
5993         struct nfs4_delegreturnres res;
5994         struct nfs_fh fh;
5995         nfs4_stateid stateid;
5996         unsigned long timestamp;
5997         struct {
5998                 struct nfs4_layoutreturn_args arg;
5999                 struct nfs4_layoutreturn_res res;
6000                 struct nfs4_xdr_opaque_data ld_private;
6001                 u32 roc_barrier;
6002                 bool roc;
6003         } lr;
6004         struct nfs_fattr fattr;
6005         int rpc_status;
6006         struct inode *inode;
6007 };
6008
6009 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6010 {
6011         struct nfs4_delegreturndata *data = calldata;
6012         struct nfs4_exception exception = {
6013                 .inode = data->inode,
6014                 .stateid = &data->stateid,
6015         };
6016
6017         if (!nfs4_sequence_done(task, &data->res.seq_res))
6018                 return;
6019
6020         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6021
6022         /* Handle Layoutreturn errors */
6023         if (data->args.lr_args && task->tk_status != 0) {
6024                 switch(data->res.lr_ret) {
6025                 default:
6026                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6027                         break;
6028                 case 0:
6029                         data->args.lr_args = NULL;
6030                         data->res.lr_res = NULL;
6031                         break;
6032                 case -NFS4ERR_OLD_STATEID:
6033                         if (nfs4_layoutreturn_refresh_stateid(&data->args.lr_args->stateid,
6034                                                 &data->args.lr_args->range,
6035                                                 data->inode))
6036                                 goto lr_restart;
6037                         /* Fallthrough */
6038                 case -NFS4ERR_ADMIN_REVOKED:
6039                 case -NFS4ERR_DELEG_REVOKED:
6040                 case -NFS4ERR_EXPIRED:
6041                 case -NFS4ERR_BAD_STATEID:
6042                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
6043                 case -NFS4ERR_WRONG_CRED:
6044                         data->args.lr_args = NULL;
6045                         data->res.lr_res = NULL;
6046                         goto lr_restart;
6047                 }
6048         }
6049
6050         switch (task->tk_status) {
6051         case 0:
6052                 renew_lease(data->res.server, data->timestamp);
6053                 break;
6054         case -NFS4ERR_ADMIN_REVOKED:
6055         case -NFS4ERR_DELEG_REVOKED:
6056         case -NFS4ERR_EXPIRED:
6057                 nfs4_free_revoked_stateid(data->res.server,
6058                                 data->args.stateid,
6059                                 task->tk_msg.rpc_cred);
6060                 /* Fallthrough */
6061         case -NFS4ERR_BAD_STATEID:
6062         case -NFS4ERR_STALE_STATEID:
6063                 task->tk_status = 0;
6064                 break;
6065         case -NFS4ERR_OLD_STATEID:
6066                 if (nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6067                         goto out_restart;
6068                 task->tk_status = 0;
6069                 break;
6070         case -NFS4ERR_ACCESS:
6071                 if (data->args.bitmask) {
6072                         data->args.bitmask = NULL;
6073                         data->res.fattr = NULL;
6074                         goto out_restart;
6075                 }
6076                 /* Fallthrough */
6077         default:
6078                 task->tk_status = nfs4_async_handle_exception(task,
6079                                 data->res.server, task->tk_status,
6080                                 &exception);
6081                 if (exception.retry)
6082                         goto out_restart;
6083         }
6084         data->rpc_status = task->tk_status;
6085         return;
6086 lr_restart:
6087         data->res.lr_ret = 0;
6088 out_restart:
6089         task->tk_status = 0;
6090         rpc_restart_call_prepare(task);
6091 }
6092
6093 static void nfs4_delegreturn_release(void *calldata)
6094 {
6095         struct nfs4_delegreturndata *data = calldata;
6096         struct inode *inode = data->inode;
6097
6098         if (inode) {
6099                 if (data->lr.roc)
6100                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6101                                         data->res.lr_ret);
6102                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6103                 nfs_iput_and_deactive(inode);
6104         }
6105         kfree(calldata);
6106 }
6107
6108 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6109 {
6110         struct nfs4_delegreturndata *d_data;
6111         struct pnfs_layout_hdr *lo;
6112
6113         d_data = (struct nfs4_delegreturndata *)data;
6114
6115         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
6116                 return;
6117
6118         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6119         if (lo && !pnfs_layout_is_valid(lo)) {
6120                 d_data->args.lr_args = NULL;
6121                 d_data->res.lr_res = NULL;
6122         }
6123
6124         nfs4_setup_sequence(d_data->res.server->nfs_client,
6125                         &d_data->args.seq_args,
6126                         &d_data->res.seq_res,
6127                         task);
6128 }
6129
6130 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6131         .rpc_call_prepare = nfs4_delegreturn_prepare,
6132         .rpc_call_done = nfs4_delegreturn_done,
6133         .rpc_release = nfs4_delegreturn_release,
6134 };
6135
6136 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6137 {
6138         struct nfs4_delegreturndata *data;
6139         struct nfs_server *server = NFS_SERVER(inode);
6140         struct rpc_task *task;
6141         struct rpc_message msg = {
6142                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6143                 .rpc_cred = cred,
6144         };
6145         struct rpc_task_setup task_setup_data = {
6146                 .rpc_client = server->client,
6147                 .rpc_message = &msg,
6148                 .callback_ops = &nfs4_delegreturn_ops,
6149                 .flags = RPC_TASK_ASYNC,
6150         };
6151         int status = 0;
6152
6153         data = kzalloc(sizeof(*data), GFP_NOFS);
6154         if (data == NULL)
6155                 return -ENOMEM;
6156         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6157
6158         nfs4_state_protect(server->nfs_client,
6159                         NFS_SP4_MACH_CRED_CLEANUP,
6160                         &task_setup_data.rpc_client, &msg);
6161
6162         data->args.fhandle = &data->fh;
6163         data->args.stateid = &data->stateid;
6164         data->args.bitmask = server->cache_consistency_bitmask;
6165         nfs_copy_fh(&data->fh, NFS_FH(inode));
6166         nfs4_stateid_copy(&data->stateid, stateid);
6167         data->res.fattr = &data->fattr;
6168         data->res.server = server;
6169         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6170         data->lr.arg.ld_private = &data->lr.ld_private;
6171         nfs_fattr_init(data->res.fattr);
6172         data->timestamp = jiffies;
6173         data->rpc_status = 0;
6174         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6175         data->inode = nfs_igrab_and_active(inode);
6176         if (data->inode) {
6177                 if (data->lr.roc) {
6178                         data->args.lr_args = &data->lr.arg;
6179                         data->res.lr_res = &data->lr.res;
6180                 }
6181         } else if (data->lr.roc) {
6182                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6183                 data->lr.roc = false;
6184         }
6185
6186         task_setup_data.callback_data = data;
6187         msg.rpc_argp = &data->args;
6188         msg.rpc_resp = &data->res;
6189         task = rpc_run_task(&task_setup_data);
6190         if (IS_ERR(task))
6191                 return PTR_ERR(task);
6192         if (!issync)
6193                 goto out;
6194         status = rpc_wait_for_completion_task(task);
6195         if (status != 0)
6196                 goto out;
6197         status = data->rpc_status;
6198 out:
6199         rpc_put_task(task);
6200         return status;
6201 }
6202
6203 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6204 {
6205         struct nfs_server *server = NFS_SERVER(inode);
6206         struct nfs4_exception exception = { };
6207         int err;
6208         do {
6209                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6210                 trace_nfs4_delegreturn(inode, stateid, err);
6211                 switch (err) {
6212                         case -NFS4ERR_STALE_STATEID:
6213                         case -NFS4ERR_EXPIRED:
6214                         case 0:
6215                                 return 0;
6216                 }
6217                 err = nfs4_handle_exception(server, err, &exception);
6218         } while (exception.retry);
6219         return err;
6220 }
6221
6222 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6223 {
6224         struct inode *inode = state->inode;
6225         struct nfs_server *server = NFS_SERVER(inode);
6226         struct nfs_client *clp = server->nfs_client;
6227         struct nfs_lockt_args arg = {
6228                 .fh = NFS_FH(inode),
6229                 .fl = request,
6230         };
6231         struct nfs_lockt_res res = {
6232                 .denied = request,
6233         };
6234         struct rpc_message msg = {
6235                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6236                 .rpc_argp       = &arg,
6237                 .rpc_resp       = &res,
6238                 .rpc_cred       = state->owner->so_cred,
6239         };
6240         struct nfs4_lock_state *lsp;
6241         int status;
6242
6243         arg.lock_owner.clientid = clp->cl_clientid;
6244         status = nfs4_set_lock_state(state, request);
6245         if (status != 0)
6246                 goto out;
6247         lsp = request->fl_u.nfs4_fl.owner;
6248         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6249         arg.lock_owner.s_dev = server->s_dev;
6250         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6251         switch (status) {
6252                 case 0:
6253                         request->fl_type = F_UNLCK;
6254                         break;
6255                 case -NFS4ERR_DENIED:
6256                         status = 0;
6257         }
6258         request->fl_ops->fl_release_private(request);
6259         request->fl_ops = NULL;
6260 out:
6261         return status;
6262 }
6263
6264 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6265 {
6266         struct nfs4_exception exception = { };
6267         int err;
6268
6269         do {
6270                 err = _nfs4_proc_getlk(state, cmd, request);
6271                 trace_nfs4_get_lock(request, state, cmd, err);
6272                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6273                                 &exception);
6274         } while (exception.retry);
6275         return err;
6276 }
6277
6278 struct nfs4_unlockdata {
6279         struct nfs_locku_args arg;
6280         struct nfs_locku_res res;
6281         struct nfs4_lock_state *lsp;
6282         struct nfs_open_context *ctx;
6283         struct nfs_lock_context *l_ctx;
6284         struct file_lock fl;
6285         struct nfs_server *server;
6286         unsigned long timestamp;
6287 };
6288
6289 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6290                 struct nfs_open_context *ctx,
6291                 struct nfs4_lock_state *lsp,
6292                 struct nfs_seqid *seqid)
6293 {
6294         struct nfs4_unlockdata *p;
6295         struct inode *inode = lsp->ls_state->inode;
6296
6297         p = kzalloc(sizeof(*p), GFP_NOFS);
6298         if (p == NULL)
6299                 return NULL;
6300         p->arg.fh = NFS_FH(inode);
6301         p->arg.fl = &p->fl;
6302         p->arg.seqid = seqid;
6303         p->res.seqid = seqid;
6304         p->lsp = lsp;
6305         refcount_inc(&lsp->ls_count);
6306         /* Ensure we don't close file until we're done freeing locks! */
6307         p->ctx = get_nfs_open_context(ctx);
6308         p->l_ctx = nfs_get_lock_context(ctx);
6309         locks_init_lock(&p->fl);
6310         locks_copy_lock(&p->fl, fl);
6311         p->server = NFS_SERVER(inode);
6312         return p;
6313 }
6314
6315 static void nfs4_locku_release_calldata(void *data)
6316 {
6317         struct nfs4_unlockdata *calldata = data;
6318         nfs_free_seqid(calldata->arg.seqid);
6319         nfs4_put_lock_state(calldata->lsp);
6320         nfs_put_lock_context(calldata->l_ctx);
6321         put_nfs_open_context(calldata->ctx);
6322         kfree(calldata);
6323 }
6324
6325 static void nfs4_locku_done(struct rpc_task *task, void *data)
6326 {
6327         struct nfs4_unlockdata *calldata = data;
6328         struct nfs4_exception exception = {
6329                 .inode = calldata->lsp->ls_state->inode,
6330                 .stateid = &calldata->arg.stateid,
6331         };
6332
6333         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6334                 return;
6335         switch (task->tk_status) {
6336                 case 0:
6337                         renew_lease(calldata->server, calldata->timestamp);
6338                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6339                         if (nfs4_update_lock_stateid(calldata->lsp,
6340                                         &calldata->res.stateid))
6341                                 break;
6342                         /* Fall through */
6343                 case -NFS4ERR_ADMIN_REVOKED:
6344                 case -NFS4ERR_EXPIRED:
6345                         nfs4_free_revoked_stateid(calldata->server,
6346                                         &calldata->arg.stateid,
6347                                         task->tk_msg.rpc_cred);
6348                         /* Fall through */
6349                 case -NFS4ERR_BAD_STATEID:
6350                 case -NFS4ERR_OLD_STATEID:
6351                 case -NFS4ERR_STALE_STATEID:
6352                         if (!nfs4_stateid_match(&calldata->arg.stateid,
6353                                                 &calldata->lsp->ls_stateid))
6354                                 rpc_restart_call_prepare(task);
6355                         break;
6356                 default:
6357                         task->tk_status = nfs4_async_handle_exception(task,
6358                                         calldata->server, task->tk_status,
6359                                         &exception);
6360                         if (exception.retry)
6361                                 rpc_restart_call_prepare(task);
6362         }
6363         nfs_release_seqid(calldata->arg.seqid);
6364 }
6365
6366 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6367 {
6368         struct nfs4_unlockdata *calldata = data;
6369
6370         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6371                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6372                 return;
6373
6374         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6375                 goto out_wait;
6376         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
6377         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6378                 /* Note: exit _without_ running nfs4_locku_done */
6379                 goto out_no_action;
6380         }
6381         calldata->timestamp = jiffies;
6382         if (nfs4_setup_sequence(calldata->server->nfs_client,
6383                                 &calldata->arg.seq_args,
6384                                 &calldata->res.seq_res,
6385                                 task) != 0)
6386                 nfs_release_seqid(calldata->arg.seqid);
6387         return;
6388 out_no_action:
6389         task->tk_action = NULL;
6390 out_wait:
6391         nfs4_sequence_done(task, &calldata->res.seq_res);
6392 }
6393
6394 static const struct rpc_call_ops nfs4_locku_ops = {
6395         .rpc_call_prepare = nfs4_locku_prepare,
6396         .rpc_call_done = nfs4_locku_done,
6397         .rpc_release = nfs4_locku_release_calldata,
6398 };
6399
6400 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6401                 struct nfs_open_context *ctx,
6402                 struct nfs4_lock_state *lsp,
6403                 struct nfs_seqid *seqid)
6404 {
6405         struct nfs4_unlockdata *data;
6406         struct rpc_message msg = {
6407                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6408                 .rpc_cred = ctx->cred,
6409         };
6410         struct rpc_task_setup task_setup_data = {
6411                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6412                 .rpc_message = &msg,
6413                 .callback_ops = &nfs4_locku_ops,
6414                 .workqueue = nfsiod_workqueue,
6415                 .flags = RPC_TASK_ASYNC,
6416         };
6417
6418         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6419                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6420
6421         /* Ensure this is an unlock - when canceling a lock, the
6422          * canceled lock is passed in, and it won't be an unlock.
6423          */
6424         fl->fl_type = F_UNLCK;
6425         if (fl->fl_flags & FL_CLOSE)
6426                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6427
6428         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6429         if (data == NULL) {
6430                 nfs_free_seqid(seqid);
6431                 return ERR_PTR(-ENOMEM);
6432         }
6433
6434         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6435         msg.rpc_argp = &data->arg;
6436         msg.rpc_resp = &data->res;
6437         task_setup_data.callback_data = data;
6438         return rpc_run_task(&task_setup_data);
6439 }
6440
6441 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6442 {
6443         struct inode *inode = state->inode;
6444         struct nfs4_state_owner *sp = state->owner;
6445         struct nfs_inode *nfsi = NFS_I(inode);
6446         struct nfs_seqid *seqid;
6447         struct nfs4_lock_state *lsp;
6448         struct rpc_task *task;
6449         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6450         int status = 0;
6451         unsigned char fl_flags = request->fl_flags;
6452
6453         status = nfs4_set_lock_state(state, request);
6454         /* Unlock _before_ we do the RPC call */
6455         request->fl_flags |= FL_EXISTS;
6456         /* Exclude nfs_delegation_claim_locks() */
6457         mutex_lock(&sp->so_delegreturn_mutex);
6458         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6459         down_read(&nfsi->rwsem);
6460         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6461                 up_read(&nfsi->rwsem);
6462                 mutex_unlock(&sp->so_delegreturn_mutex);
6463                 goto out;
6464         }
6465         up_read(&nfsi->rwsem);
6466         mutex_unlock(&sp->so_delegreturn_mutex);
6467         if (status != 0)
6468                 goto out;
6469         /* Is this a delegated lock? */
6470         lsp = request->fl_u.nfs4_fl.owner;
6471         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6472                 goto out;
6473         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6474         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6475         status = -ENOMEM;
6476         if (IS_ERR(seqid))
6477                 goto out;
6478         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6479         status = PTR_ERR(task);
6480         if (IS_ERR(task))
6481                 goto out;
6482         status = rpc_wait_for_completion_task(task);
6483         rpc_put_task(task);
6484 out:
6485         request->fl_flags = fl_flags;
6486         trace_nfs4_unlock(request, state, F_SETLK, status);
6487         return status;
6488 }
6489
6490 struct nfs4_lockdata {
6491         struct nfs_lock_args arg;
6492         struct nfs_lock_res res;
6493         struct nfs4_lock_state *lsp;
6494         struct nfs_open_context *ctx;
6495         struct file_lock fl;
6496         unsigned long timestamp;
6497         int rpc_status;
6498         int cancelled;
6499         struct nfs_server *server;
6500 };
6501
6502 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6503                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6504                 gfp_t gfp_mask)
6505 {
6506         struct nfs4_lockdata *p;
6507         struct inode *inode = lsp->ls_state->inode;
6508         struct nfs_server *server = NFS_SERVER(inode);
6509         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6510
6511         p = kzalloc(sizeof(*p), gfp_mask);
6512         if (p == NULL)
6513                 return NULL;
6514
6515         p->arg.fh = NFS_FH(inode);
6516         p->arg.fl = &p->fl;
6517         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6518         if (IS_ERR(p->arg.open_seqid))
6519                 goto out_free;
6520         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6521         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6522         if (IS_ERR(p->arg.lock_seqid))
6523                 goto out_free_seqid;
6524         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6525         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6526         p->arg.lock_owner.s_dev = server->s_dev;
6527         p->res.lock_seqid = p->arg.lock_seqid;
6528         p->lsp = lsp;
6529         p->server = server;
6530         refcount_inc(&lsp->ls_count);
6531         p->ctx = get_nfs_open_context(ctx);
6532         locks_init_lock(&p->fl);
6533         locks_copy_lock(&p->fl, fl);
6534         return p;
6535 out_free_seqid:
6536         nfs_free_seqid(p->arg.open_seqid);
6537 out_free:
6538         kfree(p);
6539         return NULL;
6540 }
6541
6542 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6543 {
6544         struct nfs4_lockdata *data = calldata;
6545         struct nfs4_state *state = data->lsp->ls_state;
6546
6547         dprintk("%s: begin!\n", __func__);
6548         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6549                 goto out_wait;
6550         /* Do we need to do an open_to_lock_owner? */
6551         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6552                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6553                         goto out_release_lock_seqid;
6554                 }
6555                 nfs4_stateid_copy(&data->arg.open_stateid,
6556                                 &state->open_stateid);
6557                 data->arg.new_lock_owner = 1;
6558                 data->res.open_seqid = data->arg.open_seqid;
6559         } else {
6560                 data->arg.new_lock_owner = 0;
6561                 nfs4_stateid_copy(&data->arg.lock_stateid,
6562                                 &data->lsp->ls_stateid);
6563         }
6564         if (!nfs4_valid_open_stateid(state)) {
6565                 data->rpc_status = -EBADF;
6566                 task->tk_action = NULL;
6567                 goto out_release_open_seqid;
6568         }
6569         data->timestamp = jiffies;
6570         if (nfs4_setup_sequence(data->server->nfs_client,
6571                                 &data->arg.seq_args,
6572                                 &data->res.seq_res,
6573                                 task) == 0)
6574                 return;
6575 out_release_open_seqid:
6576         nfs_release_seqid(data->arg.open_seqid);
6577 out_release_lock_seqid:
6578         nfs_release_seqid(data->arg.lock_seqid);
6579 out_wait:
6580         nfs4_sequence_done(task, &data->res.seq_res);
6581         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6582 }
6583
6584 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6585 {
6586         struct nfs4_lockdata *data = calldata;
6587         struct nfs4_lock_state *lsp = data->lsp;
6588
6589         dprintk("%s: begin!\n", __func__);
6590
6591         if (!nfs4_sequence_done(task, &data->res.seq_res))
6592                 return;
6593
6594         data->rpc_status = task->tk_status;
6595         switch (task->tk_status) {
6596         case 0:
6597                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6598                                 data->timestamp);
6599                 if (data->arg.new_lock && !data->cancelled) {
6600                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6601                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6602                                 goto out_restart;
6603                 }
6604                 if (data->arg.new_lock_owner != 0) {
6605                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6606                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6607                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6608                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6609                         goto out_restart;
6610                 break;
6611         case -NFS4ERR_BAD_STATEID:
6612         case -NFS4ERR_OLD_STATEID:
6613         case -NFS4ERR_STALE_STATEID:
6614         case -NFS4ERR_EXPIRED:
6615                 if (data->arg.new_lock_owner != 0) {
6616                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6617                                                 &lsp->ls_state->open_stateid))
6618                                 goto out_restart;
6619                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6620                                                 &lsp->ls_stateid))
6621                                 goto out_restart;
6622         }
6623 out_done:
6624         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6625         return;
6626 out_restart:
6627         if (!data->cancelled)
6628                 rpc_restart_call_prepare(task);
6629         goto out_done;
6630 }
6631
6632 static void nfs4_lock_release(void *calldata)
6633 {
6634         struct nfs4_lockdata *data = calldata;
6635
6636         dprintk("%s: begin!\n", __func__);
6637         nfs_free_seqid(data->arg.open_seqid);
6638         if (data->cancelled && data->rpc_status == 0) {
6639                 struct rpc_task *task;
6640                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6641                                 data->arg.lock_seqid);
6642                 if (!IS_ERR(task))
6643                         rpc_put_task_async(task);
6644                 dprintk("%s: cancelling lock!\n", __func__);
6645         } else
6646                 nfs_free_seqid(data->arg.lock_seqid);
6647         nfs4_put_lock_state(data->lsp);
6648         put_nfs_open_context(data->ctx);
6649         kfree(data);
6650         dprintk("%s: done!\n", __func__);
6651 }
6652
6653 static const struct rpc_call_ops nfs4_lock_ops = {
6654         .rpc_call_prepare = nfs4_lock_prepare,
6655         .rpc_call_done = nfs4_lock_done,
6656         .rpc_release = nfs4_lock_release,
6657 };
6658
6659 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6660 {
6661         switch (error) {
6662         case -NFS4ERR_ADMIN_REVOKED:
6663         case -NFS4ERR_EXPIRED:
6664         case -NFS4ERR_BAD_STATEID:
6665                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6666                 if (new_lock_owner != 0 ||
6667                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6668                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6669                 break;
6670         case -NFS4ERR_STALE_STATEID:
6671                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6672                 nfs4_schedule_lease_recovery(server->nfs_client);
6673         };
6674 }
6675
6676 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6677 {
6678         struct nfs4_lockdata *data;
6679         struct rpc_task *task;
6680         struct rpc_message msg = {
6681                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6682                 .rpc_cred = state->owner->so_cred,
6683         };
6684         struct rpc_task_setup task_setup_data = {
6685                 .rpc_client = NFS_CLIENT(state->inode),
6686                 .rpc_message = &msg,
6687                 .callback_ops = &nfs4_lock_ops,
6688                 .workqueue = nfsiod_workqueue,
6689                 .flags = RPC_TASK_ASYNC,
6690         };
6691         int ret;
6692
6693         dprintk("%s: begin!\n", __func__);
6694         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6695                         fl->fl_u.nfs4_fl.owner,
6696                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6697         if (data == NULL)
6698                 return -ENOMEM;
6699         if (IS_SETLKW(cmd))
6700                 data->arg.block = 1;
6701         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6702                                 recovery_type > NFS_LOCK_NEW);
6703         msg.rpc_argp = &data->arg;
6704         msg.rpc_resp = &data->res;
6705         task_setup_data.callback_data = data;
6706         if (recovery_type > NFS_LOCK_NEW) {
6707                 if (recovery_type == NFS_LOCK_RECLAIM)
6708                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6709         } else
6710                 data->arg.new_lock = 1;
6711         task = rpc_run_task(&task_setup_data);
6712         if (IS_ERR(task))
6713                 return PTR_ERR(task);
6714         ret = rpc_wait_for_completion_task(task);
6715         if (ret == 0) {
6716                 ret = data->rpc_status;
6717                 if (ret)
6718                         nfs4_handle_setlk_error(data->server, data->lsp,
6719                                         data->arg.new_lock_owner, ret);
6720         } else
6721                 data->cancelled = true;
6722         rpc_put_task(task);
6723         dprintk("%s: done, ret = %d!\n", __func__, ret);
6724         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6725         return ret;
6726 }
6727
6728 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6729 {
6730         struct nfs_server *server = NFS_SERVER(state->inode);
6731         struct nfs4_exception exception = {
6732                 .inode = state->inode,
6733         };
6734         int err;
6735
6736         do {
6737                 /* Cache the lock if possible... */
6738                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6739                         return 0;
6740                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6741                 if (err != -NFS4ERR_DELAY)
6742                         break;
6743                 nfs4_handle_exception(server, err, &exception);
6744         } while (exception.retry);
6745         return err;
6746 }
6747
6748 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6749 {
6750         struct nfs_server *server = NFS_SERVER(state->inode);
6751         struct nfs4_exception exception = {
6752                 .inode = state->inode,
6753         };
6754         int err;
6755
6756         err = nfs4_set_lock_state(state, request);
6757         if (err != 0)
6758                 return err;
6759         if (!recover_lost_locks) {
6760                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6761                 return 0;
6762         }
6763         do {
6764                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6765                         return 0;
6766                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6767                 switch (err) {
6768                 default:
6769                         goto out;
6770                 case -NFS4ERR_GRACE:
6771                 case -NFS4ERR_DELAY:
6772                         nfs4_handle_exception(server, err, &exception);
6773                         err = 0;
6774                 }
6775         } while (exception.retry);
6776 out:
6777         return err;
6778 }
6779
6780 #if defined(CONFIG_NFS_V4_1)
6781 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6782 {
6783         struct nfs4_lock_state *lsp;
6784         int status;
6785
6786         status = nfs4_set_lock_state(state, request);
6787         if (status != 0)
6788                 return status;
6789         lsp = request->fl_u.nfs4_fl.owner;
6790         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6791             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6792                 return 0;
6793         return nfs4_lock_expired(state, request);
6794 }
6795 #endif
6796
6797 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6798 {
6799         struct nfs_inode *nfsi = NFS_I(state->inode);
6800         struct nfs4_state_owner *sp = state->owner;
6801         unsigned char fl_flags = request->fl_flags;
6802         int status;
6803
6804         request->fl_flags |= FL_ACCESS;
6805         status = locks_lock_inode_wait(state->inode, request);
6806         if (status < 0)
6807                 goto out;
6808         mutex_lock(&sp->so_delegreturn_mutex);
6809         down_read(&nfsi->rwsem);
6810         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6811                 /* Yes: cache locks! */
6812                 /* ...but avoid races with delegation recall... */
6813                 request->fl_flags = fl_flags & ~FL_SLEEP;
6814                 status = locks_lock_inode_wait(state->inode, request);
6815                 up_read(&nfsi->rwsem);
6816                 mutex_unlock(&sp->so_delegreturn_mutex);
6817                 goto out;
6818         }
6819         up_read(&nfsi->rwsem);
6820         mutex_unlock(&sp->so_delegreturn_mutex);
6821         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6822 out:
6823         request->fl_flags = fl_flags;
6824         return status;
6825 }
6826
6827 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6828 {
6829         struct nfs4_exception exception = {
6830                 .state = state,
6831                 .inode = state->inode,
6832         };
6833         int err;
6834
6835         do {
6836                 err = _nfs4_proc_setlk(state, cmd, request);
6837                 if (err == -NFS4ERR_DENIED)
6838                         err = -EAGAIN;
6839                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6840                                 err, &exception);
6841         } while (exception.retry);
6842         return err;
6843 }
6844
6845 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6846 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6847
6848 static int
6849 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6850                         struct file_lock *request)
6851 {
6852         int             status = -ERESTARTSYS;
6853         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6854
6855         while(!signalled()) {
6856                 status = nfs4_proc_setlk(state, cmd, request);
6857                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6858                         break;
6859                 freezable_schedule_timeout_interruptible(timeout);
6860                 timeout *= 2;
6861                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6862                 status = -ERESTARTSYS;
6863         }
6864         return status;
6865 }
6866
6867 #ifdef CONFIG_NFS_V4_1
6868 struct nfs4_lock_waiter {
6869         struct task_struct      *task;
6870         struct inode            *inode;
6871         struct nfs_lowner       *owner;
6872         bool                    notified;
6873 };
6874
6875 static int
6876 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6877 {
6878         int ret;
6879         struct nfs4_lock_waiter *waiter = wait->private;
6880
6881         /* NULL key means to wake up everyone */
6882         if (key) {
6883                 struct cb_notify_lock_args      *cbnl = key;
6884                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
6885                                                 *wowner = waiter->owner;
6886
6887                 /* Only wake if the callback was for the same owner. */
6888                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
6889                         return 0;
6890
6891                 /* Make sure it's for the right inode */
6892                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6893                         return 0;
6894
6895                 waiter->notified = true;
6896         }
6897
6898         /* override "private" so we can use default_wake_function */
6899         wait->private = waiter->task;
6900         ret = autoremove_wake_function(wait, mode, flags, key);
6901         wait->private = waiter;
6902         return ret;
6903 }
6904
6905 static int
6906 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6907 {
6908         int status = -ERESTARTSYS;
6909         unsigned long flags;
6910         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6911         struct nfs_server *server = NFS_SERVER(state->inode);
6912         struct nfs_client *clp = server->nfs_client;
6913         wait_queue_head_t *q = &clp->cl_lock_waitq;
6914         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6915                                     .id = lsp->ls_seqid.owner_id,
6916                                     .s_dev = server->s_dev };
6917         struct nfs4_lock_waiter waiter = { .task  = current,
6918                                            .inode = state->inode,
6919                                            .owner = &owner,
6920                                            .notified = false };
6921         wait_queue_entry_t wait;
6922
6923         /* Don't bother with waitqueue if we don't expect a callback */
6924         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6925                 return nfs4_retry_setlk_simple(state, cmd, request);
6926
6927         init_wait(&wait);
6928         wait.private = &waiter;
6929         wait.func = nfs4_wake_lock_waiter;
6930         add_wait_queue(q, &wait);
6931
6932         while(!signalled()) {
6933                 waiter.notified = false;
6934                 status = nfs4_proc_setlk(state, cmd, request);
6935                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6936                         break;
6937
6938                 status = -ERESTARTSYS;
6939                 spin_lock_irqsave(&q->lock, flags);
6940                 if (waiter.notified) {
6941                         spin_unlock_irqrestore(&q->lock, flags);
6942                         continue;
6943                 }
6944                 set_current_state(TASK_INTERRUPTIBLE);
6945                 spin_unlock_irqrestore(&q->lock, flags);
6946
6947                 freezable_schedule_timeout(NFS4_LOCK_MAXTIMEOUT);
6948         }
6949
6950         finish_wait(q, &wait);
6951         return status;
6952 }
6953 #else /* !CONFIG_NFS_V4_1 */
6954 static inline int
6955 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6956 {
6957         return nfs4_retry_setlk_simple(state, cmd, request);
6958 }
6959 #endif
6960
6961 static int
6962 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6963 {
6964         struct nfs_open_context *ctx;
6965         struct nfs4_state *state;
6966         int status;
6967
6968         /* verify open state */
6969         ctx = nfs_file_open_context(filp);
6970         state = ctx->state;
6971
6972         if (IS_GETLK(cmd)) {
6973                 if (state != NULL)
6974                         return nfs4_proc_getlk(state, F_GETLK, request);
6975                 return 0;
6976         }
6977
6978         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6979                 return -EINVAL;
6980
6981         if (request->fl_type == F_UNLCK) {
6982                 if (state != NULL)
6983                         return nfs4_proc_unlck(state, cmd, request);
6984                 return 0;
6985         }
6986
6987         if (state == NULL)
6988                 return -ENOLCK;
6989
6990         if ((request->fl_flags & FL_POSIX) &&
6991             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6992                 return -ENOLCK;
6993
6994         /*
6995          * Don't rely on the VFS having checked the file open mode,
6996          * since it won't do this for flock() locks.
6997          */
6998         switch (request->fl_type) {
6999         case F_RDLCK:
7000                 if (!(filp->f_mode & FMODE_READ))
7001                         return -EBADF;
7002                 break;
7003         case F_WRLCK:
7004                 if (!(filp->f_mode & FMODE_WRITE))
7005                         return -EBADF;
7006         }
7007
7008         status = nfs4_set_lock_state(state, request);
7009         if (status != 0)
7010                 return status;
7011
7012         return nfs4_retry_setlk(state, cmd, request);
7013 }
7014
7015 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7016 {
7017         struct nfs_server *server = NFS_SERVER(state->inode);
7018         int err;
7019
7020         err = nfs4_set_lock_state(state, fl);
7021         if (err != 0)
7022                 return err;
7023         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7024         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7025 }
7026
7027 struct nfs_release_lockowner_data {
7028         struct nfs4_lock_state *lsp;
7029         struct nfs_server *server;
7030         struct nfs_release_lockowner_args args;
7031         struct nfs_release_lockowner_res res;
7032         unsigned long timestamp;
7033 };
7034
7035 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7036 {
7037         struct nfs_release_lockowner_data *data = calldata;
7038         struct nfs_server *server = data->server;
7039         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7040                            &data->res.seq_res, task);
7041         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7042         data->timestamp = jiffies;
7043 }
7044
7045 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7046 {
7047         struct nfs_release_lockowner_data *data = calldata;
7048         struct nfs_server *server = data->server;
7049
7050         nfs40_sequence_done(task, &data->res.seq_res);
7051
7052         switch (task->tk_status) {
7053         case 0:
7054                 renew_lease(server, data->timestamp);
7055                 break;
7056         case -NFS4ERR_STALE_CLIENTID:
7057         case -NFS4ERR_EXPIRED:
7058                 nfs4_schedule_lease_recovery(server->nfs_client);
7059                 break;
7060         case -NFS4ERR_LEASE_MOVED:
7061         case -NFS4ERR_DELAY:
7062                 if (nfs4_async_handle_error(task, server,
7063                                             NULL, NULL) == -EAGAIN)
7064                         rpc_restart_call_prepare(task);
7065         }
7066 }
7067
7068 static void nfs4_release_lockowner_release(void *calldata)
7069 {
7070         struct nfs_release_lockowner_data *data = calldata;
7071         nfs4_free_lock_state(data->server, data->lsp);
7072         kfree(calldata);
7073 }
7074
7075 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7076         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7077         .rpc_call_done = nfs4_release_lockowner_done,
7078         .rpc_release = nfs4_release_lockowner_release,
7079 };
7080
7081 static void
7082 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7083 {
7084         struct nfs_release_lockowner_data *data;
7085         struct rpc_message msg = {
7086                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7087         };
7088
7089         if (server->nfs_client->cl_mvops->minor_version != 0)
7090                 return;
7091
7092         data = kmalloc(sizeof(*data), GFP_NOFS);
7093         if (!data)
7094                 return;
7095         data->lsp = lsp;
7096         data->server = server;
7097         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7098         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7099         data->args.lock_owner.s_dev = server->s_dev;
7100
7101         msg.rpc_argp = &data->args;
7102         msg.rpc_resp = &data->res;
7103         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7104         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7105 }
7106
7107 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7108
7109 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7110                                    struct dentry *unused, struct inode *inode,
7111                                    const char *key, const void *buf,
7112                                    size_t buflen, int flags)
7113 {
7114         return nfs4_proc_set_acl(inode, buf, buflen);
7115 }
7116
7117 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7118                                    struct dentry *unused, struct inode *inode,
7119                                    const char *key, void *buf, size_t buflen)
7120 {
7121         return nfs4_proc_get_acl(inode, buf, buflen);
7122 }
7123
7124 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7125 {
7126         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7127 }
7128
7129 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7130
7131 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7132                                      struct dentry *unused, struct inode *inode,
7133                                      const char *key, const void *buf,
7134                                      size_t buflen, int flags)
7135 {
7136         if (security_ismaclabel(key))
7137                 return nfs4_set_security_label(inode, buf, buflen);
7138
7139         return -EOPNOTSUPP;
7140 }
7141
7142 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7143                                      struct dentry *unused, struct inode *inode,
7144                                      const char *key, void *buf, size_t buflen)
7145 {
7146         if (security_ismaclabel(key))
7147                 return nfs4_get_security_label(inode, buf, buflen);
7148         return -EOPNOTSUPP;
7149 }
7150
7151 static ssize_t
7152 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7153 {
7154         int len = 0;
7155
7156         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7157                 len = security_inode_listsecurity(inode, list, list_len);
7158                 if (list_len && len > list_len)
7159                         return -ERANGE;
7160         }
7161         return len;
7162 }
7163
7164 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7165         .prefix = XATTR_SECURITY_PREFIX,
7166         .get    = nfs4_xattr_get_nfs4_label,
7167         .set    = nfs4_xattr_set_nfs4_label,
7168 };
7169
7170 #else
7171
7172 static ssize_t
7173 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7174 {
7175         return 0;
7176 }
7177
7178 #endif
7179
7180 /*
7181  * nfs_fhget will use either the mounted_on_fileid or the fileid
7182  */
7183 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7184 {
7185         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7186                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7187               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7188               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7189                 return;
7190
7191         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7192                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7193         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7194         fattr->nlink = 2;
7195 }
7196
7197 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7198                                    const struct qstr *name,
7199                                    struct nfs4_fs_locations *fs_locations,
7200                                    struct page *page)
7201 {
7202         struct nfs_server *server = NFS_SERVER(dir);
7203         u32 bitmask[3];
7204         struct nfs4_fs_locations_arg args = {
7205                 .dir_fh = NFS_FH(dir),
7206                 .name = name,
7207                 .page = page,
7208                 .bitmask = bitmask,
7209         };
7210         struct nfs4_fs_locations_res res = {
7211                 .fs_locations = fs_locations,
7212         };
7213         struct rpc_message msg = {
7214                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7215                 .rpc_argp = &args,
7216                 .rpc_resp = &res,
7217         };
7218         int status;
7219
7220         dprintk("%s: start\n", __func__);
7221
7222         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7223         bitmask[1] = nfs4_fattr_bitmap[1];
7224
7225         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7226          * is not supported */
7227         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7228                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7229         else
7230                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7231
7232         nfs_fattr_init(&fs_locations->fattr);
7233         fs_locations->server = server;
7234         fs_locations->nlocations = 0;
7235         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7236         dprintk("%s: returned status = %d\n", __func__, status);
7237         return status;
7238 }
7239
7240 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7241                            const struct qstr *name,
7242                            struct nfs4_fs_locations *fs_locations,
7243                            struct page *page)
7244 {
7245         struct nfs4_exception exception = { };
7246         int err;
7247         do {
7248                 err = _nfs4_proc_fs_locations(client, dir, name,
7249                                 fs_locations, page);
7250                 trace_nfs4_get_fs_locations(dir, name, err);
7251                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7252                                 &exception);
7253         } while (exception.retry);
7254         return err;
7255 }
7256
7257 /*
7258  * This operation also signals the server that this client is
7259  * performing migration recovery.  The server can stop returning
7260  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7261  * appended to this compound to identify the client ID which is
7262  * performing recovery.
7263  */
7264 static int _nfs40_proc_get_locations(struct inode *inode,
7265                                      struct nfs4_fs_locations *locations,
7266                                      struct page *page, const struct cred *cred)
7267 {
7268         struct nfs_server *server = NFS_SERVER(inode);
7269         struct rpc_clnt *clnt = server->client;
7270         u32 bitmask[2] = {
7271                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7272         };
7273         struct nfs4_fs_locations_arg args = {
7274                 .clientid       = server->nfs_client->cl_clientid,
7275                 .fh             = NFS_FH(inode),
7276                 .page           = page,
7277                 .bitmask        = bitmask,
7278                 .migration      = 1,            /* skip LOOKUP */
7279                 .renew          = 1,            /* append RENEW */
7280         };
7281         struct nfs4_fs_locations_res res = {
7282                 .fs_locations   = locations,
7283                 .migration      = 1,
7284                 .renew          = 1,
7285         };
7286         struct rpc_message msg = {
7287                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7288                 .rpc_argp       = &args,
7289                 .rpc_resp       = &res,
7290                 .rpc_cred       = cred,
7291         };
7292         unsigned long now = jiffies;
7293         int status;
7294
7295         nfs_fattr_init(&locations->fattr);
7296         locations->server = server;
7297         locations->nlocations = 0;
7298
7299         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7300         status = nfs4_call_sync_sequence(clnt, server, &msg,
7301                                         &args.seq_args, &res.seq_res);
7302         if (status)
7303                 return status;
7304
7305         renew_lease(server, now);
7306         return 0;
7307 }
7308
7309 #ifdef CONFIG_NFS_V4_1
7310
7311 /*
7312  * This operation also signals the server that this client is
7313  * performing migration recovery.  The server can stop asserting
7314  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7315  * performing this operation is identified in the SEQUENCE
7316  * operation in this compound.
7317  *
7318  * When the client supports GETATTR(fs_locations_info), it can
7319  * be plumbed in here.
7320  */
7321 static int _nfs41_proc_get_locations(struct inode *inode,
7322                                      struct nfs4_fs_locations *locations,
7323                                      struct page *page, const struct cred *cred)
7324 {
7325         struct nfs_server *server = NFS_SERVER(inode);
7326         struct rpc_clnt *clnt = server->client;
7327         u32 bitmask[2] = {
7328                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7329         };
7330         struct nfs4_fs_locations_arg args = {
7331                 .fh             = NFS_FH(inode),
7332                 .page           = page,
7333                 .bitmask        = bitmask,
7334                 .migration      = 1,            /* skip LOOKUP */
7335         };
7336         struct nfs4_fs_locations_res res = {
7337                 .fs_locations   = locations,
7338                 .migration      = 1,
7339         };
7340         struct rpc_message msg = {
7341                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7342                 .rpc_argp       = &args,
7343                 .rpc_resp       = &res,
7344                 .rpc_cred       = cred,
7345         };
7346         int status;
7347
7348         nfs_fattr_init(&locations->fattr);
7349         locations->server = server;
7350         locations->nlocations = 0;
7351
7352         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7353         status = nfs4_call_sync_sequence(clnt, server, &msg,
7354                                         &args.seq_args, &res.seq_res);
7355         if (status == NFS4_OK &&
7356             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7357                 status = -NFS4ERR_LEASE_MOVED;
7358         return status;
7359 }
7360
7361 #endif  /* CONFIG_NFS_V4_1 */
7362
7363 /**
7364  * nfs4_proc_get_locations - discover locations for a migrated FSID
7365  * @inode: inode on FSID that is migrating
7366  * @locations: result of query
7367  * @page: buffer
7368  * @cred: credential to use for this operation
7369  *
7370  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7371  * operation failed, or a negative errno if a local error occurred.
7372  *
7373  * On success, "locations" is filled in, but if the server has
7374  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7375  * asserted.
7376  *
7377  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7378  * from this client that require migration recovery.
7379  */
7380 int nfs4_proc_get_locations(struct inode *inode,
7381                             struct nfs4_fs_locations *locations,
7382                             struct page *page, const struct cred *cred)
7383 {
7384         struct nfs_server *server = NFS_SERVER(inode);
7385         struct nfs_client *clp = server->nfs_client;
7386         const struct nfs4_mig_recovery_ops *ops =
7387                                         clp->cl_mvops->mig_recovery_ops;
7388         struct nfs4_exception exception = { };
7389         int status;
7390
7391         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7392                 (unsigned long long)server->fsid.major,
7393                 (unsigned long long)server->fsid.minor,
7394                 clp->cl_hostname);
7395         nfs_display_fhandle(NFS_FH(inode), __func__);
7396
7397         do {
7398                 status = ops->get_locations(inode, locations, page, cred);
7399                 if (status != -NFS4ERR_DELAY)
7400                         break;
7401                 nfs4_handle_exception(server, status, &exception);
7402         } while (exception.retry);
7403         return status;
7404 }
7405
7406 /*
7407  * This operation also signals the server that this client is
7408  * performing "lease moved" recovery.  The server can stop
7409  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7410  * is appended to this compound to identify the client ID which is
7411  * performing recovery.
7412  */
7413 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7414 {
7415         struct nfs_server *server = NFS_SERVER(inode);
7416         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7417         struct rpc_clnt *clnt = server->client;
7418         struct nfs4_fsid_present_arg args = {
7419                 .fh             = NFS_FH(inode),
7420                 .clientid       = clp->cl_clientid,
7421                 .renew          = 1,            /* append RENEW */
7422         };
7423         struct nfs4_fsid_present_res res = {
7424                 .renew          = 1,
7425         };
7426         struct rpc_message msg = {
7427                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7428                 .rpc_argp       = &args,
7429                 .rpc_resp       = &res,
7430                 .rpc_cred       = cred,
7431         };
7432         unsigned long now = jiffies;
7433         int status;
7434
7435         res.fh = nfs_alloc_fhandle();
7436         if (res.fh == NULL)
7437                 return -ENOMEM;
7438
7439         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7440         status = nfs4_call_sync_sequence(clnt, server, &msg,
7441                                                 &args.seq_args, &res.seq_res);
7442         nfs_free_fhandle(res.fh);
7443         if (status)
7444                 return status;
7445
7446         do_renew_lease(clp, now);
7447         return 0;
7448 }
7449
7450 #ifdef CONFIG_NFS_V4_1
7451
7452 /*
7453  * This operation also signals the server that this client is
7454  * performing "lease moved" recovery.  The server can stop asserting
7455  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7456  * this operation is identified in the SEQUENCE operation in this
7457  * compound.
7458  */
7459 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7460 {
7461         struct nfs_server *server = NFS_SERVER(inode);
7462         struct rpc_clnt *clnt = server->client;
7463         struct nfs4_fsid_present_arg args = {
7464                 .fh             = NFS_FH(inode),
7465         };
7466         struct nfs4_fsid_present_res res = {
7467         };
7468         struct rpc_message msg = {
7469                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7470                 .rpc_argp       = &args,
7471                 .rpc_resp       = &res,
7472                 .rpc_cred       = cred,
7473         };
7474         int status;
7475
7476         res.fh = nfs_alloc_fhandle();
7477         if (res.fh == NULL)
7478                 return -ENOMEM;
7479
7480         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7481         status = nfs4_call_sync_sequence(clnt, server, &msg,
7482                                                 &args.seq_args, &res.seq_res);
7483         nfs_free_fhandle(res.fh);
7484         if (status == NFS4_OK &&
7485             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7486                 status = -NFS4ERR_LEASE_MOVED;
7487         return status;
7488 }
7489
7490 #endif  /* CONFIG_NFS_V4_1 */
7491
7492 /**
7493  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7494  * @inode: inode on FSID to check
7495  * @cred: credential to use for this operation
7496  *
7497  * Server indicates whether the FSID is present, moved, or not
7498  * recognized.  This operation is necessary to clear a LEASE_MOVED
7499  * condition for this client ID.
7500  *
7501  * Returns NFS4_OK if the FSID is present on this server,
7502  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7503  *  NFS4ERR code if some error occurred on the server, or a
7504  *  negative errno if a local failure occurred.
7505  */
7506 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
7507 {
7508         struct nfs_server *server = NFS_SERVER(inode);
7509         struct nfs_client *clp = server->nfs_client;
7510         const struct nfs4_mig_recovery_ops *ops =
7511                                         clp->cl_mvops->mig_recovery_ops;
7512         struct nfs4_exception exception = { };
7513         int status;
7514
7515         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7516                 (unsigned long long)server->fsid.major,
7517                 (unsigned long long)server->fsid.minor,
7518                 clp->cl_hostname);
7519         nfs_display_fhandle(NFS_FH(inode), __func__);
7520
7521         do {
7522                 status = ops->fsid_present(inode, cred);
7523                 if (status != -NFS4ERR_DELAY)
7524                         break;
7525                 nfs4_handle_exception(server, status, &exception);
7526         } while (exception.retry);
7527         return status;
7528 }
7529
7530 /**
7531  * If 'use_integrity' is true and the state managment nfs_client
7532  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7533  * and the machine credential as per RFC3530bis and RFC5661 Security
7534  * Considerations sections. Otherwise, just use the user cred with the
7535  * filesystem's rpc_client.
7536  */
7537 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7538 {
7539         int status;
7540         struct nfs4_secinfo_arg args = {
7541                 .dir_fh = NFS_FH(dir),
7542                 .name   = name,
7543         };
7544         struct nfs4_secinfo_res res = {
7545                 .flavors     = flavors,
7546         };
7547         struct rpc_message msg = {
7548                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7549                 .rpc_argp = &args,
7550                 .rpc_resp = &res,
7551         };
7552         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7553         const struct cred *cred = NULL;
7554
7555         if (use_integrity) {
7556                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7557                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7558                 msg.rpc_cred = cred;
7559         }
7560
7561         dprintk("NFS call  secinfo %s\n", name->name);
7562
7563         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7564                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7565
7566         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7567                                 &res.seq_res, 0);
7568         dprintk("NFS reply  secinfo: %d\n", status);
7569
7570         put_cred(cred);
7571
7572         return status;
7573 }
7574
7575 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7576                       struct nfs4_secinfo_flavors *flavors)
7577 {
7578         struct nfs4_exception exception = { };
7579         int err;
7580         do {
7581                 err = -NFS4ERR_WRONGSEC;
7582
7583                 /* try to use integrity protection with machine cred */
7584                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7585                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7586
7587                 /*
7588                  * if unable to use integrity protection, or SECINFO with
7589                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7590                  * disallowed by spec, but exists in deployed servers) use
7591                  * the current filesystem's rpc_client and the user cred.
7592                  */
7593                 if (err == -NFS4ERR_WRONGSEC)
7594                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7595
7596                 trace_nfs4_secinfo(dir, name, err);
7597                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7598                                 &exception);
7599         } while (exception.retry);
7600         return err;
7601 }
7602
7603 #ifdef CONFIG_NFS_V4_1
7604 /*
7605  * Check the exchange flags returned by the server for invalid flags, having
7606  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7607  * DS flags set.
7608  */
7609 static int nfs4_check_cl_exchange_flags(u32 flags)
7610 {
7611         if (flags & ~EXCHGID4_FLAG_MASK_R)
7612                 goto out_inval;
7613         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7614             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7615                 goto out_inval;
7616         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7617                 goto out_inval;
7618         return NFS_OK;
7619 out_inval:
7620         return -NFS4ERR_INVAL;
7621 }
7622
7623 static bool
7624 nfs41_same_server_scope(struct nfs41_server_scope *a,
7625                         struct nfs41_server_scope *b)
7626 {
7627         if (a->server_scope_sz != b->server_scope_sz)
7628                 return false;
7629         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7630 }
7631
7632 static void
7633 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7634 {
7635 }
7636
7637 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7638         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7639 };
7640
7641 /*
7642  * nfs4_proc_bind_one_conn_to_session()
7643  *
7644  * The 4.1 client currently uses the same TCP connection for the
7645  * fore and backchannel.
7646  */
7647 static
7648 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7649                 struct rpc_xprt *xprt,
7650                 struct nfs_client *clp,
7651                 const struct cred *cred)
7652 {
7653         int status;
7654         struct nfs41_bind_conn_to_session_args args = {
7655                 .client = clp,
7656                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7657         };
7658         struct nfs41_bind_conn_to_session_res res;
7659         struct rpc_message msg = {
7660                 .rpc_proc =
7661                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7662                 .rpc_argp = &args,
7663                 .rpc_resp = &res,
7664                 .rpc_cred = cred,
7665         };
7666         struct rpc_task_setup task_setup_data = {
7667                 .rpc_client = clnt,
7668                 .rpc_xprt = xprt,
7669                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7670                 .rpc_message = &msg,
7671                 .flags = RPC_TASK_TIMEOUT,
7672         };
7673         struct rpc_task *task;
7674
7675         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7676         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7677                 args.dir = NFS4_CDFC4_FORE;
7678
7679         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7680         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7681                 args.dir = NFS4_CDFC4_FORE;
7682
7683         task = rpc_run_task(&task_setup_data);
7684         if (!IS_ERR(task)) {
7685                 status = task->tk_status;
7686                 rpc_put_task(task);
7687         } else
7688                 status = PTR_ERR(task);
7689         trace_nfs4_bind_conn_to_session(clp, status);
7690         if (status == 0) {
7691                 if (memcmp(res.sessionid.data,
7692                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7693                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7694                         return -EIO;
7695                 }
7696                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7697                         dprintk("NFS: %s: Unexpected direction from server\n",
7698                                 __func__);
7699                         return -EIO;
7700                 }
7701                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7702                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7703                                 __func__);
7704                         return -EIO;
7705                 }
7706         }
7707
7708         return status;
7709 }
7710
7711 struct rpc_bind_conn_calldata {
7712         struct nfs_client *clp;
7713         const struct cred *cred;
7714 };
7715
7716 static int
7717 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7718                 struct rpc_xprt *xprt,
7719                 void *calldata)
7720 {
7721         struct rpc_bind_conn_calldata *p = calldata;
7722
7723         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7724 }
7725
7726 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
7727 {
7728         struct rpc_bind_conn_calldata data = {
7729                 .clp = clp,
7730                 .cred = cred,
7731         };
7732         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7733                         nfs4_proc_bind_conn_to_session_callback, &data);
7734 }
7735
7736 /*
7737  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7738  * and operations we'd like to see to enable certain features in the allow map
7739  */
7740 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7741         .how = SP4_MACH_CRED,
7742         .enforce.u.words = {
7743                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7744                       1 << (OP_EXCHANGE_ID - 32) |
7745                       1 << (OP_CREATE_SESSION - 32) |
7746                       1 << (OP_DESTROY_SESSION - 32) |
7747                       1 << (OP_DESTROY_CLIENTID - 32)
7748         },
7749         .allow.u.words = {
7750                 [0] = 1 << (OP_CLOSE) |
7751                       1 << (OP_OPEN_DOWNGRADE) |
7752                       1 << (OP_LOCKU) |
7753                       1 << (OP_DELEGRETURN) |
7754                       1 << (OP_COMMIT),
7755                 [1] = 1 << (OP_SECINFO - 32) |
7756                       1 << (OP_SECINFO_NO_NAME - 32) |
7757                       1 << (OP_LAYOUTRETURN - 32) |
7758                       1 << (OP_TEST_STATEID - 32) |
7759                       1 << (OP_FREE_STATEID - 32) |
7760                       1 << (OP_WRITE - 32)
7761         }
7762 };
7763
7764 /*
7765  * Select the state protection mode for client `clp' given the server results
7766  * from exchange_id in `sp'.
7767  *
7768  * Returns 0 on success, negative errno otherwise.
7769  */
7770 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7771                                  struct nfs41_state_protection *sp)
7772 {
7773         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7774                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7775                       1 << (OP_EXCHANGE_ID - 32) |
7776                       1 << (OP_CREATE_SESSION - 32) |
7777                       1 << (OP_DESTROY_SESSION - 32) |
7778                       1 << (OP_DESTROY_CLIENTID - 32)
7779         };
7780         unsigned long flags = 0;
7781         unsigned int i;
7782         int ret = 0;
7783
7784         if (sp->how == SP4_MACH_CRED) {
7785                 /* Print state protect result */
7786                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7787                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7788                         if (test_bit(i, sp->enforce.u.longs))
7789                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7790                         if (test_bit(i, sp->allow.u.longs))
7791                                 dfprintk(MOUNT, "  allow op %d\n", i);
7792                 }
7793
7794                 /* make sure nothing is on enforce list that isn't supported */
7795                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7796                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7797                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7798                                 ret = -EINVAL;
7799                                 goto out;
7800                         }
7801                 }
7802
7803                 /*
7804                  * Minimal mode - state operations are allowed to use machine
7805                  * credential.  Note this already happens by default, so the
7806                  * client doesn't have to do anything more than the negotiation.
7807                  *
7808                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7809                  *       we're already using the machine cred for exchange_id
7810                  *       and will never use a different cred.
7811                  */
7812                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7813                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7814                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7815                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7816                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7817                         dfprintk(MOUNT, "  minimal mode enabled\n");
7818                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
7819                 } else {
7820                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7821                         ret = -EINVAL;
7822                         goto out;
7823                 }
7824
7825                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7826                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7827                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7828                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7829                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7830                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
7831                 }
7832
7833                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7834                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7835                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
7836                 }
7837
7838                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7839                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7840                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7841                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
7842                 }
7843
7844                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7845                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7846                         dfprintk(MOUNT, "  stateid mode enabled\n");
7847                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
7848                 }
7849
7850                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7851                         dfprintk(MOUNT, "  write mode enabled\n");
7852                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
7853                 }
7854
7855                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7856                         dfprintk(MOUNT, "  commit mode enabled\n");
7857                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
7858                 }
7859         }
7860 out:
7861         clp->cl_sp4_flags = flags;
7862         return ret;
7863 }
7864
7865 struct nfs41_exchange_id_data {
7866         struct nfs41_exchange_id_res res;
7867         struct nfs41_exchange_id_args args;
7868 };
7869
7870 static void nfs4_exchange_id_release(void *data)
7871 {
7872         struct nfs41_exchange_id_data *cdata =
7873                                         (struct nfs41_exchange_id_data *)data;
7874
7875         nfs_put_client(cdata->args.client);
7876         kfree(cdata->res.impl_id);
7877         kfree(cdata->res.server_scope);
7878         kfree(cdata->res.server_owner);
7879         kfree(cdata);
7880 }
7881
7882 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7883         .rpc_release = nfs4_exchange_id_release,
7884 };
7885
7886 /*
7887  * _nfs4_proc_exchange_id()
7888  *
7889  * Wrapper for EXCHANGE_ID operation.
7890  */
7891 static struct rpc_task *
7892 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
7893                         u32 sp4_how, struct rpc_xprt *xprt)
7894 {
7895         struct rpc_message msg = {
7896                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7897                 .rpc_cred = cred,
7898         };
7899         struct rpc_task_setup task_setup_data = {
7900                 .rpc_client = clp->cl_rpcclient,
7901                 .callback_ops = &nfs4_exchange_id_call_ops,
7902                 .rpc_message = &msg,
7903                 .flags = RPC_TASK_TIMEOUT,
7904         };
7905         struct nfs41_exchange_id_data *calldata;
7906         int status;
7907
7908         if (!refcount_inc_not_zero(&clp->cl_count))
7909                 return ERR_PTR(-EIO);
7910
7911         status = -ENOMEM;
7912         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7913         if (!calldata)
7914                 goto out;
7915
7916         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
7917
7918         status = nfs4_init_uniform_client_string(clp);
7919         if (status)
7920                 goto out_calldata;
7921
7922         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7923                                                 GFP_NOFS);
7924         status = -ENOMEM;
7925         if (unlikely(calldata->res.server_owner == NULL))
7926                 goto out_calldata;
7927
7928         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7929                                         GFP_NOFS);
7930         if (unlikely(calldata->res.server_scope == NULL))
7931                 goto out_server_owner;
7932
7933         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7934         if (unlikely(calldata->res.impl_id == NULL))
7935                 goto out_server_scope;
7936
7937         switch (sp4_how) {
7938         case SP4_NONE:
7939                 calldata->args.state_protect.how = SP4_NONE;
7940                 break;
7941
7942         case SP4_MACH_CRED:
7943                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7944                 break;
7945
7946         default:
7947                 /* unsupported! */
7948                 WARN_ON_ONCE(1);
7949                 status = -EINVAL;
7950                 goto out_impl_id;
7951         }
7952         if (xprt) {
7953                 task_setup_data.rpc_xprt = xprt;
7954                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
7955                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
7956                                 sizeof(calldata->args.verifier.data));
7957         }
7958         calldata->args.client = clp;
7959         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7960         EXCHGID4_FLAG_BIND_PRINC_STATEID;
7961 #ifdef CONFIG_NFS_V4_1_MIGRATION
7962         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
7963 #endif
7964         msg.rpc_argp = &calldata->args;
7965         msg.rpc_resp = &calldata->res;
7966         task_setup_data.callback_data = calldata;
7967
7968         return rpc_run_task(&task_setup_data);
7969
7970 out_impl_id:
7971         kfree(calldata->res.impl_id);
7972 out_server_scope:
7973         kfree(calldata->res.server_scope);
7974 out_server_owner:
7975         kfree(calldata->res.server_owner);
7976 out_calldata:
7977         kfree(calldata);
7978 out:
7979         nfs_put_client(clp);
7980         return ERR_PTR(status);
7981 }
7982
7983 /*
7984  * _nfs4_proc_exchange_id()
7985  *
7986  * Wrapper for EXCHANGE_ID operation.
7987  */
7988 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
7989                         u32 sp4_how)
7990 {
7991         struct rpc_task *task;
7992         struct nfs41_exchange_id_args *argp;
7993         struct nfs41_exchange_id_res *resp;
7994         int status;
7995
7996         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
7997         if (IS_ERR(task))
7998                 return PTR_ERR(task);
7999
8000         argp = task->tk_msg.rpc_argp;
8001         resp = task->tk_msg.rpc_resp;
8002         status = task->tk_status;
8003         if (status  != 0)
8004                 goto out;
8005
8006         status = nfs4_check_cl_exchange_flags(resp->flags);
8007         if (status  != 0)
8008                 goto out;
8009
8010         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8011         if (status != 0)
8012                 goto out;
8013
8014         clp->cl_clientid = resp->clientid;
8015         clp->cl_exchange_flags = resp->flags;
8016         clp->cl_seqid = resp->seqid;
8017         /* Client ID is not confirmed */
8018         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8019                 clear_bit(NFS4_SESSION_ESTABLISHED,
8020                           &clp->cl_session->session_state);
8021
8022         if (clp->cl_serverscope != NULL &&
8023             !nfs41_same_server_scope(clp->cl_serverscope,
8024                                 resp->server_scope)) {
8025                 dprintk("%s: server_scope mismatch detected\n",
8026                         __func__);
8027                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8028         }
8029
8030         swap(clp->cl_serverowner, resp->server_owner);
8031         swap(clp->cl_serverscope, resp->server_scope);
8032         swap(clp->cl_implid, resp->impl_id);
8033
8034         /* Save the EXCHANGE_ID verifier session trunk tests */
8035         memcpy(clp->cl_confirm.data, argp->verifier.data,
8036                sizeof(clp->cl_confirm.data));
8037 out:
8038         trace_nfs4_exchange_id(clp, status);
8039         rpc_put_task(task);
8040         return status;
8041 }
8042
8043 /*
8044  * nfs4_proc_exchange_id()
8045  *
8046  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8047  *
8048  * Since the clientid has expired, all compounds using sessions
8049  * associated with the stale clientid will be returning
8050  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8051  * be in some phase of session reset.
8052  *
8053  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8054  */
8055 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8056 {
8057         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8058         int status;
8059
8060         /* try SP4_MACH_CRED if krb5i/p */
8061         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8062             authflavor == RPC_AUTH_GSS_KRB5P) {
8063                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8064                 if (!status)
8065                         return 0;
8066         }
8067
8068         /* try SP4_NONE */
8069         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8070 }
8071
8072 /**
8073  * nfs4_test_session_trunk
8074  *
8075  * This is an add_xprt_test() test function called from
8076  * rpc_clnt_setup_test_and_add_xprt.
8077  *
8078  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8079  * and is dereferrenced in nfs4_exchange_id_release
8080  *
8081  * Upon success, add the new transport to the rpc_clnt
8082  *
8083  * @clnt: struct rpc_clnt to get new transport
8084  * @xprt: the rpc_xprt to test
8085  * @data: call data for _nfs4_proc_exchange_id.
8086  */
8087 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8088                             void *data)
8089 {
8090         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8091         struct rpc_task *task;
8092         int status;
8093
8094         u32 sp4_how;
8095
8096         dprintk("--> %s try %s\n", __func__,
8097                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8098
8099         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8100
8101         /* Test connection for session trunking. Async exchange_id call */
8102         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8103         if (IS_ERR(task))
8104                 return;
8105
8106         status = task->tk_status;
8107         if (status == 0)
8108                 status = nfs4_detect_session_trunking(adata->clp,
8109                                 task->tk_msg.rpc_resp, xprt);
8110
8111         if (status == 0)
8112                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8113
8114         rpc_put_task(task);
8115 }
8116 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8117
8118 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8119                 const struct cred *cred)
8120 {
8121         struct rpc_message msg = {
8122                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8123                 .rpc_argp = clp,
8124                 .rpc_cred = cred,
8125         };
8126         int status;
8127
8128         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8129         trace_nfs4_destroy_clientid(clp, status);
8130         if (status)
8131                 dprintk("NFS: Got error %d from the server %s on "
8132                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8133         return status;
8134 }
8135
8136 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8137                 const struct cred *cred)
8138 {
8139         unsigned int loop;
8140         int ret;
8141
8142         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8143                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8144                 switch (ret) {
8145                 case -NFS4ERR_DELAY:
8146                 case -NFS4ERR_CLIENTID_BUSY:
8147                         ssleep(1);
8148                         break;
8149                 default:
8150                         return ret;
8151                 }
8152         }
8153         return 0;
8154 }
8155
8156 int nfs4_destroy_clientid(struct nfs_client *clp)
8157 {
8158         const struct cred *cred;
8159         int ret = 0;
8160
8161         if (clp->cl_mvops->minor_version < 1)
8162                 goto out;
8163         if (clp->cl_exchange_flags == 0)
8164                 goto out;
8165         if (clp->cl_preserve_clid)
8166                 goto out;
8167         cred = nfs4_get_clid_cred(clp);
8168         ret = nfs4_proc_destroy_clientid(clp, cred);
8169         put_cred(cred);
8170         switch (ret) {
8171         case 0:
8172         case -NFS4ERR_STALE_CLIENTID:
8173                 clp->cl_exchange_flags = 0;
8174         }
8175 out:
8176         return ret;
8177 }
8178
8179 struct nfs4_get_lease_time_data {
8180         struct nfs4_get_lease_time_args *args;
8181         struct nfs4_get_lease_time_res *res;
8182         struct nfs_client *clp;
8183 };
8184
8185 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8186                                         void *calldata)
8187 {
8188         struct nfs4_get_lease_time_data *data =
8189                         (struct nfs4_get_lease_time_data *)calldata;
8190
8191         dprintk("--> %s\n", __func__);
8192         /* just setup sequence, do not trigger session recovery
8193            since we're invoked within one */
8194         nfs4_setup_sequence(data->clp,
8195                         &data->args->la_seq_args,
8196                         &data->res->lr_seq_res,
8197                         task);
8198         dprintk("<-- %s\n", __func__);
8199 }
8200
8201 /*
8202  * Called from nfs4_state_manager thread for session setup, so don't recover
8203  * from sequence operation or clientid errors.
8204  */
8205 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8206 {
8207         struct nfs4_get_lease_time_data *data =
8208                         (struct nfs4_get_lease_time_data *)calldata;
8209
8210         dprintk("--> %s\n", __func__);
8211         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
8212                 return;
8213         switch (task->tk_status) {
8214         case -NFS4ERR_DELAY:
8215         case -NFS4ERR_GRACE:
8216                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8217                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8218                 task->tk_status = 0;
8219                 /* fall through */
8220         case -NFS4ERR_RETRY_UNCACHED_REP:
8221                 rpc_restart_call_prepare(task);
8222                 return;
8223         }
8224         dprintk("<-- %s\n", __func__);
8225 }
8226
8227 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8228         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8229         .rpc_call_done = nfs4_get_lease_time_done,
8230 };
8231
8232 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8233 {
8234         struct rpc_task *task;
8235         struct nfs4_get_lease_time_args args;
8236         struct nfs4_get_lease_time_res res = {
8237                 .lr_fsinfo = fsinfo,
8238         };
8239         struct nfs4_get_lease_time_data data = {
8240                 .args = &args,
8241                 .res = &res,
8242                 .clp = clp,
8243         };
8244         struct rpc_message msg = {
8245                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8246                 .rpc_argp = &args,
8247                 .rpc_resp = &res,
8248         };
8249         struct rpc_task_setup task_setup = {
8250                 .rpc_client = clp->cl_rpcclient,
8251                 .rpc_message = &msg,
8252                 .callback_ops = &nfs4_get_lease_time_ops,
8253                 .callback_data = &data,
8254                 .flags = RPC_TASK_TIMEOUT,
8255         };
8256         int status;
8257
8258         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8259         task = rpc_run_task(&task_setup);
8260
8261         if (IS_ERR(task))
8262                 return PTR_ERR(task);
8263
8264         status = task->tk_status;
8265         rpc_put_task(task);
8266         return status;
8267 }
8268
8269 /*
8270  * Initialize the values to be used by the client in CREATE_SESSION
8271  * If nfs4_init_session set the fore channel request and response sizes,
8272  * use them.
8273  *
8274  * Set the back channel max_resp_sz_cached to zero to force the client to
8275  * always set csa_cachethis to FALSE because the current implementation
8276  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8277  */
8278 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8279                                     struct rpc_clnt *clnt)
8280 {
8281         unsigned int max_rqst_sz, max_resp_sz;
8282         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8283
8284         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8285         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8286
8287         /* Fore channel attributes */
8288         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8289         args->fc_attrs.max_resp_sz = max_resp_sz;
8290         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8291         args->fc_attrs.max_reqs = max_session_slots;
8292
8293         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8294                 "max_ops=%u max_reqs=%u\n",
8295                 __func__,
8296                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8297                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8298
8299         /* Back channel attributes */
8300         args->bc_attrs.max_rqst_sz = max_bc_payload;
8301         args->bc_attrs.max_resp_sz = max_bc_payload;
8302         args->bc_attrs.max_resp_sz_cached = 0;
8303         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8304         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8305
8306         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8307                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8308                 __func__,
8309                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8310                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8311                 args->bc_attrs.max_reqs);
8312 }
8313
8314 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8315                 struct nfs41_create_session_res *res)
8316 {
8317         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8318         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8319
8320         if (rcvd->max_resp_sz > sent->max_resp_sz)
8321                 return -EINVAL;
8322         /*
8323          * Our requested max_ops is the minimum we need; we're not
8324          * prepared to break up compounds into smaller pieces than that.
8325          * So, no point even trying to continue if the server won't
8326          * cooperate:
8327          */
8328         if (rcvd->max_ops < sent->max_ops)
8329                 return -EINVAL;
8330         if (rcvd->max_reqs == 0)
8331                 return -EINVAL;
8332         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8333                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8334         return 0;
8335 }
8336
8337 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8338                 struct nfs41_create_session_res *res)
8339 {
8340         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8341         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8342
8343         if (!(res->flags & SESSION4_BACK_CHAN))
8344                 goto out;
8345         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8346                 return -EINVAL;
8347         if (rcvd->max_resp_sz < sent->max_resp_sz)
8348                 return -EINVAL;
8349         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8350                 return -EINVAL;
8351         if (rcvd->max_ops > sent->max_ops)
8352                 return -EINVAL;
8353         if (rcvd->max_reqs > sent->max_reqs)
8354                 return -EINVAL;
8355 out:
8356         return 0;
8357 }
8358
8359 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8360                                      struct nfs41_create_session_res *res)
8361 {
8362         int ret;
8363
8364         ret = nfs4_verify_fore_channel_attrs(args, res);
8365         if (ret)
8366                 return ret;
8367         return nfs4_verify_back_channel_attrs(args, res);
8368 }
8369
8370 static void nfs4_update_session(struct nfs4_session *session,
8371                 struct nfs41_create_session_res *res)
8372 {
8373         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8374         /* Mark client id and session as being confirmed */
8375         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8376         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8377         session->flags = res->flags;
8378         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8379         if (res->flags & SESSION4_BACK_CHAN)
8380                 memcpy(&session->bc_attrs, &res->bc_attrs,
8381                                 sizeof(session->bc_attrs));
8382 }
8383
8384 static int _nfs4_proc_create_session(struct nfs_client *clp,
8385                 const struct cred *cred)
8386 {
8387         struct nfs4_session *session = clp->cl_session;
8388         struct nfs41_create_session_args args = {
8389                 .client = clp,
8390                 .clientid = clp->cl_clientid,
8391                 .seqid = clp->cl_seqid,
8392                 .cb_program = NFS4_CALLBACK,
8393         };
8394         struct nfs41_create_session_res res;
8395
8396         struct rpc_message msg = {
8397                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8398                 .rpc_argp = &args,
8399                 .rpc_resp = &res,
8400                 .rpc_cred = cred,
8401         };
8402         int status;
8403
8404         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8405         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8406
8407         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8408         trace_nfs4_create_session(clp, status);
8409
8410         switch (status) {
8411         case -NFS4ERR_STALE_CLIENTID:
8412         case -NFS4ERR_DELAY:
8413         case -ETIMEDOUT:
8414         case -EACCES:
8415         case -EAGAIN:
8416                 goto out;
8417         };
8418
8419         clp->cl_seqid++;
8420         if (!status) {
8421                 /* Verify the session's negotiated channel_attrs values */
8422                 status = nfs4_verify_channel_attrs(&args, &res);
8423                 /* Increment the clientid slot sequence id */
8424                 if (status)
8425                         goto out;
8426                 nfs4_update_session(session, &res);
8427         }
8428 out:
8429         return status;
8430 }
8431
8432 /*
8433  * Issues a CREATE_SESSION operation to the server.
8434  * It is the responsibility of the caller to verify the session is
8435  * expired before calling this routine.
8436  */
8437 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8438 {
8439         int status;
8440         unsigned *ptr;
8441         struct nfs4_session *session = clp->cl_session;
8442
8443         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8444
8445         status = _nfs4_proc_create_session(clp, cred);
8446         if (status)
8447                 goto out;
8448
8449         /* Init or reset the session slot tables */
8450         status = nfs4_setup_session_slot_tables(session);
8451         dprintk("slot table setup returned %d\n", status);
8452         if (status)
8453                 goto out;
8454
8455         ptr = (unsigned *)&session->sess_id.data[0];
8456         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8457                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8458 out:
8459         dprintk("<-- %s\n", __func__);
8460         return status;
8461 }
8462
8463 /*
8464  * Issue the over-the-wire RPC DESTROY_SESSION.
8465  * The caller must serialize access to this routine.
8466  */
8467 int nfs4_proc_destroy_session(struct nfs4_session *session,
8468                 const struct cred *cred)
8469 {
8470         struct rpc_message msg = {
8471                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8472                 .rpc_argp = session,
8473                 .rpc_cred = cred,
8474         };
8475         int status = 0;
8476
8477         dprintk("--> nfs4_proc_destroy_session\n");
8478
8479         /* session is still being setup */
8480         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8481                 return 0;
8482
8483         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8484         trace_nfs4_destroy_session(session->clp, status);
8485
8486         if (status)
8487                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8488                         "Session has been destroyed regardless...\n", status);
8489
8490         dprintk("<-- nfs4_proc_destroy_session\n");
8491         return status;
8492 }
8493
8494 /*
8495  * Renew the cl_session lease.
8496  */
8497 struct nfs4_sequence_data {
8498         struct nfs_client *clp;
8499         struct nfs4_sequence_args args;
8500         struct nfs4_sequence_res res;
8501 };
8502
8503 static void nfs41_sequence_release(void *data)
8504 {
8505         struct nfs4_sequence_data *calldata = data;
8506         struct nfs_client *clp = calldata->clp;
8507
8508         if (refcount_read(&clp->cl_count) > 1)
8509                 nfs4_schedule_state_renewal(clp);
8510         nfs_put_client(clp);
8511         kfree(calldata);
8512 }
8513
8514 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8515 {
8516         switch(task->tk_status) {
8517         case -NFS4ERR_DELAY:
8518                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8519                 return -EAGAIN;
8520         default:
8521                 nfs4_schedule_lease_recovery(clp);
8522         }
8523         return 0;
8524 }
8525
8526 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8527 {
8528         struct nfs4_sequence_data *calldata = data;
8529         struct nfs_client *clp = calldata->clp;
8530
8531         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8532                 return;
8533
8534         trace_nfs4_sequence(clp, task->tk_status);
8535         if (task->tk_status < 0) {
8536                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8537                 if (refcount_read(&clp->cl_count) == 1)
8538                         goto out;
8539
8540                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8541                         rpc_restart_call_prepare(task);
8542                         return;
8543                 }
8544         }
8545         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8546 out:
8547         dprintk("<-- %s\n", __func__);
8548 }
8549
8550 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8551 {
8552         struct nfs4_sequence_data *calldata = data;
8553         struct nfs_client *clp = calldata->clp;
8554         struct nfs4_sequence_args *args;
8555         struct nfs4_sequence_res *res;
8556
8557         args = task->tk_msg.rpc_argp;
8558         res = task->tk_msg.rpc_resp;
8559
8560         nfs4_setup_sequence(clp, args, res, task);
8561 }
8562
8563 static const struct rpc_call_ops nfs41_sequence_ops = {
8564         .rpc_call_done = nfs41_sequence_call_done,
8565         .rpc_call_prepare = nfs41_sequence_prepare,
8566         .rpc_release = nfs41_sequence_release,
8567 };
8568
8569 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8570                 const struct cred *cred,
8571                 struct nfs4_slot *slot,
8572                 bool is_privileged)
8573 {
8574         struct nfs4_sequence_data *calldata;
8575         struct rpc_message msg = {
8576                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8577                 .rpc_cred = cred,
8578         };
8579         struct rpc_task_setup task_setup_data = {
8580                 .rpc_client = clp->cl_rpcclient,
8581                 .rpc_message = &msg,
8582                 .callback_ops = &nfs41_sequence_ops,
8583                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8584         };
8585         struct rpc_task *ret;
8586
8587         ret = ERR_PTR(-EIO);
8588         if (!refcount_inc_not_zero(&clp->cl_count))
8589                 goto out_err;
8590
8591         ret = ERR_PTR(-ENOMEM);
8592         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8593         if (calldata == NULL)
8594                 goto out_put_clp;
8595         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8596         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8597         msg.rpc_argp = &calldata->args;
8598         msg.rpc_resp = &calldata->res;
8599         calldata->clp = clp;
8600         task_setup_data.callback_data = calldata;
8601
8602         ret = rpc_run_task(&task_setup_data);
8603         if (IS_ERR(ret))
8604                 goto out_err;
8605         return ret;
8606 out_put_clp:
8607         nfs_put_client(clp);
8608 out_err:
8609         nfs41_release_slot(slot);
8610         return ret;
8611 }
8612
8613 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
8614 {
8615         struct rpc_task *task;
8616         int ret = 0;
8617
8618         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8619                 return -EAGAIN;
8620         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8621         if (IS_ERR(task))
8622                 ret = PTR_ERR(task);
8623         else
8624                 rpc_put_task_async(task);
8625         dprintk("<-- %s status=%d\n", __func__, ret);
8626         return ret;
8627 }
8628
8629 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
8630 {
8631         struct rpc_task *task;
8632         int ret;
8633
8634         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8635         if (IS_ERR(task)) {
8636                 ret = PTR_ERR(task);
8637                 goto out;
8638         }
8639         ret = rpc_wait_for_completion_task(task);
8640         if (!ret)
8641                 ret = task->tk_status;
8642         rpc_put_task(task);
8643 out:
8644         dprintk("<-- %s status=%d\n", __func__, ret);
8645         return ret;
8646 }
8647
8648 struct nfs4_reclaim_complete_data {
8649         struct nfs_client *clp;
8650         struct nfs41_reclaim_complete_args arg;
8651         struct nfs41_reclaim_complete_res res;
8652 };
8653
8654 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8655 {
8656         struct nfs4_reclaim_complete_data *calldata = data;
8657
8658         nfs4_setup_sequence(calldata->clp,
8659                         &calldata->arg.seq_args,
8660                         &calldata->res.seq_res,
8661                         task);
8662 }
8663
8664 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8665 {
8666         switch(task->tk_status) {
8667         case 0:
8668                 wake_up_all(&clp->cl_lock_waitq);
8669                 /* Fallthrough */
8670         case -NFS4ERR_COMPLETE_ALREADY:
8671         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8672                 break;
8673         case -NFS4ERR_DELAY:
8674                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8675                 /* fall through */
8676         case -NFS4ERR_RETRY_UNCACHED_REP:
8677                 return -EAGAIN;
8678         case -NFS4ERR_BADSESSION:
8679         case -NFS4ERR_DEADSESSION:
8680         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8681                 nfs4_schedule_session_recovery(clp->cl_session,
8682                                 task->tk_status);
8683                 break;
8684         default:
8685                 nfs4_schedule_lease_recovery(clp);
8686         }
8687         return 0;
8688 }
8689
8690 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8691 {
8692         struct nfs4_reclaim_complete_data *calldata = data;
8693         struct nfs_client *clp = calldata->clp;
8694         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8695
8696         dprintk("--> %s\n", __func__);
8697         if (!nfs41_sequence_done(task, res))
8698                 return;
8699
8700         trace_nfs4_reclaim_complete(clp, task->tk_status);
8701         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8702                 rpc_restart_call_prepare(task);
8703                 return;
8704         }
8705         dprintk("<-- %s\n", __func__);
8706 }
8707
8708 static void nfs4_free_reclaim_complete_data(void *data)
8709 {
8710         struct nfs4_reclaim_complete_data *calldata = data;
8711
8712         kfree(calldata);
8713 }
8714
8715 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8716         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8717         .rpc_call_done = nfs4_reclaim_complete_done,
8718         .rpc_release = nfs4_free_reclaim_complete_data,
8719 };
8720
8721 /*
8722  * Issue a global reclaim complete.
8723  */
8724 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8725                 const struct cred *cred)
8726 {
8727         struct nfs4_reclaim_complete_data *calldata;
8728         struct rpc_task *task;
8729         struct rpc_message msg = {
8730                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8731                 .rpc_cred = cred,
8732         };
8733         struct rpc_task_setup task_setup_data = {
8734                 .rpc_client = clp->cl_rpcclient,
8735                 .rpc_message = &msg,
8736                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8737                 .flags = RPC_TASK_ASYNC,
8738         };
8739         int status = -ENOMEM;
8740
8741         dprintk("--> %s\n", __func__);
8742         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8743         if (calldata == NULL)
8744                 goto out;
8745         calldata->clp = clp;
8746         calldata->arg.one_fs = 0;
8747
8748         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
8749         msg.rpc_argp = &calldata->arg;
8750         msg.rpc_resp = &calldata->res;
8751         task_setup_data.callback_data = calldata;
8752         task = rpc_run_task(&task_setup_data);
8753         if (IS_ERR(task)) {
8754                 status = PTR_ERR(task);
8755                 goto out;
8756         }
8757         status = rpc_wait_for_completion_task(task);
8758         if (status == 0)
8759                 status = task->tk_status;
8760         rpc_put_task(task);
8761 out:
8762         dprintk("<-- %s status=%d\n", __func__, status);
8763         return status;
8764 }
8765
8766 static void
8767 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8768 {
8769         struct nfs4_layoutget *lgp = calldata;
8770         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8771
8772         dprintk("--> %s\n", __func__);
8773         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8774                                 &lgp->res.seq_res, task);
8775         dprintk("<-- %s\n", __func__);
8776 }
8777
8778 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8779 {
8780         struct nfs4_layoutget *lgp = calldata;
8781
8782         dprintk("--> %s\n", __func__);
8783         nfs41_sequence_process(task, &lgp->res.seq_res);
8784         dprintk("<-- %s\n", __func__);
8785 }
8786
8787 static int
8788 nfs4_layoutget_handle_exception(struct rpc_task *task,
8789                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8790 {
8791         struct inode *inode = lgp->args.inode;
8792         struct nfs_server *server = NFS_SERVER(inode);
8793         struct pnfs_layout_hdr *lo;
8794         int nfs4err = task->tk_status;
8795         int err, status = 0;
8796         LIST_HEAD(head);
8797
8798         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8799
8800         nfs4_sequence_free_slot(&lgp->res.seq_res);
8801
8802         switch (nfs4err) {
8803         case 0:
8804                 goto out;
8805
8806         /*
8807          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8808          * on the file. set tk_status to -ENODATA to tell upper layer to
8809          * retry go inband.
8810          */
8811         case -NFS4ERR_LAYOUTUNAVAILABLE:
8812                 status = -ENODATA;
8813                 goto out;
8814         /*
8815          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8816          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8817          */
8818         case -NFS4ERR_BADLAYOUT:
8819                 status = -EOVERFLOW;
8820                 goto out;
8821         /*
8822          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8823          * (or clients) writing to the same RAID stripe except when
8824          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8825          *
8826          * Treat it like we would RECALLCONFLICT -- we retry for a little
8827          * while, and then eventually give up.
8828          */
8829         case -NFS4ERR_LAYOUTTRYLATER:
8830                 if (lgp->args.minlength == 0) {
8831                         status = -EOVERFLOW;
8832                         goto out;
8833                 }
8834                 status = -EBUSY;
8835                 break;
8836         case -NFS4ERR_RECALLCONFLICT:
8837                 status = -ERECALLCONFLICT;
8838                 break;
8839         case -NFS4ERR_DELEG_REVOKED:
8840         case -NFS4ERR_ADMIN_REVOKED:
8841         case -NFS4ERR_EXPIRED:
8842         case -NFS4ERR_BAD_STATEID:
8843                 exception->timeout = 0;
8844                 spin_lock(&inode->i_lock);
8845                 lo = NFS_I(inode)->layout;
8846                 /* If the open stateid was bad, then recover it. */
8847                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8848                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
8849                         spin_unlock(&inode->i_lock);
8850                         exception->state = lgp->args.ctx->state;
8851                         exception->stateid = &lgp->args.stateid;
8852                         break;
8853                 }
8854
8855                 /*
8856                  * Mark the bad layout state as invalid, then retry
8857                  */
8858                 pnfs_mark_layout_stateid_invalid(lo, &head);
8859                 spin_unlock(&inode->i_lock);
8860                 nfs_commit_inode(inode, 0);
8861                 pnfs_free_lseg_list(&head);
8862                 status = -EAGAIN;
8863                 goto out;
8864         }
8865
8866         err = nfs4_handle_exception(server, nfs4err, exception);
8867         if (!status) {
8868                 if (exception->retry)
8869                         status = -EAGAIN;
8870                 else
8871                         status = err;
8872         }
8873 out:
8874         dprintk("<-- %s\n", __func__);
8875         return status;
8876 }
8877
8878 size_t max_response_pages(struct nfs_server *server)
8879 {
8880         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8881         return nfs_page_array_len(0, max_resp_sz);
8882 }
8883
8884 static void nfs4_layoutget_release(void *calldata)
8885 {
8886         struct nfs4_layoutget *lgp = calldata;
8887
8888         dprintk("--> %s\n", __func__);
8889         nfs4_sequence_free_slot(&lgp->res.seq_res);
8890         pnfs_layoutget_free(lgp);
8891         dprintk("<-- %s\n", __func__);
8892 }
8893
8894 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8895         .rpc_call_prepare = nfs4_layoutget_prepare,
8896         .rpc_call_done = nfs4_layoutget_done,
8897         .rpc_release = nfs4_layoutget_release,
8898 };
8899
8900 struct pnfs_layout_segment *
8901 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
8902 {
8903         struct inode *inode = lgp->args.inode;
8904         struct nfs_server *server = NFS_SERVER(inode);
8905         struct rpc_task *task;
8906         struct rpc_message msg = {
8907                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8908                 .rpc_argp = &lgp->args,
8909                 .rpc_resp = &lgp->res,
8910                 .rpc_cred = lgp->cred,
8911         };
8912         struct rpc_task_setup task_setup_data = {
8913                 .rpc_client = server->client,
8914                 .rpc_message = &msg,
8915                 .callback_ops = &nfs4_layoutget_call_ops,
8916                 .callback_data = lgp,
8917                 .flags = RPC_TASK_ASYNC,
8918         };
8919         struct pnfs_layout_segment *lseg = NULL;
8920         struct nfs4_exception exception = {
8921                 .inode = inode,
8922                 .timeout = *timeout,
8923         };
8924         int status = 0;
8925
8926         dprintk("--> %s\n", __func__);
8927
8928         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8929         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8930
8931         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
8932
8933         task = rpc_run_task(&task_setup_data);
8934         if (IS_ERR(task))
8935                 return ERR_CAST(task);
8936         status = rpc_wait_for_completion_task(task);
8937         if (status != 0)
8938                 goto out;
8939
8940         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8941         if (task->tk_status < 0 || lgp->res.layoutp->len == 0) {
8942                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8943                 *timeout = exception.timeout;
8944         } else
8945                 lseg = pnfs_layout_process(lgp);
8946 out:
8947         trace_nfs4_layoutget(lgp->args.ctx,
8948                         &lgp->args.range,
8949                         &lgp->res.range,
8950                         &lgp->res.stateid,
8951                         status);
8952
8953         rpc_put_task(task);
8954         dprintk("<-- %s status=%d\n", __func__, status);
8955         if (status)
8956                 return ERR_PTR(status);
8957         return lseg;
8958 }
8959
8960 static void
8961 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8962 {
8963         struct nfs4_layoutreturn *lrp = calldata;
8964
8965         dprintk("--> %s\n", __func__);
8966         nfs4_setup_sequence(lrp->clp,
8967                         &lrp->args.seq_args,
8968                         &lrp->res.seq_res,
8969                         task);
8970         if (!pnfs_layout_is_valid(lrp->args.layout))
8971                 rpc_exit(task, 0);
8972 }
8973
8974 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8975 {
8976         struct nfs4_layoutreturn *lrp = calldata;
8977         struct nfs_server *server;
8978
8979         dprintk("--> %s\n", __func__);
8980
8981         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8982                 return;
8983
8984         server = NFS_SERVER(lrp->args.inode);
8985         switch (task->tk_status) {
8986         case -NFS4ERR_OLD_STATEID:
8987                 if (nfs4_layoutreturn_refresh_stateid(&lrp->args.stateid,
8988                                         &lrp->args.range,
8989                                         lrp->args.inode))
8990                         goto out_restart;
8991                 /* Fallthrough */
8992         default:
8993                 task->tk_status = 0;
8994                 /* Fallthrough */
8995         case 0:
8996                 break;
8997         case -NFS4ERR_DELAY:
8998                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8999                         break;
9000                 goto out_restart;
9001         }
9002         dprintk("<-- %s\n", __func__);
9003         return;
9004 out_restart:
9005         task->tk_status = 0;
9006         nfs4_sequence_free_slot(&lrp->res.seq_res);
9007         rpc_restart_call_prepare(task);
9008 }
9009
9010 static void nfs4_layoutreturn_release(void *calldata)
9011 {
9012         struct nfs4_layoutreturn *lrp = calldata;
9013         struct pnfs_layout_hdr *lo = lrp->args.layout;
9014
9015         dprintk("--> %s\n", __func__);
9016         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9017                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9018         nfs4_sequence_free_slot(&lrp->res.seq_res);
9019         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9020                 lrp->ld_private.ops->free(&lrp->ld_private);
9021         pnfs_put_layout_hdr(lrp->args.layout);
9022         nfs_iput_and_deactive(lrp->inode);
9023         kfree(calldata);
9024         dprintk("<-- %s\n", __func__);
9025 }
9026
9027 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9028         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9029         .rpc_call_done = nfs4_layoutreturn_done,
9030         .rpc_release = nfs4_layoutreturn_release,
9031 };
9032
9033 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9034 {
9035         struct rpc_task *task;
9036         struct rpc_message msg = {
9037                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9038                 .rpc_argp = &lrp->args,
9039                 .rpc_resp = &lrp->res,
9040                 .rpc_cred = lrp->cred,
9041         };
9042         struct rpc_task_setup task_setup_data = {
9043                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9044                 .rpc_message = &msg,
9045                 .callback_ops = &nfs4_layoutreturn_call_ops,
9046                 .callback_data = lrp,
9047         };
9048         int status = 0;
9049
9050         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9051                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9052                         &task_setup_data.rpc_client, &msg);
9053
9054         dprintk("--> %s\n", __func__);
9055         if (!sync) {
9056                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9057                 if (!lrp->inode) {
9058                         nfs4_layoutreturn_release(lrp);
9059                         return -EAGAIN;
9060                 }
9061                 task_setup_data.flags |= RPC_TASK_ASYNC;
9062         }
9063         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9064         task = rpc_run_task(&task_setup_data);
9065         if (IS_ERR(task))
9066                 return PTR_ERR(task);
9067         if (sync)
9068                 status = task->tk_status;
9069         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9070         dprintk("<-- %s status=%d\n", __func__, status);
9071         rpc_put_task(task);
9072         return status;
9073 }
9074
9075 static int
9076 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9077                 struct pnfs_device *pdev,
9078                 const struct cred *cred)
9079 {
9080         struct nfs4_getdeviceinfo_args args = {
9081                 .pdev = pdev,
9082                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9083                         NOTIFY_DEVICEID4_DELETE,
9084         };
9085         struct nfs4_getdeviceinfo_res res = {
9086                 .pdev = pdev,
9087         };
9088         struct rpc_message msg = {
9089                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9090                 .rpc_argp = &args,
9091                 .rpc_resp = &res,
9092                 .rpc_cred = cred,
9093         };
9094         int status;
9095
9096         dprintk("--> %s\n", __func__);
9097         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9098         if (res.notification & ~args.notify_types)
9099                 dprintk("%s: unsupported notification\n", __func__);
9100         if (res.notification != args.notify_types)
9101                 pdev->nocache = 1;
9102
9103         dprintk("<-- %s status=%d\n", __func__, status);
9104
9105         return status;
9106 }
9107
9108 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9109                 struct pnfs_device *pdev,
9110                 const struct cred *cred)
9111 {
9112         struct nfs4_exception exception = { };
9113         int err;
9114
9115         do {
9116                 err = nfs4_handle_exception(server,
9117                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9118                                         &exception);
9119         } while (exception.retry);
9120         return err;
9121 }
9122 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9123
9124 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9125 {
9126         struct nfs4_layoutcommit_data *data = calldata;
9127         struct nfs_server *server = NFS_SERVER(data->args.inode);
9128
9129         nfs4_setup_sequence(server->nfs_client,
9130                         &data->args.seq_args,
9131                         &data->res.seq_res,
9132                         task);
9133 }
9134
9135 static void
9136 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9137 {
9138         struct nfs4_layoutcommit_data *data = calldata;
9139         struct nfs_server *server = NFS_SERVER(data->args.inode);
9140
9141         if (!nfs41_sequence_done(task, &data->res.seq_res))
9142                 return;
9143
9144         switch (task->tk_status) { /* Just ignore these failures */
9145         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9146         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9147         case -NFS4ERR_BADLAYOUT:     /* no layout */
9148         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9149                 task->tk_status = 0;
9150         case 0:
9151                 break;
9152         default:
9153                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9154                         rpc_restart_call_prepare(task);
9155                         return;
9156                 }
9157         }
9158 }
9159
9160 static void nfs4_layoutcommit_release(void *calldata)
9161 {
9162         struct nfs4_layoutcommit_data *data = calldata;
9163
9164         pnfs_cleanup_layoutcommit(data);
9165         nfs_post_op_update_inode_force_wcc(data->args.inode,
9166                                            data->res.fattr);
9167         put_cred(data->cred);
9168         nfs_iput_and_deactive(data->inode);
9169         kfree(data);
9170 }
9171
9172 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9173         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9174         .rpc_call_done = nfs4_layoutcommit_done,
9175         .rpc_release = nfs4_layoutcommit_release,
9176 };
9177
9178 int
9179 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9180 {
9181         struct rpc_message msg = {
9182                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9183                 .rpc_argp = &data->args,
9184                 .rpc_resp = &data->res,
9185                 .rpc_cred = data->cred,
9186         };
9187         struct rpc_task_setup task_setup_data = {
9188                 .task = &data->task,
9189                 .rpc_client = NFS_CLIENT(data->args.inode),
9190                 .rpc_message = &msg,
9191                 .callback_ops = &nfs4_layoutcommit_ops,
9192                 .callback_data = data,
9193         };
9194         struct rpc_task *task;
9195         int status = 0;
9196
9197         dprintk("NFS: initiating layoutcommit call. sync %d "
9198                 "lbw: %llu inode %lu\n", sync,
9199                 data->args.lastbytewritten,
9200                 data->args.inode->i_ino);
9201
9202         if (!sync) {
9203                 data->inode = nfs_igrab_and_active(data->args.inode);
9204                 if (data->inode == NULL) {
9205                         nfs4_layoutcommit_release(data);
9206                         return -EAGAIN;
9207                 }
9208                 task_setup_data.flags = RPC_TASK_ASYNC;
9209         }
9210         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9211         task = rpc_run_task(&task_setup_data);
9212         if (IS_ERR(task))
9213                 return PTR_ERR(task);
9214         if (sync)
9215                 status = task->tk_status;
9216         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9217         dprintk("%s: status %d\n", __func__, status);
9218         rpc_put_task(task);
9219         return status;
9220 }
9221
9222 /**
9223  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9224  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9225  */
9226 static int
9227 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9228                     struct nfs_fsinfo *info,
9229                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9230 {
9231         struct nfs41_secinfo_no_name_args args = {
9232                 .style = SECINFO_STYLE_CURRENT_FH,
9233         };
9234         struct nfs4_secinfo_res res = {
9235                 .flavors = flavors,
9236         };
9237         struct rpc_message msg = {
9238                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9239                 .rpc_argp = &args,
9240                 .rpc_resp = &res,
9241         };
9242         struct rpc_clnt *clnt = server->client;
9243         const struct cred *cred = NULL;
9244         int status;
9245
9246         if (use_integrity) {
9247                 clnt = server->nfs_client->cl_rpcclient;
9248                 cred = nfs4_get_clid_cred(server->nfs_client);
9249                 msg.rpc_cred = cred;
9250         }
9251
9252         dprintk("--> %s\n", __func__);
9253         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
9254                                 &res.seq_res, 0);
9255         dprintk("<-- %s status=%d\n", __func__, status);
9256
9257         put_cred(cred);
9258
9259         return status;
9260 }
9261
9262 static int
9263 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9264                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9265 {
9266         struct nfs4_exception exception = { };
9267         int err;
9268         do {
9269                 /* first try using integrity protection */
9270                 err = -NFS4ERR_WRONGSEC;
9271
9272                 /* try to use integrity protection with machine cred */
9273                 if (_nfs4_is_integrity_protected(server->nfs_client))
9274                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9275                                                           flavors, true);
9276
9277                 /*
9278                  * if unable to use integrity protection, or SECINFO with
9279                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9280                  * disallowed by spec, but exists in deployed servers) use
9281                  * the current filesystem's rpc_client and the user cred.
9282                  */
9283                 if (err == -NFS4ERR_WRONGSEC)
9284                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9285                                                           flavors, false);
9286
9287                 switch (err) {
9288                 case 0:
9289                 case -NFS4ERR_WRONGSEC:
9290                 case -ENOTSUPP:
9291                         goto out;
9292                 default:
9293                         err = nfs4_handle_exception(server, err, &exception);
9294                 }
9295         } while (exception.retry);
9296 out:
9297         return err;
9298 }
9299
9300 static int
9301 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9302                     struct nfs_fsinfo *info)
9303 {
9304         int err;
9305         struct page *page;
9306         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9307         struct nfs4_secinfo_flavors *flavors;
9308         struct nfs4_secinfo4 *secinfo;
9309         int i;
9310
9311         page = alloc_page(GFP_KERNEL);
9312         if (!page) {
9313                 err = -ENOMEM;
9314                 goto out;
9315         }
9316
9317         flavors = page_address(page);
9318         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9319
9320         /*
9321          * Fall back on "guess and check" method if
9322          * the server doesn't support SECINFO_NO_NAME
9323          */
9324         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9325                 err = nfs4_find_root_sec(server, fhandle, info);
9326                 goto out_freepage;
9327         }
9328         if (err)
9329                 goto out_freepage;
9330
9331         for (i = 0; i < flavors->num_flavors; i++) {
9332                 secinfo = &flavors->flavors[i];
9333
9334                 switch (secinfo->flavor) {
9335                 case RPC_AUTH_NULL:
9336                 case RPC_AUTH_UNIX:
9337                 case RPC_AUTH_GSS:
9338                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9339                                         &secinfo->flavor_info);
9340                         break;
9341                 default:
9342                         flavor = RPC_AUTH_MAXFLAVOR;
9343                         break;
9344                 }
9345
9346                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9347                         flavor = RPC_AUTH_MAXFLAVOR;
9348
9349                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9350                         err = nfs4_lookup_root_sec(server, fhandle,
9351                                                    info, flavor);
9352                         if (!err)
9353                                 break;
9354                 }
9355         }
9356
9357         if (flavor == RPC_AUTH_MAXFLAVOR)
9358                 err = -EPERM;
9359
9360 out_freepage:
9361         put_page(page);
9362         if (err == -EACCES)
9363                 return -EPERM;
9364 out:
9365         return err;
9366 }
9367
9368 static int _nfs41_test_stateid(struct nfs_server *server,
9369                 nfs4_stateid *stateid,
9370                 const struct cred *cred)
9371 {
9372         int status;
9373         struct nfs41_test_stateid_args args = {
9374                 .stateid = stateid,
9375         };
9376         struct nfs41_test_stateid_res res;
9377         struct rpc_message msg = {
9378                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9379                 .rpc_argp = &args,
9380                 .rpc_resp = &res,
9381                 .rpc_cred = cred,
9382         };
9383         struct rpc_clnt *rpc_client = server->client;
9384
9385         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9386                 &rpc_client, &msg);
9387
9388         dprintk("NFS call  test_stateid %p\n", stateid);
9389         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9390         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9391                         &args.seq_args, &res.seq_res);
9392         if (status != NFS_OK) {
9393                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9394                 return status;
9395         }
9396         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9397         return -res.status;
9398 }
9399
9400 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9401                 int err, struct nfs4_exception *exception)
9402 {
9403         exception->retry = 0;
9404         switch(err) {
9405         case -NFS4ERR_DELAY:
9406         case -NFS4ERR_RETRY_UNCACHED_REP:
9407                 nfs4_handle_exception(server, err, exception);
9408                 break;
9409         case -NFS4ERR_BADSESSION:
9410         case -NFS4ERR_BADSLOT:
9411         case -NFS4ERR_BAD_HIGH_SLOT:
9412         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9413         case -NFS4ERR_DEADSESSION:
9414                 nfs4_do_handle_exception(server, err, exception);
9415         }
9416 }
9417
9418 /**
9419  * nfs41_test_stateid - perform a TEST_STATEID operation
9420  *
9421  * @server: server / transport on which to perform the operation
9422  * @stateid: state ID to test
9423  * @cred: credential
9424  *
9425  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9426  * Otherwise a negative NFS4ERR value is returned if the operation
9427  * failed or the state ID is not currently valid.
9428  */
9429 static int nfs41_test_stateid(struct nfs_server *server,
9430                 nfs4_stateid *stateid,
9431                 const struct cred *cred)
9432 {
9433         struct nfs4_exception exception = { };
9434         int err;
9435         do {
9436                 err = _nfs41_test_stateid(server, stateid, cred);
9437                 nfs4_handle_delay_or_session_error(server, err, &exception);
9438         } while (exception.retry);
9439         return err;
9440 }
9441
9442 struct nfs_free_stateid_data {
9443         struct nfs_server *server;
9444         struct nfs41_free_stateid_args args;
9445         struct nfs41_free_stateid_res res;
9446 };
9447
9448 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9449 {
9450         struct nfs_free_stateid_data *data = calldata;
9451         nfs4_setup_sequence(data->server->nfs_client,
9452                         &data->args.seq_args,
9453                         &data->res.seq_res,
9454                         task);
9455 }
9456
9457 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9458 {
9459         struct nfs_free_stateid_data *data = calldata;
9460
9461         nfs41_sequence_done(task, &data->res.seq_res);
9462
9463         switch (task->tk_status) {
9464         case -NFS4ERR_DELAY:
9465                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9466                         rpc_restart_call_prepare(task);
9467         }
9468 }
9469
9470 static void nfs41_free_stateid_release(void *calldata)
9471 {
9472         kfree(calldata);
9473 }
9474
9475 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9476         .rpc_call_prepare = nfs41_free_stateid_prepare,
9477         .rpc_call_done = nfs41_free_stateid_done,
9478         .rpc_release = nfs41_free_stateid_release,
9479 };
9480
9481 /**
9482  * nfs41_free_stateid - perform a FREE_STATEID operation
9483  *
9484  * @server: server / transport on which to perform the operation
9485  * @stateid: state ID to release
9486  * @cred: credential
9487  * @is_recovery: set to true if this call needs to be privileged
9488  *
9489  * Note: this function is always asynchronous.
9490  */
9491 static int nfs41_free_stateid(struct nfs_server *server,
9492                 const nfs4_stateid *stateid,
9493                 const struct cred *cred,
9494                 bool privileged)
9495 {
9496         struct rpc_message msg = {
9497                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9498                 .rpc_cred = cred,
9499         };
9500         struct rpc_task_setup task_setup = {
9501                 .rpc_client = server->client,
9502                 .rpc_message = &msg,
9503                 .callback_ops = &nfs41_free_stateid_ops,
9504                 .flags = RPC_TASK_ASYNC,
9505         };
9506         struct nfs_free_stateid_data *data;
9507         struct rpc_task *task;
9508
9509         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9510                 &task_setup.rpc_client, &msg);
9511
9512         dprintk("NFS call  free_stateid %p\n", stateid);
9513         data = kmalloc(sizeof(*data), GFP_NOFS);
9514         if (!data)
9515                 return -ENOMEM;
9516         data->server = server;
9517         nfs4_stateid_copy(&data->args.stateid, stateid);
9518
9519         task_setup.callback_data = data;
9520
9521         msg.rpc_argp = &data->args;
9522         msg.rpc_resp = &data->res;
9523         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9524         task = rpc_run_task(&task_setup);
9525         if (IS_ERR(task))
9526                 return PTR_ERR(task);
9527         rpc_put_task(task);
9528         return 0;
9529 }
9530
9531 static void
9532 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9533 {
9534         const struct cred *cred = lsp->ls_state->owner->so_cred;
9535
9536         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9537         nfs4_free_lock_state(server, lsp);
9538 }
9539
9540 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9541                 const nfs4_stateid *s2)
9542 {
9543         if (s1->type != s2->type)
9544                 return false;
9545
9546         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9547                 return false;
9548
9549         if (s1->seqid == s2->seqid)
9550                 return true;
9551
9552         return s1->seqid == 0 || s2->seqid == 0;
9553 }
9554
9555 #endif /* CONFIG_NFS_V4_1 */
9556
9557 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9558                 const nfs4_stateid *s2)
9559 {
9560         return nfs4_stateid_match(s1, s2);
9561 }
9562
9563
9564 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9565         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9566         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9567         .recover_open   = nfs4_open_reclaim,
9568         .recover_lock   = nfs4_lock_reclaim,
9569         .establish_clid = nfs4_init_clientid,
9570         .detect_trunking = nfs40_discover_server_trunking,
9571 };
9572
9573 #if defined(CONFIG_NFS_V4_1)
9574 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9575         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9576         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9577         .recover_open   = nfs4_open_reclaim,
9578         .recover_lock   = nfs4_lock_reclaim,
9579         .establish_clid = nfs41_init_clientid,
9580         .reclaim_complete = nfs41_proc_reclaim_complete,
9581         .detect_trunking = nfs41_discover_server_trunking,
9582 };
9583 #endif /* CONFIG_NFS_V4_1 */
9584
9585 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9586         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9587         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9588         .recover_open   = nfs40_open_expired,
9589         .recover_lock   = nfs4_lock_expired,
9590         .establish_clid = nfs4_init_clientid,
9591 };
9592
9593 #if defined(CONFIG_NFS_V4_1)
9594 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9595         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9596         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9597         .recover_open   = nfs41_open_expired,
9598         .recover_lock   = nfs41_lock_expired,
9599         .establish_clid = nfs41_init_clientid,
9600 };
9601 #endif /* CONFIG_NFS_V4_1 */
9602
9603 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9604         .sched_state_renewal = nfs4_proc_async_renew,
9605         .get_state_renewal_cred = nfs4_get_renew_cred,
9606         .renew_lease = nfs4_proc_renew,
9607 };
9608
9609 #if defined(CONFIG_NFS_V4_1)
9610 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9611         .sched_state_renewal = nfs41_proc_async_sequence,
9612         .get_state_renewal_cred = nfs4_get_machine_cred,
9613         .renew_lease = nfs4_proc_sequence,
9614 };
9615 #endif
9616
9617 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9618         .get_locations = _nfs40_proc_get_locations,
9619         .fsid_present = _nfs40_proc_fsid_present,
9620 };
9621
9622 #if defined(CONFIG_NFS_V4_1)
9623 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9624         .get_locations = _nfs41_proc_get_locations,
9625         .fsid_present = _nfs41_proc_fsid_present,
9626 };
9627 #endif  /* CONFIG_NFS_V4_1 */
9628
9629 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9630         .minor_version = 0,
9631         .init_caps = NFS_CAP_READDIRPLUS
9632                 | NFS_CAP_ATOMIC_OPEN
9633                 | NFS_CAP_POSIX_LOCK,
9634         .init_client = nfs40_init_client,
9635         .shutdown_client = nfs40_shutdown_client,
9636         .match_stateid = nfs4_match_stateid,
9637         .find_root_sec = nfs4_find_root_sec,
9638         .free_lock_state = nfs4_release_lockowner,
9639         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9640         .alloc_seqid = nfs_alloc_seqid,
9641         .call_sync_ops = &nfs40_call_sync_ops,
9642         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9643         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9644         .state_renewal_ops = &nfs40_state_renewal_ops,
9645         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9646 };
9647
9648 #if defined(CONFIG_NFS_V4_1)
9649 static struct nfs_seqid *
9650 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9651 {
9652         return NULL;
9653 }
9654
9655 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9656         .minor_version = 1,
9657         .init_caps = NFS_CAP_READDIRPLUS
9658                 | NFS_CAP_ATOMIC_OPEN
9659                 | NFS_CAP_POSIX_LOCK
9660                 | NFS_CAP_STATEID_NFSV41
9661                 | NFS_CAP_ATOMIC_OPEN_V1
9662                 | NFS_CAP_LGOPEN,
9663         .init_client = nfs41_init_client,
9664         .shutdown_client = nfs41_shutdown_client,
9665         .match_stateid = nfs41_match_stateid,
9666         .find_root_sec = nfs41_find_root_sec,
9667         .free_lock_state = nfs41_free_lock_state,
9668         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9669         .alloc_seqid = nfs_alloc_no_seqid,
9670         .session_trunk = nfs4_test_session_trunk,
9671         .call_sync_ops = &nfs41_call_sync_ops,
9672         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9673         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9674         .state_renewal_ops = &nfs41_state_renewal_ops,
9675         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9676 };
9677 #endif
9678
9679 #if defined(CONFIG_NFS_V4_2)
9680 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9681         .minor_version = 2,
9682         .init_caps = NFS_CAP_READDIRPLUS
9683                 | NFS_CAP_ATOMIC_OPEN
9684                 | NFS_CAP_POSIX_LOCK
9685                 | NFS_CAP_STATEID_NFSV41
9686                 | NFS_CAP_ATOMIC_OPEN_V1
9687                 | NFS_CAP_LGOPEN
9688                 | NFS_CAP_ALLOCATE
9689                 | NFS_CAP_COPY
9690                 | NFS_CAP_OFFLOAD_CANCEL
9691                 | NFS_CAP_DEALLOCATE
9692                 | NFS_CAP_SEEK
9693                 | NFS_CAP_LAYOUTSTATS
9694                 | NFS_CAP_CLONE,
9695         .init_client = nfs41_init_client,
9696         .shutdown_client = nfs41_shutdown_client,
9697         .match_stateid = nfs41_match_stateid,
9698         .find_root_sec = nfs41_find_root_sec,
9699         .free_lock_state = nfs41_free_lock_state,
9700         .call_sync_ops = &nfs41_call_sync_ops,
9701         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9702         .alloc_seqid = nfs_alloc_no_seqid,
9703         .session_trunk = nfs4_test_session_trunk,
9704         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9705         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9706         .state_renewal_ops = &nfs41_state_renewal_ops,
9707         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9708 };
9709 #endif
9710
9711 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9712         [0] = &nfs_v4_0_minor_ops,
9713 #if defined(CONFIG_NFS_V4_1)
9714         [1] = &nfs_v4_1_minor_ops,
9715 #endif
9716 #if defined(CONFIG_NFS_V4_2)
9717         [2] = &nfs_v4_2_minor_ops,
9718 #endif
9719 };
9720
9721 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9722 {
9723         ssize_t error, error2;
9724
9725         error = generic_listxattr(dentry, list, size);
9726         if (error < 0)
9727                 return error;
9728         if (list) {
9729                 list += error;
9730                 size -= error;
9731         }
9732
9733         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9734         if (error2 < 0)
9735                 return error2;
9736         return error + error2;
9737 }
9738
9739 static const struct inode_operations nfs4_dir_inode_operations = {
9740         .create         = nfs_create,
9741         .lookup         = nfs_lookup,
9742         .atomic_open    = nfs_atomic_open,
9743         .link           = nfs_link,
9744         .unlink         = nfs_unlink,
9745         .symlink        = nfs_symlink,
9746         .mkdir          = nfs_mkdir,
9747         .rmdir          = nfs_rmdir,
9748         .mknod          = nfs_mknod,
9749         .rename         = nfs_rename,
9750         .permission     = nfs_permission,
9751         .getattr        = nfs_getattr,
9752         .setattr        = nfs_setattr,
9753         .listxattr      = nfs4_listxattr,
9754 };
9755
9756 static const struct inode_operations nfs4_file_inode_operations = {
9757         .permission     = nfs_permission,
9758         .getattr        = nfs_getattr,
9759         .setattr        = nfs_setattr,
9760         .listxattr      = nfs4_listxattr,
9761 };
9762
9763 const struct nfs_rpc_ops nfs_v4_clientops = {
9764         .version        = 4,                    /* protocol version */
9765         .dentry_ops     = &nfs4_dentry_operations,
9766         .dir_inode_ops  = &nfs4_dir_inode_operations,
9767         .file_inode_ops = &nfs4_file_inode_operations,
9768         .file_ops       = &nfs4_file_operations,
9769         .getroot        = nfs4_proc_get_root,
9770         .submount       = nfs4_submount,
9771         .try_mount      = nfs4_try_mount,
9772         .getattr        = nfs4_proc_getattr,
9773         .setattr        = nfs4_proc_setattr,
9774         .lookup         = nfs4_proc_lookup,
9775         .lookupp        = nfs4_proc_lookupp,
9776         .access         = nfs4_proc_access,
9777         .readlink       = nfs4_proc_readlink,
9778         .create         = nfs4_proc_create,
9779         .remove         = nfs4_proc_remove,
9780         .unlink_setup   = nfs4_proc_unlink_setup,
9781         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9782         .unlink_done    = nfs4_proc_unlink_done,
9783         .rename_setup   = nfs4_proc_rename_setup,
9784         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9785         .rename_done    = nfs4_proc_rename_done,
9786         .link           = nfs4_proc_link,
9787         .symlink        = nfs4_proc_symlink,
9788         .mkdir          = nfs4_proc_mkdir,
9789         .rmdir          = nfs4_proc_rmdir,
9790         .readdir        = nfs4_proc_readdir,
9791         .mknod          = nfs4_proc_mknod,
9792         .statfs         = nfs4_proc_statfs,
9793         .fsinfo         = nfs4_proc_fsinfo,
9794         .pathconf       = nfs4_proc_pathconf,
9795         .set_capabilities = nfs4_server_capabilities,
9796         .decode_dirent  = nfs4_decode_dirent,
9797         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9798         .read_setup     = nfs4_proc_read_setup,
9799         .read_done      = nfs4_read_done,
9800         .write_setup    = nfs4_proc_write_setup,
9801         .write_done     = nfs4_write_done,
9802         .commit_setup   = nfs4_proc_commit_setup,
9803         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9804         .commit_done    = nfs4_commit_done,
9805         .lock           = nfs4_proc_lock,
9806         .clear_acl_cache = nfs4_zap_acl_attr,
9807         .close_context  = nfs4_close_context,
9808         .open_context   = nfs4_atomic_open,
9809         .have_delegation = nfs4_have_delegation,
9810         .alloc_client   = nfs4_alloc_client,
9811         .init_client    = nfs4_init_client,
9812         .free_client    = nfs4_free_client,
9813         .create_server  = nfs4_create_server,
9814         .clone_server   = nfs_clone_server,
9815 };
9816
9817 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9818         .name   = XATTR_NAME_NFSV4_ACL,
9819         .list   = nfs4_xattr_list_nfs4_acl,
9820         .get    = nfs4_xattr_get_nfs4_acl,
9821         .set    = nfs4_xattr_set_nfs4_acl,
9822 };
9823
9824 const struct xattr_handler *nfs4_xattr_handlers[] = {
9825         &nfs4_xattr_nfs4_acl_handler,
9826 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9827         &nfs4_xattr_nfs4_label_handler,
9828 #endif
9829         NULL
9830 };
9831
9832 /*
9833  * Local variables:
9834  *  c-basic-offset: 8
9835  * End:
9836  */