fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57 #include <linux/iversion.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_BITMASK_SZ         3
  75
  76 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  77 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  78
  79 /* file attributes which can be mapped to nfs attributes */
  80 #define NFS4_VALID_ATTRS (ATTR_MODE \
  81         | ATTR_UID \
  82         | ATTR_GID \
  83         | ATTR_SIZE \
  84         | ATTR_ATIME \
  85         | ATTR_MTIME \
  86         | ATTR_CTIME \
  87         | ATTR_ATIME_SET \
  88         | ATTR_MTIME_SET)
  89
  90 struct nfs4_opendata;
  91 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  92 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  93 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  94 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label, struct inode *inode);
  95 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  96 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  97                             struct nfs_fattr *fattr, struct iattr *sattr,
  98                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  99                             struct nfs4_label *olabel);
 100 #ifdef CONFIG_NFS_V4_1
 101 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 102                 struct rpc_cred *cred,
 103                 struct nfs4_slot *slot,
 104                 bool is_privileged);
 105 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 106                 struct rpc_cred *);
 107 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 108                 struct rpc_cred *, bool);
 109 #endif
 110
 111 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 112 static inline struct nfs4_label *
 113 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 114         struct iattr *sattr, struct nfs4_label *label)
 115 {
 116         int err;
 117
 118         if (label == NULL)
 119                 return NULL;
 120
 121         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 122                 return NULL;
 123
 124         err = security_dentry_init_security(dentry, sattr->ia_mode,
 125                                 &dentry->d_name, (void **)&label->label, &label->len);
 126         if (err == 0)
 127                 return label;
 128
 129         return NULL;
 130 }
 131 static inline void
 132 nfs4_label_release_security(struct nfs4_label *label)
 133 {
 134         if (label)
 135                 security_release_secctx(label->label, label->len);
 136 }
 137 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 138 {
 139         if (label)
 140                 return server->attr_bitmask;
 141
 142         return server->attr_bitmask_nl;
 143 }
 144 #else
 145 static inline struct nfs4_label *
 146 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 147         struct iattr *sattr, struct nfs4_label *l)
 148 { return NULL; }
 149 static inline void
 150 nfs4_label_release_security(struct nfs4_label *label)
 151 { return; }
 152 static inline u32 *
 153 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 154 { return server->attr_bitmask; }
 155 #endif
 156
 157 /* Prevent leaks of NFSv4 errors into userland */
 158 static int nfs4_map_errors(int err)
 159 {
 160         if (err >= -1000)
 161                 return err;
 162         switch (err) {
 163         case -NFS4ERR_RESOURCE:
 164         case -NFS4ERR_LAYOUTTRYLATER:
 165         case -NFS4ERR_RECALLCONFLICT:
 166                 return -EREMOTEIO;
 167         case -NFS4ERR_WRONGSEC:
 168         case -NFS4ERR_WRONG_CRED:
 169                 return -EPERM;
 170         case -NFS4ERR_BADOWNER:
 171         case -NFS4ERR_BADNAME:
 172                 return -EINVAL;
 173         case -NFS4ERR_SHARE_DENIED:
 174                 return -EACCES;
 175         case -NFS4ERR_MINOR_VERS_MISMATCH:
 176                 return -EPROTONOSUPPORT;
 177         case -NFS4ERR_FILE_OPEN:
 178                 return -EBUSY;
 179         default:
 180                 dprintk("%s could not handle NFSv4 error %d\n",
 181                                 __func__, -err);
 182                 break;
 183         }
 184         return -EIO;
 185 }
 186
 187 /*
 188  * This is our standard bitmap for GETATTR requests.
 189  */
 190 const u32 nfs4_fattr_bitmap[3] = {
 191         FATTR4_WORD0_TYPE
 192         | FATTR4_WORD0_CHANGE
 193         | FATTR4_WORD0_SIZE
 194         | FATTR4_WORD0_FSID
 195         | FATTR4_WORD0_FILEID,
 196         FATTR4_WORD1_MODE
 197         | FATTR4_WORD1_NUMLINKS
 198         | FATTR4_WORD1_OWNER
 199         | FATTR4_WORD1_OWNER_GROUP
 200         | FATTR4_WORD1_RAWDEV
 201         | FATTR4_WORD1_SPACE_USED
 202         | FATTR4_WORD1_TIME_ACCESS
 203         | FATTR4_WORD1_TIME_METADATA
 204         | FATTR4_WORD1_TIME_MODIFY
 205         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 206 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 207         FATTR4_WORD2_SECURITY_LABEL
 208 #endif
 209 };
 210
 211 static const u32 nfs4_pnfs_open_bitmap[3] = {
 212         FATTR4_WORD0_TYPE
 213         | FATTR4_WORD0_CHANGE
 214         | FATTR4_WORD0_SIZE
 215         | FATTR4_WORD0_FSID
 216         | FATTR4_WORD0_FILEID,
 217         FATTR4_WORD1_MODE
 218         | FATTR4_WORD1_NUMLINKS
 219         | FATTR4_WORD1_OWNER
 220         | FATTR4_WORD1_OWNER_GROUP
 221         | FATTR4_WORD1_RAWDEV
 222         | FATTR4_WORD1_SPACE_USED
 223         | FATTR4_WORD1_TIME_ACCESS
 224         | FATTR4_WORD1_TIME_METADATA
 225         | FATTR4_WORD1_TIME_MODIFY,
 226         FATTR4_WORD2_MDSTHRESHOLD
 227 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 228         | FATTR4_WORD2_SECURITY_LABEL
 229 #endif
 230 };
 231
 232 static const u32 nfs4_open_noattr_bitmap[3] = {
 233         FATTR4_WORD0_TYPE
 234         | FATTR4_WORD0_FILEID,
 235 };
 236
 237 const u32 nfs4_statfs_bitmap[3] = {
 238         FATTR4_WORD0_FILES_AVAIL
 239         | FATTR4_WORD0_FILES_FREE
 240         | FATTR4_WORD0_FILES_TOTAL,
 241         FATTR4_WORD1_SPACE_AVAIL
 242         | FATTR4_WORD1_SPACE_FREE
 243         | FATTR4_WORD1_SPACE_TOTAL
 244 };
 245
 246 const u32 nfs4_pathconf_bitmap[3] = {
 247         FATTR4_WORD0_MAXLINK
 248         | FATTR4_WORD0_MAXNAME,
 249         0
 250 };
 251
 252 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 253                         | FATTR4_WORD0_MAXREAD
 254                         | FATTR4_WORD0_MAXWRITE
 255                         | FATTR4_WORD0_LEASE_TIME,
 256                         FATTR4_WORD1_TIME_DELTA
 257                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 258                         FATTR4_WORD2_LAYOUT_BLKSIZE
 259                         | FATTR4_WORD2_CLONE_BLKSIZE
 260 };
 261
 262 const u32 nfs4_fs_locations_bitmap[3] = {
 263         FATTR4_WORD0_CHANGE
 264         | FATTR4_WORD0_SIZE
 265         | FATTR4_WORD0_FSID
 266         | FATTR4_WORD0_FILEID
 267         | FATTR4_WORD0_FS_LOCATIONS,
 268         FATTR4_WORD1_OWNER
 269         | FATTR4_WORD1_OWNER_GROUP
 270         | FATTR4_WORD1_RAWDEV
 271         | FATTR4_WORD1_SPACE_USED
 272         | FATTR4_WORD1_TIME_ACCESS
 273         | FATTR4_WORD1_TIME_METADATA
 274         | FATTR4_WORD1_TIME_MODIFY
 275         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 276 };
 277
 278 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 279                 struct inode *inode)
 280 {
 281         unsigned long cache_validity;
 282
 283         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 284         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 285                 return;
 286
 287         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 288         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 289                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 290                                 | NFS_INO_INVALID_SIZE);
 291
 292         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 293                 dst[0] &= ~FATTR4_WORD0_SIZE;
 294
 295         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 296                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 297 }
 298
 299 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 300                 const __u32 *src, struct inode *inode)
 301 {
 302         nfs4_bitmap_copy_adjust(dst, src, inode);
 303 }
 304
 305 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 306                 struct nfs4_readdir_arg *readdir)
 307 {
 308         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 309         __be32 *start, *p;
 310
 311         if (cookie > 2) {
 312                 readdir->cookie = cookie;
 313                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 314                 return;
 315         }
 316
 317         readdir->cookie = 0;
 318         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 319         if (cookie == 2)
 320                 return;
 321
 322         /*
 323          * NFSv4 servers do not return entries for '.' and '..'
 324          * Therefore, we fake these entries here.  We let '.'
 325          * have cookie 0 and '..' have cookie 1.  Note that
 326          * when talking to the server, we always send cookie 0
 327          * instead of 1 or 2.
 328          */
 329         start = p = kmap_atomic(*readdir->pages);
 330
 331         if (cookie == 0) {
 332                 *p++ = xdr_one;                                  /* next */
 333                 *p++ = xdr_zero;                   /* cookie, first word */
 334                 *p++ = xdr_one;                   /* cookie, second word */
 335                 *p++ = xdr_one;                             /* entry len */
 336                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 337                 p++;
 338                 *p++ = xdr_one;                         /* bitmap length */
 339                 *p++ = htonl(attrs);                           /* bitmap */
 340                 *p++ = htonl(12);             /* attribute buffer length */
 341                 *p++ = htonl(NF4DIR);
 342                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 343         }
 344
 345         *p++ = xdr_one;                                  /* next */
 346         *p++ = xdr_zero;                   /* cookie, first word */
 347         *p++ = xdr_two;                   /* cookie, second word */
 348         *p++ = xdr_two;                             /* entry len */
 349         memcpy(p, "..\0\0", 4);                         /* entry */
 350         p++;
 351         *p++ = xdr_one;                         /* bitmap length */
 352         *p++ = htonl(attrs);                           /* bitmap */
 353         *p++ = htonl(12);             /* attribute buffer length */
 354         *p++ = htonl(NF4DIR);
 355         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 356
 357         readdir->pgbase = (char *)p - (char *)start;
 358         readdir->count -= readdir->pgbase;
 359         kunmap_atomic(start);
 360 }
 361
 362 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 363                 nfs4_stateid *stateid,
 364                 struct rpc_cred *cred)
 365 {
 366         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 367
 368         ops->test_and_free_expired(server, stateid, cred);
 369 }
 370
 371 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 372                 nfs4_stateid *stateid,
 373                 struct rpc_cred *cred)
 374 {
 375         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 376         nfs4_test_and_free_stateid(server, stateid, cred);
 377 }
 378
 379 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 380                 const nfs4_stateid *stateid,
 381                 struct rpc_cred *cred)
 382 {
 383         nfs4_stateid tmp;
 384
 385         nfs4_stateid_copy(&tmp, stateid);
 386         __nfs4_free_revoked_stateid(server, &tmp, cred);
 387 }
 388
 389 static long nfs4_update_delay(long *timeout)
 390 {
 391         long ret;
 392         if (!timeout)
 393                 return NFS4_POLL_RETRY_MAX;
 394         if (*timeout <= 0)
 395                 *timeout = NFS4_POLL_RETRY_MIN;
 396         if (*timeout > NFS4_POLL_RETRY_MAX)
 397                 *timeout = NFS4_POLL_RETRY_MAX;
 398         ret = *timeout;
 399         *timeout <<= 1;
 400         return ret;
 401 }
 402
 403 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 404 {
 405         int res = 0;
 406
 407         might_sleep();
 408
 409         freezable_schedule_timeout_killable_unsafe(
 410                 nfs4_update_delay(timeout));
 411         if (fatal_signal_pending(current))
 412                 res = -ERESTARTSYS;
 413         return res;
 414 }
 415
 416 /* This is the error handling routine for processes that are allowed
 417  * to sleep.
 418  */
 419 static int nfs4_do_handle_exception(struct nfs_server *server,
 420                 int errorcode, struct nfs4_exception *exception)
 421 {
 422         struct nfs_client *clp = server->nfs_client;
 423         struct nfs4_state *state = exception->state;
 424         const nfs4_stateid *stateid = exception->stateid;
 425         struct inode *inode = exception->inode;
 426         int ret = errorcode;
 427
 428         exception->delay = 0;
 429         exception->recovering = 0;
 430         exception->retry = 0;
 431
 432         if (stateid == NULL && state != NULL)
 433                 stateid = &state->stateid;
 434
 435         switch(errorcode) {
 436                 case 0:
 437                         return 0;
 438                 case -NFS4ERR_BADHANDLE:
 439                 case -ESTALE:
 440                         if (inode != NULL && S_ISREG(inode->i_mode))
 441                                 pnfs_destroy_layout(NFS_I(inode));
 442                         break;
 443                 case -NFS4ERR_DELEG_REVOKED:
 444                 case -NFS4ERR_ADMIN_REVOKED:
 445                 case -NFS4ERR_EXPIRED:
 446                 case -NFS4ERR_BAD_STATEID:
 447                         if (inode != NULL && stateid != NULL) {
 448                                 nfs_inode_find_state_and_recover(inode,
 449                                                 stateid);
 450                                 goto wait_on_recovery;
 451                         }
 452                         /* Fall through */
 453                 case -NFS4ERR_OPENMODE:
 454                         if (inode) {
 455                                 int err;
 456
 457                                 err = nfs_async_inode_return_delegation(inode,
 458                                                 stateid);
 459                                 if (err == 0)
 460                                         goto wait_on_recovery;
 461                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 462                                         exception->retry = 1;
 463                                         break;
 464                                 }
 465                         }
 466                         if (state == NULL)
 467                                 break;
 468                         ret = nfs4_schedule_stateid_recovery(server, state);
 469                         if (ret < 0)
 470                                 break;
 471                         goto wait_on_recovery;
 472                 case -NFS4ERR_STALE_STATEID:
 473                 case -NFS4ERR_STALE_CLIENTID:
 474                         nfs4_schedule_lease_recovery(clp);
 475                         goto wait_on_recovery;
 476                 case -NFS4ERR_MOVED:
 477                         ret = nfs4_schedule_migration_recovery(server);
 478                         if (ret < 0)
 479                                 break;
 480                         goto wait_on_recovery;
 481                 case -NFS4ERR_LEASE_MOVED:
 482                         nfs4_schedule_lease_moved_recovery(clp);
 483                         goto wait_on_recovery;
 484 #if defined(CONFIG_NFS_V4_1)
 485                 case -NFS4ERR_BADSESSION:
 486                 case -NFS4ERR_BADSLOT:
 487                 case -NFS4ERR_BAD_HIGH_SLOT:
 488                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 489                 case -NFS4ERR_DEADSESSION:
 490                 case -NFS4ERR_SEQ_FALSE_RETRY:
 491                 case -NFS4ERR_SEQ_MISORDERED:
 492                         dprintk("%s ERROR: %d Reset session\n", __func__,
 493                                 errorcode);
 494                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 495                         goto wait_on_recovery;
 496 #endif /* defined(CONFIG_NFS_V4_1) */
 497                 case -NFS4ERR_FILE_OPEN:
 498                         if (exception->timeout > HZ) {
 499                                 /* We have retried a decent amount, time to
 500                                  * fail
 501                                  */
 502                                 ret = -EBUSY;
 503                                 break;
 504                         }
 505                         /* Fall through */
 506                 case -NFS4ERR_DELAY:
 507                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 508                         /* Fall through */
 509                 case -NFS4ERR_GRACE:
 510                 case -NFS4ERR_LAYOUTTRYLATER:
 511                 case -NFS4ERR_RECALLCONFLICT:
 512                         exception->delay = 1;
 513                         return 0;
 514
 515                 case -NFS4ERR_RETRY_UNCACHED_REP:
 516                 case -NFS4ERR_OLD_STATEID:
 517                         exception->retry = 1;
 518                         break;
 519                 case -NFS4ERR_BADOWNER:
 520                         /* The following works around a Linux server bug! */
 521                 case -NFS4ERR_BADNAME:
 522                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 523                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 524                                 exception->retry = 1;
 525                                 printk(KERN_WARNING "NFS: v4 server %s "
 526                                                 "does not accept raw "
 527                                                 "uid/gids. "
 528                                                 "Reenabling the idmapper.\n",
 529                                                 server->nfs_client->cl_hostname);
 530                         }
 531         }
 532         /* We failed to handle the error */
 533         return nfs4_map_errors(ret);
 534 wait_on_recovery:
 535         exception->recovering = 1;
 536         return 0;
 537 }
 538
 539 /* This is the error handling routine for processes that are allowed
 540  * to sleep.
 541  */
 542 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 543 {
 544         struct nfs_client *clp = server->nfs_client;
 545         int ret;
 546
 547         ret = nfs4_do_handle_exception(server, errorcode, exception);
 548         if (exception->delay) {
 549                 ret = nfs4_delay(server->client, &exception->timeout);
 550                 goto out_retry;
 551         }
 552         if (exception->recovering) {
 553                 ret = nfs4_wait_clnt_recover(clp);
 554                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 555                         return -EIO;
 556                 goto out_retry;
 557         }
 558         return ret;
 559 out_retry:
 560         if (ret == 0)
 561                 exception->retry = 1;
 562         return ret;
 563 }
 564
 565 static int
 566 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 567                 int errorcode, struct nfs4_exception *exception)
 568 {
 569         struct nfs_client *clp = server->nfs_client;
 570         int ret;
 571
 572         ret = nfs4_do_handle_exception(server, errorcode, exception);
 573         if (exception->delay) {
 574                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 575                 goto out_retry;
 576         }
 577         if (exception->recovering) {
 578                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 579                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 580                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 581                 goto out_retry;
 582         }
 583         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 584                 ret = -EIO;
 585         return ret;
 586 out_retry:
 587         if (ret == 0) {
 588                 exception->retry = 1;
 589                 /*
 590                  * For NFS4ERR_MOVED, the client transport will need to
 591                  * be recomputed after migration recovery has completed.
 592                  */
 593                 if (errorcode == -NFS4ERR_MOVED)
 594                         rpc_task_release_transport(task);
 595         }
 596         return ret;
 597 }
 598
 599 int
 600 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 601                         struct nfs4_state *state, long *timeout)
 602 {
 603         struct nfs4_exception exception = {
 604                 .state = state,
 605         };
 606
 607         if (task->tk_status >= 0)
 608                 return 0;
 609         if (timeout)
 610                 exception.timeout = *timeout;
 611         task->tk_status = nfs4_async_handle_exception(task, server,
 612                         task->tk_status,
 613                         &exception);
 614         if (exception.delay && timeout)
 615                 *timeout = exception.timeout;
 616         if (exception.retry)
 617                 return -EAGAIN;
 618         return 0;
 619 }
 620
 621 /*
 622  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 623  * or 'false' otherwise.
 624  */
 625 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 626 {
 627         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 628         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 629 }
 630
 631 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 632 {
 633         spin_lock(&clp->cl_lock);
 634         if (time_before(clp->cl_last_renewal,timestamp))
 635                 clp->cl_last_renewal = timestamp;
 636         spin_unlock(&clp->cl_lock);
 637 }
 638
 639 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 640 {
 641         struct nfs_client *clp = server->nfs_client;
 642
 643         if (!nfs4_has_session(clp))
 644                 do_renew_lease(clp, timestamp);
 645 }
 646
 647 struct nfs4_call_sync_data {
 648         const struct nfs_server *seq_server;
 649         struct nfs4_sequence_args *seq_args;
 650         struct nfs4_sequence_res *seq_res;
 651 };
 652
 653 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 654                         struct nfs4_sequence_res *res, int cache_reply,
 655                         int privileged)
 656 {
 657         args->sa_slot = NULL;
 658         args->sa_cache_this = cache_reply;
 659         args->sa_privileged = privileged;
 660
 661         res->sr_slot = NULL;
 662 }
 663
 664 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 665 {
 666         struct nfs4_slot *slot = res->sr_slot;
 667         struct nfs4_slot_table *tbl;
 668
 669         tbl = slot->table;
 670         spin_lock(&tbl->slot_tbl_lock);
 671         if (!nfs41_wake_and_assign_slot(tbl, slot))
 672                 nfs4_free_slot(tbl, slot);
 673         spin_unlock(&tbl->slot_tbl_lock);
 674
 675         res->sr_slot = NULL;
 676 }
 677
 678 static int nfs40_sequence_done(struct rpc_task *task,
 679                                struct nfs4_sequence_res *res)
 680 {
 681         if (res->sr_slot != NULL)
 682                 nfs40_sequence_free_slot(res);
 683         return 1;
 684 }
 685
 686 #if defined(CONFIG_NFS_V4_1)
 687
 688 static void nfs41_release_slot(struct nfs4_slot *slot)
 689 {
 690         struct nfs4_session *session;
 691         struct nfs4_slot_table *tbl;
 692         bool send_new_highest_used_slotid = false;
 693
 694         if (!slot)
 695                 return;
 696         tbl = slot->table;
 697         session = tbl->session;
 698
 699         /* Bump the slot sequence number */
 700         if (slot->seq_done)
 701                 slot->seq_nr++;
 702         slot->seq_done = 0;
 703
 704         spin_lock(&tbl->slot_tbl_lock);
 705         /* Be nice to the server: try to ensure that the last transmitted
 706          * value for highest_user_slotid <= target_highest_slotid
 707          */
 708         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 709                 send_new_highest_used_slotid = true;
 710
 711         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 712                 send_new_highest_used_slotid = false;
 713                 goto out_unlock;
 714         }
 715         nfs4_free_slot(tbl, slot);
 716
 717         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 718                 send_new_highest_used_slotid = false;
 719 out_unlock:
 720         spin_unlock(&tbl->slot_tbl_lock);
 721         if (send_new_highest_used_slotid)
 722                 nfs41_notify_server(session->clp);
 723         if (waitqueue_active(&tbl->slot_waitq))
 724                 wake_up_all(&tbl->slot_waitq);
 725 }
 726
 727 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 728 {
 729         nfs41_release_slot(res->sr_slot);
 730         res->sr_slot = NULL;
 731 }
 732
 733 static int nfs41_sequence_process(struct rpc_task *task,
 734                 struct nfs4_sequence_res *res)
 735 {
 736         struct nfs4_session *session;
 737         struct nfs4_slot *slot = res->sr_slot;
 738         struct nfs_client *clp;
 739         bool interrupted = false;
 740         int ret = 1;
 741
 742         if (slot == NULL)
 743                 goto out_noaction;
 744         /* don't increment the sequence number if the task wasn't sent */
 745         if (!RPC_WAS_SENT(task))
 746                 goto out;
 747
 748         session = slot->table->session;
 749
 750         if (slot->interrupted) {
 751                 if (res->sr_status != -NFS4ERR_DELAY)
 752                         slot->interrupted = 0;
 753                 interrupted = true;
 754         }
 755
 756         trace_nfs4_sequence_done(session, res);
 757         /* Check the SEQUENCE operation status */
 758         switch (res->sr_status) {
 759         case 0:
 760                 /* Update the slot's sequence and clientid lease timer */
 761                 slot->seq_done = 1;
 762                 clp = session->clp;
 763                 do_renew_lease(clp, res->sr_timestamp);
 764                 /* Check sequence flags */
 765                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 766                                 !!slot->privileged);
 767                 nfs41_update_target_slotid(slot->table, slot, res);
 768                 break;
 769         case 1:
 770                 /*
 771                  * sr_status remains 1 if an RPC level error occurred.
 772                  * The server may or may not have processed the sequence
 773                  * operation..
 774                  * Mark the slot as having hosted an interrupted RPC call.
 775                  */
 776                 slot->interrupted = 1;
 777                 goto out;
 778         case -NFS4ERR_DELAY:
 779                 /* The server detected a resend of the RPC call and
 780                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 781                  * of RFC5661.
 782                  */
 783                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 784                         __func__,
 785                         slot->slot_nr,
 786                         slot->seq_nr);
 787                 goto out_retry;
 788         case -NFS4ERR_RETRY_UNCACHED_REP:
 789         case -NFS4ERR_SEQ_FALSE_RETRY:
 790                 /*
 791                  * The server thinks we tried to replay a request.
 792                  * Retry the call after bumping the sequence ID.
 793                  */
 794                 goto retry_new_seq;
 795         case -NFS4ERR_BADSLOT:
 796                 /*
 797                  * The slot id we used was probably retired. Try again
 798                  * using a different slot id.
 799                  */
 800                 if (slot->slot_nr < slot->table->target_highest_slotid)
 801                         goto session_recover;
 802                 goto retry_nowait;
 803         case -NFS4ERR_SEQ_MISORDERED:
 804                 /*
 805                  * Was the last operation on this sequence interrupted?
 806                  * If so, retry after bumping the sequence number.
 807                  */
 808                 if (interrupted)
 809                         goto retry_new_seq;
 810                 /*
 811                  * Could this slot have been previously retired?
 812                  * If so, then the server may be expecting seq_nr = 1!
 813                  */
 814                 if (slot->seq_nr != 1) {
 815                         slot->seq_nr = 1;
 816                         goto retry_nowait;
 817                 }
 818                 goto session_recover;
 819         default:
 820                 /* Just update the slot sequence no. */
 821                 slot->seq_done = 1;
 822         }
 823 out:
 824         /* The session may be reset by one of the error handlers. */
 825         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 826 out_noaction:
 827         return ret;
 828 session_recover:
 829         nfs4_schedule_session_recovery(session, res->sr_status);
 830         goto retry_nowait;
 831 retry_new_seq:
 832         ++slot->seq_nr;
 833 retry_nowait:
 834         if (rpc_restart_call_prepare(task)) {
 835                 nfs41_sequence_free_slot(res);
 836                 task->tk_status = 0;
 837                 ret = 0;
 838         }
 839         goto out;
 840 out_retry:
 841         if (!rpc_restart_call(task))
 842                 goto out;
 843         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 844         return 0;
 845 }
 846
 847 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 848 {
 849         if (!nfs41_sequence_process(task, res))
 850                 return 0;
 851         if (res->sr_slot != NULL)
 852                 nfs41_sequence_free_slot(res);
 853         return 1;
 854
 855 }
 856 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 857
 858 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 859 {
 860         if (res->sr_slot == NULL)
 861                 return 1;
 862         if (res->sr_slot->table->session != NULL)
 863                 return nfs41_sequence_process(task, res);
 864         return nfs40_sequence_done(task, res);
 865 }
 866
 867 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 868 {
 869         if (res->sr_slot != NULL) {
 870                 if (res->sr_slot->table->session != NULL)
 871                         nfs41_sequence_free_slot(res);
 872                 else
 873                         nfs40_sequence_free_slot(res);
 874         }
 875 }
 876
 877 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 878 {
 879         if (res->sr_slot == NULL)
 880                 return 1;
 881         if (!res->sr_slot->table->session)
 882                 return nfs40_sequence_done(task, res);
 883         return nfs41_sequence_done(task, res);
 884 }
 885 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 886
 887 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 888 {
 889         struct nfs4_call_sync_data *data = calldata;
 890
 891         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 892
 893         nfs4_setup_sequence(data->seq_server->nfs_client,
 894                             data->seq_args, data->seq_res, task);
 895 }
 896
 897 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 898 {
 899         struct nfs4_call_sync_data *data = calldata;
 900
 901         nfs41_sequence_done(task, data->seq_res);
 902 }
 903
 904 static const struct rpc_call_ops nfs41_call_sync_ops = {
 905         .rpc_call_prepare = nfs41_call_sync_prepare,
 906         .rpc_call_done = nfs41_call_sync_done,
 907 };
 908
 909 static void
 910 nfs4_sequence_process_interrupted(struct nfs_client *client,
 911                 struct nfs4_slot *slot, struct rpc_cred *cred)
 912 {
 913         struct rpc_task *task;
 914
 915         task = _nfs41_proc_sequence(client, cred, slot, true);
 916         if (!IS_ERR(task))
 917                 rpc_put_task_async(task);
 918 }
 919
 920 #else   /* !CONFIG_NFS_V4_1 */
 921
 922 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 923 {
 924         return nfs40_sequence_done(task, res);
 925 }
 926
 927 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 928 {
 929         if (res->sr_slot != NULL)
 930                 nfs40_sequence_free_slot(res);
 931 }
 932
 933 int nfs4_sequence_done(struct rpc_task *task,
 934                        struct nfs4_sequence_res *res)
 935 {
 936         return nfs40_sequence_done(task, res);
 937 }
 938 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 939
 940 static void
 941 nfs4_sequence_process_interrupted(struct nfs_client *client,
 942                 struct nfs4_slot *slot, struct rpc_cred *cred)
 943 {
 944         WARN_ON_ONCE(1);
 945         slot->interrupted = 0;
 946 }
 947
 948 #endif  /* !CONFIG_NFS_V4_1 */
 949
 950 static
 951 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
 952                 struct nfs4_sequence_res *res,
 953                 struct nfs4_slot *slot)
 954 {
 955         if (!slot)
 956                 return;
 957         slot->privileged = args->sa_privileged ? 1 : 0;
 958         args->sa_slot = slot;
 959
 960         res->sr_slot = slot;
 961         res->sr_timestamp = jiffies;
 962         res->sr_status_flags = 0;
 963         res->sr_status = 1;
 964
 965 }
 966
 967 int nfs4_setup_sequence(struct nfs_client *client,
 968                         struct nfs4_sequence_args *args,
 969                         struct nfs4_sequence_res *res,
 970                         struct rpc_task *task)
 971 {
 972         struct nfs4_session *session = nfs4_get_session(client);
 973         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
 974         struct nfs4_slot *slot;
 975
 976         /* slot already allocated? */
 977         if (res->sr_slot != NULL)
 978                 goto out_start;
 979
 980         if (session) {
 981                 tbl = &session->fc_slot_table;
 982                 task->tk_timeout = 0;
 983         }
 984
 985         for (;;) {
 986                 spin_lock(&tbl->slot_tbl_lock);
 987                 /* The state manager will wait until the slot table is empty */
 988                 if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 989                         goto out_sleep;
 990
 991                 slot = nfs4_alloc_slot(tbl);
 992                 if (IS_ERR(slot)) {
 993                         /* Try again in 1/4 second */
 994                         if (slot == ERR_PTR(-ENOMEM))
 995                                 task->tk_timeout = HZ >> 2;
 996                         goto out_sleep;
 997                 }
 998                 spin_unlock(&tbl->slot_tbl_lock);
 999
1000                 if (likely(!slot->interrupted))
1001                         break;
1002                 nfs4_sequence_process_interrupted(client,
1003                                 slot, task->tk_msg.rpc_cred);
1004         }
1005
1006         nfs4_sequence_attach_slot(args, res, slot);
1007
1008         trace_nfs4_setup_sequence(session, args);
1009 out_start:
1010         rpc_call_start(task);
1011         return 0;
1012
1013 out_sleep:
1014         if (args->sa_privileged)
1015                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1016                                 NULL, RPC_PRIORITY_PRIVILEGED);
1017         else
1018                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1019         spin_unlock(&tbl->slot_tbl_lock);
1020         return -EAGAIN;
1021 }
1022 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1023
1024 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1025 {
1026         struct nfs4_call_sync_data *data = calldata;
1027         nfs4_setup_sequence(data->seq_server->nfs_client,
1028                                 data->seq_args, data->seq_res, task);
1029 }
1030
1031 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1032 {
1033         struct nfs4_call_sync_data *data = calldata;
1034         nfs4_sequence_done(task, data->seq_res);
1035 }
1036
1037 static const struct rpc_call_ops nfs40_call_sync_ops = {
1038         .rpc_call_prepare = nfs40_call_sync_prepare,
1039         .rpc_call_done = nfs40_call_sync_done,
1040 };
1041
1042 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1043                                    struct nfs_server *server,
1044                                    struct rpc_message *msg,
1045                                    struct nfs4_sequence_args *args,
1046                                    struct nfs4_sequence_res *res)
1047 {
1048         int ret;
1049         struct rpc_task *task;
1050         struct nfs_client *clp = server->nfs_client;
1051         struct nfs4_call_sync_data data = {
1052                 .seq_server = server,
1053                 .seq_args = args,
1054                 .seq_res = res,
1055         };
1056         struct rpc_task_setup task_setup = {
1057                 .rpc_client = clnt,
1058                 .rpc_message = msg,
1059                 .callback_ops = clp->cl_mvops->call_sync_ops,
1060                 .callback_data = &data
1061         };
1062
1063         task = rpc_run_task(&task_setup);
1064         if (IS_ERR(task))
1065                 ret = PTR_ERR(task);
1066         else {
1067                 ret = task->tk_status;
1068                 rpc_put_task(task);
1069         }
1070         return ret;
1071 }
1072
1073 int nfs4_call_sync(struct rpc_clnt *clnt,
1074                    struct nfs_server *server,
1075                    struct rpc_message *msg,
1076                    struct nfs4_sequence_args *args,
1077                    struct nfs4_sequence_res *res,
1078                    int cache_reply)
1079 {
1080         nfs4_init_sequence(args, res, cache_reply, 0);
1081         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1082 }
1083
1084 static void
1085 nfs4_inc_nlink_locked(struct inode *inode)
1086 {
1087         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1088         inc_nlink(inode);
1089 }
1090
1091 static void
1092 nfs4_dec_nlink_locked(struct inode *inode)
1093 {
1094         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1095         drop_nlink(inode);
1096 }
1097
1098 static void
1099 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1100                 unsigned long timestamp, unsigned long cache_validity)
1101 {
1102         struct nfs_inode *nfsi = NFS_I(dir);
1103
1104         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1105                 | NFS_INO_INVALID_MTIME
1106                 | NFS_INO_INVALID_DATA
1107                 | cache_validity;
1108         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1109                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1110                 nfsi->attrtimeo_timestamp = jiffies;
1111         } else {
1112                 nfs_force_lookup_revalidate(dir);
1113                 if (cinfo->before != inode_peek_iversion_raw(dir))
1114                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1115                                 NFS_INO_INVALID_ACL;
1116         }
1117         inode_set_iversion_raw(dir, cinfo->after);
1118         nfsi->read_cache_jiffies = timestamp;
1119         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1120         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1121         nfs_fscache_invalidate(dir);
1122 }
1123
1124 static void
1125 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1126                 unsigned long timestamp, unsigned long cache_validity)
1127 {
1128         spin_lock(&dir->i_lock);
1129         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1130         spin_unlock(&dir->i_lock);
1131 }
1132
1133 struct nfs4_open_createattrs {
1134         struct nfs4_label *label;
1135         struct iattr *sattr;
1136         const __u32 verf[2];
1137 };
1138
1139 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1140                 int err, struct nfs4_exception *exception)
1141 {
1142         if (err != -EINVAL)
1143                 return false;
1144         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1145                 return false;
1146         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1147         exception->retry = 1;
1148         return true;
1149 }
1150
1151 static u32
1152 nfs4_map_atomic_open_share(struct nfs_server *server,
1153                 fmode_t fmode, int openflags)
1154 {
1155         u32 res = 0;
1156
1157         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1158         case FMODE_READ:
1159                 res = NFS4_SHARE_ACCESS_READ;
1160                 break;
1161         case FMODE_WRITE:
1162                 res = NFS4_SHARE_ACCESS_WRITE;
1163                 break;
1164         case FMODE_READ|FMODE_WRITE:
1165                 res = NFS4_SHARE_ACCESS_BOTH;
1166         }
1167         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1168                 goto out;
1169         /* Want no delegation if we're using O_DIRECT */
1170         if (openflags & O_DIRECT)
1171                 res |= NFS4_SHARE_WANT_NO_DELEG;
1172 out:
1173         return res;
1174 }
1175
1176 static enum open_claim_type4
1177 nfs4_map_atomic_open_claim(struct nfs_server *server,
1178                 enum open_claim_type4 claim)
1179 {
1180         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1181                 return claim;
1182         switch (claim) {
1183         default:
1184                 return claim;
1185         case NFS4_OPEN_CLAIM_FH:
1186                 return NFS4_OPEN_CLAIM_NULL;
1187         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1188                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1189         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1190                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1191         }
1192 }
1193
1194 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1195 {
1196         p->o_res.f_attr = &p->f_attr;
1197         p->o_res.f_label = p->f_label;
1198         p->o_res.seqid = p->o_arg.seqid;
1199         p->c_res.seqid = p->c_arg.seqid;
1200         p->o_res.server = p->o_arg.server;
1201         p->o_res.access_request = p->o_arg.access;
1202         nfs_fattr_init(&p->f_attr);
1203         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1204 }
1205
1206 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1207                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1208                 const struct nfs4_open_createattrs *c,
1209                 enum open_claim_type4 claim,
1210                 gfp_t gfp_mask)
1211 {
1212         struct dentry *parent = dget_parent(dentry);
1213         struct inode *dir = d_inode(parent);
1214         struct nfs_server *server = NFS_SERVER(dir);
1215         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1216         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1217         struct nfs4_opendata *p;
1218
1219         p = kzalloc(sizeof(*p), gfp_mask);
1220         if (p == NULL)
1221                 goto err;
1222
1223         p->f_label = nfs4_label_alloc(server, gfp_mask);
1224         if (IS_ERR(p->f_label))
1225                 goto err_free_p;
1226
1227         p->a_label = nfs4_label_alloc(server, gfp_mask);
1228         if (IS_ERR(p->a_label))
1229                 goto err_free_f;
1230
1231         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1232         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1233         if (IS_ERR(p->o_arg.seqid))
1234                 goto err_free_label;
1235         nfs_sb_active(dentry->d_sb);
1236         p->dentry = dget(dentry);
1237         p->dir = parent;
1238         p->owner = sp;
1239         atomic_inc(&sp->so_count);
1240         p->o_arg.open_flags = flags;
1241         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1242         p->o_arg.umask = current_umask();
1243         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1244         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1245                         fmode, flags);
1246         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1247          * will return permission denied for all bits until close */
1248         if (!(flags & O_EXCL)) {
1249                 /* ask server to check for all possible rights as results
1250                  * are cached */
1251                 switch (p->o_arg.claim) {
1252                 default:
1253                         break;
1254                 case NFS4_OPEN_CLAIM_NULL:
1255                 case NFS4_OPEN_CLAIM_FH:
1256                         p->o_arg.access = NFS4_ACCESS_READ |
1257                                 NFS4_ACCESS_MODIFY |
1258                                 NFS4_ACCESS_EXTEND |
1259                                 NFS4_ACCESS_EXECUTE;
1260                 }
1261         }
1262         p->o_arg.clientid = server->nfs_client->cl_clientid;
1263         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1264         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1265         p->o_arg.name = &dentry->d_name;
1266         p->o_arg.server = server;
1267         p->o_arg.bitmask = nfs4_bitmask(server, label);
1268         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1269         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1270         switch (p->o_arg.claim) {
1271         case NFS4_OPEN_CLAIM_NULL:
1272         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1273         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1274                 p->o_arg.fh = NFS_FH(dir);
1275                 break;
1276         case NFS4_OPEN_CLAIM_PREVIOUS:
1277         case NFS4_OPEN_CLAIM_FH:
1278         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1279         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1280                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1281         }
1282         if (c != NULL && c->sattr != NULL && c->sattr->ia_valid != 0) {
1283                 p->o_arg.u.attrs = &p->attrs;
1284                 memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1285
1286                 memcpy(p->o_arg.u.verifier.data, c->verf,
1287                                 sizeof(p->o_arg.u.verifier.data));
1288         }
1289         p->c_arg.fh = &p->o_res.fh;
1290         p->c_arg.stateid = &p->o_res.stateid;
1291         p->c_arg.seqid = p->o_arg.seqid;
1292         nfs4_init_opendata_res(p);
1293         kref_init(&p->kref);
1294         return p;
1295
1296 err_free_label:
1297         nfs4_label_free(p->a_label);
1298 err_free_f:
1299         nfs4_label_free(p->f_label);
1300 err_free_p:
1301         kfree(p);
1302 err:
1303         dput(parent);
1304         return NULL;
1305 }
1306
1307 static void nfs4_opendata_free(struct kref *kref)
1308 {
1309         struct nfs4_opendata *p = container_of(kref,
1310                         struct nfs4_opendata, kref);
1311         struct super_block *sb = p->dentry->d_sb;
1312
1313         nfs4_lgopen_release(p->lgp);
1314         nfs_free_seqid(p->o_arg.seqid);
1315         nfs4_sequence_free_slot(&p->o_res.seq_res);
1316         if (p->state != NULL)
1317                 nfs4_put_open_state(p->state);
1318         nfs4_put_state_owner(p->owner);
1319
1320         nfs4_label_free(p->a_label);
1321         nfs4_label_free(p->f_label);
1322
1323         dput(p->dir);
1324         dput(p->dentry);
1325         nfs_sb_deactive(sb);
1326         nfs_fattr_free_names(&p->f_attr);
1327         kfree(p->f_attr.mdsthreshold);
1328         kfree(p);
1329 }
1330
1331 static void nfs4_opendata_put(struct nfs4_opendata *p)
1332 {
1333         if (p != NULL)
1334                 kref_put(&p->kref, nfs4_opendata_free);
1335 }
1336
1337 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1338                 fmode_t fmode)
1339 {
1340         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1341         case FMODE_READ|FMODE_WRITE:
1342                 return state->n_rdwr != 0;
1343         case FMODE_WRITE:
1344                 return state->n_wronly != 0;
1345         case FMODE_READ:
1346                 return state->n_rdonly != 0;
1347         }
1348         WARN_ON_ONCE(1);
1349         return false;
1350 }
1351
1352 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
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(struct rpc_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                 struct rpc_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                 struct rpc_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         struct rpc_cred *cred;
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         cred = get_rpccred(delegation->cred);
2722         rcu_read_unlock();
2723         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2724         trace_nfs4_test_delegation_stateid(state, NULL, status);
2725         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2726                 nfs_finish_clear_delegation_stateid(state, &stateid);
2727
2728         put_rpccred(cred);
2729 }
2730
2731 /**
2732  * nfs41_check_expired_locks - possibly free a lock stateid
2733  *
2734  * @state: NFSv4 state for an inode
2735  *
2736  * Returns NFS_OK if recovery for this stateid is now finished.
2737  * Otherwise a negative NFS4ERR value is returned.
2738  */
2739 static int nfs41_check_expired_locks(struct nfs4_state *state)
2740 {
2741         int status, ret = NFS_OK;
2742         struct nfs4_lock_state *lsp, *prev = NULL;
2743         struct nfs_server *server = NFS_SERVER(state->inode);
2744
2745         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2746                 goto out;
2747
2748         spin_lock(&state->state_lock);
2749         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2750                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2751                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2752
2753                         refcount_inc(&lsp->ls_count);
2754                         spin_unlock(&state->state_lock);
2755
2756                         nfs4_put_lock_state(prev);
2757                         prev = lsp;
2758
2759                         status = nfs41_test_and_free_expired_stateid(server,
2760                                         &lsp->ls_stateid,
2761                                         cred);
2762                         trace_nfs4_test_lock_stateid(state, lsp, status);
2763                         if (status == -NFS4ERR_EXPIRED ||
2764                             status == -NFS4ERR_BAD_STATEID) {
2765                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2766                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2767                                 if (!recover_lost_locks)
2768                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2769                         } else if (status != NFS_OK) {
2770                                 ret = status;
2771                                 nfs4_put_lock_state(prev);
2772                                 goto out;
2773                         }
2774                         spin_lock(&state->state_lock);
2775                 }
2776         }
2777         spin_unlock(&state->state_lock);
2778         nfs4_put_lock_state(prev);
2779 out:
2780         return ret;
2781 }
2782
2783 /**
2784  * nfs41_check_open_stateid - possibly free an open stateid
2785  *
2786  * @state: NFSv4 state for an inode
2787  *
2788  * Returns NFS_OK if recovery for this stateid is now finished.
2789  * Otherwise a negative NFS4ERR value is returned.
2790  */
2791 static int nfs41_check_open_stateid(struct nfs4_state *state)
2792 {
2793         struct nfs_server *server = NFS_SERVER(state->inode);
2794         nfs4_stateid *stateid = &state->open_stateid;
2795         struct rpc_cred *cred = state->owner->so_cred;
2796         int status;
2797
2798         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2799                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2800                         if (nfs4_have_delegation(state->inode, state->state))
2801                                 return NFS_OK;
2802                         return -NFS4ERR_OPENMODE;
2803                 }
2804                 return -NFS4ERR_BAD_STATEID;
2805         }
2806         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2807         trace_nfs4_test_open_stateid(state, NULL, status);
2808         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2809                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2810                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2811                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2812                 clear_bit(NFS_OPEN_STATE, &state->flags);
2813                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2814                 return status;
2815         }
2816         if (nfs_open_stateid_recover_openmode(state))
2817                 return -NFS4ERR_OPENMODE;
2818         return NFS_OK;
2819 }
2820
2821 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2822 {
2823         int status;
2824
2825         nfs41_check_delegation_stateid(state);
2826         status = nfs41_check_expired_locks(state);
2827         if (status != NFS_OK)
2828                 return status;
2829         status = nfs41_check_open_stateid(state);
2830         if (status != NFS_OK)
2831                 status = nfs4_open_expired(sp, state);
2832         return status;
2833 }
2834 #endif
2835
2836 /*
2837  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2838  * fields corresponding to attributes that were used to store the verifier.
2839  * Make sure we clobber those fields in the later setattr call
2840  */
2841 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2842                                 struct iattr *sattr, struct nfs4_label **label)
2843 {
2844         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2845         __u32 attrset[3];
2846         unsigned ret;
2847         unsigned i;
2848
2849         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2850                 attrset[i] = opendata->o_res.attrset[i];
2851                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2852                         attrset[i] &= ~bitmask[i];
2853         }
2854
2855         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2856                 sattr->ia_valid : 0;
2857
2858         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2859                 if (sattr->ia_valid & ATTR_ATIME_SET)
2860                         ret |= ATTR_ATIME_SET;
2861                 else
2862                         ret |= ATTR_ATIME;
2863         }
2864
2865         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2866                 if (sattr->ia_valid & ATTR_MTIME_SET)
2867                         ret |= ATTR_MTIME_SET;
2868                 else
2869                         ret |= ATTR_MTIME;
2870         }
2871
2872         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2873                 *label = NULL;
2874         return ret;
2875 }
2876
2877 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2878                 fmode_t fmode,
2879                 int flags,
2880                 struct nfs_open_context *ctx)
2881 {
2882         struct nfs4_state_owner *sp = opendata->owner;
2883         struct nfs_server *server = sp->so_server;
2884         struct dentry *dentry;
2885         struct nfs4_state *state;
2886         unsigned int seq;
2887         int ret;
2888
2889         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2890
2891         ret = _nfs4_proc_open(opendata, ctx);
2892         if (ret != 0)
2893                 goto out;
2894
2895         state = _nfs4_opendata_to_nfs4_state(opendata);
2896         ret = PTR_ERR(state);
2897         if (IS_ERR(state))
2898                 goto out;
2899         ctx->state = state;
2900         if (server->caps & NFS_CAP_POSIX_LOCK)
2901                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2902         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2903                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2904
2905         dentry = opendata->dentry;
2906         if (d_really_is_negative(dentry)) {
2907                 struct dentry *alias;
2908                 d_drop(dentry);
2909                 alias = d_exact_alias(dentry, state->inode);
2910                 if (!alias)
2911                         alias = d_splice_alias(igrab(state->inode), dentry);
2912                 /* d_splice_alias() can't fail here - it's a non-directory */
2913                 if (alias) {
2914                         dput(ctx->dentry);
2915                         ctx->dentry = dentry = alias;
2916                 }
2917                 nfs_set_verifier(dentry,
2918                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2919         }
2920
2921         /* Parse layoutget results before we check for access */
2922         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
2923
2924         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2925         if (ret != 0)
2926                 goto out;
2927
2928         if (d_inode(dentry) == state->inode) {
2929                 nfs_inode_attach_open_context(ctx);
2930                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2931                         nfs4_schedule_stateid_recovery(server, state);
2932         }
2933
2934 out:
2935         nfs4_sequence_free_slot(&opendata->o_res.seq_res);
2936         return ret;
2937 }
2938
2939 /*
2940  * Returns a referenced nfs4_state
2941  */
2942 static int _nfs4_do_open(struct inode *dir,
2943                         struct nfs_open_context *ctx,
2944                         int flags,
2945                         const struct nfs4_open_createattrs *c,
2946                         int *opened)
2947 {
2948         struct nfs4_state_owner  *sp;
2949         struct nfs4_state     *state = NULL;
2950         struct nfs_server       *server = NFS_SERVER(dir);
2951         struct nfs4_opendata *opendata;
2952         struct dentry *dentry = ctx->dentry;
2953         struct rpc_cred *cred = ctx->cred;
2954         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2955         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2956         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2957         struct iattr *sattr = c->sattr;
2958         struct nfs4_label *label = c->label;
2959         struct nfs4_label *olabel = NULL;
2960         int status;
2961
2962         /* Protect against reboot recovery conflicts */
2963         status = -ENOMEM;
2964         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2965         if (sp == NULL) {
2966                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2967                 goto out_err;
2968         }
2969         status = nfs4_client_recover_expired_lease(server->nfs_client);
2970         if (status != 0)
2971                 goto err_put_state_owner;
2972         if (d_really_is_positive(dentry))
2973                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2974         status = -ENOMEM;
2975         if (d_really_is_positive(dentry))
2976                 claim = NFS4_OPEN_CLAIM_FH;
2977         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
2978                         c, claim, GFP_KERNEL);
2979         if (opendata == NULL)
2980                 goto err_put_state_owner;
2981
2982         if (label) {
2983                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2984                 if (IS_ERR(olabel)) {
2985                         status = PTR_ERR(olabel);
2986                         goto err_opendata_put;
2987                 }
2988         }
2989
2990         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2991                 if (!opendata->f_attr.mdsthreshold) {
2992                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2993                         if (!opendata->f_attr.mdsthreshold)
2994                                 goto err_free_label;
2995                 }
2996                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2997         }
2998         if (d_really_is_positive(dentry))
2999                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3000
3001         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
3002         if (status != 0)
3003                 goto err_free_label;
3004         state = ctx->state;
3005
3006         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3007             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3008                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3009                 /*
3010                  * send create attributes which was not set by open
3011                  * with an extra setattr.
3012                  */
3013                 if (attrs || label) {
3014                         unsigned ia_old = sattr->ia_valid;
3015
3016                         sattr->ia_valid = attrs;
3017                         nfs_fattr_init(opendata->o_res.f_attr);
3018                         status = nfs4_do_setattr(state->inode, cred,
3019                                         opendata->o_res.f_attr, sattr,
3020                                         ctx, label, olabel);
3021                         if (status == 0) {
3022                                 nfs_setattr_update_inode(state->inode, sattr,
3023                                                 opendata->o_res.f_attr);
3024                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3025                         }
3026                         sattr->ia_valid = ia_old;
3027                 }
3028         }
3029         if (opened && opendata->file_created)
3030                 *opened = 1;
3031
3032         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3033                 *ctx_th = opendata->f_attr.mdsthreshold;
3034                 opendata->f_attr.mdsthreshold = NULL;
3035         }
3036
3037         nfs4_label_free(olabel);
3038
3039         nfs4_opendata_put(opendata);
3040         nfs4_put_state_owner(sp);
3041         return 0;
3042 err_free_label:
3043         nfs4_label_free(olabel);
3044 err_opendata_put:
3045         nfs4_opendata_put(opendata);
3046 err_put_state_owner:
3047         nfs4_put_state_owner(sp);
3048 out_err:
3049         return status;
3050 }
3051
3052
3053 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3054                                         struct nfs_open_context *ctx,
3055                                         int flags,
3056                                         struct iattr *sattr,
3057                                         struct nfs4_label *label,
3058                                         int *opened)
3059 {
3060         struct nfs_server *server = NFS_SERVER(dir);
3061         struct nfs4_exception exception = { };
3062         struct nfs4_state *res;
3063         struct nfs4_open_createattrs c = {
3064                 .label = label,
3065                 .sattr = sattr,
3066                 .verf = {
3067                         [0] = (__u32)jiffies,
3068                         [1] = (__u32)current->pid,
3069                 },
3070         };
3071         int status;
3072
3073         do {
3074                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3075                 res = ctx->state;
3076                 trace_nfs4_open_file(ctx, flags, status);
3077                 if (status == 0)
3078                         break;
3079                 /* NOTE: BAD_SEQID means the server and client disagree about the
3080                  * book-keeping w.r.t. state-changing operations
3081                  * (OPEN/CLOSE/LOCK/LOCKU...)
3082                  * It is actually a sign of a bug on the client or on the server.
3083                  *
3084                  * If we receive a BAD_SEQID error in the particular case of
3085                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3086                  * have unhashed the old state_owner for us, and that we can
3087                  * therefore safely retry using a new one. We should still warn
3088                  * the user though...
3089                  */
3090                 if (status == -NFS4ERR_BAD_SEQID) {
3091                         pr_warn_ratelimited("NFS: v4 server %s "
3092                                         " returned a bad sequence-id error!\n",
3093                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3094                         exception.retry = 1;
3095                         continue;
3096                 }
3097                 /*
3098                  * BAD_STATEID on OPEN means that the server cancelled our
3099                  * state before it received the OPEN_CONFIRM.
3100                  * Recover by retrying the request as per the discussion
3101                  * on Page 181 of RFC3530.
3102                  */
3103                 if (status == -NFS4ERR_BAD_STATEID) {
3104                         exception.retry = 1;
3105                         continue;
3106                 }
3107                 if (status == -EAGAIN) {
3108                         /* We must have found a delegation */
3109                         exception.retry = 1;
3110                         continue;
3111                 }
3112                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3113                         continue;
3114                 res = ERR_PTR(nfs4_handle_exception(server,
3115                                         status, &exception));
3116         } while (exception.retry);
3117         return res;
3118 }
3119
3120 static int _nfs4_do_setattr(struct inode *inode,
3121                             struct nfs_setattrargs *arg,
3122                             struct nfs_setattrres *res,
3123                             struct rpc_cred *cred,
3124                             struct nfs_open_context *ctx)
3125 {
3126         struct nfs_server *server = NFS_SERVER(inode);
3127         struct rpc_message msg = {
3128                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3129                 .rpc_argp       = arg,
3130                 .rpc_resp       = res,
3131                 .rpc_cred       = cred,
3132         };
3133         struct rpc_cred *delegation_cred = NULL;
3134         unsigned long timestamp = jiffies;
3135         bool truncate;
3136         int status;
3137
3138         nfs_fattr_init(res->fattr);
3139
3140         /* Servers should only apply open mode checks for file size changes */
3141         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3142         if (!truncate)
3143                 goto zero_stateid;
3144
3145         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3146                 /* Use that stateid */
3147         } else if (ctx != NULL) {
3148                 struct nfs_lock_context *l_ctx;
3149                 if (!nfs4_valid_open_stateid(ctx->state))
3150                         return -EBADF;
3151                 l_ctx = nfs_get_lock_context(ctx);
3152                 if (IS_ERR(l_ctx))
3153                         return PTR_ERR(l_ctx);
3154                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3155                                                 &arg->stateid, &delegation_cred);
3156                 nfs_put_lock_context(l_ctx);
3157                 if (status == -EIO)
3158                         return -EBADF;
3159         } else {
3160 zero_stateid:
3161                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3162         }
3163         if (delegation_cred)
3164                 msg.rpc_cred = delegation_cred;
3165
3166         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3167
3168         put_rpccred(delegation_cred);
3169         if (status == 0 && ctx != NULL)
3170                 renew_lease(server, timestamp);
3171         trace_nfs4_setattr(inode, &arg->stateid, status);
3172         return status;
3173 }
3174
3175 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
3176                            struct nfs_fattr *fattr, struct iattr *sattr,
3177                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3178                            struct nfs4_label *olabel)
3179 {
3180         struct nfs_server *server = NFS_SERVER(inode);
3181         __u32 bitmask[NFS4_BITMASK_SZ];
3182         struct nfs4_state *state = ctx ? ctx->state : NULL;
3183         struct nfs_setattrargs  arg = {
3184                 .fh             = NFS_FH(inode),
3185                 .iap            = sattr,
3186                 .server         = server,
3187                 .bitmask = bitmask,
3188                 .label          = ilabel,
3189         };
3190         struct nfs_setattrres  res = {
3191                 .fattr          = fattr,
3192                 .label          = olabel,
3193                 .server         = server,
3194         };
3195         struct nfs4_exception exception = {
3196                 .state = state,
3197                 .inode = inode,
3198                 .stateid = &arg.stateid,
3199         };
3200         int err;
3201
3202         do {
3203                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3204                                 nfs4_bitmask(server, olabel),
3205                                 inode);
3206
3207                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3208                 switch (err) {
3209                 case -NFS4ERR_OPENMODE:
3210                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3211                                 pr_warn_once("NFSv4: server %s is incorrectly "
3212                                                 "applying open mode checks to "
3213                                                 "a SETATTR that is not "
3214                                                 "changing file size.\n",
3215                                                 server->nfs_client->cl_hostname);
3216                         }
3217                         if (state && !(state->state & FMODE_WRITE)) {
3218                                 err = -EBADF;
3219                                 if (sattr->ia_valid & ATTR_OPEN)
3220                                         err = -EACCES;
3221                                 goto out;
3222                         }
3223                 }
3224                 err = nfs4_handle_exception(server, err, &exception);
3225         } while (exception.retry);
3226 out:
3227         return err;
3228 }
3229
3230 static bool
3231 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3232 {
3233         if (inode == NULL || !nfs_have_layout(inode))
3234                 return false;
3235
3236         return pnfs_wait_on_layoutreturn(inode, task);
3237 }
3238
3239 struct nfs4_closedata {
3240         struct inode *inode;
3241         struct nfs4_state *state;
3242         struct nfs_closeargs arg;
3243         struct nfs_closeres res;
3244         struct {
3245                 struct nfs4_layoutreturn_args arg;
3246                 struct nfs4_layoutreturn_res res;
3247                 struct nfs4_xdr_opaque_data ld_private;
3248                 u32 roc_barrier;
3249                 bool roc;
3250         } lr;
3251         struct nfs_fattr fattr;
3252         unsigned long timestamp;
3253 };
3254
3255 static void nfs4_free_closedata(void *data)
3256 {
3257         struct nfs4_closedata *calldata = data;
3258         struct nfs4_state_owner *sp = calldata->state->owner;
3259         struct super_block *sb = calldata->state->inode->i_sb;
3260
3261         if (calldata->lr.roc)
3262                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3263                                 calldata->res.lr_ret);
3264         nfs4_put_open_state(calldata->state);
3265         nfs_free_seqid(calldata->arg.seqid);
3266         nfs4_put_state_owner(sp);
3267         nfs_sb_deactive(sb);
3268         kfree(calldata);
3269 }
3270
3271 static void nfs4_close_done(struct rpc_task *task, void *data)
3272 {
3273         struct nfs4_closedata *calldata = data;
3274         struct nfs4_state *state = calldata->state;
3275         struct nfs_server *server = NFS_SERVER(calldata->inode);
3276         nfs4_stateid *res_stateid = NULL;
3277         struct nfs4_exception exception = {
3278                 .state = state,
3279                 .inode = calldata->inode,
3280                 .stateid = &calldata->arg.stateid,
3281         };
3282
3283         dprintk("%s: begin!\n", __func__);
3284         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3285                 return;
3286         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3287
3288         /* Handle Layoutreturn errors */
3289         if (calldata->arg.lr_args && task->tk_status != 0) {
3290                 switch (calldata->res.lr_ret) {
3291                 default:
3292                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3293                         break;
3294                 case 0:
3295                         calldata->arg.lr_args = NULL;
3296                         calldata->res.lr_res = NULL;
3297                         break;
3298                 case -NFS4ERR_OLD_STATEID:
3299                         if (nfs4_layoutreturn_refresh_stateid(&calldata->arg.lr_args->stateid,
3300                                                 &calldata->arg.lr_args->range,
3301                                                 calldata->inode))
3302                                 goto lr_restart;
3303                         /* Fallthrough */
3304                 case -NFS4ERR_ADMIN_REVOKED:
3305                 case -NFS4ERR_DELEG_REVOKED:
3306                 case -NFS4ERR_EXPIRED:
3307                 case -NFS4ERR_BAD_STATEID:
3308                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3309                 case -NFS4ERR_WRONG_CRED:
3310                         calldata->arg.lr_args = NULL;
3311                         calldata->res.lr_res = NULL;
3312                         goto lr_restart;
3313                 }
3314         }
3315
3316         /* hmm. we are done with the inode, and in the process of freeing
3317          * the state_owner. we keep this around to process errors
3318          */
3319         switch (task->tk_status) {
3320                 case 0:
3321                         res_stateid = &calldata->res.stateid;
3322                         renew_lease(server, calldata->timestamp);
3323                         break;
3324                 case -NFS4ERR_ACCESS:
3325                         if (calldata->arg.bitmask != NULL) {
3326                                 calldata->arg.bitmask = NULL;
3327                                 calldata->res.fattr = NULL;
3328                                 goto out_restart;
3329
3330                         }
3331                         break;
3332                 case -NFS4ERR_OLD_STATEID:
3333                         /* Did we race with OPEN? */
3334                         if (nfs4_refresh_open_stateid(&calldata->arg.stateid,
3335                                                 state))
3336                                 goto out_restart;
3337                         goto out_release;
3338                 case -NFS4ERR_ADMIN_REVOKED:
3339                 case -NFS4ERR_STALE_STATEID:
3340                 case -NFS4ERR_EXPIRED:
3341                         nfs4_free_revoked_stateid(server,
3342                                         &calldata->arg.stateid,
3343                                         task->tk_msg.rpc_cred);
3344                         /* Fallthrough */
3345                 case -NFS4ERR_BAD_STATEID:
3346                         break;
3347                 default:
3348                         task->tk_status = nfs4_async_handle_exception(task,
3349                                         server, task->tk_status, &exception);
3350                         if (exception.retry)
3351                                 goto out_restart;
3352         }
3353         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3354                         res_stateid, calldata->arg.fmode);
3355 out_release:
3356         task->tk_status = 0;
3357         nfs_release_seqid(calldata->arg.seqid);
3358         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3359         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3360         return;
3361 lr_restart:
3362         calldata->res.lr_ret = 0;
3363 out_restart:
3364         task->tk_status = 0;
3365         rpc_restart_call_prepare(task);
3366         goto out_release;
3367 }
3368
3369 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3370 {
3371         struct nfs4_closedata *calldata = data;
3372         struct nfs4_state *state = calldata->state;
3373         struct inode *inode = calldata->inode;
3374         struct pnfs_layout_hdr *lo;
3375         bool is_rdonly, is_wronly, is_rdwr;
3376         int call_close = 0;
3377
3378         dprintk("%s: begin!\n", __func__);
3379         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3380                 goto out_wait;
3381
3382         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3383         spin_lock(&state->owner->so_lock);
3384         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3385         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3386         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3387         /* Calculate the change in open mode */
3388         calldata->arg.fmode = 0;
3389         if (state->n_rdwr == 0) {
3390                 if (state->n_rdonly == 0)
3391                         call_close |= is_rdonly;
3392                 else if (is_rdonly)
3393                         calldata->arg.fmode |= FMODE_READ;
3394                 if (state->n_wronly == 0)
3395                         call_close |= is_wronly;
3396                 else if (is_wronly)
3397                         calldata->arg.fmode |= FMODE_WRITE;
3398                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3399                         call_close |= is_rdwr;
3400         } else if (is_rdwr)
3401                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3402
3403         if (!nfs4_valid_open_stateid(state) ||
3404             !nfs4_refresh_open_stateid(&calldata->arg.stateid, state))
3405                 call_close = 0;
3406         spin_unlock(&state->owner->so_lock);
3407
3408         if (!call_close) {
3409                 /* Note: exit _without_ calling nfs4_close_done */
3410                 goto out_no_action;
3411         }
3412
3413         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3414                 nfs_release_seqid(calldata->arg.seqid);
3415                 goto out_wait;
3416         }
3417
3418         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3419         if (lo && !pnfs_layout_is_valid(lo)) {
3420                 calldata->arg.lr_args = NULL;
3421                 calldata->res.lr_res = NULL;
3422         }
3423
3424         if (calldata->arg.fmode == 0)
3425                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3426
3427         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3428                 /* Close-to-open cache consistency revalidation */
3429                 if (!nfs4_have_delegation(inode, FMODE_READ))
3430                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3431                 else
3432                         calldata->arg.bitmask = NULL;
3433         }
3434
3435         calldata->arg.share_access =
3436                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3437                                 calldata->arg.fmode, 0);
3438
3439         if (calldata->res.fattr == NULL)
3440                 calldata->arg.bitmask = NULL;
3441         else if (calldata->arg.bitmask == NULL)
3442                 calldata->res.fattr = NULL;
3443         calldata->timestamp = jiffies;
3444         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3445                                 &calldata->arg.seq_args,
3446                                 &calldata->res.seq_res,
3447                                 task) != 0)
3448                 nfs_release_seqid(calldata->arg.seqid);
3449         dprintk("%s: done!\n", __func__);
3450         return;
3451 out_no_action:
3452         task->tk_action = NULL;
3453 out_wait:
3454         nfs4_sequence_done(task, &calldata->res.seq_res);
3455 }
3456
3457 static const struct rpc_call_ops nfs4_close_ops = {
3458         .rpc_call_prepare = nfs4_close_prepare,
3459         .rpc_call_done = nfs4_close_done,
3460         .rpc_release = nfs4_free_closedata,
3461 };
3462
3463 /*
3464  * It is possible for data to be read/written from a mem-mapped file
3465  * after the sys_close call (which hits the vfs layer as a flush).
3466  * This means that we can't safely call nfsv4 close on a file until
3467  * the inode is cleared. This in turn means that we are not good
3468  * NFSv4 citizens - we do not indicate to the server to update the file's
3469  * share state even when we are done with one of the three share
3470  * stateid's in the inode.
3471  *
3472  * NOTE: Caller must be holding the sp->so_owner semaphore!
3473  */
3474 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3475 {
3476         struct nfs_server *server = NFS_SERVER(state->inode);
3477         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3478         struct nfs4_closedata *calldata;
3479         struct nfs4_state_owner *sp = state->owner;
3480         struct rpc_task *task;
3481         struct rpc_message msg = {
3482                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3483                 .rpc_cred = state->owner->so_cred,
3484         };
3485         struct rpc_task_setup task_setup_data = {
3486                 .rpc_client = server->client,
3487                 .rpc_message = &msg,
3488                 .callback_ops = &nfs4_close_ops,
3489                 .workqueue = nfsiod_workqueue,
3490                 .flags = RPC_TASK_ASYNC,
3491         };
3492         int status = -ENOMEM;
3493
3494         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3495                 &task_setup_data.rpc_client, &msg);
3496
3497         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3498         if (calldata == NULL)
3499                 goto out;
3500         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3501         calldata->inode = state->inode;
3502         calldata->state = state;
3503         calldata->arg.fh = NFS_FH(state->inode);
3504         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3505                 goto out_free_calldata;
3506         /* Serialization for the sequence id */
3507         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3508         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3509         if (IS_ERR(calldata->arg.seqid))
3510                 goto out_free_calldata;
3511         nfs_fattr_init(&calldata->fattr);
3512         calldata->arg.fmode = 0;
3513         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3514         calldata->res.fattr = &calldata->fattr;
3515         calldata->res.seqid = calldata->arg.seqid;
3516         calldata->res.server = server;
3517         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3518         calldata->lr.roc = pnfs_roc(state->inode,
3519                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3520         if (calldata->lr.roc) {
3521                 calldata->arg.lr_args = &calldata->lr.arg;
3522                 calldata->res.lr_res = &calldata->lr.res;
3523         }
3524         nfs_sb_active(calldata->inode->i_sb);
3525
3526         msg.rpc_argp = &calldata->arg;
3527         msg.rpc_resp = &calldata->res;
3528         task_setup_data.callback_data = calldata;
3529         task = rpc_run_task(&task_setup_data);
3530         if (IS_ERR(task))
3531                 return PTR_ERR(task);
3532         status = 0;
3533         if (wait)
3534                 status = rpc_wait_for_completion_task(task);
3535         rpc_put_task(task);
3536         return status;
3537 out_free_calldata:
3538         kfree(calldata);
3539 out:
3540         nfs4_put_open_state(state);
3541         nfs4_put_state_owner(sp);
3542         return status;
3543 }
3544
3545 static struct inode *
3546 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3547                 int open_flags, struct iattr *attr, int *opened)
3548 {
3549         struct nfs4_state *state;
3550         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3551
3552         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3553
3554         /* Protect against concurrent sillydeletes */
3555         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3556
3557         nfs4_label_release_security(label);
3558
3559         if (IS_ERR(state))
3560                 return ERR_CAST(state);
3561         return state->inode;
3562 }
3563
3564 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3565 {
3566         if (ctx->state == NULL)
3567                 return;
3568         if (is_sync)
3569                 nfs4_close_sync(ctx->state, ctx->mode);
3570         else
3571                 nfs4_close_state(ctx->state, ctx->mode);
3572 }
3573
3574 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3575 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3576 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3577
3578 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3579 {
3580         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3581         struct nfs4_server_caps_arg args = {
3582                 .fhandle = fhandle,
3583                 .bitmask = bitmask,
3584         };
3585         struct nfs4_server_caps_res res = {};
3586         struct rpc_message msg = {
3587                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3588                 .rpc_argp = &args,
3589                 .rpc_resp = &res,
3590         };
3591         int status;
3592         int i;
3593
3594         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3595                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3596                      FATTR4_WORD0_LINK_SUPPORT |
3597                      FATTR4_WORD0_SYMLINK_SUPPORT |
3598                      FATTR4_WORD0_ACLSUPPORT;
3599         if (minorversion)
3600                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3601
3602         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3603         if (status == 0) {
3604                 /* Sanity check the server answers */
3605                 switch (minorversion) {
3606                 case 0:
3607                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3608                         res.attr_bitmask[2] = 0;
3609                         break;
3610                 case 1:
3611                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3612                         break;
3613                 case 2:
3614                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3615                 }
3616                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3617                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3618                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3619                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3620                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3621                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3622                                 NFS_CAP_SECURITY_LABEL);
3623                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3624                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3625                         server->caps |= NFS_CAP_ACLS;
3626                 if (res.has_links != 0)
3627                         server->caps |= NFS_CAP_HARDLINKS;
3628                 if (res.has_symlinks != 0)
3629                         server->caps |= NFS_CAP_SYMLINKS;
3630                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3631                         server->caps |= NFS_CAP_FILEID;
3632                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3633                         server->caps |= NFS_CAP_MODE;
3634                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3635                         server->caps |= NFS_CAP_NLINK;
3636                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3637                         server->caps |= NFS_CAP_OWNER;
3638                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3639                         server->caps |= NFS_CAP_OWNER_GROUP;
3640                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3641                         server->caps |= NFS_CAP_ATIME;
3642                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3643                         server->caps |= NFS_CAP_CTIME;
3644                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3645                         server->caps |= NFS_CAP_MTIME;
3646 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3647                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3648                         server->caps |= NFS_CAP_SECURITY_LABEL;
3649 #endif
3650                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3651                                 sizeof(server->attr_bitmask));
3652                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3653
3654                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3655                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3656                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3657                 server->cache_consistency_bitmask[2] = 0;
3658
3659                 /* Avoid a regression due to buggy server */
3660                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3661                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3662                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3663                         sizeof(server->exclcreat_bitmask));
3664
3665                 server->acl_bitmask = res.acl_bitmask;
3666                 server->fh_expire_type = res.fh_expire_type;
3667         }
3668
3669         return status;
3670 }
3671
3672 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3673 {
3674         struct nfs4_exception exception = { };
3675         int err;
3676         do {
3677                 err = nfs4_handle_exception(server,
3678                                 _nfs4_server_capabilities(server, fhandle),
3679                                 &exception);
3680         } while (exception.retry);
3681         return err;
3682 }
3683
3684 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3685                 struct nfs_fsinfo *info)
3686 {
3687         u32 bitmask[3];
3688         struct nfs4_lookup_root_arg args = {
3689                 .bitmask = bitmask,
3690         };
3691         struct nfs4_lookup_res res = {
3692                 .server = server,
3693                 .fattr = info->fattr,
3694                 .fh = fhandle,
3695         };
3696         struct rpc_message msg = {
3697                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3698                 .rpc_argp = &args,
3699                 .rpc_resp = &res,
3700         };
3701
3702         bitmask[0] = nfs4_fattr_bitmap[0];
3703         bitmask[1] = nfs4_fattr_bitmap[1];
3704         /*
3705          * Process the label in the upcoming getfattr
3706          */
3707         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3708
3709         nfs_fattr_init(info->fattr);
3710         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3711 }
3712
3713 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3714                 struct nfs_fsinfo *info)
3715 {
3716         struct nfs4_exception exception = { };
3717         int err;
3718         do {
3719                 err = _nfs4_lookup_root(server, fhandle, info);
3720                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3721                 switch (err) {
3722                 case 0:
3723                 case -NFS4ERR_WRONGSEC:
3724                         goto out;
3725                 default:
3726                         err = nfs4_handle_exception(server, err, &exception);
3727                 }
3728         } while (exception.retry);
3729 out:
3730         return err;
3731 }
3732
3733 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3734                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3735 {
3736         struct rpc_auth_create_args auth_args = {
3737                 .pseudoflavor = flavor,
3738         };
3739         struct rpc_auth *auth;
3740
3741         auth = rpcauth_create(&auth_args, server->client);
3742         if (IS_ERR(auth))
3743                 return -EACCES;
3744         return nfs4_lookup_root(server, fhandle, info);
3745 }
3746
3747 /*
3748  * Retry pseudoroot lookup with various security flavors.  We do this when:
3749  *
3750  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3751  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3752  *
3753  * Returns zero on success, or a negative NFS4ERR value, or a
3754  * negative errno value.
3755  */
3756 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3757                               struct nfs_fsinfo *info)
3758 {
3759         /* Per 3530bis 15.33.5 */
3760         static const rpc_authflavor_t flav_array[] = {
3761                 RPC_AUTH_GSS_KRB5P,
3762                 RPC_AUTH_GSS_KRB5I,
3763                 RPC_AUTH_GSS_KRB5,
3764                 RPC_AUTH_UNIX,                  /* courtesy */
3765                 RPC_AUTH_NULL,
3766         };
3767         int status = -EPERM;
3768         size_t i;
3769
3770         if (server->auth_info.flavor_len > 0) {
3771                 /* try each flavor specified by user */
3772                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3773                         status = nfs4_lookup_root_sec(server, fhandle, info,
3774                                                 server->auth_info.flavors[i]);
3775                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3776                                 continue;
3777                         break;
3778                 }
3779         } else {
3780                 /* no flavors specified by user, try default list */
3781                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3782                         status = nfs4_lookup_root_sec(server, fhandle, info,
3783                                                       flav_array[i]);
3784                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3785                                 continue;
3786                         break;
3787                 }
3788         }
3789
3790         /*
3791          * -EACCES could mean that the user doesn't have correct permissions
3792          * to access the mount.  It could also mean that we tried to mount
3793          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3794          * existing mount programs don't handle -EACCES very well so it should
3795          * be mapped to -EPERM instead.
3796          */
3797         if (status == -EACCES)
3798                 status = -EPERM;
3799         return status;
3800 }
3801
3802 /**
3803  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3804  * @server: initialized nfs_server handle
3805  * @fhandle: we fill in the pseudo-fs root file handle
3806  * @info: we fill in an FSINFO struct
3807  * @auth_probe: probe the auth flavours
3808  *
3809  * Returns zero on success, or a negative errno.
3810  */
3811 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3812                          struct nfs_fsinfo *info,
3813                          bool auth_probe)
3814 {
3815         int status = 0;
3816
3817         if (!auth_probe)
3818                 status = nfs4_lookup_root(server, fhandle, info);
3819
3820         if (auth_probe || status == NFS4ERR_WRONGSEC)
3821                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3822                                 fhandle, info);
3823
3824         if (status == 0)
3825                 status = nfs4_server_capabilities(server, fhandle);
3826         if (status == 0)
3827                 status = nfs4_do_fsinfo(server, fhandle, info);
3828
3829         return nfs4_map_errors(status);
3830 }
3831
3832 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3833                               struct nfs_fsinfo *info)
3834 {
3835         int error;
3836         struct nfs_fattr *fattr = info->fattr;
3837         struct nfs4_label *label = NULL;
3838
3839         error = nfs4_server_capabilities(server, mntfh);
3840         if (error < 0) {
3841                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3842                 return error;
3843         }
3844
3845         label = nfs4_label_alloc(server, GFP_KERNEL);
3846         if (IS_ERR(label))
3847                 return PTR_ERR(label);
3848
3849         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
3850         if (error < 0) {
3851                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3852                 goto err_free_label;
3853         }
3854
3855         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3856             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3857                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3858
3859 err_free_label:
3860         nfs4_label_free(label);
3861
3862         return error;
3863 }
3864
3865 /*
3866  * Get locations and (maybe) other attributes of a referral.
3867  * Note that we'll actually follow the referral later when
3868  * we detect fsid mismatch in inode revalidation
3869  */
3870 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3871                              const struct qstr *name, struct nfs_fattr *fattr,
3872                              struct nfs_fh *fhandle)
3873 {
3874         int status = -ENOMEM;
3875         struct page *page = NULL;
3876         struct nfs4_fs_locations *locations = NULL;
3877
3878         page = alloc_page(GFP_KERNEL);
3879         if (page == NULL)
3880                 goto out;
3881         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3882         if (locations == NULL)
3883                 goto out;
3884
3885         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3886         if (status != 0)
3887                 goto out;
3888
3889         /*
3890          * If the fsid didn't change, this is a migration event, not a
3891          * referral.  Cause us to drop into the exception handler, which
3892          * will kick off migration recovery.
3893          */
3894         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3895                 dprintk("%s: server did not return a different fsid for"
3896                         " a referral at %s\n", __func__, name->name);
3897                 status = -NFS4ERR_MOVED;
3898                 goto out;
3899         }
3900         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3901         nfs_fixup_referral_attributes(&locations->fattr);
3902
3903         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3904         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3905         memset(fhandle, 0, sizeof(struct nfs_fh));
3906 out:
3907         if (page)
3908                 __free_page(page);
3909         kfree(locations);
3910         return status;
3911 }
3912
3913 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3914                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3915                                 struct inode *inode)
3916 {
3917         __u32 bitmask[NFS4_BITMASK_SZ];
3918         struct nfs4_getattr_arg args = {
3919                 .fh = fhandle,
3920                 .bitmask = bitmask,
3921         };
3922         struct nfs4_getattr_res res = {
3923                 .fattr = fattr,
3924                 .label = label,
3925                 .server = server,
3926         };
3927         struct rpc_message msg = {
3928                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3929                 .rpc_argp = &args,
3930                 .rpc_resp = &res,
3931         };
3932
3933         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
3934
3935         nfs_fattr_init(fattr);
3936         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3937 }
3938
3939 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3940                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3941                                 struct inode *inode)
3942 {
3943         struct nfs4_exception exception = { };
3944         int err;
3945         do {
3946                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
3947                 trace_nfs4_getattr(server, fhandle, fattr, err);
3948                 err = nfs4_handle_exception(server, err,
3949                                 &exception);
3950         } while (exception.retry);
3951         return err;
3952 }
3953
3954 /*
3955  * The file is not closed if it is opened due to the a request to change
3956  * the size of the file. The open call will not be needed once the
3957  * VFS layer lookup-intents are implemented.
3958  *
3959  * Close is called when the inode is destroyed.
3960  * If we haven't opened the file for O_WRONLY, we
3961  * need to in the size_change case to obtain a stateid.
3962  *
3963  * Got race?
3964  * Because OPEN is always done by name in nfsv4, it is
3965  * possible that we opened a different file by the same
3966  * name.  We can recognize this race condition, but we
3967  * can't do anything about it besides returning an error.
3968  *
3969  * This will be fixed with VFS changes (lookup-intent).
3970  */
3971 static int
3972 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3973                   struct iattr *sattr)
3974 {
3975         struct inode *inode = d_inode(dentry);
3976         struct rpc_cred *cred = NULL;
3977         struct nfs_open_context *ctx = NULL;
3978         struct nfs4_label *label = NULL;
3979         int status;
3980
3981         if (pnfs_ld_layoutret_on_setattr(inode) &&
3982             sattr->ia_valid & ATTR_SIZE &&
3983             sattr->ia_size < i_size_read(inode))
3984                 pnfs_commit_and_return_layout(inode);
3985
3986         nfs_fattr_init(fattr);
3987
3988         /* Deal with open(O_TRUNC) */
3989         if (sattr->ia_valid & ATTR_OPEN)
3990                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3991
3992         /* Optimization: if the end result is no change, don't RPC */
3993         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3994                 return 0;
3995
3996         /* Search for an existing open(O_WRITE) file */
3997         if (sattr->ia_valid & ATTR_FILE) {
3998
3999                 ctx = nfs_file_open_context(sattr->ia_file);
4000                 if (ctx)
4001                         cred = ctx->cred;
4002         }
4003
4004         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4005         if (IS_ERR(label))
4006                 return PTR_ERR(label);
4007
4008         /* Return any delegations if we're going to change ACLs */
4009         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4010                 nfs4_inode_make_writeable(inode);
4011
4012         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4013         if (status == 0) {
4014                 nfs_setattr_update_inode(inode, sattr, fattr);
4015                 nfs_setsecurity(inode, fattr, label);
4016         }
4017         nfs4_label_free(label);
4018         return status;
4019 }
4020
4021 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4022                 const struct qstr *name, struct nfs_fh *fhandle,
4023                 struct nfs_fattr *fattr, struct nfs4_label *label)
4024 {
4025         struct nfs_server *server = NFS_SERVER(dir);
4026         int                    status;
4027         struct nfs4_lookup_arg args = {
4028                 .bitmask = server->attr_bitmask,
4029                 .dir_fh = NFS_FH(dir),
4030                 .name = name,
4031         };
4032         struct nfs4_lookup_res res = {
4033                 .server = server,
4034                 .fattr = fattr,
4035                 .label = label,
4036                 .fh = fhandle,
4037         };
4038         struct rpc_message msg = {
4039                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4040                 .rpc_argp = &args,
4041                 .rpc_resp = &res,
4042         };
4043
4044         args.bitmask = nfs4_bitmask(server, label);
4045
4046         nfs_fattr_init(fattr);
4047
4048         dprintk("NFS call  lookup %s\n", name->name);
4049         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
4050         dprintk("NFS reply lookup: %d\n", status);
4051         return status;
4052 }
4053
4054 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4055 {
4056         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4057                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4058         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4059         fattr->nlink = 2;
4060 }
4061
4062 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4063                                    const struct qstr *name, struct nfs_fh *fhandle,
4064                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4065 {
4066         struct nfs4_exception exception = { };
4067         struct rpc_clnt *client = *clnt;
4068         int err;
4069         do {
4070                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
4071                 trace_nfs4_lookup(dir, name, err);
4072                 switch (err) {
4073                 case -NFS4ERR_BADNAME:
4074                         err = -ENOENT;
4075                         goto out;
4076                 case -NFS4ERR_MOVED:
4077                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4078                         if (err == -NFS4ERR_MOVED)
4079                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4080                         goto out;
4081                 case -NFS4ERR_WRONGSEC:
4082                         err = -EPERM;
4083                         if (client != *clnt)
4084                                 goto out;
4085                         client = nfs4_negotiate_security(client, dir, name);
4086                         if (IS_ERR(client))
4087                                 return PTR_ERR(client);
4088
4089                         exception.retry = 1;
4090                         break;
4091                 default:
4092                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4093                 }
4094         } while (exception.retry);
4095
4096 out:
4097         if (err == 0)
4098                 *clnt = client;
4099         else if (client != *clnt)
4100                 rpc_shutdown_client(client);
4101
4102         return err;
4103 }
4104
4105 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
4106                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4107                             struct nfs4_label *label)
4108 {
4109         int status;
4110         struct rpc_clnt *client = NFS_CLIENT(dir);
4111
4112         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
4113         if (client != NFS_CLIENT(dir)) {
4114                 rpc_shutdown_client(client);
4115                 nfs_fixup_secinfo_attributes(fattr);
4116         }
4117         return status;
4118 }
4119
4120 struct rpc_clnt *
4121 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
4122                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4123 {
4124         struct rpc_clnt *client = NFS_CLIENT(dir);
4125         int status;
4126
4127         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
4128         if (status < 0)
4129                 return ERR_PTR(status);
4130         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4131 }
4132
4133 static int _nfs4_proc_lookupp(struct inode *inode,
4134                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4135                 struct nfs4_label *label)
4136 {
4137         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4138         struct nfs_server *server = NFS_SERVER(inode);
4139         int                    status;
4140         struct nfs4_lookupp_arg args = {
4141                 .bitmask = server->attr_bitmask,
4142                 .fh = NFS_FH(inode),
4143         };
4144         struct nfs4_lookupp_res res = {
4145                 .server = server,
4146                 .fattr = fattr,
4147                 .label = label,
4148                 .fh = fhandle,
4149         };
4150         struct rpc_message msg = {
4151                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4152                 .rpc_argp = &args,
4153                 .rpc_resp = &res,
4154         };
4155
4156         args.bitmask = nfs4_bitmask(server, label);
4157
4158         nfs_fattr_init(fattr);
4159
4160         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4161         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4162                                 &res.seq_res, 0);
4163         dprintk("NFS reply lookupp: %d\n", status);
4164         return status;
4165 }
4166
4167 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4168                              struct nfs_fattr *fattr, struct nfs4_label *label)
4169 {
4170         struct nfs4_exception exception = { };
4171         int err;
4172         do {
4173                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4174                 trace_nfs4_lookupp(inode, err);
4175                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4176                                 &exception);
4177         } while (exception.retry);
4178         return err;
4179 }
4180
4181 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4182 {
4183         struct nfs_server *server = NFS_SERVER(inode);
4184         struct nfs4_accessargs args = {
4185                 .fh = NFS_FH(inode),
4186                 .access = entry->mask,
4187         };
4188         struct nfs4_accessres res = {
4189                 .server = server,
4190         };
4191         struct rpc_message msg = {
4192                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4193                 .rpc_argp = &args,
4194                 .rpc_resp = &res,
4195                 .rpc_cred = entry->cred,
4196         };
4197         int status = 0;
4198
4199         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4200                 res.fattr = nfs_alloc_fattr();
4201                 if (res.fattr == NULL)
4202                         return -ENOMEM;
4203                 args.bitmask = server->cache_consistency_bitmask;
4204         }
4205
4206         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4207         if (!status) {
4208                 nfs_access_set_mask(entry, res.access);
4209                 if (res.fattr)
4210                         nfs_refresh_inode(inode, res.fattr);
4211         }
4212         nfs_free_fattr(res.fattr);
4213         return status;
4214 }
4215
4216 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4217 {
4218         struct nfs4_exception exception = { };
4219         int err;
4220         do {
4221                 err = _nfs4_proc_access(inode, entry);
4222                 trace_nfs4_access(inode, err);
4223                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4224                                 &exception);
4225         } while (exception.retry);
4226         return err;
4227 }
4228
4229 /*
4230  * TODO: For the time being, we don't try to get any attributes
4231  * along with any of the zero-copy operations READ, READDIR,
4232  * READLINK, WRITE.
4233  *
4234  * In the case of the first three, we want to put the GETATTR
4235  * after the read-type operation -- this is because it is hard
4236  * to predict the length of a GETATTR response in v4, and thus
4237  * align the READ data correctly.  This means that the GETATTR
4238  * may end up partially falling into the page cache, and we should
4239  * shift it into the 'tail' of the xdr_buf before processing.
4240  * To do this efficiently, we need to know the total length
4241  * of data received, which doesn't seem to be available outside
4242  * of the RPC layer.
4243  *
4244  * In the case of WRITE, we also want to put the GETATTR after
4245  * the operation -- in this case because we want to make sure
4246  * we get the post-operation mtime and size.
4247  *
4248  * Both of these changes to the XDR layer would in fact be quite
4249  * minor, but I decided to leave them for a subsequent patch.
4250  */
4251 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4252                 unsigned int pgbase, unsigned int pglen)
4253 {
4254         struct nfs4_readlink args = {
4255                 .fh       = NFS_FH(inode),
4256                 .pgbase   = pgbase,
4257                 .pglen    = pglen,
4258                 .pages    = &page,
4259         };
4260         struct nfs4_readlink_res res;
4261         struct rpc_message msg = {
4262                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4263                 .rpc_argp = &args,
4264                 .rpc_resp = &res,
4265         };
4266
4267         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4268 }
4269
4270 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4271                 unsigned int pgbase, unsigned int pglen)
4272 {
4273         struct nfs4_exception exception = { };
4274         int err;
4275         do {
4276                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4277                 trace_nfs4_readlink(inode, err);
4278                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4279                                 &exception);
4280         } while (exception.retry);
4281         return err;
4282 }
4283
4284 /*
4285  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4286  */
4287 static int
4288 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4289                  int flags)
4290 {
4291         struct nfs_server *server = NFS_SERVER(dir);
4292         struct nfs4_label l, *ilabel = NULL;
4293         struct nfs_open_context *ctx;
4294         struct nfs4_state *state;
4295         int status = 0;
4296
4297         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4298         if (IS_ERR(ctx))
4299                 return PTR_ERR(ctx);
4300
4301         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4302
4303         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4304                 sattr->ia_mode &= ~current_umask();
4305         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4306         if (IS_ERR(state)) {
4307                 status = PTR_ERR(state);
4308                 goto out;
4309         }
4310 out:
4311         nfs4_label_release_security(ilabel);
4312         put_nfs_open_context(ctx);
4313         return status;
4314 }
4315
4316 static int
4317 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4318 {
4319         struct nfs_server *server = NFS_SERVER(dir);
4320         struct nfs_removeargs args = {
4321                 .fh = NFS_FH(dir),
4322                 .name = *name,
4323         };
4324         struct nfs_removeres res = {
4325                 .server = server,
4326         };
4327         struct rpc_message msg = {
4328                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4329                 .rpc_argp = &args,
4330                 .rpc_resp = &res,
4331         };
4332         unsigned long timestamp = jiffies;
4333         int status;
4334
4335         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4336         if (status == 0) {
4337                 spin_lock(&dir->i_lock);
4338                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4339                 /* Removing a directory decrements nlink in the parent */
4340                 if (ftype == NF4DIR && dir->i_nlink > 2)
4341                         nfs4_dec_nlink_locked(dir);
4342                 spin_unlock(&dir->i_lock);
4343         }
4344         return status;
4345 }
4346
4347 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4348 {
4349         struct nfs4_exception exception = { };
4350         struct inode *inode = d_inode(dentry);
4351         int err;
4352
4353         if (inode) {
4354                 if (inode->i_nlink == 1)
4355                         nfs4_inode_return_delegation(inode);
4356                 else
4357                         nfs4_inode_make_writeable(inode);
4358         }
4359         do {
4360                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4361                 trace_nfs4_remove(dir, &dentry->d_name, err);
4362                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4363                                 &exception);
4364         } while (exception.retry);
4365         return err;
4366 }
4367
4368 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4369 {
4370         struct nfs4_exception exception = { };
4371         int err;
4372
4373         do {
4374                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4375                 trace_nfs4_remove(dir, name, err);
4376                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4377                                 &exception);
4378         } while (exception.retry);
4379         return err;
4380 }
4381
4382 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4383                 struct dentry *dentry,
4384                 struct inode *inode)
4385 {
4386         struct nfs_removeargs *args = msg->rpc_argp;
4387         struct nfs_removeres *res = msg->rpc_resp;
4388
4389         res->server = NFS_SB(dentry->d_sb);
4390         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4391         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4392
4393         nfs_fattr_init(res->dir_attr);
4394
4395         if (inode)
4396                 nfs4_inode_return_delegation(inode);
4397 }
4398
4399 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4400 {
4401         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4402                         &data->args.seq_args,
4403                         &data->res.seq_res,
4404                         task);
4405 }
4406
4407 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4408 {
4409         struct nfs_unlinkdata *data = task->tk_calldata;
4410         struct nfs_removeres *res = &data->res;
4411
4412         if (!nfs4_sequence_done(task, &res->seq_res))
4413                 return 0;
4414         if (nfs4_async_handle_error(task, res->server, NULL,
4415                                     &data->timeout) == -EAGAIN)
4416                 return 0;
4417         if (task->tk_status == 0)
4418                 update_changeattr(dir, &res->cinfo,
4419                                 res->dir_attr->time_start, 0);
4420         return 1;
4421 }
4422
4423 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4424                 struct dentry *old_dentry,
4425                 struct dentry *new_dentry)
4426 {
4427         struct nfs_renameargs *arg = msg->rpc_argp;
4428         struct nfs_renameres *res = msg->rpc_resp;
4429         struct inode *old_inode = d_inode(old_dentry);
4430         struct inode *new_inode = d_inode(new_dentry);
4431
4432         if (old_inode)
4433                 nfs4_inode_make_writeable(old_inode);
4434         if (new_inode)
4435                 nfs4_inode_return_delegation(new_inode);
4436         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4437         res->server = NFS_SB(old_dentry->d_sb);
4438         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4439 }
4440
4441 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4442 {
4443         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4444                         &data->args.seq_args,
4445                         &data->res.seq_res,
4446                         task);
4447 }
4448
4449 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4450                                  struct inode *new_dir)
4451 {
4452         struct nfs_renamedata *data = task->tk_calldata;
4453         struct nfs_renameres *res = &data->res;
4454
4455         if (!nfs4_sequence_done(task, &res->seq_res))
4456                 return 0;
4457         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4458                 return 0;
4459
4460         if (task->tk_status == 0) {
4461                 if (new_dir != old_dir) {
4462                         /* Note: If we moved a directory, nlink will change */
4463                         update_changeattr(old_dir, &res->old_cinfo,
4464                                         res->old_fattr->time_start,
4465                                         NFS_INO_INVALID_OTHER);
4466                         update_changeattr(new_dir, &res->new_cinfo,
4467                                         res->new_fattr->time_start,
4468                                         NFS_INO_INVALID_OTHER);
4469                 } else
4470                         update_changeattr(old_dir, &res->old_cinfo,
4471                                         res->old_fattr->time_start,
4472                                         0);
4473         }
4474         return 1;
4475 }
4476
4477 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4478 {
4479         struct nfs_server *server = NFS_SERVER(inode);
4480         __u32 bitmask[NFS4_BITMASK_SZ];
4481         struct nfs4_link_arg arg = {
4482                 .fh     = NFS_FH(inode),
4483                 .dir_fh = NFS_FH(dir),
4484                 .name   = name,
4485                 .bitmask = bitmask,
4486         };
4487         struct nfs4_link_res res = {
4488                 .server = server,
4489                 .label = NULL,
4490         };
4491         struct rpc_message msg = {
4492                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4493                 .rpc_argp = &arg,
4494                 .rpc_resp = &res,
4495         };
4496         int status = -ENOMEM;
4497
4498         res.fattr = nfs_alloc_fattr();
4499         if (res.fattr == NULL)
4500                 goto out;
4501
4502         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4503         if (IS_ERR(res.label)) {
4504                 status = PTR_ERR(res.label);
4505                 goto out;
4506         }
4507
4508         nfs4_inode_make_writeable(inode);
4509         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4510
4511         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4512         if (!status) {
4513                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4514                 status = nfs_post_op_update_inode(inode, res.fattr);
4515                 if (!status)
4516                         nfs_setsecurity(inode, res.fattr, res.label);
4517         }
4518
4519
4520         nfs4_label_free(res.label);
4521
4522 out:
4523         nfs_free_fattr(res.fattr);
4524         return status;
4525 }
4526
4527 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4528 {
4529         struct nfs4_exception exception = { };
4530         int err;
4531         do {
4532                 err = nfs4_handle_exception(NFS_SERVER(inode),
4533                                 _nfs4_proc_link(inode, dir, name),
4534                                 &exception);
4535         } while (exception.retry);
4536         return err;
4537 }
4538
4539 struct nfs4_createdata {
4540         struct rpc_message msg;
4541         struct nfs4_create_arg arg;
4542         struct nfs4_create_res res;
4543         struct nfs_fh fh;
4544         struct nfs_fattr fattr;
4545         struct nfs4_label *label;
4546 };
4547
4548 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4549                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4550 {
4551         struct nfs4_createdata *data;
4552
4553         data = kzalloc(sizeof(*data), GFP_KERNEL);
4554         if (data != NULL) {
4555                 struct nfs_server *server = NFS_SERVER(dir);
4556
4557                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4558                 if (IS_ERR(data->label))
4559                         goto out_free;
4560
4561                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4562                 data->msg.rpc_argp = &data->arg;
4563                 data->msg.rpc_resp = &data->res;
4564                 data->arg.dir_fh = NFS_FH(dir);
4565                 data->arg.server = server;
4566                 data->arg.name = name;
4567                 data->arg.attrs = sattr;
4568                 data->arg.ftype = ftype;
4569                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4570                 data->arg.umask = current_umask();
4571                 data->res.server = server;
4572                 data->res.fh = &data->fh;
4573                 data->res.fattr = &data->fattr;
4574                 data->res.label = data->label;
4575                 nfs_fattr_init(data->res.fattr);
4576         }
4577         return data;
4578 out_free:
4579         kfree(data);
4580         return NULL;
4581 }
4582
4583 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4584 {
4585         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4586                                     &data->arg.seq_args, &data->res.seq_res, 1);
4587         if (status == 0) {
4588                 spin_lock(&dir->i_lock);
4589                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4590                                 data->res.fattr->time_start, 0);
4591                 /* Creating a directory bumps nlink in the parent */
4592                 if (data->arg.ftype == NF4DIR)
4593                         nfs4_inc_nlink_locked(dir);
4594                 spin_unlock(&dir->i_lock);
4595                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4596         }
4597         return status;
4598 }
4599
4600 static void nfs4_free_createdata(struct nfs4_createdata *data)
4601 {
4602         nfs4_label_free(data->label);
4603         kfree(data);
4604 }
4605
4606 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4607                 struct page *page, unsigned int len, struct iattr *sattr,
4608                 struct nfs4_label *label)
4609 {
4610         struct nfs4_createdata *data;
4611         int status = -ENAMETOOLONG;
4612
4613         if (len > NFS4_MAXPATHLEN)
4614                 goto out;
4615
4616         status = -ENOMEM;
4617         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4618         if (data == NULL)
4619                 goto out;
4620
4621         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4622         data->arg.u.symlink.pages = &page;
4623         data->arg.u.symlink.len = len;
4624         data->arg.label = label;
4625
4626         status = nfs4_do_create(dir, dentry, data);
4627
4628         nfs4_free_createdata(data);
4629 out:
4630         return status;
4631 }
4632
4633 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4634                 struct page *page, unsigned int len, struct iattr *sattr)
4635 {
4636         struct nfs4_exception exception = { };
4637         struct nfs4_label l, *label = NULL;
4638         int err;
4639
4640         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4641
4642         do {
4643                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4644                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4645                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4646                                 &exception);
4647         } while (exception.retry);
4648
4649         nfs4_label_release_security(label);
4650         return err;
4651 }
4652
4653 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4654                 struct iattr *sattr, struct nfs4_label *label)
4655 {
4656         struct nfs4_createdata *data;
4657         int status = -ENOMEM;
4658
4659         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4660         if (data == NULL)
4661                 goto out;
4662
4663         data->arg.label = label;
4664         status = nfs4_do_create(dir, dentry, data);
4665
4666         nfs4_free_createdata(data);
4667 out:
4668         return status;
4669 }
4670
4671 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4672                 struct iattr *sattr)
4673 {
4674         struct nfs_server *server = NFS_SERVER(dir);
4675         struct nfs4_exception exception = { };
4676         struct nfs4_label l, *label = NULL;
4677         int err;
4678
4679         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4680
4681         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4682                 sattr->ia_mode &= ~current_umask();
4683         do {
4684                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4685                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4686                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4687                                 &exception);
4688         } while (exception.retry);
4689         nfs4_label_release_security(label);
4690
4691         return err;
4692 }
4693
4694 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4695                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4696 {
4697         struct inode            *dir = d_inode(dentry);
4698         struct nfs4_readdir_arg args = {
4699                 .fh = NFS_FH(dir),
4700                 .pages = pages,
4701                 .pgbase = 0,
4702                 .count = count,
4703                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4704                 .plus = plus,
4705         };
4706         struct nfs4_readdir_res res;
4707         struct rpc_message msg = {
4708                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4709                 .rpc_argp = &args,
4710                 .rpc_resp = &res,
4711                 .rpc_cred = cred,
4712         };
4713         int                     status;
4714
4715         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4716                         dentry,
4717                         (unsigned long long)cookie);
4718         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4719         res.pgbase = args.pgbase;
4720         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4721         if (status >= 0) {
4722                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4723                 status += args.pgbase;
4724         }
4725
4726         nfs_invalidate_atime(dir);
4727
4728         dprintk("%s: returns %d\n", __func__, status);
4729         return status;
4730 }
4731
4732 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4733                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4734 {
4735         struct nfs4_exception exception = { };
4736         int err;
4737         do {
4738                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4739                                 pages, count, plus);
4740                 trace_nfs4_readdir(d_inode(dentry), err);
4741                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4742                                 &exception);
4743         } while (exception.retry);
4744         return err;
4745 }
4746
4747 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4748                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4749 {
4750         struct nfs4_createdata *data;
4751         int mode = sattr->ia_mode;
4752         int status = -ENOMEM;
4753
4754         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4755         if (data == NULL)
4756                 goto out;
4757
4758         if (S_ISFIFO(mode))
4759                 data->arg.ftype = NF4FIFO;
4760         else if (S_ISBLK(mode)) {
4761                 data->arg.ftype = NF4BLK;
4762                 data->arg.u.device.specdata1 = MAJOR(rdev);
4763                 data->arg.u.device.specdata2 = MINOR(rdev);
4764         }
4765         else if (S_ISCHR(mode)) {
4766                 data->arg.ftype = NF4CHR;
4767                 data->arg.u.device.specdata1 = MAJOR(rdev);
4768                 data->arg.u.device.specdata2 = MINOR(rdev);
4769         } else if (!S_ISSOCK(mode)) {
4770                 status = -EINVAL;
4771                 goto out_free;
4772         }
4773
4774         data->arg.label = label;
4775         status = nfs4_do_create(dir, dentry, data);
4776 out_free:
4777         nfs4_free_createdata(data);
4778 out:
4779         return status;
4780 }
4781
4782 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4783                 struct iattr *sattr, dev_t rdev)
4784 {
4785         struct nfs_server *server = NFS_SERVER(dir);
4786         struct nfs4_exception exception = { };
4787         struct nfs4_label l, *label = NULL;
4788         int err;
4789
4790         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4791
4792         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4793                 sattr->ia_mode &= ~current_umask();
4794         do {
4795                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4796                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4797                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4798                                 &exception);
4799         } while (exception.retry);
4800
4801         nfs4_label_release_security(label);
4802
4803         return err;
4804 }
4805
4806 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4807                  struct nfs_fsstat *fsstat)
4808 {
4809         struct nfs4_statfs_arg args = {
4810                 .fh = fhandle,
4811                 .bitmask = server->attr_bitmask,
4812         };
4813         struct nfs4_statfs_res res = {
4814                 .fsstat = fsstat,
4815         };
4816         struct rpc_message msg = {
4817                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4818                 .rpc_argp = &args,
4819                 .rpc_resp = &res,
4820         };
4821
4822         nfs_fattr_init(fsstat->fattr);
4823         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4824 }
4825
4826 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4827 {
4828         struct nfs4_exception exception = { };
4829         int err;
4830         do {
4831                 err = nfs4_handle_exception(server,
4832                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4833                                 &exception);
4834         } while (exception.retry);
4835         return err;
4836 }
4837
4838 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4839                 struct nfs_fsinfo *fsinfo)
4840 {
4841         struct nfs4_fsinfo_arg args = {
4842                 .fh = fhandle,
4843                 .bitmask = server->attr_bitmask,
4844         };
4845         struct nfs4_fsinfo_res res = {
4846                 .fsinfo = fsinfo,
4847         };
4848         struct rpc_message msg = {
4849                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4850                 .rpc_argp = &args,
4851                 .rpc_resp = &res,
4852         };
4853
4854         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4855 }
4856
4857 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4858 {
4859         struct nfs4_exception exception = { };
4860         unsigned long now = jiffies;
4861         int err;
4862
4863         do {
4864                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4865                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4866                 if (err == 0) {
4867                         nfs4_set_lease_period(server->nfs_client,
4868                                         fsinfo->lease_time * HZ,
4869                                         now);
4870                         break;
4871                 }
4872                 err = nfs4_handle_exception(server, err, &exception);
4873         } while (exception.retry);
4874         return err;
4875 }
4876
4877 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4878 {
4879         int error;
4880
4881         nfs_fattr_init(fsinfo->fattr);
4882         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4883         if (error == 0) {
4884                 /* block layout checks this! */
4885                 server->pnfs_blksize = fsinfo->blksize;
4886                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4887         }
4888
4889         return error;
4890 }
4891
4892 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4893                 struct nfs_pathconf *pathconf)
4894 {
4895         struct nfs4_pathconf_arg args = {
4896                 .fh = fhandle,
4897                 .bitmask = server->attr_bitmask,
4898         };
4899         struct nfs4_pathconf_res res = {
4900                 .pathconf = pathconf,
4901         };
4902         struct rpc_message msg = {
4903                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4904                 .rpc_argp = &args,
4905                 .rpc_resp = &res,
4906         };
4907
4908         /* None of the pathconf attributes are mandatory to implement */
4909         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4910                 memset(pathconf, 0, sizeof(*pathconf));
4911                 return 0;
4912         }
4913
4914         nfs_fattr_init(pathconf->fattr);
4915         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4916 }
4917
4918 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4919                 struct nfs_pathconf *pathconf)
4920 {
4921         struct nfs4_exception exception = { };
4922         int err;
4923
4924         do {
4925                 err = nfs4_handle_exception(server,
4926                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4927                                 &exception);
4928         } while (exception.retry);
4929         return err;
4930 }
4931
4932 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4933                 const struct nfs_open_context *ctx,
4934                 const struct nfs_lock_context *l_ctx,
4935                 fmode_t fmode)
4936 {
4937         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4938 }
4939 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4940
4941 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4942                 const struct nfs_open_context *ctx,
4943                 const struct nfs_lock_context *l_ctx,
4944                 fmode_t fmode)
4945 {
4946         nfs4_stateid current_stateid;
4947
4948         /* If the current stateid represents a lost lock, then exit */
4949         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4950                 return true;
4951         return nfs4_stateid_match(stateid, &current_stateid);
4952 }
4953
4954 static bool nfs4_error_stateid_expired(int err)
4955 {
4956         switch (err) {
4957         case -NFS4ERR_DELEG_REVOKED:
4958         case -NFS4ERR_ADMIN_REVOKED:
4959         case -NFS4ERR_BAD_STATEID:
4960         case -NFS4ERR_STALE_STATEID:
4961         case -NFS4ERR_OLD_STATEID:
4962         case -NFS4ERR_OPENMODE:
4963         case -NFS4ERR_EXPIRED:
4964                 return true;
4965         }
4966         return false;
4967 }
4968
4969 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4970 {
4971         struct nfs_server *server = NFS_SERVER(hdr->inode);
4972
4973         trace_nfs4_read(hdr, task->tk_status);
4974         if (task->tk_status < 0) {
4975                 struct nfs4_exception exception = {
4976                         .inode = hdr->inode,
4977                         .state = hdr->args.context->state,
4978                         .stateid = &hdr->args.stateid,
4979                 };
4980                 task->tk_status = nfs4_async_handle_exception(task,
4981                                 server, task->tk_status, &exception);
4982                 if (exception.retry) {
4983                         rpc_restart_call_prepare(task);
4984                         return -EAGAIN;
4985                 }
4986         }
4987
4988         if (task->tk_status > 0)
4989                 renew_lease(server, hdr->timestamp);
4990         return 0;
4991 }
4992
4993 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4994                 struct nfs_pgio_args *args)
4995 {
4996
4997         if (!nfs4_error_stateid_expired(task->tk_status) ||
4998                 nfs4_stateid_is_current(&args->stateid,
4999                                 args->context,
5000                                 args->lock_context,
5001                                 FMODE_READ))
5002                 return false;
5003         rpc_restart_call_prepare(task);
5004         return true;
5005 }
5006
5007 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5008 {
5009
5010         dprintk("--> %s\n", __func__);
5011
5012         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5013                 return -EAGAIN;
5014         if (nfs4_read_stateid_changed(task, &hdr->args))
5015                 return -EAGAIN;
5016         if (task->tk_status > 0)
5017                 nfs_invalidate_atime(hdr->inode);
5018         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5019                                     nfs4_read_done_cb(task, hdr);
5020 }
5021
5022 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5023                                  struct rpc_message *msg)
5024 {
5025         hdr->timestamp   = jiffies;
5026         if (!hdr->pgio_done_cb)
5027                 hdr->pgio_done_cb = nfs4_read_done_cb;
5028         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5029         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5030 }
5031
5032 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5033                                       struct nfs_pgio_header *hdr)
5034 {
5035         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5036                         &hdr->args.seq_args,
5037                         &hdr->res.seq_res,
5038                         task))
5039                 return 0;
5040         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5041                                 hdr->args.lock_context,
5042                                 hdr->rw_mode) == -EIO)
5043                 return -EIO;
5044         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5045                 return -EIO;
5046         return 0;
5047 }
5048
5049 static int nfs4_write_done_cb(struct rpc_task *task,
5050                               struct nfs_pgio_header *hdr)
5051 {
5052         struct inode *inode = hdr->inode;
5053
5054         trace_nfs4_write(hdr, task->tk_status);
5055         if (task->tk_status < 0) {
5056                 struct nfs4_exception exception = {
5057                         .inode = hdr->inode,
5058                         .state = hdr->args.context->state,
5059                         .stateid = &hdr->args.stateid,
5060                 };
5061                 task->tk_status = nfs4_async_handle_exception(task,
5062                                 NFS_SERVER(inode), task->tk_status,
5063                                 &exception);
5064                 if (exception.retry) {
5065                         rpc_restart_call_prepare(task);
5066                         return -EAGAIN;
5067                 }
5068         }
5069         if (task->tk_status >= 0) {
5070                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5071                 nfs_writeback_update_inode(hdr);
5072         }
5073         return 0;
5074 }
5075
5076 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5077                 struct nfs_pgio_args *args)
5078 {
5079
5080         if (!nfs4_error_stateid_expired(task->tk_status) ||
5081                 nfs4_stateid_is_current(&args->stateid,
5082                                 args->context,
5083                                 args->lock_context,
5084                                 FMODE_WRITE))
5085                 return false;
5086         rpc_restart_call_prepare(task);
5087         return true;
5088 }
5089
5090 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5091 {
5092         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5093                 return -EAGAIN;
5094         if (nfs4_write_stateid_changed(task, &hdr->args))
5095                 return -EAGAIN;
5096         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5097                 nfs4_write_done_cb(task, hdr);
5098 }
5099
5100 static
5101 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5102 {
5103         /* Don't request attributes for pNFS or O_DIRECT writes */
5104         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5105                 return false;
5106         /* Otherwise, request attributes if and only if we don't hold
5107          * a delegation
5108          */
5109         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5110 }
5111
5112 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5113                                   struct rpc_message *msg,
5114                                   struct rpc_clnt **clnt)
5115 {
5116         struct nfs_server *server = NFS_SERVER(hdr->inode);
5117
5118         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5119                 hdr->args.bitmask = NULL;
5120                 hdr->res.fattr = NULL;
5121         } else
5122                 hdr->args.bitmask = server->cache_consistency_bitmask;
5123
5124         if (!hdr->pgio_done_cb)
5125                 hdr->pgio_done_cb = nfs4_write_done_cb;
5126         hdr->res.server = server;
5127         hdr->timestamp   = jiffies;
5128
5129         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5130         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1, 0);
5131         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5132 }
5133
5134 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5135 {
5136         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5137                         &data->args.seq_args,
5138                         &data->res.seq_res,
5139                         task);
5140 }
5141
5142 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5143 {
5144         struct inode *inode = data->inode;
5145
5146         trace_nfs4_commit(data, task->tk_status);
5147         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5148                                     NULL, NULL) == -EAGAIN) {
5149                 rpc_restart_call_prepare(task);
5150                 return -EAGAIN;
5151         }
5152         return 0;
5153 }
5154
5155 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5156 {
5157         if (!nfs4_sequence_done(task, &data->res.seq_res))
5158                 return -EAGAIN;
5159         return data->commit_done_cb(task, data);
5160 }
5161
5162 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5163                                    struct rpc_clnt **clnt)
5164 {
5165         struct nfs_server *server = NFS_SERVER(data->inode);
5166
5167         if (data->commit_done_cb == NULL)
5168                 data->commit_done_cb = nfs4_commit_done_cb;
5169         data->res.server = server;
5170         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5171         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5172         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5173 }
5174
5175 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5176                                 struct nfs_commitres *res)
5177 {
5178         struct inode *dst_inode = file_inode(dst);
5179         struct nfs_server *server = NFS_SERVER(dst_inode);
5180         struct rpc_message msg = {
5181                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5182                 .rpc_argp = args,
5183                 .rpc_resp = res,
5184         };
5185
5186         args->fh = NFS_FH(dst_inode);
5187         return nfs4_call_sync(server->client, server, &msg,
5188                         &args->seq_args, &res->seq_res, 1);
5189 }
5190
5191 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5192 {
5193         struct nfs_commitargs args = {
5194                 .offset = offset,
5195                 .count = count,
5196         };
5197         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5198         struct nfs4_exception exception = { };
5199         int status;
5200
5201         do {
5202                 status = _nfs4_proc_commit(dst, &args, res);
5203                 status = nfs4_handle_exception(dst_server, status, &exception);
5204         } while (exception.retry);
5205
5206         return status;
5207 }
5208
5209 struct nfs4_renewdata {
5210         struct nfs_client       *client;
5211         unsigned long           timestamp;
5212 };
5213
5214 /*
5215  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5216  * standalone procedure for queueing an asynchronous RENEW.
5217  */
5218 static void nfs4_renew_release(void *calldata)
5219 {
5220         struct nfs4_renewdata *data = calldata;
5221         struct nfs_client *clp = data->client;
5222
5223         if (refcount_read(&clp->cl_count) > 1)
5224                 nfs4_schedule_state_renewal(clp);
5225         nfs_put_client(clp);
5226         kfree(data);
5227 }
5228
5229 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5230 {
5231         struct nfs4_renewdata *data = calldata;
5232         struct nfs_client *clp = data->client;
5233         unsigned long timestamp = data->timestamp;
5234
5235         trace_nfs4_renew_async(clp, task->tk_status);
5236         switch (task->tk_status) {
5237         case 0:
5238                 break;
5239         case -NFS4ERR_LEASE_MOVED:
5240                 nfs4_schedule_lease_moved_recovery(clp);
5241                 break;
5242         default:
5243                 /* Unless we're shutting down, schedule state recovery! */
5244                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5245                         return;
5246                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5247                         nfs4_schedule_lease_recovery(clp);
5248                         return;
5249                 }
5250                 nfs4_schedule_path_down_recovery(clp);
5251         }
5252         do_renew_lease(clp, timestamp);
5253 }
5254
5255 static const struct rpc_call_ops nfs4_renew_ops = {
5256         .rpc_call_done = nfs4_renew_done,
5257         .rpc_release = nfs4_renew_release,
5258 };
5259
5260 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5261 {
5262         struct rpc_message msg = {
5263                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5264                 .rpc_argp       = clp,
5265                 .rpc_cred       = cred,
5266         };
5267         struct nfs4_renewdata *data;
5268
5269         if (renew_flags == 0)
5270                 return 0;
5271         if (!refcount_inc_not_zero(&clp->cl_count))
5272                 return -EIO;
5273         data = kmalloc(sizeof(*data), GFP_NOFS);
5274         if (data == NULL) {
5275                 nfs_put_client(clp);
5276                 return -ENOMEM;
5277         }
5278         data->client = clp;
5279         data->timestamp = jiffies;
5280         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5281                         &nfs4_renew_ops, data);
5282 }
5283
5284 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
5285 {
5286         struct rpc_message msg = {
5287                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5288                 .rpc_argp       = clp,
5289                 .rpc_cred       = cred,
5290         };
5291         unsigned long now = jiffies;
5292         int status;
5293
5294         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5295         if (status < 0)
5296                 return status;
5297         do_renew_lease(clp, now);
5298         return 0;
5299 }
5300
5301 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5302 {
5303         return server->caps & NFS_CAP_ACLS;
5304 }
5305
5306 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5307  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5308  * the stack.
5309  */
5310 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5311
5312 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5313                 struct page **pages)
5314 {
5315         struct page *newpage, **spages;
5316         int rc = 0;
5317         size_t len;
5318         spages = pages;
5319
5320         do {
5321                 len = min_t(size_t, PAGE_SIZE, buflen);
5322                 newpage = alloc_page(GFP_KERNEL);
5323
5324                 if (newpage == NULL)
5325                         goto unwind;
5326                 memcpy(page_address(newpage), buf, len);
5327                 buf += len;
5328                 buflen -= len;
5329                 *pages++ = newpage;
5330                 rc++;
5331         } while (buflen != 0);
5332
5333         return rc;
5334
5335 unwind:
5336         for(; rc > 0; rc--)
5337                 __free_page(spages[rc-1]);
5338         return -ENOMEM;
5339 }
5340
5341 struct nfs4_cached_acl {
5342         int cached;
5343         size_t len;
5344         char data[0];
5345 };
5346
5347 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5348 {
5349         struct nfs_inode *nfsi = NFS_I(inode);
5350
5351         spin_lock(&inode->i_lock);
5352         kfree(nfsi->nfs4_acl);
5353         nfsi->nfs4_acl = acl;
5354         spin_unlock(&inode->i_lock);
5355 }
5356
5357 static void nfs4_zap_acl_attr(struct inode *inode)
5358 {
5359         nfs4_set_cached_acl(inode, NULL);
5360 }
5361
5362 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5363 {
5364         struct nfs_inode *nfsi = NFS_I(inode);
5365         struct nfs4_cached_acl *acl;
5366         int ret = -ENOENT;
5367
5368         spin_lock(&inode->i_lock);
5369         acl = nfsi->nfs4_acl;
5370         if (acl == NULL)
5371                 goto out;
5372         if (buf == NULL) /* user is just asking for length */
5373                 goto out_len;
5374         if (acl->cached == 0)
5375                 goto out;
5376         ret = -ERANGE; /* see getxattr(2) man page */
5377         if (acl->len > buflen)
5378                 goto out;
5379         memcpy(buf, acl->data, acl->len);
5380 out_len:
5381         ret = acl->len;
5382 out:
5383         spin_unlock(&inode->i_lock);
5384         return ret;
5385 }
5386
5387 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5388 {
5389         struct nfs4_cached_acl *acl;
5390         size_t buflen = sizeof(*acl) + acl_len;
5391
5392         if (buflen <= PAGE_SIZE) {
5393                 acl = kmalloc(buflen, GFP_KERNEL);
5394                 if (acl == NULL)
5395                         goto out;
5396                 acl->cached = 1;
5397                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5398         } else {
5399                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5400                 if (acl == NULL)
5401                         goto out;
5402                 acl->cached = 0;
5403         }
5404         acl->len = acl_len;
5405 out:
5406         nfs4_set_cached_acl(inode, acl);
5407 }
5408
5409 /*
5410  * The getxattr API returns the required buffer length when called with a
5411  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5412  * the required buf.  On a NULL buf, we send a page of data to the server
5413  * guessing that the ACL request can be serviced by a page. If so, we cache
5414  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5415  * the cache. If not so, we throw away the page, and cache the required
5416  * length. The next getxattr call will then produce another round trip to
5417  * the server, this time with the input buf of the required size.
5418  */
5419 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5420 {
5421         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5422         struct nfs_getaclargs args = {
5423                 .fh = NFS_FH(inode),
5424                 .acl_pages = pages,
5425                 .acl_len = buflen,
5426         };
5427         struct nfs_getaclres res = {
5428                 .acl_len = buflen,
5429         };
5430         struct rpc_message msg = {
5431                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5432                 .rpc_argp = &args,
5433                 .rpc_resp = &res,
5434         };
5435         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5436         int ret = -ENOMEM, i;
5437
5438         if (npages > ARRAY_SIZE(pages))
5439                 return -ERANGE;
5440
5441         for (i = 0; i < npages; i++) {
5442                 pages[i] = alloc_page(GFP_KERNEL);
5443                 if (!pages[i])
5444                         goto out_free;
5445         }
5446
5447         /* for decoding across pages */
5448         res.acl_scratch = alloc_page(GFP_KERNEL);
5449         if (!res.acl_scratch)
5450                 goto out_free;
5451
5452         args.acl_len = npages * PAGE_SIZE;
5453
5454         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5455                 __func__, buf, buflen, npages, args.acl_len);
5456         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5457                              &msg, &args.seq_args, &res.seq_res, 0);
5458         if (ret)
5459                 goto out_free;
5460
5461         /* Handle the case where the passed-in buffer is too short */
5462         if (res.acl_flags & NFS4_ACL_TRUNC) {
5463                 /* Did the user only issue a request for the acl length? */
5464                 if (buf == NULL)
5465                         goto out_ok;
5466                 ret = -ERANGE;
5467                 goto out_free;
5468         }
5469         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5470         if (buf) {
5471                 if (res.acl_len > buflen) {
5472                         ret = -ERANGE;
5473                         goto out_free;
5474                 }
5475                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5476         }
5477 out_ok:
5478         ret = res.acl_len;
5479 out_free:
5480         for (i = 0; i < npages; i++)
5481                 if (pages[i])
5482                         __free_page(pages[i]);
5483         if (res.acl_scratch)
5484                 __free_page(res.acl_scratch);
5485         return ret;
5486 }
5487
5488 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5489 {
5490         struct nfs4_exception exception = { };
5491         ssize_t ret;
5492         do {
5493                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5494                 trace_nfs4_get_acl(inode, ret);
5495                 if (ret >= 0)
5496                         break;
5497                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5498         } while (exception.retry);
5499         return ret;
5500 }
5501
5502 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5503 {
5504         struct nfs_server *server = NFS_SERVER(inode);
5505         int ret;
5506
5507         if (!nfs4_server_supports_acls(server))
5508                 return -EOPNOTSUPP;
5509         ret = nfs_revalidate_inode(server, inode);
5510         if (ret < 0)
5511                 return ret;
5512         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5513                 nfs_zap_acl_cache(inode);
5514         ret = nfs4_read_cached_acl(inode, buf, buflen);
5515         if (ret != -ENOENT)
5516                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5517                  * but no cached acl data, just the acl length */
5518                 return ret;
5519         return nfs4_get_acl_uncached(inode, buf, buflen);
5520 }
5521
5522 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5523 {
5524         struct nfs_server *server = NFS_SERVER(inode);
5525         struct page *pages[NFS4ACL_MAXPAGES];
5526         struct nfs_setaclargs arg = {
5527                 .fh             = NFS_FH(inode),
5528                 .acl_pages      = pages,
5529                 .acl_len        = buflen,
5530         };
5531         struct nfs_setaclres res;
5532         struct rpc_message msg = {
5533                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5534                 .rpc_argp       = &arg,
5535                 .rpc_resp       = &res,
5536         };
5537         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5538         int ret, i;
5539
5540         if (!nfs4_server_supports_acls(server))
5541                 return -EOPNOTSUPP;
5542         if (npages > ARRAY_SIZE(pages))
5543                 return -ERANGE;
5544         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5545         if (i < 0)
5546                 return i;
5547         nfs4_inode_make_writeable(inode);
5548         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5549
5550         /*
5551          * Free each page after tx, so the only ref left is
5552          * held by the network stack
5553          */
5554         for (; i > 0; i--)
5555                 put_page(pages[i-1]);
5556
5557         /*
5558          * Acl update can result in inode attribute update.
5559          * so mark the attribute cache invalid.
5560          */
5561         spin_lock(&inode->i_lock);
5562         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5563                 | NFS_INO_INVALID_CTIME
5564                 | NFS_INO_REVAL_FORCED;
5565         spin_unlock(&inode->i_lock);
5566         nfs_access_zap_cache(inode);
5567         nfs_zap_acl_cache(inode);
5568         return ret;
5569 }
5570
5571 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5572 {
5573         struct nfs4_exception exception = { };
5574         int err;
5575         do {
5576                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5577                 trace_nfs4_set_acl(inode, err);
5578                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5579                                 &exception);
5580         } while (exception.retry);
5581         return err;
5582 }
5583
5584 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5585 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5586                                         size_t buflen)
5587 {
5588         struct nfs_server *server = NFS_SERVER(inode);
5589         struct nfs_fattr fattr;
5590         struct nfs4_label label = {0, 0, buflen, buf};
5591
5592         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5593         struct nfs4_getattr_arg arg = {
5594                 .fh             = NFS_FH(inode),
5595                 .bitmask        = bitmask,
5596         };
5597         struct nfs4_getattr_res res = {
5598                 .fattr          = &fattr,
5599                 .label          = &label,
5600                 .server         = server,
5601         };
5602         struct rpc_message msg = {
5603                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5604                 .rpc_argp       = &arg,
5605                 .rpc_resp       = &res,
5606         };
5607         int ret;
5608
5609         nfs_fattr_init(&fattr);
5610
5611         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5612         if (ret)
5613                 return ret;
5614         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5615                 return -ENOENT;
5616         if (buflen < label.len)
5617                 return -ERANGE;
5618         return 0;
5619 }
5620
5621 static int nfs4_get_security_label(struct inode *inode, void *buf,
5622                                         size_t buflen)
5623 {
5624         struct nfs4_exception exception = { };
5625         int err;
5626
5627         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5628                 return -EOPNOTSUPP;
5629
5630         do {
5631                 err = _nfs4_get_security_label(inode, buf, buflen);
5632                 trace_nfs4_get_security_label(inode, err);
5633                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5634                                 &exception);
5635         } while (exception.retry);
5636         return err;
5637 }
5638
5639 static int _nfs4_do_set_security_label(struct inode *inode,
5640                 struct nfs4_label *ilabel,
5641                 struct nfs_fattr *fattr,
5642                 struct nfs4_label *olabel)
5643 {
5644
5645         struct iattr sattr = {0};
5646         struct nfs_server *server = NFS_SERVER(inode);
5647         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5648         struct nfs_setattrargs arg = {
5649                 .fh             = NFS_FH(inode),
5650                 .iap            = &sattr,
5651                 .server         = server,
5652                 .bitmask        = bitmask,
5653                 .label          = ilabel,
5654         };
5655         struct nfs_setattrres res = {
5656                 .fattr          = fattr,
5657                 .label          = olabel,
5658                 .server         = server,
5659         };
5660         struct rpc_message msg = {
5661                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5662                 .rpc_argp       = &arg,
5663                 .rpc_resp       = &res,
5664         };
5665         int status;
5666
5667         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5668
5669         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5670         if (status)
5671                 dprintk("%s failed: %d\n", __func__, status);
5672
5673         return status;
5674 }
5675
5676 static int nfs4_do_set_security_label(struct inode *inode,
5677                 struct nfs4_label *ilabel,
5678                 struct nfs_fattr *fattr,
5679                 struct nfs4_label *olabel)
5680 {
5681         struct nfs4_exception exception = { };
5682         int err;
5683
5684         do {
5685                 err = _nfs4_do_set_security_label(inode, ilabel,
5686                                 fattr, olabel);
5687                 trace_nfs4_set_security_label(inode, err);
5688                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5689                                 &exception);
5690         } while (exception.retry);
5691         return err;
5692 }
5693
5694 static int
5695 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5696 {
5697         struct nfs4_label ilabel, *olabel = NULL;
5698         struct nfs_fattr fattr;
5699         struct rpc_cred *cred;
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         cred = rpc_lookup_cred();
5713         if (IS_ERR(cred))
5714                 return PTR_ERR(cred);
5715
5716         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5717         if (IS_ERR(olabel)) {
5718                 status = -PTR_ERR(olabel);
5719                 goto out;
5720         }
5721
5722         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5723         if (status == 0)
5724                 nfs_setsecurity(inode, &fattr, olabel);
5725
5726         nfs4_label_free(olabel);
5727 out:
5728         put_rpccred(cred);
5729         return status;
5730 }
5731 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5732
5733
5734 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5735                                     nfs4_verifier *bootverf)
5736 {
5737         __be32 verf[2];
5738
5739         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5740                 /* An impossible timestamp guarantees this value
5741                  * will never match a generated boot time. */
5742                 verf[0] = cpu_to_be32(U32_MAX);
5743                 verf[1] = cpu_to_be32(U32_MAX);
5744         } else {
5745                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5746                 u64 ns = ktime_to_ns(nn->boot_time);
5747
5748                 verf[0] = cpu_to_be32(ns >> 32);
5749                 verf[1] = cpu_to_be32(ns);
5750         }
5751         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5752 }
5753
5754 static int
5755 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5756 {
5757         size_t len;
5758         char *str;
5759
5760         if (clp->cl_owner_id != NULL)
5761                 return 0;
5762
5763         rcu_read_lock();
5764         len = 14 +
5765                 strlen(clp->cl_rpcclient->cl_nodename) +
5766                 1 +
5767                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5768                 1;
5769         rcu_read_unlock();
5770         if (nfs4_client_id_uniquifier[0] != '\0')
5771                 len += strlen(nfs4_client_id_uniquifier) + 1;
5772         if (len > NFS4_OPAQUE_LIMIT + 1)
5773                 return -EINVAL;
5774
5775         /*
5776          * Since this string is allocated at mount time, and held until the
5777          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5778          * about a memory-reclaim deadlock.
5779          */
5780         str = kmalloc(len, GFP_KERNEL);
5781         if (!str)
5782                 return -ENOMEM;
5783
5784         rcu_read_lock();
5785         if (nfs4_client_id_uniquifier[0] != '\0')
5786                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5787                           clp->cl_rpcclient->cl_nodename,
5788                           nfs4_client_id_uniquifier,
5789                           rpc_peeraddr2str(clp->cl_rpcclient,
5790                                            RPC_DISPLAY_ADDR));
5791         else
5792                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
5793                           clp->cl_rpcclient->cl_nodename,
5794                           rpc_peeraddr2str(clp->cl_rpcclient,
5795                                            RPC_DISPLAY_ADDR));
5796         rcu_read_unlock();
5797
5798         clp->cl_owner_id = str;
5799         return 0;
5800 }
5801
5802 static int
5803 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5804 {
5805         size_t len;
5806         char *str;
5807
5808         len = 10 + 10 + 1 + 10 + 1 +
5809                 strlen(nfs4_client_id_uniquifier) + 1 +
5810                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5811
5812         if (len > NFS4_OPAQUE_LIMIT + 1)
5813                 return -EINVAL;
5814
5815         /*
5816          * Since this string is allocated at mount time, and held until the
5817          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5818          * about a memory-reclaim deadlock.
5819          */
5820         str = kmalloc(len, GFP_KERNEL);
5821         if (!str)
5822                 return -ENOMEM;
5823
5824         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5825                         clp->rpc_ops->version, clp->cl_minorversion,
5826                         nfs4_client_id_uniquifier,
5827                         clp->cl_rpcclient->cl_nodename);
5828         clp->cl_owner_id = str;
5829         return 0;
5830 }
5831
5832 static int
5833 nfs4_init_uniform_client_string(struct nfs_client *clp)
5834 {
5835         size_t len;
5836         char *str;
5837
5838         if (clp->cl_owner_id != NULL)
5839                 return 0;
5840
5841         if (nfs4_client_id_uniquifier[0] != '\0')
5842                 return nfs4_init_uniquifier_client_string(clp);
5843
5844         len = 10 + 10 + 1 + 10 + 1 +
5845                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5846
5847         if (len > NFS4_OPAQUE_LIMIT + 1)
5848                 return -EINVAL;
5849
5850         /*
5851          * Since this string is allocated at mount time, and held until the
5852          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5853          * about a memory-reclaim deadlock.
5854          */
5855         str = kmalloc(len, GFP_KERNEL);
5856         if (!str)
5857                 return -ENOMEM;
5858
5859         scnprintf(str, len, "Linux NFSv%u.%u %s",
5860                         clp->rpc_ops->version, clp->cl_minorversion,
5861                         clp->cl_rpcclient->cl_nodename);
5862         clp->cl_owner_id = str;
5863         return 0;
5864 }
5865
5866 /*
5867  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5868  * services.  Advertise one based on the address family of the
5869  * clientaddr.
5870  */
5871 static unsigned int
5872 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5873 {
5874         if (strchr(clp->cl_ipaddr, ':') != NULL)
5875                 return scnprintf(buf, len, "tcp6");
5876         else
5877                 return scnprintf(buf, len, "tcp");
5878 }
5879
5880 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5881 {
5882         struct nfs4_setclientid *sc = calldata;
5883
5884         if (task->tk_status == 0)
5885                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5886 }
5887
5888 static const struct rpc_call_ops nfs4_setclientid_ops = {
5889         .rpc_call_done = nfs4_setclientid_done,
5890 };
5891
5892 /**
5893  * nfs4_proc_setclientid - Negotiate client ID
5894  * @clp: state data structure
5895  * @program: RPC program for NFSv4 callback service
5896  * @port: IP port number for NFS4 callback service
5897  * @cred: RPC credential to use for this call
5898  * @res: where to place the result
5899  *
5900  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5901  */
5902 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5903                 unsigned short port, struct rpc_cred *cred,
5904                 struct nfs4_setclientid_res *res)
5905 {
5906         nfs4_verifier sc_verifier;
5907         struct nfs4_setclientid setclientid = {
5908                 .sc_verifier = &sc_verifier,
5909                 .sc_prog = program,
5910                 .sc_clnt = clp,
5911         };
5912         struct rpc_message msg = {
5913                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5914                 .rpc_argp = &setclientid,
5915                 .rpc_resp = res,
5916                 .rpc_cred = cred,
5917         };
5918         struct rpc_task *task;
5919         struct rpc_task_setup task_setup_data = {
5920                 .rpc_client = clp->cl_rpcclient,
5921                 .rpc_message = &msg,
5922                 .callback_ops = &nfs4_setclientid_ops,
5923                 .callback_data = &setclientid,
5924                 .flags = RPC_TASK_TIMEOUT,
5925         };
5926         int status;
5927
5928         /* nfs_client_id4 */
5929         nfs4_init_boot_verifier(clp, &sc_verifier);
5930
5931         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5932                 status = nfs4_init_uniform_client_string(clp);
5933         else
5934                 status = nfs4_init_nonuniform_client_string(clp);
5935
5936         if (status)
5937                 goto out;
5938
5939         /* cb_client4 */
5940         setclientid.sc_netid_len =
5941                                 nfs4_init_callback_netid(clp,
5942                                                 setclientid.sc_netid,
5943                                                 sizeof(setclientid.sc_netid));
5944         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5945                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5946                                 clp->cl_ipaddr, port >> 8, port & 255);
5947
5948         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5949                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5950                 clp->cl_owner_id);
5951         task = rpc_run_task(&task_setup_data);
5952         if (IS_ERR(task)) {
5953                 status = PTR_ERR(task);
5954                 goto out;
5955         }
5956         status = task->tk_status;
5957         if (setclientid.sc_cred) {
5958                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5959                 put_rpccred(setclientid.sc_cred);
5960         }
5961         rpc_put_task(task);
5962 out:
5963         trace_nfs4_setclientid(clp, status);
5964         dprintk("NFS reply setclientid: %d\n", status);
5965         return status;
5966 }
5967
5968 /**
5969  * nfs4_proc_setclientid_confirm - Confirm client ID
5970  * @clp: state data structure
5971  * @res: result of a previous SETCLIENTID
5972  * @cred: RPC credential to use for this call
5973  *
5974  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5975  */
5976 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5977                 struct nfs4_setclientid_res *arg,
5978                 struct rpc_cred *cred)
5979 {
5980         struct rpc_message msg = {
5981                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5982                 .rpc_argp = arg,
5983                 .rpc_cred = cred,
5984         };
5985         int status;
5986
5987         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5988                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5989                 clp->cl_clientid);
5990         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5991         trace_nfs4_setclientid_confirm(clp, status);
5992         dprintk("NFS reply setclientid_confirm: %d\n", status);
5993         return status;
5994 }
5995
5996 struct nfs4_delegreturndata {
5997         struct nfs4_delegreturnargs args;
5998         struct nfs4_delegreturnres res;
5999         struct nfs_fh fh;
6000         nfs4_stateid stateid;
6001         unsigned long timestamp;
6002         struct {
6003                 struct nfs4_layoutreturn_args arg;
6004                 struct nfs4_layoutreturn_res res;
6005                 struct nfs4_xdr_opaque_data ld_private;
6006                 u32 roc_barrier;
6007                 bool roc;
6008         } lr;
6009         struct nfs_fattr fattr;
6010         int rpc_status;
6011         struct inode *inode;
6012 };
6013
6014 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6015 {
6016         struct nfs4_delegreturndata *data = calldata;
6017         struct nfs4_exception exception = {
6018                 .inode = data->inode,
6019                 .stateid = &data->stateid,
6020         };
6021
6022         if (!nfs4_sequence_done(task, &data->res.seq_res))
6023                 return;
6024
6025         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6026
6027         /* Handle Layoutreturn errors */
6028         if (data->args.lr_args && task->tk_status != 0) {
6029                 switch(data->res.lr_ret) {
6030                 default:
6031                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6032                         break;
6033                 case 0:
6034                         data->args.lr_args = NULL;
6035                         data->res.lr_res = NULL;
6036                         break;
6037                 case -NFS4ERR_OLD_STATEID:
6038                         if (nfs4_layoutreturn_refresh_stateid(&data->args.lr_args->stateid,
6039                                                 &data->args.lr_args->range,
6040                                                 data->inode))
6041                                 goto lr_restart;
6042                         /* Fallthrough */
6043                 case -NFS4ERR_ADMIN_REVOKED:
6044                 case -NFS4ERR_DELEG_REVOKED:
6045                 case -NFS4ERR_EXPIRED:
6046                 case -NFS4ERR_BAD_STATEID:
6047                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
6048                 case -NFS4ERR_WRONG_CRED:
6049                         data->args.lr_args = NULL;
6050                         data->res.lr_res = NULL;
6051                         goto lr_restart;
6052                 }
6053         }
6054
6055         switch (task->tk_status) {
6056         case 0:
6057                 renew_lease(data->res.server, data->timestamp);
6058                 break;
6059         case -NFS4ERR_ADMIN_REVOKED:
6060         case -NFS4ERR_DELEG_REVOKED:
6061         case -NFS4ERR_EXPIRED:
6062                 nfs4_free_revoked_stateid(data->res.server,
6063                                 data->args.stateid,
6064                                 task->tk_msg.rpc_cred);
6065                 /* Fallthrough */
6066         case -NFS4ERR_BAD_STATEID:
6067         case -NFS4ERR_STALE_STATEID:
6068                 task->tk_status = 0;
6069                 break;
6070         case -NFS4ERR_OLD_STATEID:
6071                 if (nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6072                         goto out_restart;
6073                 task->tk_status = 0;
6074                 break;
6075         case -NFS4ERR_ACCESS:
6076                 if (data->args.bitmask) {
6077                         data->args.bitmask = NULL;
6078                         data->res.fattr = NULL;
6079                         goto out_restart;
6080                 }
6081                 /* Fallthrough */
6082         default:
6083                 task->tk_status = nfs4_async_handle_exception(task,
6084                                 data->res.server, task->tk_status,
6085                                 &exception);
6086                 if (exception.retry)
6087                         goto out_restart;
6088         }
6089         data->rpc_status = task->tk_status;
6090         return;
6091 lr_restart:
6092         data->res.lr_ret = 0;
6093 out_restart:
6094         task->tk_status = 0;
6095         rpc_restart_call_prepare(task);
6096 }
6097
6098 static void nfs4_delegreturn_release(void *calldata)
6099 {
6100         struct nfs4_delegreturndata *data = calldata;
6101         struct inode *inode = data->inode;
6102
6103         if (inode) {
6104                 if (data->lr.roc)
6105                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6106                                         data->res.lr_ret);
6107                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6108                 nfs_iput_and_deactive(inode);
6109         }
6110         kfree(calldata);
6111 }
6112
6113 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6114 {
6115         struct nfs4_delegreturndata *d_data;
6116         struct pnfs_layout_hdr *lo;
6117
6118         d_data = (struct nfs4_delegreturndata *)data;
6119
6120         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
6121                 return;
6122
6123         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6124         if (lo && !pnfs_layout_is_valid(lo)) {
6125                 d_data->args.lr_args = NULL;
6126                 d_data->res.lr_res = NULL;
6127         }
6128
6129         nfs4_setup_sequence(d_data->res.server->nfs_client,
6130                         &d_data->args.seq_args,
6131                         &d_data->res.seq_res,
6132                         task);
6133 }
6134
6135 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6136         .rpc_call_prepare = nfs4_delegreturn_prepare,
6137         .rpc_call_done = nfs4_delegreturn_done,
6138         .rpc_release = nfs4_delegreturn_release,
6139 };
6140
6141 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
6142 {
6143         struct nfs4_delegreturndata *data;
6144         struct nfs_server *server = NFS_SERVER(inode);
6145         struct rpc_task *task;
6146         struct rpc_message msg = {
6147                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6148                 .rpc_cred = cred,
6149         };
6150         struct rpc_task_setup task_setup_data = {
6151                 .rpc_client = server->client,
6152                 .rpc_message = &msg,
6153                 .callback_ops = &nfs4_delegreturn_ops,
6154                 .flags = RPC_TASK_ASYNC,
6155         };
6156         int status = 0;
6157
6158         data = kzalloc(sizeof(*data), GFP_NOFS);
6159         if (data == NULL)
6160                 return -ENOMEM;
6161         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6162
6163         nfs4_state_protect(server->nfs_client,
6164                         NFS_SP4_MACH_CRED_CLEANUP,
6165                         &task_setup_data.rpc_client, &msg);
6166
6167         data->args.fhandle = &data->fh;
6168         data->args.stateid = &data->stateid;
6169         data->args.bitmask = server->cache_consistency_bitmask;
6170         nfs_copy_fh(&data->fh, NFS_FH(inode));
6171         nfs4_stateid_copy(&data->stateid, stateid);
6172         data->res.fattr = &data->fattr;
6173         data->res.server = server;
6174         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6175         data->lr.arg.ld_private = &data->lr.ld_private;
6176         nfs_fattr_init(data->res.fattr);
6177         data->timestamp = jiffies;
6178         data->rpc_status = 0;
6179         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6180         data->inode = nfs_igrab_and_active(inode);
6181         if (data->inode) {
6182                 if (data->lr.roc) {
6183                         data->args.lr_args = &data->lr.arg;
6184                         data->res.lr_res = &data->lr.res;
6185                 }
6186         } else if (data->lr.roc) {
6187                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6188                 data->lr.roc = false;
6189         }
6190
6191         task_setup_data.callback_data = data;
6192         msg.rpc_argp = &data->args;
6193         msg.rpc_resp = &data->res;
6194         task = rpc_run_task(&task_setup_data);
6195         if (IS_ERR(task))
6196                 return PTR_ERR(task);
6197         if (!issync)
6198                 goto out;
6199         status = rpc_wait_for_completion_task(task);
6200         if (status != 0)
6201                 goto out;
6202         status = data->rpc_status;
6203 out:
6204         rpc_put_task(task);
6205         return status;
6206 }
6207
6208 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
6209 {
6210         struct nfs_server *server = NFS_SERVER(inode);
6211         struct nfs4_exception exception = { };
6212         int err;
6213         do {
6214                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6215                 trace_nfs4_delegreturn(inode, stateid, err);
6216                 switch (err) {
6217                         case -NFS4ERR_STALE_STATEID:
6218                         case -NFS4ERR_EXPIRED:
6219                         case 0:
6220                                 return 0;
6221                 }
6222                 err = nfs4_handle_exception(server, err, &exception);
6223         } while (exception.retry);
6224         return err;
6225 }
6226
6227 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6228 {
6229         struct inode *inode = state->inode;
6230         struct nfs_server *server = NFS_SERVER(inode);
6231         struct nfs_client *clp = server->nfs_client;
6232         struct nfs_lockt_args arg = {
6233                 .fh = NFS_FH(inode),
6234                 .fl = request,
6235         };
6236         struct nfs_lockt_res res = {
6237                 .denied = request,
6238         };
6239         struct rpc_message msg = {
6240                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6241                 .rpc_argp       = &arg,
6242                 .rpc_resp       = &res,
6243                 .rpc_cred       = state->owner->so_cred,
6244         };
6245         struct nfs4_lock_state *lsp;
6246         int status;
6247
6248         arg.lock_owner.clientid = clp->cl_clientid;
6249         status = nfs4_set_lock_state(state, request);
6250         if (status != 0)
6251                 goto out;
6252         lsp = request->fl_u.nfs4_fl.owner;
6253         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6254         arg.lock_owner.s_dev = server->s_dev;
6255         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6256         switch (status) {
6257                 case 0:
6258                         request->fl_type = F_UNLCK;
6259                         break;
6260                 case -NFS4ERR_DENIED:
6261                         status = 0;
6262         }
6263         request->fl_ops->fl_release_private(request);
6264         request->fl_ops = NULL;
6265 out:
6266         return status;
6267 }
6268
6269 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6270 {
6271         struct nfs4_exception exception = { };
6272         int err;
6273
6274         do {
6275                 err = _nfs4_proc_getlk(state, cmd, request);
6276                 trace_nfs4_get_lock(request, state, cmd, err);
6277                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6278                                 &exception);
6279         } while (exception.retry);
6280         return err;
6281 }
6282
6283 struct nfs4_unlockdata {
6284         struct nfs_locku_args arg;
6285         struct nfs_locku_res res;
6286         struct nfs4_lock_state *lsp;
6287         struct nfs_open_context *ctx;
6288         struct nfs_lock_context *l_ctx;
6289         struct file_lock fl;
6290         struct nfs_server *server;
6291         unsigned long timestamp;
6292 };
6293
6294 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6295                 struct nfs_open_context *ctx,
6296                 struct nfs4_lock_state *lsp,
6297                 struct nfs_seqid *seqid)
6298 {
6299         struct nfs4_unlockdata *p;
6300         struct inode *inode = lsp->ls_state->inode;
6301
6302         p = kzalloc(sizeof(*p), GFP_NOFS);
6303         if (p == NULL)
6304                 return NULL;
6305         p->arg.fh = NFS_FH(inode);
6306         p->arg.fl = &p->fl;
6307         p->arg.seqid = seqid;
6308         p->res.seqid = seqid;
6309         p->lsp = lsp;
6310         refcount_inc(&lsp->ls_count);
6311         /* Ensure we don't close file until we're done freeing locks! */
6312         p->ctx = get_nfs_open_context(ctx);
6313         p->l_ctx = nfs_get_lock_context(ctx);
6314         locks_init_lock(&p->fl);
6315         locks_copy_lock(&p->fl, fl);
6316         p->server = NFS_SERVER(inode);
6317         return p;
6318 }
6319
6320 static void nfs4_locku_release_calldata(void *data)
6321 {
6322         struct nfs4_unlockdata *calldata = data;
6323         nfs_free_seqid(calldata->arg.seqid);
6324         nfs4_put_lock_state(calldata->lsp);
6325         nfs_put_lock_context(calldata->l_ctx);
6326         put_nfs_open_context(calldata->ctx);
6327         kfree(calldata);
6328 }
6329
6330 static void nfs4_locku_done(struct rpc_task *task, void *data)
6331 {
6332         struct nfs4_unlockdata *calldata = data;
6333         struct nfs4_exception exception = {
6334                 .inode = calldata->lsp->ls_state->inode,
6335                 .stateid = &calldata->arg.stateid,
6336         };
6337
6338         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6339                 return;
6340         switch (task->tk_status) {
6341                 case 0:
6342                         renew_lease(calldata->server, calldata->timestamp);
6343                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6344                         if (nfs4_update_lock_stateid(calldata->lsp,
6345                                         &calldata->res.stateid))
6346                                 break;
6347                         /* Fall through */
6348                 case -NFS4ERR_ADMIN_REVOKED:
6349                 case -NFS4ERR_EXPIRED:
6350                         nfs4_free_revoked_stateid(calldata->server,
6351                                         &calldata->arg.stateid,
6352                                         task->tk_msg.rpc_cred);
6353                         /* Fall through */
6354                 case -NFS4ERR_BAD_STATEID:
6355                 case -NFS4ERR_OLD_STATEID:
6356                 case -NFS4ERR_STALE_STATEID:
6357                         if (!nfs4_stateid_match(&calldata->arg.stateid,
6358                                                 &calldata->lsp->ls_stateid))
6359                                 rpc_restart_call_prepare(task);
6360                         break;
6361                 default:
6362                         task->tk_status = nfs4_async_handle_exception(task,
6363                                         calldata->server, task->tk_status,
6364                                         &exception);
6365                         if (exception.retry)
6366                                 rpc_restart_call_prepare(task);
6367         }
6368         nfs_release_seqid(calldata->arg.seqid);
6369 }
6370
6371 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6372 {
6373         struct nfs4_unlockdata *calldata = data;
6374
6375         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6376                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6377                 return;
6378
6379         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6380                 goto out_wait;
6381         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
6382         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6383                 /* Note: exit _without_ running nfs4_locku_done */
6384                 goto out_no_action;
6385         }
6386         calldata->timestamp = jiffies;
6387         if (nfs4_setup_sequence(calldata->server->nfs_client,
6388                                 &calldata->arg.seq_args,
6389                                 &calldata->res.seq_res,
6390                                 task) != 0)
6391                 nfs_release_seqid(calldata->arg.seqid);
6392         return;
6393 out_no_action:
6394         task->tk_action = NULL;
6395 out_wait:
6396         nfs4_sequence_done(task, &calldata->res.seq_res);
6397 }
6398
6399 static const struct rpc_call_ops nfs4_locku_ops = {
6400         .rpc_call_prepare = nfs4_locku_prepare,
6401         .rpc_call_done = nfs4_locku_done,
6402         .rpc_release = nfs4_locku_release_calldata,
6403 };
6404
6405 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6406                 struct nfs_open_context *ctx,
6407                 struct nfs4_lock_state *lsp,
6408                 struct nfs_seqid *seqid)
6409 {
6410         struct nfs4_unlockdata *data;
6411         struct rpc_message msg = {
6412                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6413                 .rpc_cred = ctx->cred,
6414         };
6415         struct rpc_task_setup task_setup_data = {
6416                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6417                 .rpc_message = &msg,
6418                 .callback_ops = &nfs4_locku_ops,
6419                 .workqueue = nfsiod_workqueue,
6420                 .flags = RPC_TASK_ASYNC,
6421         };
6422
6423         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6424                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6425
6426         /* Ensure this is an unlock - when canceling a lock, the
6427          * canceled lock is passed in, and it won't be an unlock.
6428          */
6429         fl->fl_type = F_UNLCK;
6430         if (fl->fl_flags & FL_CLOSE)
6431                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6432
6433         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6434         if (data == NULL) {
6435                 nfs_free_seqid(seqid);
6436                 return ERR_PTR(-ENOMEM);
6437         }
6438
6439         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6440         msg.rpc_argp = &data->arg;
6441         msg.rpc_resp = &data->res;
6442         task_setup_data.callback_data = data;
6443         return rpc_run_task(&task_setup_data);
6444 }
6445
6446 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6447 {
6448         struct inode *inode = state->inode;
6449         struct nfs4_state_owner *sp = state->owner;
6450         struct nfs_inode *nfsi = NFS_I(inode);
6451         struct nfs_seqid *seqid;
6452         struct nfs4_lock_state *lsp;
6453         struct rpc_task *task;
6454         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6455         int status = 0;
6456         unsigned char fl_flags = request->fl_flags;
6457
6458         status = nfs4_set_lock_state(state, request);
6459         /* Unlock _before_ we do the RPC call */
6460         request->fl_flags |= FL_EXISTS;
6461         /* Exclude nfs_delegation_claim_locks() */
6462         mutex_lock(&sp->so_delegreturn_mutex);
6463         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6464         down_read(&nfsi->rwsem);
6465         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6466                 up_read(&nfsi->rwsem);
6467                 mutex_unlock(&sp->so_delegreturn_mutex);
6468                 goto out;
6469         }
6470         up_read(&nfsi->rwsem);
6471         mutex_unlock(&sp->so_delegreturn_mutex);
6472         if (status != 0)
6473                 goto out;
6474         /* Is this a delegated lock? */
6475         lsp = request->fl_u.nfs4_fl.owner;
6476         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6477                 goto out;
6478         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6479         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6480         status = -ENOMEM;
6481         if (IS_ERR(seqid))
6482                 goto out;
6483         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6484         status = PTR_ERR(task);
6485         if (IS_ERR(task))
6486                 goto out;
6487         status = rpc_wait_for_completion_task(task);
6488         rpc_put_task(task);
6489 out:
6490         request->fl_flags = fl_flags;
6491         trace_nfs4_unlock(request, state, F_SETLK, status);
6492         return status;
6493 }
6494
6495 struct nfs4_lockdata {
6496         struct nfs_lock_args arg;
6497         struct nfs_lock_res res;
6498         struct nfs4_lock_state *lsp;
6499         struct nfs_open_context *ctx;
6500         struct file_lock fl;
6501         unsigned long timestamp;
6502         int rpc_status;
6503         int cancelled;
6504         struct nfs_server *server;
6505 };
6506
6507 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6508                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6509                 gfp_t gfp_mask)
6510 {
6511         struct nfs4_lockdata *p;
6512         struct inode *inode = lsp->ls_state->inode;
6513         struct nfs_server *server = NFS_SERVER(inode);
6514         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6515
6516         p = kzalloc(sizeof(*p), gfp_mask);
6517         if (p == NULL)
6518                 return NULL;
6519
6520         p->arg.fh = NFS_FH(inode);
6521         p->arg.fl = &p->fl;
6522         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6523         if (IS_ERR(p->arg.open_seqid))
6524                 goto out_free;
6525         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6526         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6527         if (IS_ERR(p->arg.lock_seqid))
6528                 goto out_free_seqid;
6529         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6530         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6531         p->arg.lock_owner.s_dev = server->s_dev;
6532         p->res.lock_seqid = p->arg.lock_seqid;
6533         p->lsp = lsp;
6534         p->server = server;
6535         refcount_inc(&lsp->ls_count);
6536         p->ctx = get_nfs_open_context(ctx);
6537         locks_init_lock(&p->fl);
6538         locks_copy_lock(&p->fl, fl);
6539         return p;
6540 out_free_seqid:
6541         nfs_free_seqid(p->arg.open_seqid);
6542 out_free:
6543         kfree(p);
6544         return NULL;
6545 }
6546
6547 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6548 {
6549         struct nfs4_lockdata *data = calldata;
6550         struct nfs4_state *state = data->lsp->ls_state;
6551
6552         dprintk("%s: begin!\n", __func__);
6553         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6554                 goto out_wait;
6555         /* Do we need to do an open_to_lock_owner? */
6556         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6557                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6558                         goto out_release_lock_seqid;
6559                 }
6560                 nfs4_stateid_copy(&data->arg.open_stateid,
6561                                 &state->open_stateid);
6562                 data->arg.new_lock_owner = 1;
6563                 data->res.open_seqid = data->arg.open_seqid;
6564         } else {
6565                 data->arg.new_lock_owner = 0;
6566                 nfs4_stateid_copy(&data->arg.lock_stateid,
6567                                 &data->lsp->ls_stateid);
6568         }
6569         if (!nfs4_valid_open_stateid(state)) {
6570                 data->rpc_status = -EBADF;
6571                 task->tk_action = NULL;
6572                 goto out_release_open_seqid;
6573         }
6574         data->timestamp = jiffies;
6575         if (nfs4_setup_sequence(data->server->nfs_client,
6576                                 &data->arg.seq_args,
6577                                 &data->res.seq_res,
6578                                 task) == 0)
6579                 return;
6580 out_release_open_seqid:
6581         nfs_release_seqid(data->arg.open_seqid);
6582 out_release_lock_seqid:
6583         nfs_release_seqid(data->arg.lock_seqid);
6584 out_wait:
6585         nfs4_sequence_done(task, &data->res.seq_res);
6586         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6587 }
6588
6589 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6590 {
6591         struct nfs4_lockdata *data = calldata;
6592         struct nfs4_lock_state *lsp = data->lsp;
6593
6594         dprintk("%s: begin!\n", __func__);
6595
6596         if (!nfs4_sequence_done(task, &data->res.seq_res))
6597                 return;
6598
6599         data->rpc_status = task->tk_status;
6600         switch (task->tk_status) {
6601         case 0:
6602                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6603                                 data->timestamp);
6604                 if (data->arg.new_lock && !data->cancelled) {
6605                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6606                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6607                                 goto out_restart;
6608                 }
6609                 if (data->arg.new_lock_owner != 0) {
6610                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6611                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6612                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6613                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6614                         goto out_restart;
6615                 break;
6616         case -NFS4ERR_BAD_STATEID:
6617         case -NFS4ERR_OLD_STATEID:
6618         case -NFS4ERR_STALE_STATEID:
6619         case -NFS4ERR_EXPIRED:
6620                 if (data->arg.new_lock_owner != 0) {
6621                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6622                                                 &lsp->ls_state->open_stateid))
6623                                 goto out_restart;
6624                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6625                                                 &lsp->ls_stateid))
6626                                 goto out_restart;
6627         }
6628 out_done:
6629         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6630         return;
6631 out_restart:
6632         if (!data->cancelled)
6633                 rpc_restart_call_prepare(task);
6634         goto out_done;
6635 }
6636
6637 static void nfs4_lock_release(void *calldata)
6638 {
6639         struct nfs4_lockdata *data = calldata;
6640
6641         dprintk("%s: begin!\n", __func__);
6642         nfs_free_seqid(data->arg.open_seqid);
6643         if (data->cancelled && data->rpc_status == 0) {
6644                 struct rpc_task *task;
6645                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6646                                 data->arg.lock_seqid);
6647                 if (!IS_ERR(task))
6648                         rpc_put_task_async(task);
6649                 dprintk("%s: cancelling lock!\n", __func__);
6650         } else
6651                 nfs_free_seqid(data->arg.lock_seqid);
6652         nfs4_put_lock_state(data->lsp);
6653         put_nfs_open_context(data->ctx);
6654         kfree(data);
6655         dprintk("%s: done!\n", __func__);
6656 }
6657
6658 static const struct rpc_call_ops nfs4_lock_ops = {
6659         .rpc_call_prepare = nfs4_lock_prepare,
6660         .rpc_call_done = nfs4_lock_done,
6661         .rpc_release = nfs4_lock_release,
6662 };
6663
6664 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6665 {
6666         switch (error) {
6667         case -NFS4ERR_ADMIN_REVOKED:
6668         case -NFS4ERR_EXPIRED:
6669         case -NFS4ERR_BAD_STATEID:
6670                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6671                 if (new_lock_owner != 0 ||
6672                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6673                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6674                 break;
6675         case -NFS4ERR_STALE_STATEID:
6676                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6677                 nfs4_schedule_lease_recovery(server->nfs_client);
6678         };
6679 }
6680
6681 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6682 {
6683         struct nfs4_lockdata *data;
6684         struct rpc_task *task;
6685         struct rpc_message msg = {
6686                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6687                 .rpc_cred = state->owner->so_cred,
6688         };
6689         struct rpc_task_setup task_setup_data = {
6690                 .rpc_client = NFS_CLIENT(state->inode),
6691                 .rpc_message = &msg,
6692                 .callback_ops = &nfs4_lock_ops,
6693                 .workqueue = nfsiod_workqueue,
6694                 .flags = RPC_TASK_ASYNC,
6695         };
6696         int ret;
6697
6698         dprintk("%s: begin!\n", __func__);
6699         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6700                         fl->fl_u.nfs4_fl.owner,
6701                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6702         if (data == NULL)
6703                 return -ENOMEM;
6704         if (IS_SETLKW(cmd))
6705                 data->arg.block = 1;
6706         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6707                                 recovery_type > NFS_LOCK_NEW);
6708         msg.rpc_argp = &data->arg;
6709         msg.rpc_resp = &data->res;
6710         task_setup_data.callback_data = data;
6711         if (recovery_type > NFS_LOCK_NEW) {
6712                 if (recovery_type == NFS_LOCK_RECLAIM)
6713                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6714         } else
6715                 data->arg.new_lock = 1;
6716         task = rpc_run_task(&task_setup_data);
6717         if (IS_ERR(task))
6718                 return PTR_ERR(task);
6719         ret = rpc_wait_for_completion_task(task);
6720         if (ret == 0) {
6721                 ret = data->rpc_status;
6722                 if (ret)
6723                         nfs4_handle_setlk_error(data->server, data->lsp,
6724                                         data->arg.new_lock_owner, ret);
6725         } else
6726                 data->cancelled = true;
6727         rpc_put_task(task);
6728         dprintk("%s: done, ret = %d!\n", __func__, ret);
6729         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6730         return ret;
6731 }
6732
6733 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6734 {
6735         struct nfs_server *server = NFS_SERVER(state->inode);
6736         struct nfs4_exception exception = {
6737                 .inode = state->inode,
6738         };
6739         int err;
6740
6741         do {
6742                 /* Cache the lock if possible... */
6743                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6744                         return 0;
6745                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6746                 if (err != -NFS4ERR_DELAY)
6747                         break;
6748                 nfs4_handle_exception(server, err, &exception);
6749         } while (exception.retry);
6750         return err;
6751 }
6752
6753 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6754 {
6755         struct nfs_server *server = NFS_SERVER(state->inode);
6756         struct nfs4_exception exception = {
6757                 .inode = state->inode,
6758         };
6759         int err;
6760
6761         err = nfs4_set_lock_state(state, request);
6762         if (err != 0)
6763                 return err;
6764         if (!recover_lost_locks) {
6765                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6766                 return 0;
6767         }
6768         do {
6769                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6770                         return 0;
6771                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6772                 switch (err) {
6773                 default:
6774                         goto out;
6775                 case -NFS4ERR_GRACE:
6776                 case -NFS4ERR_DELAY:
6777                         nfs4_handle_exception(server, err, &exception);
6778                         err = 0;
6779                 }
6780         } while (exception.retry);
6781 out:
6782         return err;
6783 }
6784
6785 #if defined(CONFIG_NFS_V4_1)
6786 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6787 {
6788         struct nfs4_lock_state *lsp;
6789         int status;
6790
6791         status = nfs4_set_lock_state(state, request);
6792         if (status != 0)
6793                 return status;
6794         lsp = request->fl_u.nfs4_fl.owner;
6795         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6796             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6797                 return 0;
6798         return nfs4_lock_expired(state, request);
6799 }
6800 #endif
6801
6802 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6803 {
6804         struct nfs_inode *nfsi = NFS_I(state->inode);
6805         struct nfs4_state_owner *sp = state->owner;
6806         unsigned char fl_flags = request->fl_flags;
6807         int status;
6808
6809         request->fl_flags |= FL_ACCESS;
6810         status = locks_lock_inode_wait(state->inode, request);
6811         if (status < 0)
6812                 goto out;
6813         mutex_lock(&sp->so_delegreturn_mutex);
6814         down_read(&nfsi->rwsem);
6815         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6816                 /* Yes: cache locks! */
6817                 /* ...but avoid races with delegation recall... */
6818                 request->fl_flags = fl_flags & ~FL_SLEEP;
6819                 status = locks_lock_inode_wait(state->inode, request);
6820                 up_read(&nfsi->rwsem);
6821                 mutex_unlock(&sp->so_delegreturn_mutex);
6822                 goto out;
6823         }
6824         up_read(&nfsi->rwsem);
6825         mutex_unlock(&sp->so_delegreturn_mutex);
6826         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6827 out:
6828         request->fl_flags = fl_flags;
6829         return status;
6830 }
6831
6832 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6833 {
6834         struct nfs4_exception exception = {
6835                 .state = state,
6836                 .inode = state->inode,
6837         };
6838         int err;
6839
6840         do {
6841                 err = _nfs4_proc_setlk(state, cmd, request);
6842                 if (err == -NFS4ERR_DENIED)
6843                         err = -EAGAIN;
6844                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6845                                 err, &exception);
6846         } while (exception.retry);
6847         return err;
6848 }
6849
6850 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6851 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6852
6853 static int
6854 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6855                         struct file_lock *request)
6856 {
6857         int             status = -ERESTARTSYS;
6858         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6859
6860         while(!signalled()) {
6861                 status = nfs4_proc_setlk(state, cmd, request);
6862                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6863                         break;
6864                 freezable_schedule_timeout_interruptible(timeout);
6865                 timeout *= 2;
6866                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6867                 status = -ERESTARTSYS;
6868         }
6869         return status;
6870 }
6871
6872 #ifdef CONFIG_NFS_V4_1
6873 struct nfs4_lock_waiter {
6874         struct task_struct      *task;
6875         struct inode            *inode;
6876         struct nfs_lowner       *owner;
6877         bool                    notified;
6878 };
6879
6880 static int
6881 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6882 {
6883         int ret;
6884         struct nfs4_lock_waiter *waiter = wait->private;
6885
6886         /* NULL key means to wake up everyone */
6887         if (key) {
6888                 struct cb_notify_lock_args      *cbnl = key;
6889                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
6890                                                 *wowner = waiter->owner;
6891
6892                 /* Only wake if the callback was for the same owner. */
6893                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
6894                         return 0;
6895
6896                 /* Make sure it's for the right inode */
6897                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6898                         return 0;
6899
6900                 waiter->notified = true;
6901         }
6902
6903         /* override "private" so we can use default_wake_function */
6904         wait->private = waiter->task;
6905         ret = autoremove_wake_function(wait, mode, flags, key);
6906         wait->private = waiter;
6907         return ret;
6908 }
6909
6910 static int
6911 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6912 {
6913         int status = -ERESTARTSYS;
6914         unsigned long flags;
6915         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6916         struct nfs_server *server = NFS_SERVER(state->inode);
6917         struct nfs_client *clp = server->nfs_client;
6918         wait_queue_head_t *q = &clp->cl_lock_waitq;
6919         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6920                                     .id = lsp->ls_seqid.owner_id,
6921                                     .s_dev = server->s_dev };
6922         struct nfs4_lock_waiter waiter = { .task  = current,
6923                                            .inode = state->inode,
6924                                            .owner = &owner,
6925                                            .notified = false };
6926         wait_queue_entry_t wait;
6927
6928         /* Don't bother with waitqueue if we don't expect a callback */
6929         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6930                 return nfs4_retry_setlk_simple(state, cmd, request);
6931
6932         init_wait(&wait);
6933         wait.private = &waiter;
6934         wait.func = nfs4_wake_lock_waiter;
6935         add_wait_queue(q, &wait);
6936
6937         while(!signalled()) {
6938                 waiter.notified = false;
6939                 status = nfs4_proc_setlk(state, cmd, request);
6940                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6941                         break;
6942
6943                 status = -ERESTARTSYS;
6944                 spin_lock_irqsave(&q->lock, flags);
6945                 if (waiter.notified) {
6946                         spin_unlock_irqrestore(&q->lock, flags);
6947                         continue;
6948                 }
6949                 set_current_state(TASK_INTERRUPTIBLE);
6950                 spin_unlock_irqrestore(&q->lock, flags);
6951
6952                 freezable_schedule_timeout(NFS4_LOCK_MAXTIMEOUT);
6953         }
6954
6955         finish_wait(q, &wait);
6956         return status;
6957 }
6958 #else /* !CONFIG_NFS_V4_1 */
6959 static inline int
6960 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6961 {
6962         return nfs4_retry_setlk_simple(state, cmd, request);
6963 }
6964 #endif
6965
6966 static int
6967 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6968 {
6969         struct nfs_open_context *ctx;
6970         struct nfs4_state *state;
6971         int status;
6972
6973         /* verify open state */
6974         ctx = nfs_file_open_context(filp);
6975         state = ctx->state;
6976
6977         if (IS_GETLK(cmd)) {
6978                 if (state != NULL)
6979                         return nfs4_proc_getlk(state, F_GETLK, request);
6980                 return 0;
6981         }
6982
6983         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6984                 return -EINVAL;
6985
6986         if (request->fl_type == F_UNLCK) {
6987                 if (state != NULL)
6988                         return nfs4_proc_unlck(state, cmd, request);
6989                 return 0;
6990         }
6991
6992         if (state == NULL)
6993                 return -ENOLCK;
6994
6995         if ((request->fl_flags & FL_POSIX) &&
6996             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6997                 return -ENOLCK;
6998
6999         /*
7000          * Don't rely on the VFS having checked the file open mode,
7001          * since it won't do this for flock() locks.
7002          */
7003         switch (request->fl_type) {
7004         case F_RDLCK:
7005                 if (!(filp->f_mode & FMODE_READ))
7006                         return -EBADF;
7007                 break;
7008         case F_WRLCK:
7009                 if (!(filp->f_mode & FMODE_WRITE))
7010                         return -EBADF;
7011         }
7012
7013         status = nfs4_set_lock_state(state, request);
7014         if (status != 0)
7015                 return status;
7016
7017         return nfs4_retry_setlk(state, cmd, request);
7018 }
7019
7020 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7021 {
7022         struct nfs_server *server = NFS_SERVER(state->inode);
7023         int err;
7024
7025         err = nfs4_set_lock_state(state, fl);
7026         if (err != 0)
7027                 return err;
7028         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7029         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7030 }
7031
7032 struct nfs_release_lockowner_data {
7033         struct nfs4_lock_state *lsp;
7034         struct nfs_server *server;
7035         struct nfs_release_lockowner_args args;
7036         struct nfs_release_lockowner_res res;
7037         unsigned long timestamp;
7038 };
7039
7040 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7041 {
7042         struct nfs_release_lockowner_data *data = calldata;
7043         struct nfs_server *server = data->server;
7044         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7045                            &data->res.seq_res, task);
7046         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7047         data->timestamp = jiffies;
7048 }
7049
7050 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7051 {
7052         struct nfs_release_lockowner_data *data = calldata;
7053         struct nfs_server *server = data->server;
7054
7055         nfs40_sequence_done(task, &data->res.seq_res);
7056
7057         switch (task->tk_status) {
7058         case 0:
7059                 renew_lease(server, data->timestamp);
7060                 break;
7061         case -NFS4ERR_STALE_CLIENTID:
7062         case -NFS4ERR_EXPIRED:
7063                 nfs4_schedule_lease_recovery(server->nfs_client);
7064                 break;
7065         case -NFS4ERR_LEASE_MOVED:
7066         case -NFS4ERR_DELAY:
7067                 if (nfs4_async_handle_error(task, server,
7068                                             NULL, NULL) == -EAGAIN)
7069                         rpc_restart_call_prepare(task);
7070         }
7071 }
7072
7073 static void nfs4_release_lockowner_release(void *calldata)
7074 {
7075         struct nfs_release_lockowner_data *data = calldata;
7076         nfs4_free_lock_state(data->server, data->lsp);
7077         kfree(calldata);
7078 }
7079
7080 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7081         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7082         .rpc_call_done = nfs4_release_lockowner_done,
7083         .rpc_release = nfs4_release_lockowner_release,
7084 };
7085
7086 static void
7087 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7088 {
7089         struct nfs_release_lockowner_data *data;
7090         struct rpc_message msg = {
7091                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7092         };
7093
7094         if (server->nfs_client->cl_mvops->minor_version != 0)
7095                 return;
7096
7097         data = kmalloc(sizeof(*data), GFP_NOFS);
7098         if (!data)
7099                 return;
7100         data->lsp = lsp;
7101         data->server = server;
7102         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7103         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7104         data->args.lock_owner.s_dev = server->s_dev;
7105
7106         msg.rpc_argp = &data->args;
7107         msg.rpc_resp = &data->res;
7108         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7109         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7110 }
7111
7112 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7113
7114 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7115                                    struct dentry *unused, struct inode *inode,
7116                                    const char *key, const void *buf,
7117                                    size_t buflen, int flags)
7118 {
7119         return nfs4_proc_set_acl(inode, buf, buflen);
7120 }
7121
7122 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7123                                    struct dentry *unused, struct inode *inode,
7124                                    const char *key, void *buf, size_t buflen)
7125 {
7126         return nfs4_proc_get_acl(inode, buf, buflen);
7127 }
7128
7129 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7130 {
7131         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7132 }
7133
7134 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7135
7136 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7137                                      struct dentry *unused, struct inode *inode,
7138                                      const char *key, const void *buf,
7139                                      size_t buflen, int flags)
7140 {
7141         if (security_ismaclabel(key))
7142                 return nfs4_set_security_label(inode, buf, buflen);
7143
7144         return -EOPNOTSUPP;
7145 }
7146
7147 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7148                                      struct dentry *unused, struct inode *inode,
7149                                      const char *key, void *buf, size_t buflen)
7150 {
7151         if (security_ismaclabel(key))
7152                 return nfs4_get_security_label(inode, buf, buflen);
7153         return -EOPNOTSUPP;
7154 }
7155
7156 static ssize_t
7157 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7158 {
7159         int len = 0;
7160
7161         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7162                 len = security_inode_listsecurity(inode, list, list_len);
7163                 if (list_len && len > list_len)
7164                         return -ERANGE;
7165         }
7166         return len;
7167 }
7168
7169 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7170         .prefix = XATTR_SECURITY_PREFIX,
7171         .get    = nfs4_xattr_get_nfs4_label,
7172         .set    = nfs4_xattr_set_nfs4_label,
7173 };
7174
7175 #else
7176
7177 static ssize_t
7178 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7179 {
7180         return 0;
7181 }
7182
7183 #endif
7184
7185 /*
7186  * nfs_fhget will use either the mounted_on_fileid or the fileid
7187  */
7188 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7189 {
7190         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7191                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7192               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7193               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7194                 return;
7195
7196         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7197                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7198         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7199         fattr->nlink = 2;
7200 }
7201
7202 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7203                                    const struct qstr *name,
7204                                    struct nfs4_fs_locations *fs_locations,
7205                                    struct page *page)
7206 {
7207         struct nfs_server *server = NFS_SERVER(dir);
7208         u32 bitmask[3];
7209         struct nfs4_fs_locations_arg args = {
7210                 .dir_fh = NFS_FH(dir),
7211                 .name = name,
7212                 .page = page,
7213                 .bitmask = bitmask,
7214         };
7215         struct nfs4_fs_locations_res res = {
7216                 .fs_locations = fs_locations,
7217         };
7218         struct rpc_message msg = {
7219                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7220                 .rpc_argp = &args,
7221                 .rpc_resp = &res,
7222         };
7223         int status;
7224
7225         dprintk("%s: start\n", __func__);
7226
7227         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7228         bitmask[1] = nfs4_fattr_bitmap[1];
7229
7230         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7231          * is not supported */
7232         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7233                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7234         else
7235                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7236
7237         nfs_fattr_init(&fs_locations->fattr);
7238         fs_locations->server = server;
7239         fs_locations->nlocations = 0;
7240         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7241         dprintk("%s: returned status = %d\n", __func__, status);
7242         return status;
7243 }
7244
7245 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7246                            const struct qstr *name,
7247                            struct nfs4_fs_locations *fs_locations,
7248                            struct page *page)
7249 {
7250         struct nfs4_exception exception = { };
7251         int err;
7252         do {
7253                 err = _nfs4_proc_fs_locations(client, dir, name,
7254                                 fs_locations, page);
7255                 trace_nfs4_get_fs_locations(dir, name, err);
7256                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7257                                 &exception);
7258         } while (exception.retry);
7259         return err;
7260 }
7261
7262 /*
7263  * This operation also signals the server that this client is
7264  * performing migration recovery.  The server can stop returning
7265  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7266  * appended to this compound to identify the client ID which is
7267  * performing recovery.
7268  */
7269 static int _nfs40_proc_get_locations(struct inode *inode,
7270                                      struct nfs4_fs_locations *locations,
7271                                      struct page *page, struct rpc_cred *cred)
7272 {
7273         struct nfs_server *server = NFS_SERVER(inode);
7274         struct rpc_clnt *clnt = server->client;
7275         u32 bitmask[2] = {
7276                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7277         };
7278         struct nfs4_fs_locations_arg args = {
7279                 .clientid       = server->nfs_client->cl_clientid,
7280                 .fh             = NFS_FH(inode),
7281                 .page           = page,
7282                 .bitmask        = bitmask,
7283                 .migration      = 1,            /* skip LOOKUP */
7284                 .renew          = 1,            /* append RENEW */
7285         };
7286         struct nfs4_fs_locations_res res = {
7287                 .fs_locations   = locations,
7288                 .migration      = 1,
7289                 .renew          = 1,
7290         };
7291         struct rpc_message msg = {
7292                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7293                 .rpc_argp       = &args,
7294                 .rpc_resp       = &res,
7295                 .rpc_cred       = cred,
7296         };
7297         unsigned long now = jiffies;
7298         int status;
7299
7300         nfs_fattr_init(&locations->fattr);
7301         locations->server = server;
7302         locations->nlocations = 0;
7303
7304         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7305         status = nfs4_call_sync_sequence(clnt, server, &msg,
7306                                         &args.seq_args, &res.seq_res);
7307         if (status)
7308                 return status;
7309
7310         renew_lease(server, now);
7311         return 0;
7312 }
7313
7314 #ifdef CONFIG_NFS_V4_1
7315
7316 /*
7317  * This operation also signals the server that this client is
7318  * performing migration recovery.  The server can stop asserting
7319  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7320  * performing this operation is identified in the SEQUENCE
7321  * operation in this compound.
7322  *
7323  * When the client supports GETATTR(fs_locations_info), it can
7324  * be plumbed in here.
7325  */
7326 static int _nfs41_proc_get_locations(struct inode *inode,
7327                                      struct nfs4_fs_locations *locations,
7328                                      struct page *page, struct rpc_cred *cred)
7329 {
7330         struct nfs_server *server = NFS_SERVER(inode);
7331         struct rpc_clnt *clnt = server->client;
7332         u32 bitmask[2] = {
7333                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7334         };
7335         struct nfs4_fs_locations_arg args = {
7336                 .fh             = NFS_FH(inode),
7337                 .page           = page,
7338                 .bitmask        = bitmask,
7339                 .migration      = 1,            /* skip LOOKUP */
7340         };
7341         struct nfs4_fs_locations_res res = {
7342                 .fs_locations   = locations,
7343                 .migration      = 1,
7344         };
7345         struct rpc_message msg = {
7346                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7347                 .rpc_argp       = &args,
7348                 .rpc_resp       = &res,
7349                 .rpc_cred       = cred,
7350         };
7351         int status;
7352
7353         nfs_fattr_init(&locations->fattr);
7354         locations->server = server;
7355         locations->nlocations = 0;
7356
7357         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7358         status = nfs4_call_sync_sequence(clnt, server, &msg,
7359                                         &args.seq_args, &res.seq_res);
7360         if (status == NFS4_OK &&
7361             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7362                 status = -NFS4ERR_LEASE_MOVED;
7363         return status;
7364 }
7365
7366 #endif  /* CONFIG_NFS_V4_1 */
7367
7368 /**
7369  * nfs4_proc_get_locations - discover locations for a migrated FSID
7370  * @inode: inode on FSID that is migrating
7371  * @locations: result of query
7372  * @page: buffer
7373  * @cred: credential to use for this operation
7374  *
7375  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7376  * operation failed, or a negative errno if a local error occurred.
7377  *
7378  * On success, "locations" is filled in, but if the server has
7379  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7380  * asserted.
7381  *
7382  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7383  * from this client that require migration recovery.
7384  */
7385 int nfs4_proc_get_locations(struct inode *inode,
7386                             struct nfs4_fs_locations *locations,
7387                             struct page *page, struct rpc_cred *cred)
7388 {
7389         struct nfs_server *server = NFS_SERVER(inode);
7390         struct nfs_client *clp = server->nfs_client;
7391         const struct nfs4_mig_recovery_ops *ops =
7392                                         clp->cl_mvops->mig_recovery_ops;
7393         struct nfs4_exception exception = { };
7394         int status;
7395
7396         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7397                 (unsigned long long)server->fsid.major,
7398                 (unsigned long long)server->fsid.minor,
7399                 clp->cl_hostname);
7400         nfs_display_fhandle(NFS_FH(inode), __func__);
7401
7402         do {
7403                 status = ops->get_locations(inode, locations, page, cred);
7404                 if (status != -NFS4ERR_DELAY)
7405                         break;
7406                 nfs4_handle_exception(server, status, &exception);
7407         } while (exception.retry);
7408         return status;
7409 }
7410
7411 /*
7412  * This operation also signals the server that this client is
7413  * performing "lease moved" recovery.  The server can stop
7414  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7415  * is appended to this compound to identify the client ID which is
7416  * performing recovery.
7417  */
7418 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7419 {
7420         struct nfs_server *server = NFS_SERVER(inode);
7421         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7422         struct rpc_clnt *clnt = server->client;
7423         struct nfs4_fsid_present_arg args = {
7424                 .fh             = NFS_FH(inode),
7425                 .clientid       = clp->cl_clientid,
7426                 .renew          = 1,            /* append RENEW */
7427         };
7428         struct nfs4_fsid_present_res res = {
7429                 .renew          = 1,
7430         };
7431         struct rpc_message msg = {
7432                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7433                 .rpc_argp       = &args,
7434                 .rpc_resp       = &res,
7435                 .rpc_cred       = cred,
7436         };
7437         unsigned long now = jiffies;
7438         int status;
7439
7440         res.fh = nfs_alloc_fhandle();
7441         if (res.fh == NULL)
7442                 return -ENOMEM;
7443
7444         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7445         status = nfs4_call_sync_sequence(clnt, server, &msg,
7446                                                 &args.seq_args, &res.seq_res);
7447         nfs_free_fhandle(res.fh);
7448         if (status)
7449                 return status;
7450
7451         do_renew_lease(clp, now);
7452         return 0;
7453 }
7454
7455 #ifdef CONFIG_NFS_V4_1
7456
7457 /*
7458  * This operation also signals the server that this client is
7459  * performing "lease moved" recovery.  The server can stop asserting
7460  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7461  * this operation is identified in the SEQUENCE operation in this
7462  * compound.
7463  */
7464 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7465 {
7466         struct nfs_server *server = NFS_SERVER(inode);
7467         struct rpc_clnt *clnt = server->client;
7468         struct nfs4_fsid_present_arg args = {
7469                 .fh             = NFS_FH(inode),
7470         };
7471         struct nfs4_fsid_present_res res = {
7472         };
7473         struct rpc_message msg = {
7474                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7475                 .rpc_argp       = &args,
7476                 .rpc_resp       = &res,
7477                 .rpc_cred       = cred,
7478         };
7479         int status;
7480
7481         res.fh = nfs_alloc_fhandle();
7482         if (res.fh == NULL)
7483                 return -ENOMEM;
7484
7485         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7486         status = nfs4_call_sync_sequence(clnt, server, &msg,
7487                                                 &args.seq_args, &res.seq_res);
7488         nfs_free_fhandle(res.fh);
7489         if (status == NFS4_OK &&
7490             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7491                 status = -NFS4ERR_LEASE_MOVED;
7492         return status;
7493 }
7494
7495 #endif  /* CONFIG_NFS_V4_1 */
7496
7497 /**
7498  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7499  * @inode: inode on FSID to check
7500  * @cred: credential to use for this operation
7501  *
7502  * Server indicates whether the FSID is present, moved, or not
7503  * recognized.  This operation is necessary to clear a LEASE_MOVED
7504  * condition for this client ID.
7505  *
7506  * Returns NFS4_OK if the FSID is present on this server,
7507  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7508  *  NFS4ERR code if some error occurred on the server, or a
7509  *  negative errno if a local failure occurred.
7510  */
7511 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7512 {
7513         struct nfs_server *server = NFS_SERVER(inode);
7514         struct nfs_client *clp = server->nfs_client;
7515         const struct nfs4_mig_recovery_ops *ops =
7516                                         clp->cl_mvops->mig_recovery_ops;
7517         struct nfs4_exception exception = { };
7518         int status;
7519
7520         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7521                 (unsigned long long)server->fsid.major,
7522                 (unsigned long long)server->fsid.minor,
7523                 clp->cl_hostname);
7524         nfs_display_fhandle(NFS_FH(inode), __func__);
7525
7526         do {
7527                 status = ops->fsid_present(inode, cred);
7528                 if (status != -NFS4ERR_DELAY)
7529                         break;
7530                 nfs4_handle_exception(server, status, &exception);
7531         } while (exception.retry);
7532         return status;
7533 }
7534
7535 /**
7536  * If 'use_integrity' is true and the state managment nfs_client
7537  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7538  * and the machine credential as per RFC3530bis and RFC5661 Security
7539  * Considerations sections. Otherwise, just use the user cred with the
7540  * filesystem's rpc_client.
7541  */
7542 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7543 {
7544         int status;
7545         struct nfs4_secinfo_arg args = {
7546                 .dir_fh = NFS_FH(dir),
7547                 .name   = name,
7548         };
7549         struct nfs4_secinfo_res res = {
7550                 .flavors     = flavors,
7551         };
7552         struct rpc_message msg = {
7553                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7554                 .rpc_argp = &args,
7555                 .rpc_resp = &res,
7556         };
7557         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7558         struct rpc_cred *cred = NULL;
7559
7560         if (use_integrity) {
7561                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7562                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7563                 msg.rpc_cred = cred;
7564         }
7565
7566         dprintk("NFS call  secinfo %s\n", name->name);
7567
7568         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7569                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7570
7571         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7572                                 &res.seq_res, 0);
7573         dprintk("NFS reply  secinfo: %d\n", status);
7574
7575         if (cred)
7576                 put_rpccred(cred);
7577
7578         return status;
7579 }
7580
7581 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7582                       struct nfs4_secinfo_flavors *flavors)
7583 {
7584         struct nfs4_exception exception = { };
7585         int err;
7586         do {
7587                 err = -NFS4ERR_WRONGSEC;
7588
7589                 /* try to use integrity protection with machine cred */
7590                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7591                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7592
7593                 /*
7594                  * if unable to use integrity protection, or SECINFO with
7595                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7596                  * disallowed by spec, but exists in deployed servers) use
7597                  * the current filesystem's rpc_client and the user cred.
7598                  */
7599                 if (err == -NFS4ERR_WRONGSEC)
7600                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7601
7602                 trace_nfs4_secinfo(dir, name, err);
7603                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7604                                 &exception);
7605         } while (exception.retry);
7606         return err;
7607 }
7608
7609 #ifdef CONFIG_NFS_V4_1
7610 /*
7611  * Check the exchange flags returned by the server for invalid flags, having
7612  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7613  * DS flags set.
7614  */
7615 static int nfs4_check_cl_exchange_flags(u32 flags)
7616 {
7617         if (flags & ~EXCHGID4_FLAG_MASK_R)
7618                 goto out_inval;
7619         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7620             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7621                 goto out_inval;
7622         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7623                 goto out_inval;
7624         return NFS_OK;
7625 out_inval:
7626         return -NFS4ERR_INVAL;
7627 }
7628
7629 static bool
7630 nfs41_same_server_scope(struct nfs41_server_scope *a,
7631                         struct nfs41_server_scope *b)
7632 {
7633         if (a->server_scope_sz != b->server_scope_sz)
7634                 return false;
7635         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7636 }
7637
7638 static void
7639 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7640 {
7641 }
7642
7643 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7644         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7645 };
7646
7647 /*
7648  * nfs4_proc_bind_one_conn_to_session()
7649  *
7650  * The 4.1 client currently uses the same TCP connection for the
7651  * fore and backchannel.
7652  */
7653 static
7654 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7655                 struct rpc_xprt *xprt,
7656                 struct nfs_client *clp,
7657                 struct rpc_cred *cred)
7658 {
7659         int status;
7660         struct nfs41_bind_conn_to_session_args args = {
7661                 .client = clp,
7662                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7663         };
7664         struct nfs41_bind_conn_to_session_res res;
7665         struct rpc_message msg = {
7666                 .rpc_proc =
7667                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7668                 .rpc_argp = &args,
7669                 .rpc_resp = &res,
7670                 .rpc_cred = cred,
7671         };
7672         struct rpc_task_setup task_setup_data = {
7673                 .rpc_client = clnt,
7674                 .rpc_xprt = xprt,
7675                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7676                 .rpc_message = &msg,
7677                 .flags = RPC_TASK_TIMEOUT,
7678         };
7679         struct rpc_task *task;
7680
7681         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7682         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7683                 args.dir = NFS4_CDFC4_FORE;
7684
7685         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7686         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7687                 args.dir = NFS4_CDFC4_FORE;
7688
7689         task = rpc_run_task(&task_setup_data);
7690         if (!IS_ERR(task)) {
7691                 status = task->tk_status;
7692                 rpc_put_task(task);
7693         } else
7694                 status = PTR_ERR(task);
7695         trace_nfs4_bind_conn_to_session(clp, status);
7696         if (status == 0) {
7697                 if (memcmp(res.sessionid.data,
7698                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7699                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7700                         return -EIO;
7701                 }
7702                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7703                         dprintk("NFS: %s: Unexpected direction from server\n",
7704                                 __func__);
7705                         return -EIO;
7706                 }
7707                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7708                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7709                                 __func__);
7710                         return -EIO;
7711                 }
7712         }
7713
7714         return status;
7715 }
7716
7717 struct rpc_bind_conn_calldata {
7718         struct nfs_client *clp;
7719         struct rpc_cred *cred;
7720 };
7721
7722 static int
7723 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7724                 struct rpc_xprt *xprt,
7725                 void *calldata)
7726 {
7727         struct rpc_bind_conn_calldata *p = calldata;
7728
7729         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7730 }
7731
7732 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7733 {
7734         struct rpc_bind_conn_calldata data = {
7735                 .clp = clp,
7736                 .cred = cred,
7737         };
7738         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7739                         nfs4_proc_bind_conn_to_session_callback, &data);
7740 }
7741
7742 /*
7743  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7744  * and operations we'd like to see to enable certain features in the allow map
7745  */
7746 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7747         .how = SP4_MACH_CRED,
7748         .enforce.u.words = {
7749                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7750                       1 << (OP_EXCHANGE_ID - 32) |
7751                       1 << (OP_CREATE_SESSION - 32) |
7752                       1 << (OP_DESTROY_SESSION - 32) |
7753                       1 << (OP_DESTROY_CLIENTID - 32)
7754         },
7755         .allow.u.words = {
7756                 [0] = 1 << (OP_CLOSE) |
7757                       1 << (OP_OPEN_DOWNGRADE) |
7758                       1 << (OP_LOCKU) |
7759                       1 << (OP_DELEGRETURN) |
7760                       1 << (OP_COMMIT),
7761                 [1] = 1 << (OP_SECINFO - 32) |
7762                       1 << (OP_SECINFO_NO_NAME - 32) |
7763                       1 << (OP_LAYOUTRETURN - 32) |
7764                       1 << (OP_TEST_STATEID - 32) |
7765                       1 << (OP_FREE_STATEID - 32) |
7766                       1 << (OP_WRITE - 32)
7767         }
7768 };
7769
7770 /*
7771  * Select the state protection mode for client `clp' given the server results
7772  * from exchange_id in `sp'.
7773  *
7774  * Returns 0 on success, negative errno otherwise.
7775  */
7776 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7777                                  struct nfs41_state_protection *sp)
7778 {
7779         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7780                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7781                       1 << (OP_EXCHANGE_ID - 32) |
7782                       1 << (OP_CREATE_SESSION - 32) |
7783                       1 << (OP_DESTROY_SESSION - 32) |
7784                       1 << (OP_DESTROY_CLIENTID - 32)
7785         };
7786         unsigned long flags = 0;
7787         unsigned int i;
7788         int ret = 0;
7789
7790         if (sp->how == SP4_MACH_CRED) {
7791                 /* Print state protect result */
7792                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7793                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7794                         if (test_bit(i, sp->enforce.u.longs))
7795                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7796                         if (test_bit(i, sp->allow.u.longs))
7797                                 dfprintk(MOUNT, "  allow op %d\n", i);
7798                 }
7799
7800                 /* make sure nothing is on enforce list that isn't supported */
7801                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7802                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7803                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7804                                 ret = -EINVAL;
7805                                 goto out;
7806                         }
7807                 }
7808
7809                 /*
7810                  * Minimal mode - state operations are allowed to use machine
7811                  * credential.  Note this already happens by default, so the
7812                  * client doesn't have to do anything more than the negotiation.
7813                  *
7814                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7815                  *       we're already using the machine cred for exchange_id
7816                  *       and will never use a different cred.
7817                  */
7818                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7819                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7820                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7821                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7822                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7823                         dfprintk(MOUNT, "  minimal mode enabled\n");
7824                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
7825                 } else {
7826                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7827                         ret = -EINVAL;
7828                         goto out;
7829                 }
7830
7831                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7832                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7833                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7834                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7835                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7836                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
7837                 }
7838
7839                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7840                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7841                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
7842                 }
7843
7844                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7845                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7846                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7847                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
7848                 }
7849
7850                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7851                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7852                         dfprintk(MOUNT, "  stateid mode enabled\n");
7853                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
7854                 }
7855
7856                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7857                         dfprintk(MOUNT, "  write mode enabled\n");
7858                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
7859                 }
7860
7861                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7862                         dfprintk(MOUNT, "  commit mode enabled\n");
7863                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
7864                 }
7865         }
7866 out:
7867         clp->cl_sp4_flags = flags;
7868         return ret;
7869 }
7870
7871 struct nfs41_exchange_id_data {
7872         struct nfs41_exchange_id_res res;
7873         struct nfs41_exchange_id_args args;
7874 };
7875
7876 static void nfs4_exchange_id_release(void *data)
7877 {
7878         struct nfs41_exchange_id_data *cdata =
7879                                         (struct nfs41_exchange_id_data *)data;
7880
7881         nfs_put_client(cdata->args.client);
7882         kfree(cdata->res.impl_id);
7883         kfree(cdata->res.server_scope);
7884         kfree(cdata->res.server_owner);
7885         kfree(cdata);
7886 }
7887
7888 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7889         .rpc_release = nfs4_exchange_id_release,
7890 };
7891
7892 /*
7893  * _nfs4_proc_exchange_id()
7894  *
7895  * Wrapper for EXCHANGE_ID operation.
7896  */
7897 static struct rpc_task *
7898 nfs4_run_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7899                         u32 sp4_how, struct rpc_xprt *xprt)
7900 {
7901         struct rpc_message msg = {
7902                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7903                 .rpc_cred = cred,
7904         };
7905         struct rpc_task_setup task_setup_data = {
7906                 .rpc_client = clp->cl_rpcclient,
7907                 .callback_ops = &nfs4_exchange_id_call_ops,
7908                 .rpc_message = &msg,
7909                 .flags = RPC_TASK_TIMEOUT,
7910         };
7911         struct nfs41_exchange_id_data *calldata;
7912         int status;
7913
7914         if (!refcount_inc_not_zero(&clp->cl_count))
7915                 return ERR_PTR(-EIO);
7916
7917         status = -ENOMEM;
7918         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7919         if (!calldata)
7920                 goto out;
7921
7922         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
7923
7924         status = nfs4_init_uniform_client_string(clp);
7925         if (status)
7926                 goto out_calldata;
7927
7928         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7929                                                 GFP_NOFS);
7930         status = -ENOMEM;
7931         if (unlikely(calldata->res.server_owner == NULL))
7932                 goto out_calldata;
7933
7934         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7935                                         GFP_NOFS);
7936         if (unlikely(calldata->res.server_scope == NULL))
7937                 goto out_server_owner;
7938
7939         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7940         if (unlikely(calldata->res.impl_id == NULL))
7941                 goto out_server_scope;
7942
7943         switch (sp4_how) {
7944         case SP4_NONE:
7945                 calldata->args.state_protect.how = SP4_NONE;
7946                 break;
7947
7948         case SP4_MACH_CRED:
7949                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7950                 break;
7951
7952         default:
7953                 /* unsupported! */
7954                 WARN_ON_ONCE(1);
7955                 status = -EINVAL;
7956                 goto out_impl_id;
7957         }
7958         if (xprt) {
7959                 task_setup_data.rpc_xprt = xprt;
7960                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
7961                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
7962                                 sizeof(calldata->args.verifier.data));
7963         }
7964         calldata->args.client = clp;
7965         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7966         EXCHGID4_FLAG_BIND_PRINC_STATEID;
7967 #ifdef CONFIG_NFS_V4_1_MIGRATION
7968         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
7969 #endif
7970         msg.rpc_argp = &calldata->args;
7971         msg.rpc_resp = &calldata->res;
7972         task_setup_data.callback_data = calldata;
7973
7974         return rpc_run_task(&task_setup_data);
7975
7976 out_impl_id:
7977         kfree(calldata->res.impl_id);
7978 out_server_scope:
7979         kfree(calldata->res.server_scope);
7980 out_server_owner:
7981         kfree(calldata->res.server_owner);
7982 out_calldata:
7983         kfree(calldata);
7984 out:
7985         nfs_put_client(clp);
7986         return ERR_PTR(status);
7987 }
7988
7989 /*
7990  * _nfs4_proc_exchange_id()
7991  *
7992  * Wrapper for EXCHANGE_ID operation.
7993  */
7994 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7995                         u32 sp4_how)
7996 {
7997         struct rpc_task *task;
7998         struct nfs41_exchange_id_args *argp;
7999         struct nfs41_exchange_id_res *resp;
8000         int status;
8001
8002         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8003         if (IS_ERR(task))
8004                 return PTR_ERR(task);
8005
8006         argp = task->tk_msg.rpc_argp;
8007         resp = task->tk_msg.rpc_resp;
8008         status = task->tk_status;
8009         if (status  != 0)
8010                 goto out;
8011
8012         status = nfs4_check_cl_exchange_flags(resp->flags);
8013         if (status  != 0)
8014                 goto out;
8015
8016         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8017         if (status != 0)
8018                 goto out;
8019
8020         clp->cl_clientid = resp->clientid;
8021         clp->cl_exchange_flags = resp->flags;
8022         clp->cl_seqid = resp->seqid;
8023         /* Client ID is not confirmed */
8024         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8025                 clear_bit(NFS4_SESSION_ESTABLISHED,
8026                           &clp->cl_session->session_state);
8027
8028         if (clp->cl_serverscope != NULL &&
8029             !nfs41_same_server_scope(clp->cl_serverscope,
8030                                 resp->server_scope)) {
8031                 dprintk("%s: server_scope mismatch detected\n",
8032                         __func__);
8033                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8034         }
8035
8036         swap(clp->cl_serverowner, resp->server_owner);
8037         swap(clp->cl_serverscope, resp->server_scope);
8038         swap(clp->cl_implid, resp->impl_id);
8039
8040         /* Save the EXCHANGE_ID verifier session trunk tests */
8041         memcpy(clp->cl_confirm.data, argp->verifier.data,
8042                sizeof(clp->cl_confirm.data));
8043 out:
8044         trace_nfs4_exchange_id(clp, status);
8045         rpc_put_task(task);
8046         return status;
8047 }
8048
8049 /*
8050  * nfs4_proc_exchange_id()
8051  *
8052  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8053  *
8054  * Since the clientid has expired, all compounds using sessions
8055  * associated with the stale clientid will be returning
8056  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8057  * be in some phase of session reset.
8058  *
8059  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8060  */
8061 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
8062 {
8063         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8064         int status;
8065
8066         /* try SP4_MACH_CRED if krb5i/p */
8067         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8068             authflavor == RPC_AUTH_GSS_KRB5P) {
8069                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8070                 if (!status)
8071                         return 0;
8072         }
8073
8074         /* try SP4_NONE */
8075         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8076 }
8077
8078 /**
8079  * nfs4_test_session_trunk
8080  *
8081  * This is an add_xprt_test() test function called from
8082  * rpc_clnt_setup_test_and_add_xprt.
8083  *
8084  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8085  * and is dereferrenced in nfs4_exchange_id_release
8086  *
8087  * Upon success, add the new transport to the rpc_clnt
8088  *
8089  * @clnt: struct rpc_clnt to get new transport
8090  * @xprt: the rpc_xprt to test
8091  * @data: call data for _nfs4_proc_exchange_id.
8092  */
8093 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8094                             void *data)
8095 {
8096         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8097         struct rpc_task *task;
8098         int status;
8099
8100         u32 sp4_how;
8101
8102         dprintk("--> %s try %s\n", __func__,
8103                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8104
8105         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8106
8107         /* Test connection for session trunking. Async exchange_id call */
8108         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8109         if (IS_ERR(task))
8110                 return PTR_ERR(task);
8111
8112         status = task->tk_status;
8113         if (status == 0)
8114                 status = nfs4_detect_session_trunking(adata->clp,
8115                                 task->tk_msg.rpc_resp, xprt);
8116
8117         rpc_put_task(task);
8118         return status;
8119 }
8120 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8121
8122 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8123                 struct rpc_cred *cred)
8124 {
8125         struct rpc_message msg = {
8126                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8127                 .rpc_argp = clp,
8128                 .rpc_cred = cred,
8129         };
8130         int status;
8131
8132         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8133         trace_nfs4_destroy_clientid(clp, status);
8134         if (status)
8135                 dprintk("NFS: Got error %d from the server %s on "
8136                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8137         return status;
8138 }
8139
8140 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8141                 struct rpc_cred *cred)
8142 {
8143         unsigned int loop;
8144         int ret;
8145
8146         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8147                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8148                 switch (ret) {
8149                 case -NFS4ERR_DELAY:
8150                 case -NFS4ERR_CLIENTID_BUSY:
8151                         ssleep(1);
8152                         break;
8153                 default:
8154                         return ret;
8155                 }
8156         }
8157         return 0;
8158 }
8159
8160 int nfs4_destroy_clientid(struct nfs_client *clp)
8161 {
8162         struct rpc_cred *cred;
8163         int ret = 0;
8164
8165         if (clp->cl_mvops->minor_version < 1)
8166                 goto out;
8167         if (clp->cl_exchange_flags == 0)
8168                 goto out;
8169         if (clp->cl_preserve_clid)
8170                 goto out;
8171         cred = nfs4_get_clid_cred(clp);
8172         ret = nfs4_proc_destroy_clientid(clp, cred);
8173         if (cred)
8174                 put_rpccred(cred);
8175         switch (ret) {
8176         case 0:
8177         case -NFS4ERR_STALE_CLIENTID:
8178                 clp->cl_exchange_flags = 0;
8179         }
8180 out:
8181         return ret;
8182 }
8183
8184 struct nfs4_get_lease_time_data {
8185         struct nfs4_get_lease_time_args *args;
8186         struct nfs4_get_lease_time_res *res;
8187         struct nfs_client *clp;
8188 };
8189
8190 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8191                                         void *calldata)
8192 {
8193         struct nfs4_get_lease_time_data *data =
8194                         (struct nfs4_get_lease_time_data *)calldata;
8195
8196         dprintk("--> %s\n", __func__);
8197         /* just setup sequence, do not trigger session recovery
8198            since we're invoked within one */
8199         nfs4_setup_sequence(data->clp,
8200                         &data->args->la_seq_args,
8201                         &data->res->lr_seq_res,
8202                         task);
8203         dprintk("<-- %s\n", __func__);
8204 }
8205
8206 /*
8207  * Called from nfs4_state_manager thread for session setup, so don't recover
8208  * from sequence operation or clientid errors.
8209  */
8210 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8211 {
8212         struct nfs4_get_lease_time_data *data =
8213                         (struct nfs4_get_lease_time_data *)calldata;
8214
8215         dprintk("--> %s\n", __func__);
8216         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
8217                 return;
8218         switch (task->tk_status) {
8219         case -NFS4ERR_DELAY:
8220         case -NFS4ERR_GRACE:
8221                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8222                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8223                 task->tk_status = 0;
8224                 /* fall through */
8225         case -NFS4ERR_RETRY_UNCACHED_REP:
8226                 rpc_restart_call_prepare(task);
8227                 return;
8228         }
8229         dprintk("<-- %s\n", __func__);
8230 }
8231
8232 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8233         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8234         .rpc_call_done = nfs4_get_lease_time_done,
8235 };
8236
8237 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8238 {
8239         struct rpc_task *task;
8240         struct nfs4_get_lease_time_args args;
8241         struct nfs4_get_lease_time_res res = {
8242                 .lr_fsinfo = fsinfo,
8243         };
8244         struct nfs4_get_lease_time_data data = {
8245                 .args = &args,
8246                 .res = &res,
8247                 .clp = clp,
8248         };
8249         struct rpc_message msg = {
8250                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8251                 .rpc_argp = &args,
8252                 .rpc_resp = &res,
8253         };
8254         struct rpc_task_setup task_setup = {
8255                 .rpc_client = clp->cl_rpcclient,
8256                 .rpc_message = &msg,
8257                 .callback_ops = &nfs4_get_lease_time_ops,
8258                 .callback_data = &data,
8259                 .flags = RPC_TASK_TIMEOUT,
8260         };
8261         int status;
8262
8263         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8264         task = rpc_run_task(&task_setup);
8265
8266         if (IS_ERR(task))
8267                 return PTR_ERR(task);
8268
8269         status = task->tk_status;
8270         rpc_put_task(task);
8271         return status;
8272 }
8273
8274 /*
8275  * Initialize the values to be used by the client in CREATE_SESSION
8276  * If nfs4_init_session set the fore channel request and response sizes,
8277  * use them.
8278  *
8279  * Set the back channel max_resp_sz_cached to zero to force the client to
8280  * always set csa_cachethis to FALSE because the current implementation
8281  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8282  */
8283 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8284                                     struct rpc_clnt *clnt)
8285 {
8286         unsigned int max_rqst_sz, max_resp_sz;
8287         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8288
8289         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8290         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8291
8292         /* Fore channel attributes */
8293         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8294         args->fc_attrs.max_resp_sz = max_resp_sz;
8295         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8296         args->fc_attrs.max_reqs = max_session_slots;
8297
8298         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8299                 "max_ops=%u max_reqs=%u\n",
8300                 __func__,
8301                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8302                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8303
8304         /* Back channel attributes */
8305         args->bc_attrs.max_rqst_sz = max_bc_payload;
8306         args->bc_attrs.max_resp_sz = max_bc_payload;
8307         args->bc_attrs.max_resp_sz_cached = 0;
8308         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8309         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8310
8311         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8312                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8313                 __func__,
8314                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8315                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8316                 args->bc_attrs.max_reqs);
8317 }
8318
8319 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8320                 struct nfs41_create_session_res *res)
8321 {
8322         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8323         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8324
8325         if (rcvd->max_resp_sz > sent->max_resp_sz)
8326                 return -EINVAL;
8327         /*
8328          * Our requested max_ops is the minimum we need; we're not
8329          * prepared to break up compounds into smaller pieces than that.
8330          * So, no point even trying to continue if the server won't
8331          * cooperate:
8332          */
8333         if (rcvd->max_ops < sent->max_ops)
8334                 return -EINVAL;
8335         if (rcvd->max_reqs == 0)
8336                 return -EINVAL;
8337         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8338                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8339         return 0;
8340 }
8341
8342 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8343                 struct nfs41_create_session_res *res)
8344 {
8345         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8346         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8347
8348         if (!(res->flags & SESSION4_BACK_CHAN))
8349                 goto out;
8350         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8351                 return -EINVAL;
8352         if (rcvd->max_resp_sz < sent->max_resp_sz)
8353                 return -EINVAL;
8354         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8355                 return -EINVAL;
8356         if (rcvd->max_ops > sent->max_ops)
8357                 return -EINVAL;
8358         if (rcvd->max_reqs > sent->max_reqs)
8359                 return -EINVAL;
8360 out:
8361         return 0;
8362 }
8363
8364 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8365                                      struct nfs41_create_session_res *res)
8366 {
8367         int ret;
8368
8369         ret = nfs4_verify_fore_channel_attrs(args, res);
8370         if (ret)
8371                 return ret;
8372         return nfs4_verify_back_channel_attrs(args, res);
8373 }
8374
8375 static void nfs4_update_session(struct nfs4_session *session,
8376                 struct nfs41_create_session_res *res)
8377 {
8378         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8379         /* Mark client id and session as being confirmed */
8380         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8381         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8382         session->flags = res->flags;
8383         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8384         if (res->flags & SESSION4_BACK_CHAN)
8385                 memcpy(&session->bc_attrs, &res->bc_attrs,
8386                                 sizeof(session->bc_attrs));
8387 }
8388
8389 static int _nfs4_proc_create_session(struct nfs_client *clp,
8390                 struct rpc_cred *cred)
8391 {
8392         struct nfs4_session *session = clp->cl_session;
8393         struct nfs41_create_session_args args = {
8394                 .client = clp,
8395                 .clientid = clp->cl_clientid,
8396                 .seqid = clp->cl_seqid,
8397                 .cb_program = NFS4_CALLBACK,
8398         };
8399         struct nfs41_create_session_res res;
8400
8401         struct rpc_message msg = {
8402                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8403                 .rpc_argp = &args,
8404                 .rpc_resp = &res,
8405                 .rpc_cred = cred,
8406         };
8407         int status;
8408
8409         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8410         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8411
8412         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8413         trace_nfs4_create_session(clp, status);
8414
8415         switch (status) {
8416         case -NFS4ERR_STALE_CLIENTID:
8417         case -NFS4ERR_DELAY:
8418         case -ETIMEDOUT:
8419         case -EACCES:
8420         case -EAGAIN:
8421                 goto out;
8422         };
8423
8424         clp->cl_seqid++;
8425         if (!status) {
8426                 /* Verify the session's negotiated channel_attrs values */
8427                 status = nfs4_verify_channel_attrs(&args, &res);
8428                 /* Increment the clientid slot sequence id */
8429                 if (status)
8430                         goto out;
8431                 nfs4_update_session(session, &res);
8432         }
8433 out:
8434         return status;
8435 }
8436
8437 /*
8438  * Issues a CREATE_SESSION operation to the server.
8439  * It is the responsibility of the caller to verify the session is
8440  * expired before calling this routine.
8441  */
8442 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
8443 {
8444         int status;
8445         unsigned *ptr;
8446         struct nfs4_session *session = clp->cl_session;
8447
8448         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8449
8450         status = _nfs4_proc_create_session(clp, cred);
8451         if (status)
8452                 goto out;
8453
8454         /* Init or reset the session slot tables */
8455         status = nfs4_setup_session_slot_tables(session);
8456         dprintk("slot table setup returned %d\n", status);
8457         if (status)
8458                 goto out;
8459
8460         ptr = (unsigned *)&session->sess_id.data[0];
8461         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8462                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8463 out:
8464         dprintk("<-- %s\n", __func__);
8465         return status;
8466 }
8467
8468 /*
8469  * Issue the over-the-wire RPC DESTROY_SESSION.
8470  * The caller must serialize access to this routine.
8471  */
8472 int nfs4_proc_destroy_session(struct nfs4_session *session,
8473                 struct rpc_cred *cred)
8474 {
8475         struct rpc_message msg = {
8476                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8477                 .rpc_argp = session,
8478                 .rpc_cred = cred,
8479         };
8480         int status = 0;
8481
8482         dprintk("--> nfs4_proc_destroy_session\n");
8483
8484         /* session is still being setup */
8485         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8486                 return 0;
8487
8488         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8489         trace_nfs4_destroy_session(session->clp, status);
8490
8491         if (status)
8492                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8493                         "Session has been destroyed regardless...\n", status);
8494
8495         dprintk("<-- nfs4_proc_destroy_session\n");
8496         return status;
8497 }
8498
8499 /*
8500  * Renew the cl_session lease.
8501  */
8502 struct nfs4_sequence_data {
8503         struct nfs_client *clp;
8504         struct nfs4_sequence_args args;
8505         struct nfs4_sequence_res res;
8506 };
8507
8508 static void nfs41_sequence_release(void *data)
8509 {
8510         struct nfs4_sequence_data *calldata = data;
8511         struct nfs_client *clp = calldata->clp;
8512
8513         if (refcount_read(&clp->cl_count) > 1)
8514                 nfs4_schedule_state_renewal(clp);
8515         nfs_put_client(clp);
8516         kfree(calldata);
8517 }
8518
8519 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8520 {
8521         switch(task->tk_status) {
8522         case -NFS4ERR_DELAY:
8523                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8524                 return -EAGAIN;
8525         default:
8526                 nfs4_schedule_lease_recovery(clp);
8527         }
8528         return 0;
8529 }
8530
8531 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8532 {
8533         struct nfs4_sequence_data *calldata = data;
8534         struct nfs_client *clp = calldata->clp;
8535
8536         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8537                 return;
8538
8539         trace_nfs4_sequence(clp, task->tk_status);
8540         if (task->tk_status < 0) {
8541                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8542                 if (refcount_read(&clp->cl_count) == 1)
8543                         goto out;
8544
8545                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8546                         rpc_restart_call_prepare(task);
8547                         return;
8548                 }
8549         }
8550         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8551 out:
8552         dprintk("<-- %s\n", __func__);
8553 }
8554
8555 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8556 {
8557         struct nfs4_sequence_data *calldata = data;
8558         struct nfs_client *clp = calldata->clp;
8559         struct nfs4_sequence_args *args;
8560         struct nfs4_sequence_res *res;
8561
8562         args = task->tk_msg.rpc_argp;
8563         res = task->tk_msg.rpc_resp;
8564
8565         nfs4_setup_sequence(clp, args, res, task);
8566 }
8567
8568 static const struct rpc_call_ops nfs41_sequence_ops = {
8569         .rpc_call_done = nfs41_sequence_call_done,
8570         .rpc_call_prepare = nfs41_sequence_prepare,
8571         .rpc_release = nfs41_sequence_release,
8572 };
8573
8574 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8575                 struct rpc_cred *cred,
8576                 struct nfs4_slot *slot,
8577                 bool is_privileged)
8578 {
8579         struct nfs4_sequence_data *calldata;
8580         struct rpc_message msg = {
8581                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8582                 .rpc_cred = cred,
8583         };
8584         struct rpc_task_setup task_setup_data = {
8585                 .rpc_client = clp->cl_rpcclient,
8586                 .rpc_message = &msg,
8587                 .callback_ops = &nfs41_sequence_ops,
8588                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8589         };
8590         struct rpc_task *ret;
8591
8592         ret = ERR_PTR(-EIO);
8593         if (!refcount_inc_not_zero(&clp->cl_count))
8594                 goto out_err;
8595
8596         ret = ERR_PTR(-ENOMEM);
8597         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8598         if (calldata == NULL)
8599                 goto out_put_clp;
8600         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8601         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8602         msg.rpc_argp = &calldata->args;
8603         msg.rpc_resp = &calldata->res;
8604         calldata->clp = clp;
8605         task_setup_data.callback_data = calldata;
8606
8607         ret = rpc_run_task(&task_setup_data);
8608         if (IS_ERR(ret))
8609                 goto out_err;
8610         return ret;
8611 out_put_clp:
8612         nfs_put_client(clp);
8613 out_err:
8614         nfs41_release_slot(slot);
8615         return ret;
8616 }
8617
8618 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8619 {
8620         struct rpc_task *task;
8621         int ret = 0;
8622
8623         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8624                 return -EAGAIN;
8625         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8626         if (IS_ERR(task))
8627                 ret = PTR_ERR(task);
8628         else
8629                 rpc_put_task_async(task);
8630         dprintk("<-- %s status=%d\n", __func__, ret);
8631         return ret;
8632 }
8633
8634 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8635 {
8636         struct rpc_task *task;
8637         int ret;
8638
8639         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8640         if (IS_ERR(task)) {
8641                 ret = PTR_ERR(task);
8642                 goto out;
8643         }
8644         ret = rpc_wait_for_completion_task(task);
8645         if (!ret)
8646                 ret = task->tk_status;
8647         rpc_put_task(task);
8648 out:
8649         dprintk("<-- %s status=%d\n", __func__, ret);
8650         return ret;
8651 }
8652
8653 struct nfs4_reclaim_complete_data {
8654         struct nfs_client *clp;
8655         struct nfs41_reclaim_complete_args arg;
8656         struct nfs41_reclaim_complete_res res;
8657 };
8658
8659 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8660 {
8661         struct nfs4_reclaim_complete_data *calldata = data;
8662
8663         nfs4_setup_sequence(calldata->clp,
8664                         &calldata->arg.seq_args,
8665                         &calldata->res.seq_res,
8666                         task);
8667 }
8668
8669 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8670 {
8671         switch(task->tk_status) {
8672         case 0:
8673                 wake_up_all(&clp->cl_lock_waitq);
8674                 /* Fallthrough */
8675         case -NFS4ERR_COMPLETE_ALREADY:
8676         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8677                 break;
8678         case -NFS4ERR_DELAY:
8679                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8680                 /* fall through */
8681         case -NFS4ERR_RETRY_UNCACHED_REP:
8682                 return -EAGAIN;
8683         case -NFS4ERR_BADSESSION:
8684         case -NFS4ERR_DEADSESSION:
8685         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8686                 nfs4_schedule_session_recovery(clp->cl_session,
8687                                 task->tk_status);
8688                 break;
8689         default:
8690                 nfs4_schedule_lease_recovery(clp);
8691         }
8692         return 0;
8693 }
8694
8695 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8696 {
8697         struct nfs4_reclaim_complete_data *calldata = data;
8698         struct nfs_client *clp = calldata->clp;
8699         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8700
8701         dprintk("--> %s\n", __func__);
8702         if (!nfs41_sequence_done(task, res))
8703                 return;
8704
8705         trace_nfs4_reclaim_complete(clp, task->tk_status);
8706         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8707                 rpc_restart_call_prepare(task);
8708                 return;
8709         }
8710         dprintk("<-- %s\n", __func__);
8711 }
8712
8713 static void nfs4_free_reclaim_complete_data(void *data)
8714 {
8715         struct nfs4_reclaim_complete_data *calldata = data;
8716
8717         kfree(calldata);
8718 }
8719
8720 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8721         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8722         .rpc_call_done = nfs4_reclaim_complete_done,
8723         .rpc_release = nfs4_free_reclaim_complete_data,
8724 };
8725
8726 /*
8727  * Issue a global reclaim complete.
8728  */
8729 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8730                 struct rpc_cred *cred)
8731 {
8732         struct nfs4_reclaim_complete_data *calldata;
8733         struct rpc_task *task;
8734         struct rpc_message msg = {
8735                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8736                 .rpc_cred = cred,
8737         };
8738         struct rpc_task_setup task_setup_data = {
8739                 .rpc_client = clp->cl_rpcclient,
8740                 .rpc_message = &msg,
8741                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8742                 .flags = RPC_TASK_ASYNC,
8743         };
8744         int status = -ENOMEM;
8745
8746         dprintk("--> %s\n", __func__);
8747         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8748         if (calldata == NULL)
8749                 goto out;
8750         calldata->clp = clp;
8751         calldata->arg.one_fs = 0;
8752
8753         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
8754         msg.rpc_argp = &calldata->arg;
8755         msg.rpc_resp = &calldata->res;
8756         task_setup_data.callback_data = calldata;
8757         task = rpc_run_task(&task_setup_data);
8758         if (IS_ERR(task)) {
8759                 status = PTR_ERR(task);
8760                 goto out;
8761         }
8762         status = rpc_wait_for_completion_task(task);
8763         if (status == 0)
8764                 status = task->tk_status;
8765         rpc_put_task(task);
8766 out:
8767         dprintk("<-- %s status=%d\n", __func__, status);
8768         return status;
8769 }
8770
8771 static void
8772 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8773 {
8774         struct nfs4_layoutget *lgp = calldata;
8775         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8776
8777         dprintk("--> %s\n", __func__);
8778         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8779                                 &lgp->res.seq_res, task);
8780         dprintk("<-- %s\n", __func__);
8781 }
8782
8783 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8784 {
8785         struct nfs4_layoutget *lgp = calldata;
8786
8787         dprintk("--> %s\n", __func__);
8788         nfs41_sequence_process(task, &lgp->res.seq_res);
8789         dprintk("<-- %s\n", __func__);
8790 }
8791
8792 static int
8793 nfs4_layoutget_handle_exception(struct rpc_task *task,
8794                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8795 {
8796         struct inode *inode = lgp->args.inode;
8797         struct nfs_server *server = NFS_SERVER(inode);
8798         struct pnfs_layout_hdr *lo;
8799         int nfs4err = task->tk_status;
8800         int err, status = 0;
8801         LIST_HEAD(head);
8802
8803         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8804
8805         nfs4_sequence_free_slot(&lgp->res.seq_res);
8806
8807         switch (nfs4err) {
8808         case 0:
8809                 goto out;
8810
8811         /*
8812          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8813          * on the file. set tk_status to -ENODATA to tell upper layer to
8814          * retry go inband.
8815          */
8816         case -NFS4ERR_LAYOUTUNAVAILABLE:
8817                 status = -ENODATA;
8818                 goto out;
8819         /*
8820          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8821          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8822          */
8823         case -NFS4ERR_BADLAYOUT:
8824                 status = -EOVERFLOW;
8825                 goto out;
8826         /*
8827          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8828          * (or clients) writing to the same RAID stripe except when
8829          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8830          *
8831          * Treat it like we would RECALLCONFLICT -- we retry for a little
8832          * while, and then eventually give up.
8833          */
8834         case -NFS4ERR_LAYOUTTRYLATER:
8835                 if (lgp->args.minlength == 0) {
8836                         status = -EOVERFLOW;
8837                         goto out;
8838                 }
8839                 status = -EBUSY;
8840                 break;
8841         case -NFS4ERR_RECALLCONFLICT:
8842                 status = -ERECALLCONFLICT;
8843                 break;
8844         case -NFS4ERR_DELEG_REVOKED:
8845         case -NFS4ERR_ADMIN_REVOKED:
8846         case -NFS4ERR_EXPIRED:
8847         case -NFS4ERR_BAD_STATEID:
8848                 exception->timeout = 0;
8849                 spin_lock(&inode->i_lock);
8850                 lo = NFS_I(inode)->layout;
8851                 /* If the open stateid was bad, then recover it. */
8852                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8853                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
8854                         spin_unlock(&inode->i_lock);
8855                         exception->state = lgp->args.ctx->state;
8856                         exception->stateid = &lgp->args.stateid;
8857                         break;
8858                 }
8859
8860                 /*
8861                  * Mark the bad layout state as invalid, then retry
8862                  */
8863                 pnfs_mark_layout_stateid_invalid(lo, &head);
8864                 spin_unlock(&inode->i_lock);
8865                 nfs_commit_inode(inode, 0);
8866                 pnfs_free_lseg_list(&head);
8867                 status = -EAGAIN;
8868                 goto out;
8869         }
8870
8871         err = nfs4_handle_exception(server, nfs4err, exception);
8872         if (!status) {
8873                 if (exception->retry)
8874                         status = -EAGAIN;
8875                 else
8876                         status = err;
8877         }
8878 out:
8879         dprintk("<-- %s\n", __func__);
8880         return status;
8881 }
8882
8883 size_t max_response_pages(struct nfs_server *server)
8884 {
8885         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8886         return nfs_page_array_len(0, max_resp_sz);
8887 }
8888
8889 static void nfs4_layoutget_release(void *calldata)
8890 {
8891         struct nfs4_layoutget *lgp = calldata;
8892
8893         dprintk("--> %s\n", __func__);
8894         nfs4_sequence_free_slot(&lgp->res.seq_res);
8895         pnfs_layoutget_free(lgp);
8896         dprintk("<-- %s\n", __func__);
8897 }
8898
8899 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8900         .rpc_call_prepare = nfs4_layoutget_prepare,
8901         .rpc_call_done = nfs4_layoutget_done,
8902         .rpc_release = nfs4_layoutget_release,
8903 };
8904
8905 struct pnfs_layout_segment *
8906 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
8907 {
8908         struct inode *inode = lgp->args.inode;
8909         struct nfs_server *server = NFS_SERVER(inode);
8910         struct rpc_task *task;
8911         struct rpc_message msg = {
8912                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8913                 .rpc_argp = &lgp->args,
8914                 .rpc_resp = &lgp->res,
8915                 .rpc_cred = lgp->cred,
8916         };
8917         struct rpc_task_setup task_setup_data = {
8918                 .rpc_client = server->client,
8919                 .rpc_message = &msg,
8920                 .callback_ops = &nfs4_layoutget_call_ops,
8921                 .callback_data = lgp,
8922                 .flags = RPC_TASK_ASYNC,
8923         };
8924         struct pnfs_layout_segment *lseg = NULL;
8925         struct nfs4_exception exception = {
8926                 .inode = inode,
8927                 .timeout = *timeout,
8928         };
8929         int status = 0;
8930
8931         dprintk("--> %s\n", __func__);
8932
8933         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8934         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8935
8936         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
8937
8938         task = rpc_run_task(&task_setup_data);
8939         if (IS_ERR(task))
8940                 return ERR_CAST(task);
8941         status = rpc_wait_for_completion_task(task);
8942         if (status != 0)
8943                 goto out;
8944
8945         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8946         if (task->tk_status < 0 || lgp->res.layoutp->len == 0) {
8947                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8948                 *timeout = exception.timeout;
8949         } else
8950                 lseg = pnfs_layout_process(lgp);
8951 out:
8952         trace_nfs4_layoutget(lgp->args.ctx,
8953                         &lgp->args.range,
8954                         &lgp->res.range,
8955                         &lgp->res.stateid,
8956                         status);
8957
8958         rpc_put_task(task);
8959         dprintk("<-- %s status=%d\n", __func__, status);
8960         if (status)
8961                 return ERR_PTR(status);
8962         return lseg;
8963 }
8964
8965 static void
8966 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8967 {
8968         struct nfs4_layoutreturn *lrp = calldata;
8969
8970         dprintk("--> %s\n", __func__);
8971         nfs4_setup_sequence(lrp->clp,
8972                         &lrp->args.seq_args,
8973                         &lrp->res.seq_res,
8974                         task);
8975         if (!pnfs_layout_is_valid(lrp->args.layout))
8976                 rpc_exit(task, 0);
8977 }
8978
8979 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8980 {
8981         struct nfs4_layoutreturn *lrp = calldata;
8982         struct nfs_server *server;
8983
8984         dprintk("--> %s\n", __func__);
8985
8986         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8987                 return;
8988
8989         server = NFS_SERVER(lrp->args.inode);
8990         switch (task->tk_status) {
8991         case -NFS4ERR_OLD_STATEID:
8992                 if (nfs4_layoutreturn_refresh_stateid(&lrp->args.stateid,
8993                                         &lrp->args.range,
8994                                         lrp->args.inode))
8995                         goto out_restart;
8996                 /* Fallthrough */
8997         default:
8998                 task->tk_status = 0;
8999                 /* Fallthrough */
9000         case 0:
9001                 break;
9002         case -NFS4ERR_DELAY:
9003                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9004                         break;
9005                 goto out_restart;
9006         }
9007         dprintk("<-- %s\n", __func__);
9008         return;
9009 out_restart:
9010         task->tk_status = 0;
9011         nfs4_sequence_free_slot(&lrp->res.seq_res);
9012         rpc_restart_call_prepare(task);
9013 }
9014
9015 static void nfs4_layoutreturn_release(void *calldata)
9016 {
9017         struct nfs4_layoutreturn *lrp = calldata;
9018         struct pnfs_layout_hdr *lo = lrp->args.layout;
9019
9020         dprintk("--> %s\n", __func__);
9021         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9022                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9023         nfs4_sequence_free_slot(&lrp->res.seq_res);
9024         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9025                 lrp->ld_private.ops->free(&lrp->ld_private);
9026         pnfs_put_layout_hdr(lrp->args.layout);
9027         nfs_iput_and_deactive(lrp->inode);
9028         kfree(calldata);
9029         dprintk("<-- %s\n", __func__);
9030 }
9031
9032 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9033         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9034         .rpc_call_done = nfs4_layoutreturn_done,
9035         .rpc_release = nfs4_layoutreturn_release,
9036 };
9037
9038 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9039 {
9040         struct rpc_task *task;
9041         struct rpc_message msg = {
9042                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9043                 .rpc_argp = &lrp->args,
9044                 .rpc_resp = &lrp->res,
9045                 .rpc_cred = lrp->cred,
9046         };
9047         struct rpc_task_setup task_setup_data = {
9048                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9049                 .rpc_message = &msg,
9050                 .callback_ops = &nfs4_layoutreturn_call_ops,
9051                 .callback_data = lrp,
9052         };
9053         int status = 0;
9054
9055         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9056                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9057                         &task_setup_data.rpc_client, &msg);
9058
9059         dprintk("--> %s\n", __func__);
9060         if (!sync) {
9061                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9062                 if (!lrp->inode) {
9063                         nfs4_layoutreturn_release(lrp);
9064                         return -EAGAIN;
9065                 }
9066                 task_setup_data.flags |= RPC_TASK_ASYNC;
9067         }
9068         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9069         task = rpc_run_task(&task_setup_data);
9070         if (IS_ERR(task))
9071                 return PTR_ERR(task);
9072         if (sync)
9073                 status = task->tk_status;
9074         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9075         dprintk("<-- %s status=%d\n", __func__, status);
9076         rpc_put_task(task);
9077         return status;
9078 }
9079
9080 static int
9081 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9082                 struct pnfs_device *pdev,
9083                 struct rpc_cred *cred)
9084 {
9085         struct nfs4_getdeviceinfo_args args = {
9086                 .pdev = pdev,
9087                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9088                         NOTIFY_DEVICEID4_DELETE,
9089         };
9090         struct nfs4_getdeviceinfo_res res = {
9091                 .pdev = pdev,
9092         };
9093         struct rpc_message msg = {
9094                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9095                 .rpc_argp = &args,
9096                 .rpc_resp = &res,
9097                 .rpc_cred = cred,
9098         };
9099         int status;
9100
9101         dprintk("--> %s\n", __func__);
9102         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9103         if (res.notification & ~args.notify_types)
9104                 dprintk("%s: unsupported notification\n", __func__);
9105         if (res.notification != args.notify_types)
9106                 pdev->nocache = 1;
9107
9108         dprintk("<-- %s status=%d\n", __func__, status);
9109
9110         return status;
9111 }
9112
9113 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9114                 struct pnfs_device *pdev,
9115                 struct rpc_cred *cred)
9116 {
9117         struct nfs4_exception exception = { };
9118         int err;
9119
9120         do {
9121                 err = nfs4_handle_exception(server,
9122                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9123                                         &exception);
9124         } while (exception.retry);
9125         return err;
9126 }
9127 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9128
9129 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9130 {
9131         struct nfs4_layoutcommit_data *data = calldata;
9132         struct nfs_server *server = NFS_SERVER(data->args.inode);
9133
9134         nfs4_setup_sequence(server->nfs_client,
9135                         &data->args.seq_args,
9136                         &data->res.seq_res,
9137                         task);
9138 }
9139
9140 static void
9141 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9142 {
9143         struct nfs4_layoutcommit_data *data = calldata;
9144         struct nfs_server *server = NFS_SERVER(data->args.inode);
9145
9146         if (!nfs41_sequence_done(task, &data->res.seq_res))
9147                 return;
9148
9149         switch (task->tk_status) { /* Just ignore these failures */
9150         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9151         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9152         case -NFS4ERR_BADLAYOUT:     /* no layout */
9153         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9154                 task->tk_status = 0;
9155         case 0:
9156                 break;
9157         default:
9158                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9159                         rpc_restart_call_prepare(task);
9160                         return;
9161                 }
9162         }
9163 }
9164
9165 static void nfs4_layoutcommit_release(void *calldata)
9166 {
9167         struct nfs4_layoutcommit_data *data = calldata;
9168
9169         pnfs_cleanup_layoutcommit(data);
9170         nfs_post_op_update_inode_force_wcc(data->args.inode,
9171                                            data->res.fattr);
9172         put_rpccred(data->cred);
9173         nfs_iput_and_deactive(data->inode);
9174         kfree(data);
9175 }
9176
9177 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9178         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9179         .rpc_call_done = nfs4_layoutcommit_done,
9180         .rpc_release = nfs4_layoutcommit_release,
9181 };
9182
9183 int
9184 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9185 {
9186         struct rpc_message msg = {
9187                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9188                 .rpc_argp = &data->args,
9189                 .rpc_resp = &data->res,
9190                 .rpc_cred = data->cred,
9191         };
9192         struct rpc_task_setup task_setup_data = {
9193                 .task = &data->task,
9194                 .rpc_client = NFS_CLIENT(data->args.inode),
9195                 .rpc_message = &msg,
9196                 .callback_ops = &nfs4_layoutcommit_ops,
9197                 .callback_data = data,
9198         };
9199         struct rpc_task *task;
9200         int status = 0;
9201
9202         dprintk("NFS: initiating layoutcommit call. sync %d "
9203                 "lbw: %llu inode %lu\n", sync,
9204                 data->args.lastbytewritten,
9205                 data->args.inode->i_ino);
9206
9207         if (!sync) {
9208                 data->inode = nfs_igrab_and_active(data->args.inode);
9209                 if (data->inode == NULL) {
9210                         nfs4_layoutcommit_release(data);
9211                         return -EAGAIN;
9212                 }
9213                 task_setup_data.flags = RPC_TASK_ASYNC;
9214         }
9215         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9216         task = rpc_run_task(&task_setup_data);
9217         if (IS_ERR(task))
9218                 return PTR_ERR(task);
9219         if (sync)
9220                 status = task->tk_status;
9221         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9222         dprintk("%s: status %d\n", __func__, status);
9223         rpc_put_task(task);
9224         return status;
9225 }
9226
9227 /**
9228  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9229  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9230  */
9231 static int
9232 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9233                     struct nfs_fsinfo *info,
9234                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9235 {
9236         struct nfs41_secinfo_no_name_args args = {
9237                 .style = SECINFO_STYLE_CURRENT_FH,
9238         };
9239         struct nfs4_secinfo_res res = {
9240                 .flavors = flavors,
9241         };
9242         struct rpc_message msg = {
9243                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9244                 .rpc_argp = &args,
9245                 .rpc_resp = &res,
9246         };
9247         struct rpc_clnt *clnt = server->client;
9248         struct rpc_cred *cred = NULL;
9249         int status;
9250
9251         if (use_integrity) {
9252                 clnt = server->nfs_client->cl_rpcclient;
9253                 cred = nfs4_get_clid_cred(server->nfs_client);
9254                 msg.rpc_cred = cred;
9255         }
9256
9257         dprintk("--> %s\n", __func__);
9258         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
9259                                 &res.seq_res, 0);
9260         dprintk("<-- %s status=%d\n", __func__, status);
9261
9262         if (cred)
9263                 put_rpccred(cred);
9264
9265         return status;
9266 }
9267
9268 static int
9269 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9270                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9271 {
9272         struct nfs4_exception exception = { };
9273         int err;
9274         do {
9275                 /* first try using integrity protection */
9276                 err = -NFS4ERR_WRONGSEC;
9277
9278                 /* try to use integrity protection with machine cred */
9279                 if (_nfs4_is_integrity_protected(server->nfs_client))
9280                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9281                                                           flavors, true);
9282
9283                 /*
9284                  * if unable to use integrity protection, or SECINFO with
9285                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9286                  * disallowed by spec, but exists in deployed servers) use
9287                  * the current filesystem's rpc_client and the user cred.
9288                  */
9289                 if (err == -NFS4ERR_WRONGSEC)
9290                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9291                                                           flavors, false);
9292
9293                 switch (err) {
9294                 case 0:
9295                 case -NFS4ERR_WRONGSEC:
9296                 case -ENOTSUPP:
9297                         goto out;
9298                 default:
9299                         err = nfs4_handle_exception(server, err, &exception);
9300                 }
9301         } while (exception.retry);
9302 out:
9303         return err;
9304 }
9305
9306 static int
9307 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9308                     struct nfs_fsinfo *info)
9309 {
9310         int err;
9311         struct page *page;
9312         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9313         struct nfs4_secinfo_flavors *flavors;
9314         struct nfs4_secinfo4 *secinfo;
9315         int i;
9316
9317         page = alloc_page(GFP_KERNEL);
9318         if (!page) {
9319                 err = -ENOMEM;
9320                 goto out;
9321         }
9322
9323         flavors = page_address(page);
9324         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9325
9326         /*
9327          * Fall back on "guess and check" method if
9328          * the server doesn't support SECINFO_NO_NAME
9329          */
9330         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9331                 err = nfs4_find_root_sec(server, fhandle, info);
9332                 goto out_freepage;
9333         }
9334         if (err)
9335                 goto out_freepage;
9336
9337         for (i = 0; i < flavors->num_flavors; i++) {
9338                 secinfo = &flavors->flavors[i];
9339
9340                 switch (secinfo->flavor) {
9341                 case RPC_AUTH_NULL:
9342                 case RPC_AUTH_UNIX:
9343                 case RPC_AUTH_GSS:
9344                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9345                                         &secinfo->flavor_info);
9346                         break;
9347                 default:
9348                         flavor = RPC_AUTH_MAXFLAVOR;
9349                         break;
9350                 }
9351
9352                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9353                         flavor = RPC_AUTH_MAXFLAVOR;
9354
9355                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9356                         err = nfs4_lookup_root_sec(server, fhandle,
9357                                                    info, flavor);
9358                         if (!err)
9359                                 break;
9360                 }
9361         }
9362
9363         if (flavor == RPC_AUTH_MAXFLAVOR)
9364                 err = -EPERM;
9365
9366 out_freepage:
9367         put_page(page);
9368         if (err == -EACCES)
9369                 return -EPERM;
9370 out:
9371         return err;
9372 }
9373
9374 static int _nfs41_test_stateid(struct nfs_server *server,
9375                 nfs4_stateid *stateid,
9376                 struct rpc_cred *cred)
9377 {
9378         int status;
9379         struct nfs41_test_stateid_args args = {
9380                 .stateid = stateid,
9381         };
9382         struct nfs41_test_stateid_res res;
9383         struct rpc_message msg = {
9384                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9385                 .rpc_argp = &args,
9386                 .rpc_resp = &res,
9387                 .rpc_cred = cred,
9388         };
9389         struct rpc_clnt *rpc_client = server->client;
9390
9391         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9392                 &rpc_client, &msg);
9393
9394         dprintk("NFS call  test_stateid %p\n", stateid);
9395         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9396         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9397                         &args.seq_args, &res.seq_res);
9398         if (status != NFS_OK) {
9399                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9400                 return status;
9401         }
9402         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9403         return -res.status;
9404 }
9405
9406 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9407                 int err, struct nfs4_exception *exception)
9408 {
9409         exception->retry = 0;
9410         switch(err) {
9411         case -NFS4ERR_DELAY:
9412         case -NFS4ERR_RETRY_UNCACHED_REP:
9413                 nfs4_handle_exception(server, err, exception);
9414                 break;
9415         case -NFS4ERR_BADSESSION:
9416         case -NFS4ERR_BADSLOT:
9417         case -NFS4ERR_BAD_HIGH_SLOT:
9418         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9419         case -NFS4ERR_DEADSESSION:
9420                 nfs4_do_handle_exception(server, err, exception);
9421         }
9422 }
9423
9424 /**
9425  * nfs41_test_stateid - perform a TEST_STATEID operation
9426  *
9427  * @server: server / transport on which to perform the operation
9428  * @stateid: state ID to test
9429  * @cred: credential
9430  *
9431  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9432  * Otherwise a negative NFS4ERR value is returned if the operation
9433  * failed or the state ID is not currently valid.
9434  */
9435 static int nfs41_test_stateid(struct nfs_server *server,
9436                 nfs4_stateid *stateid,
9437                 struct rpc_cred *cred)
9438 {
9439         struct nfs4_exception exception = { };
9440         int err;
9441         do {
9442                 err = _nfs41_test_stateid(server, stateid, cred);
9443                 nfs4_handle_delay_or_session_error(server, err, &exception);
9444         } while (exception.retry);
9445         return err;
9446 }
9447
9448 struct nfs_free_stateid_data {
9449         struct nfs_server *server;
9450         struct nfs41_free_stateid_args args;
9451         struct nfs41_free_stateid_res res;
9452 };
9453
9454 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9455 {
9456         struct nfs_free_stateid_data *data = calldata;
9457         nfs4_setup_sequence(data->server->nfs_client,
9458                         &data->args.seq_args,
9459                         &data->res.seq_res,
9460                         task);
9461 }
9462
9463 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9464 {
9465         struct nfs_free_stateid_data *data = calldata;
9466
9467         nfs41_sequence_done(task, &data->res.seq_res);
9468
9469         switch (task->tk_status) {
9470         case -NFS4ERR_DELAY:
9471                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9472                         rpc_restart_call_prepare(task);
9473         }
9474 }
9475
9476 static void nfs41_free_stateid_release(void *calldata)
9477 {
9478         kfree(calldata);
9479 }
9480
9481 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9482         .rpc_call_prepare = nfs41_free_stateid_prepare,
9483         .rpc_call_done = nfs41_free_stateid_done,
9484         .rpc_release = nfs41_free_stateid_release,
9485 };
9486
9487 /**
9488  * nfs41_free_stateid - perform a FREE_STATEID operation
9489  *
9490  * @server: server / transport on which to perform the operation
9491  * @stateid: state ID to release
9492  * @cred: credential
9493  * @is_recovery: set to true if this call needs to be privileged
9494  *
9495  * Note: this function is always asynchronous.
9496  */
9497 static int nfs41_free_stateid(struct nfs_server *server,
9498                 const nfs4_stateid *stateid,
9499                 struct rpc_cred *cred,
9500                 bool privileged)
9501 {
9502         struct rpc_message msg = {
9503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9504                 .rpc_cred = cred,
9505         };
9506         struct rpc_task_setup task_setup = {
9507                 .rpc_client = server->client,
9508                 .rpc_message = &msg,
9509                 .callback_ops = &nfs41_free_stateid_ops,
9510                 .flags = RPC_TASK_ASYNC,
9511         };
9512         struct nfs_free_stateid_data *data;
9513         struct rpc_task *task;
9514
9515         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9516                 &task_setup.rpc_client, &msg);
9517
9518         dprintk("NFS call  free_stateid %p\n", stateid);
9519         data = kmalloc(sizeof(*data), GFP_NOFS);
9520         if (!data)
9521                 return -ENOMEM;
9522         data->server = server;
9523         nfs4_stateid_copy(&data->args.stateid, stateid);
9524
9525         task_setup.callback_data = data;
9526
9527         msg.rpc_argp = &data->args;
9528         msg.rpc_resp = &data->res;
9529         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9530         task = rpc_run_task(&task_setup);
9531         if (IS_ERR(task))
9532                 return PTR_ERR(task);
9533         rpc_put_task(task);
9534         return 0;
9535 }
9536
9537 static void
9538 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9539 {
9540         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9541
9542         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9543         nfs4_free_lock_state(server, lsp);
9544 }
9545
9546 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9547                 const nfs4_stateid *s2)
9548 {
9549         if (s1->type != s2->type)
9550                 return false;
9551
9552         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9553                 return false;
9554
9555         if (s1->seqid == s2->seqid)
9556                 return true;
9557
9558         return s1->seqid == 0 || s2->seqid == 0;
9559 }
9560
9561 #endif /* CONFIG_NFS_V4_1 */
9562
9563 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9564                 const nfs4_stateid *s2)
9565 {
9566         return nfs4_stateid_match(s1, s2);
9567 }
9568
9569
9570 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9571         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9572         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9573         .recover_open   = nfs4_open_reclaim,
9574         .recover_lock   = nfs4_lock_reclaim,
9575         .establish_clid = nfs4_init_clientid,
9576         .detect_trunking = nfs40_discover_server_trunking,
9577 };
9578
9579 #if defined(CONFIG_NFS_V4_1)
9580 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9581         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9582         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9583         .recover_open   = nfs4_open_reclaim,
9584         .recover_lock   = nfs4_lock_reclaim,
9585         .establish_clid = nfs41_init_clientid,
9586         .reclaim_complete = nfs41_proc_reclaim_complete,
9587         .detect_trunking = nfs41_discover_server_trunking,
9588 };
9589 #endif /* CONFIG_NFS_V4_1 */
9590
9591 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9592         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9593         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9594         .recover_open   = nfs40_open_expired,
9595         .recover_lock   = nfs4_lock_expired,
9596         .establish_clid = nfs4_init_clientid,
9597 };
9598
9599 #if defined(CONFIG_NFS_V4_1)
9600 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9601         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9602         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9603         .recover_open   = nfs41_open_expired,
9604         .recover_lock   = nfs41_lock_expired,
9605         .establish_clid = nfs41_init_clientid,
9606 };
9607 #endif /* CONFIG_NFS_V4_1 */
9608
9609 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9610         .sched_state_renewal = nfs4_proc_async_renew,
9611         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9612         .renew_lease = nfs4_proc_renew,
9613 };
9614
9615 #if defined(CONFIG_NFS_V4_1)
9616 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9617         .sched_state_renewal = nfs41_proc_async_sequence,
9618         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9619         .renew_lease = nfs4_proc_sequence,
9620 };
9621 #endif
9622
9623 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9624         .get_locations = _nfs40_proc_get_locations,
9625         .fsid_present = _nfs40_proc_fsid_present,
9626 };
9627
9628 #if defined(CONFIG_NFS_V4_1)
9629 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9630         .get_locations = _nfs41_proc_get_locations,
9631         .fsid_present = _nfs41_proc_fsid_present,
9632 };
9633 #endif  /* CONFIG_NFS_V4_1 */
9634
9635 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9636         .minor_version = 0,
9637         .init_caps = NFS_CAP_READDIRPLUS
9638                 | NFS_CAP_ATOMIC_OPEN
9639                 | NFS_CAP_POSIX_LOCK,
9640         .init_client = nfs40_init_client,
9641         .shutdown_client = nfs40_shutdown_client,
9642         .match_stateid = nfs4_match_stateid,
9643         .find_root_sec = nfs4_find_root_sec,
9644         .free_lock_state = nfs4_release_lockowner,
9645         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9646         .alloc_seqid = nfs_alloc_seqid,
9647         .call_sync_ops = &nfs40_call_sync_ops,
9648         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9649         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9650         .state_renewal_ops = &nfs40_state_renewal_ops,
9651         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9652 };
9653
9654 #if defined(CONFIG_NFS_V4_1)
9655 static struct nfs_seqid *
9656 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9657 {
9658         return NULL;
9659 }
9660
9661 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9662         .minor_version = 1,
9663         .init_caps = NFS_CAP_READDIRPLUS
9664                 | NFS_CAP_ATOMIC_OPEN
9665                 | NFS_CAP_POSIX_LOCK
9666                 | NFS_CAP_STATEID_NFSV41
9667                 | NFS_CAP_ATOMIC_OPEN_V1
9668                 | NFS_CAP_LGOPEN,
9669         .init_client = nfs41_init_client,
9670         .shutdown_client = nfs41_shutdown_client,
9671         .match_stateid = nfs41_match_stateid,
9672         .find_root_sec = nfs41_find_root_sec,
9673         .free_lock_state = nfs41_free_lock_state,
9674         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9675         .alloc_seqid = nfs_alloc_no_seqid,
9676         .session_trunk = nfs4_test_session_trunk,
9677         .call_sync_ops = &nfs41_call_sync_ops,
9678         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9679         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9680         .state_renewal_ops = &nfs41_state_renewal_ops,
9681         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9682 };
9683 #endif
9684
9685 #if defined(CONFIG_NFS_V4_2)
9686 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9687         .minor_version = 2,
9688         .init_caps = NFS_CAP_READDIRPLUS
9689                 | NFS_CAP_ATOMIC_OPEN
9690                 | NFS_CAP_POSIX_LOCK
9691                 | NFS_CAP_STATEID_NFSV41
9692                 | NFS_CAP_ATOMIC_OPEN_V1
9693                 | NFS_CAP_LGOPEN
9694                 | NFS_CAP_ALLOCATE
9695                 | NFS_CAP_COPY
9696                 | NFS_CAP_OFFLOAD_CANCEL
9697                 | NFS_CAP_DEALLOCATE
9698                 | NFS_CAP_SEEK
9699                 | NFS_CAP_LAYOUTSTATS
9700                 | NFS_CAP_CLONE,
9701         .init_client = nfs41_init_client,
9702         .shutdown_client = nfs41_shutdown_client,
9703         .match_stateid = nfs41_match_stateid,
9704         .find_root_sec = nfs41_find_root_sec,
9705         .free_lock_state = nfs41_free_lock_state,
9706         .call_sync_ops = &nfs41_call_sync_ops,
9707         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9708         .alloc_seqid = nfs_alloc_no_seqid,
9709         .session_trunk = nfs4_test_session_trunk,
9710         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9711         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9712         .state_renewal_ops = &nfs41_state_renewal_ops,
9713         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9714 };
9715 #endif
9716
9717 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9718         [0] = &nfs_v4_0_minor_ops,
9719 #if defined(CONFIG_NFS_V4_1)
9720         [1] = &nfs_v4_1_minor_ops,
9721 #endif
9722 #if defined(CONFIG_NFS_V4_2)
9723         [2] = &nfs_v4_2_minor_ops,
9724 #endif
9725 };
9726
9727 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9728 {
9729         ssize_t error, error2;
9730
9731         error = generic_listxattr(dentry, list, size);
9732         if (error < 0)
9733                 return error;
9734         if (list) {
9735                 list += error;
9736                 size -= error;
9737         }
9738
9739         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9740         if (error2 < 0)
9741                 return error2;
9742         return error + error2;
9743 }
9744
9745 static const struct inode_operations nfs4_dir_inode_operations = {
9746         .create         = nfs_create,
9747         .lookup         = nfs_lookup,
9748         .atomic_open    = nfs_atomic_open,
9749         .link           = nfs_link,
9750         .unlink         = nfs_unlink,
9751         .symlink        = nfs_symlink,
9752         .mkdir          = nfs_mkdir,
9753         .rmdir          = nfs_rmdir,
9754         .mknod          = nfs_mknod,
9755         .rename         = nfs_rename,
9756         .permission     = nfs_permission,
9757         .getattr        = nfs_getattr,
9758         .setattr        = nfs_setattr,
9759         .listxattr      = nfs4_listxattr,
9760 };
9761
9762 static const struct inode_operations nfs4_file_inode_operations = {
9763         .permission     = nfs_permission,
9764         .getattr        = nfs_getattr,
9765         .setattr        = nfs_setattr,
9766         .listxattr      = nfs4_listxattr,
9767 };
9768
9769 const struct nfs_rpc_ops nfs_v4_clientops = {
9770         .version        = 4,                    /* protocol version */
9771         .dentry_ops     = &nfs4_dentry_operations,
9772         .dir_inode_ops  = &nfs4_dir_inode_operations,
9773         .file_inode_ops = &nfs4_file_inode_operations,
9774         .file_ops       = &nfs4_file_operations,
9775         .getroot        = nfs4_proc_get_root,
9776         .submount       = nfs4_submount,
9777         .try_mount      = nfs4_try_mount,
9778         .getattr        = nfs4_proc_getattr,
9779         .setattr        = nfs4_proc_setattr,
9780         .lookup         = nfs4_proc_lookup,
9781         .lookupp        = nfs4_proc_lookupp,
9782         .access         = nfs4_proc_access,
9783         .readlink       = nfs4_proc_readlink,
9784         .create         = nfs4_proc_create,
9785         .remove         = nfs4_proc_remove,
9786         .unlink_setup   = nfs4_proc_unlink_setup,
9787         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9788         .unlink_done    = nfs4_proc_unlink_done,
9789         .rename_setup   = nfs4_proc_rename_setup,
9790         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9791         .rename_done    = nfs4_proc_rename_done,
9792         .link           = nfs4_proc_link,
9793         .symlink        = nfs4_proc_symlink,
9794         .mkdir          = nfs4_proc_mkdir,
9795         .rmdir          = nfs4_proc_rmdir,
9796         .readdir        = nfs4_proc_readdir,
9797         .mknod          = nfs4_proc_mknod,
9798         .statfs         = nfs4_proc_statfs,
9799         .fsinfo         = nfs4_proc_fsinfo,
9800         .pathconf       = nfs4_proc_pathconf,
9801         .set_capabilities = nfs4_server_capabilities,
9802         .decode_dirent  = nfs4_decode_dirent,
9803         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9804         .read_setup     = nfs4_proc_read_setup,
9805         .read_done      = nfs4_read_done,
9806         .write_setup    = nfs4_proc_write_setup,
9807         .write_done     = nfs4_write_done,
9808         .commit_setup   = nfs4_proc_commit_setup,
9809         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9810         .commit_done    = nfs4_commit_done,
9811         .lock           = nfs4_proc_lock,
9812         .clear_acl_cache = nfs4_zap_acl_attr,
9813         .close_context  = nfs4_close_context,
9814         .open_context   = nfs4_atomic_open,
9815         .have_delegation = nfs4_have_delegation,
9816         .alloc_client   = nfs4_alloc_client,
9817         .init_client    = nfs4_init_client,
9818         .free_client    = nfs4_free_client,
9819         .create_server  = nfs4_create_server,
9820         .clone_server   = nfs_clone_server,
9821 };
9822
9823 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9824         .name   = XATTR_NAME_NFSV4_ACL,
9825         .list   = nfs4_xattr_list_nfs4_acl,
9826         .get    = nfs4_xattr_get_nfs4_acl,
9827         .set    = nfs4_xattr_set_nfs4_acl,
9828 };
9829
9830 const struct xattr_handler *nfs4_xattr_handlers[] = {
9831         &nfs4_xattr_nfs4_acl_handler,
9832 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9833         &nfs4_xattr_nfs4_label_handler,
9834 #endif
9835         NULL
9836 };
9837
9838 /*
9839  * Local variables:
9840  *  c-basic-offset: 8
9841  * End:
9842  */