fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <[email protected]>
  10  *  Andy Adamson   <[email protected]>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57 #include <linux/iversion.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_BITMASK_SZ         3
  75
  76 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  77 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  78
  79 /* file attributes which can be mapped to nfs attributes */
  80 #define NFS4_VALID_ATTRS (ATTR_MODE \
  81         | ATTR_UID \
  82         | ATTR_GID \
  83         | ATTR_SIZE \
  84         | ATTR_ATIME \
  85         | ATTR_MTIME \
  86         | ATTR_CTIME \
  87         | ATTR_ATIME_SET \
  88         | ATTR_MTIME_SET)
  89
  90 struct nfs4_opendata;
  91 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  92 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  93 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  94 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label, struct inode *inode);
  95 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  96 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  97                             struct nfs_fattr *fattr, struct iattr *sattr,
  98                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  99                             struct nfs4_label *olabel);
 100 #ifdef CONFIG_NFS_V4_1
 101 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 102                 struct rpc_cred *cred,
 103                 struct nfs4_slot *slot,
 104                 bool is_privileged);
 105 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 106                 struct rpc_cred *);
 107 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 108                 struct rpc_cred *, bool);
 109 #endif
 110
 111 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 112 static inline struct nfs4_label *
 113 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 114         struct iattr *sattr, struct nfs4_label *label)
 115 {
 116         int err;
 117
 118         if (label == NULL)
 119                 return NULL;
 120
 121         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 122                 return NULL;
 123
 124         err = security_dentry_init_security(dentry, sattr->ia_mode,
 125                                 &dentry->d_name, (void **)&label->label, &label->len);
 126         if (err == 0)
 127                 return label;
 128
 129         return NULL;
 130 }
 131 static inline void
 132 nfs4_label_release_security(struct nfs4_label *label)
 133 {
 134         if (label)
 135                 security_release_secctx(label->label, label->len);
 136 }
 137 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 138 {
 139         if (label)
 140                 return server->attr_bitmask;
 141
 142         return server->attr_bitmask_nl;
 143 }
 144 #else
 145 static inline struct nfs4_label *
 146 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 147         struct iattr *sattr, struct nfs4_label *l)
 148 { return NULL; }
 149 static inline void
 150 nfs4_label_release_security(struct nfs4_label *label)
 151 { return; }
 152 static inline u32 *
 153 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 154 { return server->attr_bitmask; }
 155 #endif
 156
 157 /* Prevent leaks of NFSv4 errors into userland */
 158 static int nfs4_map_errors(int err)
 159 {
 160         if (err >= -1000)
 161                 return err;
 162         switch (err) {
 163         case -NFS4ERR_RESOURCE:
 164         case -NFS4ERR_LAYOUTTRYLATER:
 165         case -NFS4ERR_RECALLCONFLICT:
 166                 return -EREMOTEIO;
 167         case -NFS4ERR_WRONGSEC:
 168         case -NFS4ERR_WRONG_CRED:
 169                 return -EPERM;
 170         case -NFS4ERR_BADOWNER:
 171         case -NFS4ERR_BADNAME:
 172                 return -EINVAL;
 173         case -NFS4ERR_SHARE_DENIED:
 174                 return -EACCES;
 175         case -NFS4ERR_MINOR_VERS_MISMATCH:
 176                 return -EPROTONOSUPPORT;
 177         case -NFS4ERR_FILE_OPEN:
 178                 return -EBUSY;
 179         default:
 180                 dprintk("%s could not handle NFSv4 error %d\n",
 181                                 __func__, -err);
 182                 break;
 183         }
 184         return -EIO;
 185 }
 186
 187 /*
 188  * This is our standard bitmap for GETATTR requests.
 189  */
 190 const u32 nfs4_fattr_bitmap[3] = {
 191         FATTR4_WORD0_TYPE
 192         | FATTR4_WORD0_CHANGE
 193         | FATTR4_WORD0_SIZE
 194         | FATTR4_WORD0_FSID
 195         | FATTR4_WORD0_FILEID,
 196         FATTR4_WORD1_MODE
 197         | FATTR4_WORD1_NUMLINKS
 198         | FATTR4_WORD1_OWNER
 199         | FATTR4_WORD1_OWNER_GROUP
 200         | FATTR4_WORD1_RAWDEV
 201         | FATTR4_WORD1_SPACE_USED
 202         | FATTR4_WORD1_TIME_ACCESS
 203         | FATTR4_WORD1_TIME_METADATA
 204         | FATTR4_WORD1_TIME_MODIFY
 205         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 206 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 207         FATTR4_WORD2_SECURITY_LABEL
 208 #endif
 209 };
 210
 211 static const u32 nfs4_pnfs_open_bitmap[3] = {
 212         FATTR4_WORD0_TYPE
 213         | FATTR4_WORD0_CHANGE
 214         | FATTR4_WORD0_SIZE
 215         | FATTR4_WORD0_FSID
 216         | FATTR4_WORD0_FILEID,
 217         FATTR4_WORD1_MODE
 218         | FATTR4_WORD1_NUMLINKS
 219         | FATTR4_WORD1_OWNER
 220         | FATTR4_WORD1_OWNER_GROUP
 221         | FATTR4_WORD1_RAWDEV
 222         | FATTR4_WORD1_SPACE_USED
 223         | FATTR4_WORD1_TIME_ACCESS
 224         | FATTR4_WORD1_TIME_METADATA
 225         | FATTR4_WORD1_TIME_MODIFY,
 226         FATTR4_WORD2_MDSTHRESHOLD
 227 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 228         | FATTR4_WORD2_SECURITY_LABEL
 229 #endif
 230 };
 231
 232 static const u32 nfs4_open_noattr_bitmap[3] = {
 233         FATTR4_WORD0_TYPE
 234         | FATTR4_WORD0_FILEID,
 235 };
 236
 237 const u32 nfs4_statfs_bitmap[3] = {
 238         FATTR4_WORD0_FILES_AVAIL
 239         | FATTR4_WORD0_FILES_FREE
 240         | FATTR4_WORD0_FILES_TOTAL,
 241         FATTR4_WORD1_SPACE_AVAIL
 242         | FATTR4_WORD1_SPACE_FREE
 243         | FATTR4_WORD1_SPACE_TOTAL
 244 };
 245
 246 const u32 nfs4_pathconf_bitmap[3] = {
 247         FATTR4_WORD0_MAXLINK
 248         | FATTR4_WORD0_MAXNAME,
 249         0
 250 };
 251
 252 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 253                         | FATTR4_WORD0_MAXREAD
 254                         | FATTR4_WORD0_MAXWRITE
 255                         | FATTR4_WORD0_LEASE_TIME,
 256                         FATTR4_WORD1_TIME_DELTA
 257                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 258                         FATTR4_WORD2_LAYOUT_BLKSIZE
 259                         | FATTR4_WORD2_CLONE_BLKSIZE
 260 };
 261
 262 const u32 nfs4_fs_locations_bitmap[3] = {
 263         FATTR4_WORD0_CHANGE
 264         | FATTR4_WORD0_SIZE
 265         | FATTR4_WORD0_FSID
 266         | FATTR4_WORD0_FILEID
 267         | FATTR4_WORD0_FS_LOCATIONS,
 268         FATTR4_WORD1_OWNER
 269         | FATTR4_WORD1_OWNER_GROUP
 270         | FATTR4_WORD1_RAWDEV
 271         | FATTR4_WORD1_SPACE_USED
 272         | FATTR4_WORD1_TIME_ACCESS
 273         | FATTR4_WORD1_TIME_METADATA
 274         | FATTR4_WORD1_TIME_MODIFY
 275         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 276 };
 277
 278 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 279                 struct inode *inode)
 280 {
 281         unsigned long cache_validity;
 282
 283         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 284         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 285                 return;
 286
 287         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 288         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 289                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 290                                 | NFS_INO_INVALID_SIZE);
 291
 292         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 293                 dst[0] &= ~FATTR4_WORD0_SIZE;
 294
 295         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 296                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 297 }
 298
 299 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 300                 const __u32 *src, struct inode *inode)
 301 {
 302         nfs4_bitmap_copy_adjust(dst, src, inode);
 303 }
 304
 305 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 306                 struct nfs4_readdir_arg *readdir)
 307 {
 308         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 309         __be32 *start, *p;
 310
 311         if (cookie > 2) {
 312                 readdir->cookie = cookie;
 313                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 314                 return;
 315         }
 316
 317         readdir->cookie = 0;
 318         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 319         if (cookie == 2)
 320                 return;
 321
 322         /*
 323          * NFSv4 servers do not return entries for '.' and '..'
 324          * Therefore, we fake these entries here.  We let '.'
 325          * have cookie 0 and '..' have cookie 1.  Note that
 326          * when talking to the server, we always send cookie 0
 327          * instead of 1 or 2.
 328          */
 329         start = p = kmap_atomic(*readdir->pages);
 330
 331         if (cookie == 0) {
 332                 *p++ = xdr_one;                                  /* next */
 333                 *p++ = xdr_zero;                   /* cookie, first word */
 334                 *p++ = xdr_one;                   /* cookie, second word */
 335                 *p++ = xdr_one;                             /* entry len */
 336                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 337                 p++;
 338                 *p++ = xdr_one;                         /* bitmap length */
 339                 *p++ = htonl(attrs);                           /* bitmap */
 340                 *p++ = htonl(12);             /* attribute buffer length */
 341                 *p++ = htonl(NF4DIR);
 342                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 343         }
 344
 345         *p++ = xdr_one;                                  /* next */
 346         *p++ = xdr_zero;                   /* cookie, first word */
 347         *p++ = xdr_two;                   /* cookie, second word */
 348         *p++ = xdr_two;                             /* entry len */
 349         memcpy(p, "..\0\0", 4);                         /* entry */
 350         p++;
 351         *p++ = xdr_one;                         /* bitmap length */
 352         *p++ = htonl(attrs);                           /* bitmap */
 353         *p++ = htonl(12);             /* attribute buffer length */
 354         *p++ = htonl(NF4DIR);
 355         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 356
 357         readdir->pgbase = (char *)p - (char *)start;
 358         readdir->count -= readdir->pgbase;
 359         kunmap_atomic(start);
 360 }
 361
 362 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 363                 nfs4_stateid *stateid,
 364                 struct rpc_cred *cred)
 365 {
 366         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 367
 368         ops->test_and_free_expired(server, stateid, cred);
 369 }
 370
 371 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 372                 nfs4_stateid *stateid,
 373                 struct rpc_cred *cred)
 374 {
 375         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 376         nfs4_test_and_free_stateid(server, stateid, cred);
 377 }
 378
 379 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 380                 const nfs4_stateid *stateid,
 381                 struct rpc_cred *cred)
 382 {
 383         nfs4_stateid tmp;
 384
 385         nfs4_stateid_copy(&tmp, stateid);
 386         __nfs4_free_revoked_stateid(server, &tmp, cred);
 387 }
 388
 389 static long nfs4_update_delay(long *timeout)
 390 {
 391         long ret;
 392         if (!timeout)
 393                 return NFS4_POLL_RETRY_MAX;
 394         if (*timeout <= 0)
 395                 *timeout = NFS4_POLL_RETRY_MIN;
 396         if (*timeout > NFS4_POLL_RETRY_MAX)
 397                 *timeout = NFS4_POLL_RETRY_MAX;
 398         ret = *timeout;
 399         *timeout <<= 1;
 400         return ret;
 401 }
 402
 403 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 404 {
 405         int res = 0;
 406
 407         might_sleep();
 408
 409         freezable_schedule_timeout_killable_unsafe(
 410                 nfs4_update_delay(timeout));
 411         if (fatal_signal_pending(current))
 412                 res = -ERESTARTSYS;
 413         return res;
 414 }
 415
 416 /* This is the error handling routine for processes that are allowed
 417  * to sleep.
 418  */
 419 static int nfs4_do_handle_exception(struct nfs_server *server,
 420                 int errorcode, struct nfs4_exception *exception)
 421 {
 422         struct nfs_client *clp = server->nfs_client;
 423         struct nfs4_state *state = exception->state;
 424         const nfs4_stateid *stateid = exception->stateid;
 425         struct inode *inode = exception->inode;
 426         int ret = errorcode;
 427
 428         exception->delay = 0;
 429         exception->recovering = 0;
 430         exception->retry = 0;
 431
 432         if (stateid == NULL && state != NULL)
 433                 stateid = &state->stateid;
 434
 435         switch(errorcode) {
 436                 case 0:
 437                         return 0;
 438                 case -NFS4ERR_BADHANDLE:
 439                 case -ESTALE:
 440                         if (inode != NULL && S_ISREG(inode->i_mode))
 441                                 pnfs_destroy_layout(NFS_I(inode));
 442                         break;
 443                 case -NFS4ERR_DELEG_REVOKED:
 444                 case -NFS4ERR_ADMIN_REVOKED:
 445                 case -NFS4ERR_EXPIRED:
 446                 case -NFS4ERR_BAD_STATEID:
 447                         if (inode != NULL && stateid != NULL) {
 448                                 nfs_inode_find_state_and_recover(inode,
 449                                                 stateid);
 450                                 goto wait_on_recovery;
 451                         }
 452                         /* Fall through */
 453                 case -NFS4ERR_OPENMODE:
 454                         if (inode) {
 455                                 int err;
 456
 457                                 err = nfs_async_inode_return_delegation(inode,
 458                                                 stateid);
 459                                 if (err == 0)
 460                                         goto wait_on_recovery;
 461                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 462                                         exception->retry = 1;
 463                                         break;
 464                                 }
 465                         }
 466                         if (state == NULL)
 467                                 break;
 468                         ret = nfs4_schedule_stateid_recovery(server, state);
 469                         if (ret < 0)
 470                                 break;
 471                         goto wait_on_recovery;
 472                 case -NFS4ERR_STALE_STATEID:
 473                 case -NFS4ERR_STALE_CLIENTID:
 474                         nfs4_schedule_lease_recovery(clp);
 475                         goto wait_on_recovery;
 476                 case -NFS4ERR_MOVED:
 477                         ret = nfs4_schedule_migration_recovery(server);
 478                         if (ret < 0)
 479                                 break;
 480                         goto wait_on_recovery;
 481                 case -NFS4ERR_LEASE_MOVED:
 482                         nfs4_schedule_lease_moved_recovery(clp);
 483                         goto wait_on_recovery;
 484 #if defined(CONFIG_NFS_V4_1)
 485                 case -NFS4ERR_BADSESSION:
 486                 case -NFS4ERR_BADSLOT:
 487                 case -NFS4ERR_BAD_HIGH_SLOT:
 488                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 489                 case -NFS4ERR_DEADSESSION:
 490                 case -NFS4ERR_SEQ_FALSE_RETRY:
 491                 case -NFS4ERR_SEQ_MISORDERED:
 492                         dprintk("%s ERROR: %d Reset session\n", __func__,
 493                                 errorcode);
 494                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 495                         goto wait_on_recovery;
 496 #endif /* defined(CONFIG_NFS_V4_1) */
 497                 case -NFS4ERR_FILE_OPEN:
 498                         if (exception->timeout > HZ) {
 499                                 /* We have retried a decent amount, time to
 500                                  * fail
 501                                  */
 502                                 ret = -EBUSY;
 503                                 break;
 504                         }
 505                         /* Fall through */
 506                 case -NFS4ERR_DELAY:
 507                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 508                         /* Fall through */
 509                 case -NFS4ERR_GRACE:
 510                 case -NFS4ERR_LAYOUTTRYLATER:
 511                 case -NFS4ERR_RECALLCONFLICT:
 512                         exception->delay = 1;
 513                         return 0;
 514
 515                 case -NFS4ERR_RETRY_UNCACHED_REP:
 516                 case -NFS4ERR_OLD_STATEID:
 517                         exception->retry = 1;
 518                         break;
 519                 case -NFS4ERR_BADOWNER:
 520                         /* The following works around a Linux server bug! */
 521                 case -NFS4ERR_BADNAME:
 522                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 523                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 524                                 exception->retry = 1;
 525                                 printk(KERN_WARNING "NFS: v4 server %s "
 526                                                 "does not accept raw "
 527                                                 "uid/gids. "
 528                                                 "Reenabling the idmapper.\n",
 529                                                 server->nfs_client->cl_hostname);
 530                         }
 531         }
 532         /* We failed to handle the error */
 533         return nfs4_map_errors(ret);
 534 wait_on_recovery:
 535         exception->recovering = 1;
 536         return 0;
 537 }
 538
 539 /* This is the error handling routine for processes that are allowed
 540  * to sleep.
 541  */
 542 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 543 {
 544         struct nfs_client *clp = server->nfs_client;
 545         int ret;
 546
 547         ret = nfs4_do_handle_exception(server, errorcode, exception);
 548         if (exception->delay) {
 549                 ret = nfs4_delay(server->client, &exception->timeout);
 550                 goto out_retry;
 551         }
 552         if (exception->recovering) {
 553                 ret = nfs4_wait_clnt_recover(clp);
 554                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 555                         return -EIO;
 556                 goto out_retry;
 557         }
 558         return ret;
 559 out_retry:
 560         if (ret == 0)
 561                 exception->retry = 1;
 562         return ret;
 563 }
 564
 565 static int
 566 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 567                 int errorcode, struct nfs4_exception *exception)
 568 {
 569         struct nfs_client *clp = server->nfs_client;
 570         int ret;
 571
 572         ret = nfs4_do_handle_exception(server, errorcode, exception);
 573         if (exception->delay) {
 574                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 575                 goto out_retry;
 576         }
 577         if (exception->recovering) {
 578                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 579                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 580                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 581                 goto out_retry;
 582         }
 583         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 584                 ret = -EIO;
 585         return ret;
 586 out_retry:
 587         if (ret == 0) {
 588                 exception->retry = 1;
 589                 /*
 590                  * For NFS4ERR_MOVED, the client transport will need to
 591                  * be recomputed after migration recovery has completed.
 592                  */
 593                 if (errorcode == -NFS4ERR_MOVED)
 594                         rpc_task_release_transport(task);
 595         }
 596         return ret;
 597 }
 598
 599 int
 600 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 601                         struct nfs4_state *state, long *timeout)
 602 {
 603         struct nfs4_exception exception = {
 604                 .state = state,
 605         };
 606
 607         if (task->tk_status >= 0)
 608                 return 0;
 609         if (timeout)
 610                 exception.timeout = *timeout;
 611         task->tk_status = nfs4_async_handle_exception(task, server,
 612                         task->tk_status,
 613                         &exception);
 614         if (exception.delay && timeout)
 615                 *timeout = exception.timeout;
 616         if (exception.retry)
 617                 return -EAGAIN;
 618         return 0;
 619 }
 620
 621 /*
 622  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 623  * or 'false' otherwise.
 624  */
 625 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 626 {
 627         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 628         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 629 }
 630
 631 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 632 {
 633         spin_lock(&clp->cl_lock);
 634         if (time_before(clp->cl_last_renewal,timestamp))
 635                 clp->cl_last_renewal = timestamp;
 636         spin_unlock(&clp->cl_lock);
 637 }
 638
 639 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 640 {
 641         struct nfs_client *clp = server->nfs_client;
 642
 643         if (!nfs4_has_session(clp))
 644                 do_renew_lease(clp, timestamp);
 645 }
 646
 647 struct nfs4_call_sync_data {
 648         const struct nfs_server *seq_server;
 649         struct nfs4_sequence_args *seq_args;
 650         struct nfs4_sequence_res *seq_res;
 651 };
 652
 653 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 654                         struct nfs4_sequence_res *res, int cache_reply,
 655                         int privileged)
 656 {
 657         args->sa_slot = NULL;
 658         args->sa_cache_this = cache_reply;
 659         args->sa_privileged = privileged;
 660
 661         res->sr_slot = NULL;
 662 }
 663
 664 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 665 {
 666         struct nfs4_slot *slot = res->sr_slot;
 667         struct nfs4_slot_table *tbl;
 668
 669         tbl = slot->table;
 670         spin_lock(&tbl->slot_tbl_lock);
 671         if (!nfs41_wake_and_assign_slot(tbl, slot))
 672                 nfs4_free_slot(tbl, slot);
 673         spin_unlock(&tbl->slot_tbl_lock);
 674
 675         res->sr_slot = NULL;
 676 }
 677
 678 static int nfs40_sequence_done(struct rpc_task *task,
 679                                struct nfs4_sequence_res *res)
 680 {
 681         if (res->sr_slot != NULL)
 682                 nfs40_sequence_free_slot(res);
 683         return 1;
 684 }
 685
 686 #if defined(CONFIG_NFS_V4_1)
 687
 688 static void nfs41_release_slot(struct nfs4_slot *slot)
 689 {
 690         struct nfs4_session *session;
 691         struct nfs4_slot_table *tbl;
 692         bool send_new_highest_used_slotid = false;
 693
 694         if (!slot)
 695                 return;
 696         tbl = slot->table;
 697         session = tbl->session;
 698
 699         /* Bump the slot sequence number */
 700         if (slot->seq_done)
 701                 slot->seq_nr++;
 702         slot->seq_done = 0;
 703
 704         spin_lock(&tbl->slot_tbl_lock);
 705         /* Be nice to the server: try to ensure that the last transmitted
 706          * value for highest_user_slotid <= target_highest_slotid
 707          */
 708         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 709                 send_new_highest_used_slotid = true;
 710
 711         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 712                 send_new_highest_used_slotid = false;
 713                 goto out_unlock;
 714         }
 715         nfs4_free_slot(tbl, slot);
 716
 717         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 718                 send_new_highest_used_slotid = false;
 719 out_unlock:
 720         spin_unlock(&tbl->slot_tbl_lock);
 721         if (send_new_highest_used_slotid)
 722                 nfs41_notify_server(session->clp);
 723         if (waitqueue_active(&tbl->slot_waitq))
 724                 wake_up_all(&tbl->slot_waitq);
 725 }
 726
 727 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 728 {
 729         nfs41_release_slot(res->sr_slot);
 730         res->sr_slot = NULL;
 731 }
 732
 733 static int nfs41_sequence_process(struct rpc_task *task,
 734                 struct nfs4_sequence_res *res)
 735 {
 736         struct nfs4_session *session;
 737         struct nfs4_slot *slot = res->sr_slot;
 738         struct nfs_client *clp;
 739         bool interrupted = false;
 740         int ret = 1;
 741
 742         if (slot == NULL)
 743                 goto out_noaction;
 744         /* don't increment the sequence number if the task wasn't sent */
 745         if (!RPC_WAS_SENT(task))
 746                 goto out;
 747
 748         session = slot->table->session;
 749
 750         if (slot->interrupted) {
 751                 if (res->sr_status != -NFS4ERR_DELAY)
 752                         slot->interrupted = 0;
 753                 interrupted = true;
 754         }
 755
 756         trace_nfs4_sequence_done(session, res);
 757         /* Check the SEQUENCE operation status */
 758         switch (res->sr_status) {
 759         case 0:
 760                 /* Update the slot's sequence and clientid lease timer */
 761                 slot->seq_done = 1;
 762                 clp = session->clp;
 763                 do_renew_lease(clp, res->sr_timestamp);
 764                 /* Check sequence flags */
 765                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 766                                 !!slot->privileged);
 767                 nfs41_update_target_slotid(slot->table, slot, res);
 768                 break;
 769         case 1:
 770                 /*
 771                  * sr_status remains 1 if an RPC level error occurred.
 772                  * The server may or may not have processed the sequence
 773                  * operation..
 774                  * Mark the slot as having hosted an interrupted RPC call.
 775                  */
 776                 slot->interrupted = 1;
 777                 goto out;
 778         case -NFS4ERR_DELAY:
 779                 /* The server detected a resend of the RPC call and
 780                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 781                  * of RFC5661.
 782                  */
 783                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 784                         __func__,
 785                         slot->slot_nr,
 786                         slot->seq_nr);
 787                 goto out_retry;
 788         case -NFS4ERR_RETRY_UNCACHED_REP:
 789         case -NFS4ERR_SEQ_FALSE_RETRY:
 790                 /*
 791                  * The server thinks we tried to replay a request.
 792                  * Retry the call after bumping the sequence ID.
 793                  */
 794                 goto retry_new_seq;
 795         case -NFS4ERR_BADSLOT:
 796                 /*
 797                  * The slot id we used was probably retired. Try again
 798                  * using a different slot id.
 799                  */
 800                 if (slot->slot_nr < slot->table->target_highest_slotid)
 801                         goto session_recover;
 802                 goto retry_nowait;
 803         case -NFS4ERR_SEQ_MISORDERED:
 804                 /*
 805                  * Was the last operation on this sequence interrupted?
 806                  * If so, retry after bumping the sequence number.
 807                  */
 808                 if (interrupted)
 809                         goto retry_new_seq;
 810                 /*
 811                  * Could this slot have been previously retired?
 812                  * If so, then the server may be expecting seq_nr = 1!
 813                  */
 814                 if (slot->seq_nr != 1) {
 815                         slot->seq_nr = 1;
 816                         goto retry_nowait;
 817                 }
 818                 goto session_recover;
 819         default:
 820                 /* Just update the slot sequence no. */
 821                 slot->seq_done = 1;
 822         }
 823 out:
 824         /* The session may be reset by one of the error handlers. */
 825         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 826 out_noaction:
 827         return ret;
 828 session_recover:
 829         nfs4_schedule_session_recovery(session, res->sr_status);
 830         goto retry_nowait;
 831 retry_new_seq:
 832         ++slot->seq_nr;
 833 retry_nowait:
 834         if (rpc_restart_call_prepare(task)) {
 835                 nfs41_sequence_free_slot(res);
 836                 task->tk_status = 0;
 837                 ret = 0;
 838         }
 839         goto out;
 840 out_retry:
 841         if (!rpc_restart_call(task))
 842                 goto out;
 843         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 844         return 0;
 845 }
 846
 847 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 848 {
 849         if (!nfs41_sequence_process(task, res))
 850                 return 0;
 851         if (res->sr_slot != NULL)
 852                 nfs41_sequence_free_slot(res);
 853         return 1;
 854
 855 }
 856 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 857
 858 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 859 {
 860         if (res->sr_slot == NULL)
 861                 return 1;
 862         if (res->sr_slot->table->session != NULL)
 863                 return nfs41_sequence_process(task, res);
 864         return nfs40_sequence_done(task, res);
 865 }
 866
 867 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 868 {
 869         if (res->sr_slot != NULL) {
 870                 if (res->sr_slot->table->session != NULL)
 871                         nfs41_sequence_free_slot(res);
 872                 else
 873                         nfs40_sequence_free_slot(res);
 874         }
 875 }
 876
 877 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 878 {
 879         if (res->sr_slot == NULL)
 880                 return 1;
 881         if (!res->sr_slot->table->session)
 882                 return nfs40_sequence_done(task, res);
 883         return nfs41_sequence_done(task, res);
 884 }
 885 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 886
 887 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 888 {
 889         struct nfs4_call_sync_data *data = calldata;
 890
 891         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 892
 893         nfs4_setup_sequence(data->seq_server->nfs_client,
 894                             data->seq_args, data->seq_res, task);
 895 }
 896
 897 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 898 {
 899         struct nfs4_call_sync_data *data = calldata;
 900
 901         nfs41_sequence_done(task, data->seq_res);
 902 }
 903
 904 static const struct rpc_call_ops nfs41_call_sync_ops = {
 905         .rpc_call_prepare = nfs41_call_sync_prepare,
 906         .rpc_call_done = nfs41_call_sync_done,
 907 };
 908
 909 static void
 910 nfs4_sequence_process_interrupted(struct nfs_client *client,
 911                 struct nfs4_slot *slot, struct rpc_cred *cred)
 912 {
 913         struct rpc_task *task;
 914
 915         task = _nfs41_proc_sequence(client, cred, slot, true);
 916         if (!IS_ERR(task))
 917                 rpc_put_task_async(task);
 918 }
 919
 920 #else   /* !CONFIG_NFS_V4_1 */
 921
 922 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 923 {
 924         return nfs40_sequence_done(task, res);
 925 }
 926
 927 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 928 {
 929         if (res->sr_slot != NULL)
 930                 nfs40_sequence_free_slot(res);
 931 }
 932
 933 int nfs4_sequence_done(struct rpc_task *task,
 934                        struct nfs4_sequence_res *res)
 935 {
 936         return nfs40_sequence_done(task, res);
 937 }
 938 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 939
 940 static void
 941 nfs4_sequence_process_interrupted(struct nfs_client *client,
 942                 struct nfs4_slot *slot, struct rpc_cred *cred)
 943 {
 944         WARN_ON_ONCE(1);
 945         slot->interrupted = 0;
 946 }
 947
 948 #endif  /* !CONFIG_NFS_V4_1 */
 949
 950 static
 951 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
 952                 struct nfs4_sequence_res *res,
 953                 struct nfs4_slot *slot)
 954 {
 955         if (!slot)
 956                 return;
 957         slot->privileged = args->sa_privileged ? 1 : 0;
 958         args->sa_slot = slot;
 959
 960         res->sr_slot = slot;
 961         res->sr_timestamp = jiffies;
 962         res->sr_status_flags = 0;
 963         res->sr_status = 1;
 964
 965 }
 966
 967 int nfs4_setup_sequence(struct nfs_client *client,
 968                         struct nfs4_sequence_args *args,
 969                         struct nfs4_sequence_res *res,
 970                         struct rpc_task *task)
 971 {
 972         struct nfs4_session *session = nfs4_get_session(client);
 973         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
 974         struct nfs4_slot *slot;
 975
 976         /* slot already allocated? */
 977         if (res->sr_slot != NULL)
 978                 goto out_start;
 979
 980         if (session) {
 981                 tbl = &session->fc_slot_table;
 982                 task->tk_timeout = 0;
 983         }
 984
 985         for (;;) {
 986                 spin_lock(&tbl->slot_tbl_lock);
 987                 /* The state manager will wait until the slot table is empty */
 988                 if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 989                         goto out_sleep;
 990
 991                 slot = nfs4_alloc_slot(tbl);
 992                 if (IS_ERR(slot)) {
 993                         /* Try again in 1/4 second */
 994                         if (slot == ERR_PTR(-ENOMEM))
 995                                 task->tk_timeout = HZ >> 2;
 996                         goto out_sleep;
 997                 }
 998                 spin_unlock(&tbl->slot_tbl_lock);
 999
1000                 if (likely(!slot->interrupted))
1001                         break;
1002                 nfs4_sequence_process_interrupted(client,
1003                                 slot, task->tk_msg.rpc_cred);
1004         }
1005
1006         nfs4_sequence_attach_slot(args, res, slot);
1007
1008         trace_nfs4_setup_sequence(session, args);
1009 out_start:
1010         rpc_call_start(task);
1011         return 0;
1012
1013 out_sleep:
1014         if (args->sa_privileged)
1015                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1016                                 NULL, RPC_PRIORITY_PRIVILEGED);
1017         else
1018                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1019         spin_unlock(&tbl->slot_tbl_lock);
1020         return -EAGAIN;
1021 }
1022 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1023
1024 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1025 {
1026         struct nfs4_call_sync_data *data = calldata;
1027         nfs4_setup_sequence(data->seq_server->nfs_client,
1028                                 data->seq_args, data->seq_res, task);
1029 }
1030
1031 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1032 {
1033         struct nfs4_call_sync_data *data = calldata;
1034         nfs4_sequence_done(task, data->seq_res);
1035 }
1036
1037 static const struct rpc_call_ops nfs40_call_sync_ops = {
1038         .rpc_call_prepare = nfs40_call_sync_prepare,
1039         .rpc_call_done = nfs40_call_sync_done,
1040 };
1041
1042 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1043                                    struct nfs_server *server,
1044                                    struct rpc_message *msg,
1045                                    struct nfs4_sequence_args *args,
1046                                    struct nfs4_sequence_res *res)
1047 {
1048         int ret;
1049         struct rpc_task *task;
1050         struct nfs_client *clp = server->nfs_client;
1051         struct nfs4_call_sync_data data = {
1052                 .seq_server = server,
1053                 .seq_args = args,
1054                 .seq_res = res,
1055         };
1056         struct rpc_task_setup task_setup = {
1057                 .rpc_client = clnt,
1058                 .rpc_message = msg,
1059                 .callback_ops = clp->cl_mvops->call_sync_ops,
1060                 .callback_data = &data
1061         };
1062
1063         task = rpc_run_task(&task_setup);
1064         if (IS_ERR(task))
1065                 ret = PTR_ERR(task);
1066         else {
1067                 ret = task->tk_status;
1068                 rpc_put_task(task);
1069         }
1070         return ret;
1071 }
1072
1073 int nfs4_call_sync(struct rpc_clnt *clnt,
1074                    struct nfs_server *server,
1075                    struct rpc_message *msg,
1076                    struct nfs4_sequence_args *args,
1077                    struct nfs4_sequence_res *res,
1078                    int cache_reply)
1079 {
1080         nfs4_init_sequence(args, res, cache_reply, 0);
1081         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1082 }
1083
1084 static void
1085 nfs4_inc_nlink_locked(struct inode *inode)
1086 {
1087         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1088         inc_nlink(inode);
1089 }
1090
1091 static void
1092 nfs4_dec_nlink_locked(struct inode *inode)
1093 {
1094         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1095         drop_nlink(inode);
1096 }
1097
1098 static void
1099 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1100                 unsigned long timestamp, unsigned long cache_validity)
1101 {
1102         struct nfs_inode *nfsi = NFS_I(dir);
1103
1104         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1105                 | NFS_INO_INVALID_MTIME
1106                 | NFS_INO_INVALID_DATA
1107                 | cache_validity;
1108         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1109                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1110                 nfsi->attrtimeo_timestamp = jiffies;
1111         } else {
1112                 nfs_force_lookup_revalidate(dir);
1113                 if (cinfo->before != inode_peek_iversion_raw(dir))
1114                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1115                                 NFS_INO_INVALID_ACL;
1116         }
1117         inode_set_iversion_raw(dir, cinfo->after);
1118         nfsi->read_cache_jiffies = timestamp;
1119         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1120         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1121         nfs_fscache_invalidate(dir);
1122 }
1123
1124 static void
1125 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1126                 unsigned long timestamp, unsigned long cache_validity)
1127 {
1128         spin_lock(&dir->i_lock);
1129         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1130         spin_unlock(&dir->i_lock);
1131 }
1132
1133 struct nfs4_open_createattrs {
1134         struct nfs4_label *label;
1135         struct iattr *sattr;
1136         const __u32 verf[2];
1137 };
1138
1139 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1140                 int err, struct nfs4_exception *exception)
1141 {
1142         if (err != -EINVAL)
1143                 return false;
1144         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1145                 return false;
1146         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1147         exception->retry = 1;
1148         return true;
1149 }
1150
1151 static u32
1152 nfs4_map_atomic_open_share(struct nfs_server *server,
1153                 fmode_t fmode, int openflags)
1154 {
1155         u32 res = 0;
1156
1157         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1158         case FMODE_READ:
1159                 res = NFS4_SHARE_ACCESS_READ;
1160                 break;
1161         case FMODE_WRITE:
1162                 res = NFS4_SHARE_ACCESS_WRITE;
1163                 break;
1164         case FMODE_READ|FMODE_WRITE:
1165                 res = NFS4_SHARE_ACCESS_BOTH;
1166         }
1167         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1168                 goto out;
1169         /* Want no delegation if we're using O_DIRECT */
1170         if (openflags & O_DIRECT)
1171                 res |= NFS4_SHARE_WANT_NO_DELEG;
1172 out:
1173         return res;
1174 }
1175
1176 static enum open_claim_type4
1177 nfs4_map_atomic_open_claim(struct nfs_server *server,
1178                 enum open_claim_type4 claim)
1179 {
1180         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1181                 return claim;
1182         switch (claim) {
1183         default:
1184                 return claim;
1185         case NFS4_OPEN_CLAIM_FH:
1186                 return NFS4_OPEN_CLAIM_NULL;
1187         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1188                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1189         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1190                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1191         }
1192 }
1193
1194 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1195 {
1196         p->o_res.f_attr = &p->f_attr;
1197         p->o_res.f_label = p->f_label;
1198         p->o_res.seqid = p->o_arg.seqid;
1199         p->c_res.seqid = p->c_arg.seqid;
1200         p->o_res.server = p->o_arg.server;
1201         p->o_res.access_request = p->o_arg.access;
1202         nfs_fattr_init(&p->f_attr);
1203         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1204 }
1205
1206 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1207                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1208                 const struct nfs4_open_createattrs *c,
1209                 enum open_claim_type4 claim,
1210                 gfp_t gfp_mask)
1211 {
1212         struct dentry *parent = dget_parent(dentry);
1213         struct inode *dir = d_inode(parent);
1214         struct nfs_server *server = NFS_SERVER(dir);
1215         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1216         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1217         struct nfs4_opendata *p;
1218
1219         p = kzalloc(sizeof(*p), gfp_mask);
1220         if (p == NULL)
1221                 goto err;
1222
1223         p->f_label = nfs4_label_alloc(server, gfp_mask);
1224         if (IS_ERR(p->f_label))
1225                 goto err_free_p;
1226
1227         p->a_label = nfs4_label_alloc(server, gfp_mask);
1228         if (IS_ERR(p->a_label))
1229                 goto err_free_f;
1230
1231         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1232         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1233         if (IS_ERR(p->o_arg.seqid))
1234                 goto err_free_label;
1235         nfs_sb_active(dentry->d_sb);
1236         p->dentry = dget(dentry);
1237         p->dir = parent;
1238         p->owner = sp;
1239         atomic_inc(&sp->so_count);
1240         p->o_arg.open_flags = flags;
1241         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1242         p->o_arg.umask = current_umask();
1243         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1244         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1245                         fmode, flags);
1246         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1247          * will return permission denied for all bits until close */
1248         if (!(flags & O_EXCL)) {
1249                 /* ask server to check for all possible rights as results
1250                  * are cached */
1251                 switch (p->o_arg.claim) {
1252                 default:
1253                         break;
1254                 case NFS4_OPEN_CLAIM_NULL:
1255                 case NFS4_OPEN_CLAIM_FH:
1256                         p->o_arg.access = NFS4_ACCESS_READ |
1257                                 NFS4_ACCESS_MODIFY |
1258                                 NFS4_ACCESS_EXTEND |
1259                                 NFS4_ACCESS_EXECUTE;
1260                 }
1261         }
1262         p->o_arg.clientid = server->nfs_client->cl_clientid;
1263         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1264         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1265         p->o_arg.name = &dentry->d_name;
1266         p->o_arg.server = server;
1267         p->o_arg.bitmask = nfs4_bitmask(server, label);
1268         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1269         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1270         switch (p->o_arg.claim) {
1271         case NFS4_OPEN_CLAIM_NULL:
1272         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1273         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1274                 p->o_arg.fh = NFS_FH(dir);
1275                 break;
1276         case NFS4_OPEN_CLAIM_PREVIOUS:
1277         case NFS4_OPEN_CLAIM_FH:
1278         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1279         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1280                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1281         }
1282         if (c != NULL && c->sattr != NULL && c->sattr->ia_valid != 0) {
1283                 p->o_arg.u.attrs = &p->attrs;
1284                 memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1285
1286                 memcpy(p->o_arg.u.verifier.data, c->verf,
1287                                 sizeof(p->o_arg.u.verifier.data));
1288         }
1289         p->c_arg.fh = &p->o_res.fh;
1290         p->c_arg.stateid = &p->o_res.stateid;
1291         p->c_arg.seqid = p->o_arg.seqid;
1292         nfs4_init_opendata_res(p);
1293         kref_init(&p->kref);
1294         return p;
1295
1296 err_free_label:
1297         nfs4_label_free(p->a_label);
1298 err_free_f:
1299         nfs4_label_free(p->f_label);
1300 err_free_p:
1301         kfree(p);
1302 err:
1303         dput(parent);
1304         return NULL;
1305 }
1306
1307 static void nfs4_opendata_free(struct kref *kref)
1308 {
1309         struct nfs4_opendata *p = container_of(kref,
1310                         struct nfs4_opendata, kref);
1311         struct super_block *sb = p->dentry->d_sb;
1312
1313         nfs4_lgopen_release(p->lgp);
1314         nfs_free_seqid(p->o_arg.seqid);
1315         nfs4_sequence_free_slot(&p->o_res.seq_res);
1316         if (p->state != NULL)
1317                 nfs4_put_open_state(p->state);
1318         nfs4_put_state_owner(p->owner);
1319
1320         nfs4_label_free(p->a_label);
1321         nfs4_label_free(p->f_label);
1322
1323         dput(p->dir);
1324         dput(p->dentry);
1325         nfs_sb_deactive(sb);
1326         nfs_fattr_free_names(&p->f_attr);
1327         kfree(p->f_attr.mdsthreshold);
1328         kfree(p);
1329 }
1330
1331 static void nfs4_opendata_put(struct nfs4_opendata *p)
1332 {
1333         if (p != NULL)
1334                 kref_put(&p->kref, nfs4_opendata_free);
1335 }
1336
1337 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1338                 fmode_t fmode)
1339 {
1340         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1341         case FMODE_READ|FMODE_WRITE:
1342                 return state->n_rdwr != 0;
1343         case FMODE_WRITE:
1344                 return state->n_wronly != 0;
1345         case FMODE_READ:
1346                 return state->n_rdonly != 0;
1347         }
1348         WARN_ON_ONCE(1);
1349         return false;
1350 }
1351
1352 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1353 {
1354         int ret = 0;
1355
1356         if (open_mode & (O_EXCL|O_TRUNC))
1357                 goto out;
1358         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1359                 case FMODE_READ:
1360                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1361                                 && state->n_rdonly != 0;
1362                         break;
1363                 case FMODE_WRITE:
1364                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1365                                 && state->n_wronly != 0;
1366                         break;
1367                 case FMODE_READ|FMODE_WRITE:
1368                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1369                                 && state->n_rdwr != 0;
1370         }
1371 out:
1372         return ret;
1373 }
1374
1375 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1376                 enum open_claim_type4 claim)
1377 {
1378         if (delegation == NULL)
1379                 return 0;
1380         if ((delegation->type & fmode) != fmode)
1381                 return 0;
1382         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1383                 return 0;
1384         switch (claim) {
1385         case NFS4_OPEN_CLAIM_NULL:
1386         case NFS4_OPEN_CLAIM_FH:
1387                 break;
1388         case NFS4_OPEN_CLAIM_PREVIOUS:
1389                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1390                         break;
1391                 /* Fall through */
1392         default:
1393                 return 0;
1394         }
1395         nfs_mark_delegation_referenced(delegation);
1396         return 1;
1397 }
1398
1399 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1400 {
1401         switch (fmode) {
1402                 case FMODE_WRITE:
1403                         state->n_wronly++;
1404                         break;
1405                 case FMODE_READ:
1406                         state->n_rdonly++;
1407                         break;
1408                 case FMODE_READ|FMODE_WRITE:
1409                         state->n_rdwr++;
1410         }
1411         nfs4_state_set_mode_locked(state, state->state | fmode);
1412 }
1413
1414 #ifdef CONFIG_NFS_V4_1
1415 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1416 {
1417         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1418                 return true;
1419         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1420                 return true;
1421         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1422                 return true;
1423         return false;
1424 }
1425 #endif /* CONFIG_NFS_V4_1 */
1426
1427 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1428 {
1429         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1430                 wake_up_all(&state->waitq);
1431 }
1432
1433 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1434                 const nfs4_stateid *stateid)
1435 {
1436         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1437         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1438
1439         if (stateid_seqid == state_seqid + 1U ||
1440             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1441                 nfs_state_log_update_open_stateid(state);
1442         else
1443                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1444 }
1445
1446 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1447 {
1448         struct nfs_client *clp = state->owner->so_server->nfs_client;
1449         bool need_recover = false;
1450
1451         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1452                 need_recover = true;
1453         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1454                 need_recover = true;
1455         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1456                 need_recover = true;
1457         if (need_recover)
1458                 nfs4_state_mark_reclaim_nograce(clp, state);
1459 }
1460
1461 /*
1462  * Check for whether or not the caller may update the open stateid
1463  * to the value passed in by stateid.
1464  *
1465  * Note: This function relies heavily on the server implementing
1466  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1467  * correctly.
1468  * i.e. The stateid seqids have to be initialised to 1, and
1469  * are then incremented on every state transition.
1470  */
1471 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1472                 const nfs4_stateid *stateid)
1473 {
1474         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1475             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1476                 if (stateid->seqid == cpu_to_be32(1))
1477                         nfs_state_log_update_open_stateid(state);
1478                 else
1479                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1480                 return true;
1481         }
1482
1483         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1484                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1485                 return true;
1486         }
1487         return false;
1488 }
1489
1490 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1491 {
1492         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1493                 return;
1494         if (state->n_wronly)
1495                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1496         if (state->n_rdonly)
1497                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1498         if (state->n_rdwr)
1499                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1500         set_bit(NFS_OPEN_STATE, &state->flags);
1501 }
1502
1503 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1504                 nfs4_stateid *stateid, fmode_t fmode)
1505 {
1506         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1507         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1508         case FMODE_WRITE:
1509                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1510                 break;
1511         case FMODE_READ:
1512                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1513                 break;
1514         case 0:
1515                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1516                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1517                 clear_bit(NFS_OPEN_STATE, &state->flags);
1518         }
1519         if (stateid == NULL)
1520                 return;
1521         /* Handle OPEN+OPEN_DOWNGRADE races */
1522         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1523             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1524                 nfs_resync_open_stateid_locked(state);
1525                 goto out;
1526         }
1527         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1528                 nfs4_stateid_copy(&state->stateid, stateid);
1529         nfs4_stateid_copy(&state->open_stateid, stateid);
1530         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1531 out:
1532         nfs_state_log_update_open_stateid(state);
1533 }
1534
1535 static void nfs_clear_open_stateid(struct nfs4_state *state,
1536         nfs4_stateid *arg_stateid,
1537         nfs4_stateid *stateid, fmode_t fmode)
1538 {
1539         write_seqlock(&state->seqlock);
1540         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1541         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1542                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1543         write_sequnlock(&state->seqlock);
1544         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1545                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1546 }
1547
1548 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1549                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1550 {
1551         DEFINE_WAIT(wait);
1552         int status = 0;
1553         for (;;) {
1554
1555                 if (!nfs_need_update_open_stateid(state, stateid))
1556                         return;
1557                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1558                         break;
1559                 if (status)
1560                         break;
1561                 /* Rely on seqids for serialisation with NFSv4.0 */
1562                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1563                         break;
1564
1565                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1566                 /*
1567                  * Ensure we process the state changes in the same order
1568                  * in which the server processed them by delaying the
1569                  * update of the stateid until we are in sequence.
1570                  */
1571                 write_sequnlock(&state->seqlock);
1572                 spin_unlock(&state->owner->so_lock);
1573                 rcu_read_unlock();
1574                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1575                 if (!signal_pending(current)) {
1576                         if (schedule_timeout(5*HZ) == 0)
1577                                 status = -EAGAIN;
1578                         else
1579                                 status = 0;
1580                 } else
1581                         status = -EINTR;
1582                 finish_wait(&state->waitq, &wait);
1583                 rcu_read_lock();
1584                 spin_lock(&state->owner->so_lock);
1585                 write_seqlock(&state->seqlock);
1586         }
1587
1588         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1589             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1590                 nfs4_stateid_copy(freeme, &state->open_stateid);
1591                 nfs_test_and_clear_all_open_stateid(state);
1592         }
1593
1594         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1595                 nfs4_stateid_copy(&state->stateid, stateid);
1596         nfs4_stateid_copy(&state->open_stateid, stateid);
1597         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1598         nfs_state_log_update_open_stateid(state);
1599 }
1600
1601 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1602                 const nfs4_stateid *open_stateid,
1603                 fmode_t fmode,
1604                 nfs4_stateid *freeme)
1605 {
1606         /*
1607          * Protect the call to nfs4_state_set_mode_locked and
1608          * serialise the stateid update
1609          */
1610         write_seqlock(&state->seqlock);
1611         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1612         switch (fmode) {
1613         case FMODE_READ:
1614                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1615                 break;
1616         case FMODE_WRITE:
1617                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1618                 break;
1619         case FMODE_READ|FMODE_WRITE:
1620                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1621         }
1622         set_bit(NFS_OPEN_STATE, &state->flags);
1623         write_sequnlock(&state->seqlock);
1624 }
1625
1626 static void nfs_state_set_delegation(struct nfs4_state *state,
1627                 const nfs4_stateid *deleg_stateid,
1628                 fmode_t fmode)
1629 {
1630         /*
1631          * Protect the call to nfs4_state_set_mode_locked and
1632          * serialise the stateid update
1633          */
1634         write_seqlock(&state->seqlock);
1635         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1636         set_bit(NFS_DELEGATED_STATE, &state->flags);
1637         write_sequnlock(&state->seqlock);
1638 }
1639
1640 static void nfs_state_clear_delegation(struct nfs4_state *state)
1641 {
1642         write_seqlock(&state->seqlock);
1643         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1644         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1645         write_sequnlock(&state->seqlock);
1646 }
1647
1648 static int update_open_stateid(struct nfs4_state *state,
1649                 const nfs4_stateid *open_stateid,
1650                 const nfs4_stateid *delegation,
1651                 fmode_t fmode)
1652 {
1653         struct nfs_server *server = NFS_SERVER(state->inode);
1654         struct nfs_client *clp = server->nfs_client;
1655         struct nfs_inode *nfsi = NFS_I(state->inode);
1656         struct nfs_delegation *deleg_cur;
1657         nfs4_stateid freeme = { };
1658         int ret = 0;
1659
1660         fmode &= (FMODE_READ|FMODE_WRITE);
1661
1662         rcu_read_lock();
1663         spin_lock(&state->owner->so_lock);
1664         if (open_stateid != NULL) {
1665                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1666                 ret = 1;
1667         }
1668
1669         deleg_cur = rcu_dereference(nfsi->delegation);
1670         if (deleg_cur == NULL)
1671                 goto no_delegation;
1672
1673         spin_lock(&deleg_cur->lock);
1674         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1675            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1676             (deleg_cur->type & fmode) != fmode)
1677                 goto no_delegation_unlock;
1678
1679         if (delegation == NULL)
1680                 delegation = &deleg_cur->stateid;
1681         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1682                 goto no_delegation_unlock;
1683
1684         nfs_mark_delegation_referenced(deleg_cur);
1685         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1686         ret = 1;
1687 no_delegation_unlock:
1688         spin_unlock(&deleg_cur->lock);
1689 no_delegation:
1690         if (ret)
1691                 update_open_stateflags(state, fmode);
1692         spin_unlock(&state->owner->so_lock);
1693         rcu_read_unlock();
1694
1695         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1696                 nfs4_schedule_state_manager(clp);
1697         if (freeme.type != 0)
1698                 nfs4_test_and_free_stateid(server, &freeme,
1699                                 state->owner->so_cred);
1700
1701         return ret;
1702 }
1703
1704 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1705                 const nfs4_stateid *stateid)
1706 {
1707         struct nfs4_state *state = lsp->ls_state;
1708         bool ret = false;
1709
1710         spin_lock(&state->state_lock);
1711         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1712                 goto out_noupdate;
1713         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1714                 goto out_noupdate;
1715         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1716         ret = true;
1717 out_noupdate:
1718         spin_unlock(&state->state_lock);
1719         return ret;
1720 }
1721
1722 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1723 {
1724         struct nfs_delegation *delegation;
1725
1726         fmode &= FMODE_READ|FMODE_WRITE;
1727         rcu_read_lock();
1728         delegation = rcu_dereference(NFS_I(inode)->delegation);
1729         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1730                 rcu_read_unlock();
1731                 return;
1732         }
1733         rcu_read_unlock();
1734         nfs4_inode_return_delegation(inode);
1735 }
1736
1737 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1738 {
1739         struct nfs4_state *state = opendata->state;
1740         struct nfs_inode *nfsi = NFS_I(state->inode);
1741         struct nfs_delegation *delegation;
1742         int open_mode = opendata->o_arg.open_flags;
1743         fmode_t fmode = opendata->o_arg.fmode;
1744         enum open_claim_type4 claim = opendata->o_arg.claim;
1745         nfs4_stateid stateid;
1746         int ret = -EAGAIN;
1747
1748         for (;;) {
1749                 spin_lock(&state->owner->so_lock);
1750                 if (can_open_cached(state, fmode, open_mode)) {
1751                         update_open_stateflags(state, fmode);
1752                         spin_unlock(&state->owner->so_lock);
1753                         goto out_return_state;
1754                 }
1755                 spin_unlock(&state->owner->so_lock);
1756                 rcu_read_lock();
1757                 delegation = rcu_dereference(nfsi->delegation);
1758                 if (!can_open_delegated(delegation, fmode, claim)) {
1759                         rcu_read_unlock();
1760                         break;
1761                 }
1762                 /* Save the delegation */
1763                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1764                 rcu_read_unlock();
1765                 nfs_release_seqid(opendata->o_arg.seqid);
1766                 if (!opendata->is_recover) {
1767                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1768                         if (ret != 0)
1769                                 goto out;
1770                 }
1771                 ret = -EAGAIN;
1772
1773                 /* Try to update the stateid using the delegation */
1774                 if (update_open_stateid(state, NULL, &stateid, fmode))
1775                         goto out_return_state;
1776         }
1777 out:
1778         return ERR_PTR(ret);
1779 out_return_state:
1780         atomic_inc(&state->count);
1781         return state;
1782 }
1783
1784 static void
1785 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1786 {
1787         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1788         struct nfs_delegation *delegation;
1789         int delegation_flags = 0;
1790
1791         rcu_read_lock();
1792         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1793         if (delegation)
1794                 delegation_flags = delegation->flags;
1795         rcu_read_unlock();
1796         switch (data->o_arg.claim) {
1797         default:
1798                 break;
1799         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1800         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1801                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1802                                    "returning a delegation for "
1803                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1804                                    clp->cl_hostname);
1805                 return;
1806         }
1807         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1808                 nfs_inode_set_delegation(state->inode,
1809                                 data->owner->so_cred,
1810                                 data->o_res.delegation_type,
1811                                 &data->o_res.delegation,
1812                                 data->o_res.pagemod_limit);
1813         else
1814                 nfs_inode_reclaim_delegation(state->inode,
1815                                 data->owner->so_cred,
1816                                 data->o_res.delegation_type,
1817                                 &data->o_res.delegation,
1818                                 data->o_res.pagemod_limit);
1819
1820         if (data->o_res.do_recall)
1821                 nfs_async_inode_return_delegation(state->inode,
1822                                                   &data->o_res.delegation);
1823 }
1824
1825 /*
1826  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1827  * and update the nfs4_state.
1828  */
1829 static struct nfs4_state *
1830 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1831 {
1832         struct inode *inode = data->state->inode;
1833         struct nfs4_state *state = data->state;
1834         int ret;
1835
1836         if (!data->rpc_done) {
1837                 if (data->rpc_status)
1838                         return ERR_PTR(data->rpc_status);
1839                 /* cached opens have already been processed */
1840                 goto update;
1841         }
1842
1843         ret = nfs_refresh_inode(inode, &data->f_attr);
1844         if (ret)
1845                 return ERR_PTR(ret);
1846
1847         if (data->o_res.delegation_type != 0)
1848                 nfs4_opendata_check_deleg(data, state);
1849 update:
1850         update_open_stateid(state, &data->o_res.stateid, NULL,
1851                             data->o_arg.fmode);
1852         atomic_inc(&state->count);
1853
1854         return state;
1855 }
1856
1857 static struct inode *
1858 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1859 {
1860         struct inode *inode;
1861
1862         switch (data->o_arg.claim) {
1863         case NFS4_OPEN_CLAIM_NULL:
1864         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1865         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1866                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1867                         return ERR_PTR(-EAGAIN);
1868                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1869                                 &data->f_attr, data->f_label);
1870                 break;
1871         default:
1872                 inode = d_inode(data->dentry);
1873                 ihold(inode);
1874                 nfs_refresh_inode(inode, &data->f_attr);
1875         }
1876         return inode;
1877 }
1878
1879 static struct nfs4_state *
1880 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1881 {
1882         struct nfs4_state *state;
1883         struct inode *inode;
1884
1885         inode = nfs4_opendata_get_inode(data);
1886         if (IS_ERR(inode))
1887                 return ERR_CAST(inode);
1888         if (data->state != NULL && data->state->inode == inode) {
1889                 state = data->state;
1890                 atomic_inc(&state->count);
1891         } else
1892                 state = nfs4_get_open_state(inode, data->owner);
1893         iput(inode);
1894         if (state == NULL)
1895                 state = ERR_PTR(-ENOMEM);
1896         return state;
1897 }
1898
1899 static struct nfs4_state *
1900 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1901 {
1902         struct nfs4_state *state;
1903
1904         if (!data->rpc_done) {
1905                 state = nfs4_try_open_cached(data);
1906                 trace_nfs4_cached_open(data->state);
1907                 goto out;
1908         }
1909
1910         state = nfs4_opendata_find_nfs4_state(data);
1911         if (IS_ERR(state))
1912                 goto out;
1913
1914         if (data->o_res.delegation_type != 0)
1915                 nfs4_opendata_check_deleg(data, state);
1916         update_open_stateid(state, &data->o_res.stateid, NULL,
1917                         data->o_arg.fmode);
1918 out:
1919         nfs_release_seqid(data->o_arg.seqid);
1920         return state;
1921 }
1922
1923 static struct nfs4_state *
1924 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1925 {
1926         struct nfs4_state *ret;
1927
1928         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1929                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1930         else
1931                 ret = _nfs4_opendata_to_nfs4_state(data);
1932         nfs4_sequence_free_slot(&data->o_res.seq_res);
1933         return ret;
1934 }
1935
1936 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1937 {
1938         struct nfs_inode *nfsi = NFS_I(state->inode);
1939         struct nfs_open_context *ctx;
1940
1941         spin_lock(&state->inode->i_lock);
1942         list_for_each_entry(ctx, &nfsi->open_files, list) {
1943                 if (ctx->state != state)
1944                         continue;
1945                 get_nfs_open_context(ctx);
1946                 spin_unlock(&state->inode->i_lock);
1947                 return ctx;
1948         }
1949         spin_unlock(&state->inode->i_lock);
1950         return ERR_PTR(-ENOENT);
1951 }
1952
1953 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1954                 struct nfs4_state *state, enum open_claim_type4 claim)
1955 {
1956         struct nfs4_opendata *opendata;
1957
1958         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1959                         NULL, claim, GFP_NOFS);
1960         if (opendata == NULL)
1961                 return ERR_PTR(-ENOMEM);
1962         opendata->state = state;
1963         atomic_inc(&state->count);
1964         return opendata;
1965 }
1966
1967 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1968                 fmode_t fmode)
1969 {
1970         struct nfs4_state *newstate;
1971         int ret;
1972
1973         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1974                 return 0;
1975         opendata->o_arg.open_flags = 0;
1976         opendata->o_arg.fmode = fmode;
1977         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1978                         NFS_SB(opendata->dentry->d_sb),
1979                         fmode, 0);
1980         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1981         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1982         nfs4_init_opendata_res(opendata);
1983         ret = _nfs4_recover_proc_open(opendata);
1984         if (ret != 0)
1985                 return ret;
1986         newstate = nfs4_opendata_to_nfs4_state(opendata);
1987         if (IS_ERR(newstate))
1988                 return PTR_ERR(newstate);
1989         if (newstate != opendata->state)
1990                 ret = -ESTALE;
1991         nfs4_close_state(newstate, fmode);
1992         return ret;
1993 }
1994
1995 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1996 {
1997         int ret;
1998
1999         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
2000         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2001         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2002         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2003         /* memory barrier prior to reading state->n_* */
2004         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2005         clear_bit(NFS_OPEN_STATE, &state->flags);
2006         smp_rmb();
2007         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2008         if (ret != 0)
2009                 return ret;
2010         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2011         if (ret != 0)
2012                 return ret;
2013         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2014         if (ret != 0)
2015                 return ret;
2016         /*
2017          * We may have performed cached opens for all three recoveries.
2018          * Check if we need to update the current stateid.
2019          */
2020         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2021             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2022                 write_seqlock(&state->seqlock);
2023                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2024                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2025                 write_sequnlock(&state->seqlock);
2026         }
2027         return 0;
2028 }
2029
2030 /*
2031  * OPEN_RECLAIM:
2032  *      reclaim state on the server after a reboot.
2033  */
2034 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2035 {
2036         struct nfs_delegation *delegation;
2037         struct nfs4_opendata *opendata;
2038         fmode_t delegation_type = 0;
2039         int status;
2040
2041         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2042                         NFS4_OPEN_CLAIM_PREVIOUS);
2043         if (IS_ERR(opendata))
2044                 return PTR_ERR(opendata);
2045         rcu_read_lock();
2046         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2047         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2048                 delegation_type = delegation->type;
2049         rcu_read_unlock();
2050         opendata->o_arg.u.delegation_type = delegation_type;
2051         status = nfs4_open_recover(opendata, state);
2052         nfs4_opendata_put(opendata);
2053         return status;
2054 }
2055
2056 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2057 {
2058         struct nfs_server *server = NFS_SERVER(state->inode);
2059         struct nfs4_exception exception = { };
2060         int err;
2061         do {
2062                 err = _nfs4_do_open_reclaim(ctx, state);
2063                 trace_nfs4_open_reclaim(ctx, 0, err);
2064                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2065                         continue;
2066                 if (err != -NFS4ERR_DELAY)
2067                         break;
2068                 nfs4_handle_exception(server, err, &exception);
2069         } while (exception.retry);
2070         return err;
2071 }
2072
2073 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2074 {
2075         struct nfs_open_context *ctx;
2076         int ret;
2077
2078         ctx = nfs4_state_find_open_context(state);
2079         if (IS_ERR(ctx))
2080                 return -EAGAIN;
2081         ret = nfs4_do_open_reclaim(ctx, state);
2082         put_nfs_open_context(ctx);
2083         return ret;
2084 }
2085
2086 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)
2087 {
2088         switch (err) {
2089                 default:
2090                         printk(KERN_ERR "NFS: %s: unhandled error "
2091                                         "%d.\n", __func__, err);
2092                 case 0:
2093                 case -ENOENT:
2094                 case -EAGAIN:
2095                 case -ESTALE:
2096                         break;
2097                 case -NFS4ERR_BADSESSION:
2098                 case -NFS4ERR_BADSLOT:
2099                 case -NFS4ERR_BAD_HIGH_SLOT:
2100                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2101                 case -NFS4ERR_DEADSESSION:
2102                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2103                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
2104                         return -EAGAIN;
2105                 case -NFS4ERR_STALE_CLIENTID:
2106                 case -NFS4ERR_STALE_STATEID:
2107                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2108                         /* Don't recall a delegation if it was lost */
2109                         nfs4_schedule_lease_recovery(server->nfs_client);
2110                         return -EAGAIN;
2111                 case -NFS4ERR_MOVED:
2112                         nfs4_schedule_migration_recovery(server);
2113                         return -EAGAIN;
2114                 case -NFS4ERR_LEASE_MOVED:
2115                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2116                         return -EAGAIN;
2117                 case -NFS4ERR_DELEG_REVOKED:
2118                 case -NFS4ERR_ADMIN_REVOKED:
2119                 case -NFS4ERR_EXPIRED:
2120                 case -NFS4ERR_BAD_STATEID:
2121                 case -NFS4ERR_OPENMODE:
2122                         nfs_inode_find_state_and_recover(state->inode,
2123                                         stateid);
2124                         nfs4_schedule_stateid_recovery(server, state);
2125                         return -EAGAIN;
2126                 case -NFS4ERR_DELAY:
2127                 case -NFS4ERR_GRACE:
2128                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2129                         ssleep(1);
2130                         return -EAGAIN;
2131                 case -ENOMEM:
2132                 case -NFS4ERR_DENIED:
2133                         if (fl) {
2134                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2135                                 if (lsp)
2136                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2137                         }
2138                         return 0;
2139         }
2140         return err;
2141 }
2142
2143 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2144                 struct nfs4_state *state, const nfs4_stateid *stateid,
2145                 fmode_t type)
2146 {
2147         struct nfs_server *server = NFS_SERVER(state->inode);
2148         struct nfs4_opendata *opendata;
2149         int err = 0;
2150
2151         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2152                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2153         if (IS_ERR(opendata))
2154                 return PTR_ERR(opendata);
2155         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2156         nfs_state_clear_delegation(state);
2157         switch (type & (FMODE_READ|FMODE_WRITE)) {
2158         case FMODE_READ|FMODE_WRITE:
2159         case FMODE_WRITE:
2160                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2161                 if (err)
2162                         break;
2163                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2164                 if (err)
2165                         break;
2166                 /* Fall through */
2167         case FMODE_READ:
2168                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2169         }
2170         nfs4_opendata_put(opendata);
2171         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2172 }
2173
2174 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2175 {
2176         struct nfs4_opendata *data = calldata;
2177
2178         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2179                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2180 }
2181
2182 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2183 {
2184         struct nfs4_opendata *data = calldata;
2185
2186         nfs40_sequence_done(task, &data->c_res.seq_res);
2187
2188         data->rpc_status = task->tk_status;
2189         if (data->rpc_status == 0) {
2190                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2191                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2192                 renew_lease(data->o_res.server, data->timestamp);
2193                 data->rpc_done = true;
2194         }
2195 }
2196
2197 static void nfs4_open_confirm_release(void *calldata)
2198 {
2199         struct nfs4_opendata *data = calldata;
2200         struct nfs4_state *state = NULL;
2201
2202         /* If this request hasn't been cancelled, do nothing */
2203         if (!data->cancelled)
2204                 goto out_free;
2205         /* In case of error, no cleanup! */
2206         if (!data->rpc_done)
2207                 goto out_free;
2208         state = nfs4_opendata_to_nfs4_state(data);
2209         if (!IS_ERR(state))
2210                 nfs4_close_state(state, data->o_arg.fmode);
2211 out_free:
2212         nfs4_opendata_put(data);
2213 }
2214
2215 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2216         .rpc_call_prepare = nfs4_open_confirm_prepare,
2217         .rpc_call_done = nfs4_open_confirm_done,
2218         .rpc_release = nfs4_open_confirm_release,
2219 };
2220
2221 /*
2222  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2223  */
2224 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2225 {
2226         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2227         struct rpc_task *task;
2228         struct  rpc_message msg = {
2229                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2230                 .rpc_argp = &data->c_arg,
2231                 .rpc_resp = &data->c_res,
2232                 .rpc_cred = data->owner->so_cred,
2233         };
2234         struct rpc_task_setup task_setup_data = {
2235                 .rpc_client = server->client,
2236                 .rpc_message = &msg,
2237                 .callback_ops = &nfs4_open_confirm_ops,
2238                 .callback_data = data,
2239                 .workqueue = nfsiod_workqueue,
2240                 .flags = RPC_TASK_ASYNC,
2241         };
2242         int status;
2243
2244         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2245                                 data->is_recover);
2246         kref_get(&data->kref);
2247         data->rpc_done = false;
2248         data->rpc_status = 0;
2249         data->timestamp = jiffies;
2250         task = rpc_run_task(&task_setup_data);
2251         if (IS_ERR(task))
2252                 return PTR_ERR(task);
2253         status = rpc_wait_for_completion_task(task);
2254         if (status != 0) {
2255                 data->cancelled = true;
2256                 smp_wmb();
2257         } else
2258                 status = data->rpc_status;
2259         rpc_put_task(task);
2260         return status;
2261 }
2262
2263 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2264 {
2265         struct nfs4_opendata *data = calldata;
2266         struct nfs4_state_owner *sp = data->owner;
2267         struct nfs_client *clp = sp->so_server->nfs_client;
2268         enum open_claim_type4 claim = data->o_arg.claim;
2269
2270         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2271                 goto out_wait;
2272         /*
2273          * Check if we still need to send an OPEN call, or if we can use
2274          * a delegation instead.
2275          */
2276         if (data->state != NULL) {
2277                 struct nfs_delegation *delegation;
2278
2279                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2280                         goto out_no_action;
2281                 rcu_read_lock();
2282                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2283                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2284                         goto unlock_no_action;
2285                 rcu_read_unlock();
2286         }
2287         /* Update client id. */
2288         data->o_arg.clientid = clp->cl_clientid;
2289         switch (claim) {
2290         default:
2291                 break;
2292         case NFS4_OPEN_CLAIM_PREVIOUS:
2293         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2294         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2295                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2296                 /* Fall through */
2297         case NFS4_OPEN_CLAIM_FH:
2298                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2299         }
2300         data->timestamp = jiffies;
2301         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2302                                 &data->o_arg.seq_args,
2303                                 &data->o_res.seq_res,
2304                                 task) != 0)
2305                 nfs_release_seqid(data->o_arg.seqid);
2306
2307         /* Set the create mode (note dependency on the session type) */
2308         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2309         if (data->o_arg.open_flags & O_EXCL) {
2310                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2311                 if (nfs4_has_persistent_session(clp))
2312                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2313                 else if (clp->cl_mvops->minor_version > 0)
2314                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2315         }
2316         return;
2317 unlock_no_action:
2318         trace_nfs4_cached_open(data->state);
2319         rcu_read_unlock();
2320 out_no_action:
2321         task->tk_action = NULL;
2322 out_wait:
2323         nfs4_sequence_done(task, &data->o_res.seq_res);
2324 }
2325
2326 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2327 {
2328         struct nfs4_opendata *data = calldata;
2329
2330         data->rpc_status = task->tk_status;
2331
2332         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2333                 return;
2334
2335         if (task->tk_status == 0) {
2336                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2337                         switch (data->o_res.f_attr->mode & S_IFMT) {
2338                         case S_IFREG:
2339                                 break;
2340                         case S_IFLNK:
2341                                 data->rpc_status = -ELOOP;
2342                                 break;
2343                         case S_IFDIR:
2344                                 data->rpc_status = -EISDIR;
2345                                 break;
2346                         default:
2347                                 data->rpc_status = -ENOTDIR;
2348                         }
2349                 }
2350                 renew_lease(data->o_res.server, data->timestamp);
2351                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2352                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2353         }
2354         data->rpc_done = true;
2355 }
2356
2357 static void nfs4_open_release(void *calldata)
2358 {
2359         struct nfs4_opendata *data = calldata;
2360         struct nfs4_state *state = NULL;
2361
2362         /* If this request hasn't been cancelled, do nothing */
2363         if (!data->cancelled)
2364                 goto out_free;
2365         /* In case of error, no cleanup! */
2366         if (data->rpc_status != 0 || !data->rpc_done)
2367                 goto out_free;
2368         /* In case we need an open_confirm, no cleanup! */
2369         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2370                 goto out_free;
2371         state = nfs4_opendata_to_nfs4_state(data);
2372         if (!IS_ERR(state))
2373                 nfs4_close_state(state, data->o_arg.fmode);
2374 out_free:
2375         nfs4_opendata_put(data);
2376 }
2377
2378 static const struct rpc_call_ops nfs4_open_ops = {
2379         .rpc_call_prepare = nfs4_open_prepare,
2380         .rpc_call_done = nfs4_open_done,
2381         .rpc_release = nfs4_open_release,
2382 };
2383
2384 static int nfs4_run_open_task(struct nfs4_opendata *data,
2385                               struct nfs_open_context *ctx)
2386 {
2387         struct inode *dir = d_inode(data->dir);
2388         struct nfs_server *server = NFS_SERVER(dir);
2389         struct nfs_openargs *o_arg = &data->o_arg;
2390         struct nfs_openres *o_res = &data->o_res;
2391         struct rpc_task *task;
2392         struct rpc_message msg = {
2393                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2394                 .rpc_argp = o_arg,
2395                 .rpc_resp = o_res,
2396                 .rpc_cred = data->owner->so_cred,
2397         };
2398         struct rpc_task_setup task_setup_data = {
2399                 .rpc_client = server->client,
2400                 .rpc_message = &msg,
2401                 .callback_ops = &nfs4_open_ops,
2402                 .callback_data = data,
2403                 .workqueue = nfsiod_workqueue,
2404                 .flags = RPC_TASK_ASYNC,
2405         };
2406         int status;
2407
2408         kref_get(&data->kref);
2409         data->rpc_done = false;
2410         data->rpc_status = 0;
2411         data->cancelled = false;
2412         data->is_recover = false;
2413         if (!ctx) {
2414                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2415                 data->is_recover = true;
2416         } else {
2417                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2418                 pnfs_lgopen_prepare(data, ctx);
2419         }
2420         task = rpc_run_task(&task_setup_data);
2421         if (IS_ERR(task))
2422                 return PTR_ERR(task);
2423         status = rpc_wait_for_completion_task(task);
2424         if (status != 0) {
2425                 data->cancelled = true;
2426                 smp_wmb();
2427         } else
2428                 status = data->rpc_status;
2429         rpc_put_task(task);
2430
2431         return status;
2432 }
2433
2434 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2435 {
2436         struct inode *dir = d_inode(data->dir);
2437         struct nfs_openres *o_res = &data->o_res;
2438         int status;
2439
2440         status = nfs4_run_open_task(data, NULL);
2441         if (status != 0 || !data->rpc_done)
2442                 return status;
2443
2444         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2445
2446         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2447                 status = _nfs4_proc_open_confirm(data);
2448
2449         return status;
2450 }
2451
2452 /*
2453  * Additional permission checks in order to distinguish between an
2454  * open for read, and an open for execute. This works around the
2455  * fact that NFSv4 OPEN treats read and execute permissions as being
2456  * the same.
2457  * Note that in the non-execute case, we want to turn off permission
2458  * checking if we just created a new file (POSIX open() semantics).
2459  */
2460 static int nfs4_opendata_access(struct rpc_cred *cred,
2461                                 struct nfs4_opendata *opendata,
2462                                 struct nfs4_state *state, fmode_t fmode,
2463                                 int openflags)
2464 {
2465         struct nfs_access_entry cache;
2466         u32 mask, flags;
2467
2468         /* access call failed or for some reason the server doesn't
2469          * support any access modes -- defer access call until later */
2470         if (opendata->o_res.access_supported == 0)
2471                 return 0;
2472
2473         mask = 0;
2474         /*
2475          * Use openflags to check for exec, because fmode won't
2476          * always have FMODE_EXEC set when file open for exec.
2477          */
2478         if (openflags & __FMODE_EXEC) {
2479                 /* ONLY check for exec rights */
2480                 if (S_ISDIR(state->inode->i_mode))
2481                         mask = NFS4_ACCESS_LOOKUP;
2482                 else
2483                         mask = NFS4_ACCESS_EXECUTE;
2484         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2485                 mask = NFS4_ACCESS_READ;
2486
2487         cache.cred = cred;
2488         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2489         nfs_access_add_cache(state->inode, &cache);
2490
2491         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2492         if ((mask & ~cache.mask & flags) == 0)
2493                 return 0;
2494
2495         return -EACCES;
2496 }
2497
2498 /*
2499  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2500  */
2501 static int _nfs4_proc_open(struct nfs4_opendata *data,
2502                            struct nfs_open_context *ctx)
2503 {
2504         struct inode *dir = d_inode(data->dir);
2505         struct nfs_server *server = NFS_SERVER(dir);
2506         struct nfs_openargs *o_arg = &data->o_arg;
2507         struct nfs_openres *o_res = &data->o_res;
2508         int status;
2509
2510         status = nfs4_run_open_task(data, ctx);
2511         if (!data->rpc_done)
2512                 return status;
2513         if (status != 0) {
2514                 if (status == -NFS4ERR_BADNAME &&
2515                                 !(o_arg->open_flags & O_CREAT))
2516                         return -ENOENT;
2517                 return status;
2518         }
2519
2520         nfs_fattr_map_and_free_names(server, &data->f_attr);
2521
2522         if (o_arg->open_flags & O_CREAT) {
2523                 if (o_arg->open_flags & O_EXCL)
2524                         data->file_created = true;
2525                 else if (o_res->cinfo.before != o_res->cinfo.after)
2526                         data->file_created = true;
2527                 if (data->file_created ||
2528                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2529                         update_changeattr(dir, &o_res->cinfo,
2530                                         o_res->f_attr->time_start, 0);
2531         }
2532         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2533                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2534         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2535                 status = _nfs4_proc_open_confirm(data);
2536                 if (status != 0)
2537                         return status;
2538         }
2539         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2540                 nfs4_sequence_free_slot(&o_res->seq_res);
2541                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2542                                 o_res->f_label, NULL);
2543         }
2544         return 0;
2545 }
2546
2547 /*
2548  * OPEN_EXPIRED:
2549  *      reclaim state on the server after a network partition.
2550  *      Assumes caller holds the appropriate lock
2551  */
2552 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2553 {
2554         struct nfs4_opendata *opendata;
2555         int ret;
2556
2557         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2558                         NFS4_OPEN_CLAIM_FH);
2559         if (IS_ERR(opendata))
2560                 return PTR_ERR(opendata);
2561         ret = nfs4_open_recover(opendata, state);
2562         if (ret == -ESTALE)
2563                 d_drop(ctx->dentry);
2564         nfs4_opendata_put(opendata);
2565         return ret;
2566 }
2567
2568 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2569 {
2570         struct nfs_server *server = NFS_SERVER(state->inode);
2571         struct nfs4_exception exception = { };
2572         int err;
2573
2574         do {
2575                 err = _nfs4_open_expired(ctx, state);
2576                 trace_nfs4_open_expired(ctx, 0, err);
2577                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2578                         continue;
2579                 switch (err) {
2580                 default:
2581                         goto out;
2582                 case -NFS4ERR_GRACE:
2583                 case -NFS4ERR_DELAY:
2584                         nfs4_handle_exception(server, err, &exception);
2585                         err = 0;
2586                 }
2587         } while (exception.retry);
2588 out:
2589         return err;
2590 }
2591
2592 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2593 {
2594         struct nfs_open_context *ctx;
2595         int ret;
2596
2597         ctx = nfs4_state_find_open_context(state);
2598         if (IS_ERR(ctx))
2599                 return -EAGAIN;
2600         ret = nfs4_do_open_expired(ctx, state);
2601         put_nfs_open_context(ctx);
2602         return ret;
2603 }
2604
2605 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2606                 const nfs4_stateid *stateid)
2607 {
2608         nfs_remove_bad_delegation(state->inode, stateid);
2609         nfs_state_clear_delegation(state);
2610 }
2611
2612 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2613 {
2614         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2615                 nfs_finish_clear_delegation_stateid(state, NULL);
2616 }
2617
2618 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2619 {
2620         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2621         nfs40_clear_delegation_stateid(state);
2622         return nfs4_open_expired(sp, state);
2623 }
2624
2625 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2626                 nfs4_stateid *stateid,
2627                 struct rpc_cred *cred)
2628 {
2629         return -NFS4ERR_BAD_STATEID;
2630 }
2631
2632 #if defined(CONFIG_NFS_V4_1)
2633 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2634                 nfs4_stateid *stateid,
2635                 struct rpc_cred *cred)
2636 {
2637         int status;
2638
2639         switch (stateid->type) {
2640         default:
2641                 break;
2642         case NFS4_INVALID_STATEID_TYPE:
2643         case NFS4_SPECIAL_STATEID_TYPE:
2644                 return -NFS4ERR_BAD_STATEID;
2645         case NFS4_REVOKED_STATEID_TYPE:
2646                 goto out_free;
2647         }
2648
2649         status = nfs41_test_stateid(server, stateid, cred);
2650         switch (status) {
2651         case -NFS4ERR_EXPIRED:
2652         case -NFS4ERR_ADMIN_REVOKED:
2653         case -NFS4ERR_DELEG_REVOKED:
2654                 break;
2655         default:
2656                 return status;
2657         }
2658 out_free:
2659         /* Ack the revoked state to the server */
2660         nfs41_free_stateid(server, stateid, cred, true);
2661         return -NFS4ERR_EXPIRED;
2662 }
2663
2664 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2665 {
2666         struct nfs_server *server = NFS_SERVER(state->inode);
2667         nfs4_stateid stateid;
2668         struct nfs_delegation *delegation;
2669         struct rpc_cred *cred;
2670         int status;
2671
2672         /* Get the delegation credential for use by test/free_stateid */
2673         rcu_read_lock();
2674         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2675         if (delegation == NULL) {
2676                 rcu_read_unlock();
2677                 nfs_state_clear_delegation(state);
2678                 return;
2679         }
2680
2681         nfs4_stateid_copy(&stateid, &delegation->stateid);
2682         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2683                 rcu_read_unlock();
2684                 nfs_state_clear_delegation(state);
2685                 return;
2686         }
2687
2688         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2689                                 &delegation->flags)) {
2690                 rcu_read_unlock();
2691                 return;
2692         }
2693
2694         cred = get_rpccred(delegation->cred);
2695         rcu_read_unlock();
2696         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2697         trace_nfs4_test_delegation_stateid(state, NULL, status);
2698         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2699                 nfs_finish_clear_delegation_stateid(state, &stateid);
2700
2701         put_rpccred(cred);
2702 }
2703
2704 /**
2705  * nfs41_check_expired_locks - possibly free a lock stateid
2706  *
2707  * @state: NFSv4 state for an inode
2708  *
2709  * Returns NFS_OK if recovery for this stateid is now finished.
2710  * Otherwise a negative NFS4ERR value is returned.
2711  */
2712 static int nfs41_check_expired_locks(struct nfs4_state *state)
2713 {
2714         int status, ret = NFS_OK;
2715         struct nfs4_lock_state *lsp, *prev = NULL;
2716         struct nfs_server *server = NFS_SERVER(state->inode);
2717
2718         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2719                 goto out;
2720
2721         spin_lock(&state->state_lock);
2722         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2723                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2724                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2725
2726                         refcount_inc(&lsp->ls_count);
2727                         spin_unlock(&state->state_lock);
2728
2729                         nfs4_put_lock_state(prev);
2730                         prev = lsp;
2731
2732                         status = nfs41_test_and_free_expired_stateid(server,
2733                                         &lsp->ls_stateid,
2734                                         cred);
2735                         trace_nfs4_test_lock_stateid(state, lsp, status);
2736                         if (status == -NFS4ERR_EXPIRED ||
2737                             status == -NFS4ERR_BAD_STATEID) {
2738                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2739                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2740                                 if (!recover_lost_locks)
2741                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2742                         } else if (status != NFS_OK) {
2743                                 ret = status;
2744                                 nfs4_put_lock_state(prev);
2745                                 goto out;
2746                         }
2747                         spin_lock(&state->state_lock);
2748                 }
2749         }
2750         spin_unlock(&state->state_lock);
2751         nfs4_put_lock_state(prev);
2752 out:
2753         return ret;
2754 }
2755
2756 /**
2757  * nfs41_check_open_stateid - possibly free an open stateid
2758  *
2759  * @state: NFSv4 state for an inode
2760  *
2761  * Returns NFS_OK if recovery for this stateid is now finished.
2762  * Otherwise a negative NFS4ERR value is returned.
2763  */
2764 static int nfs41_check_open_stateid(struct nfs4_state *state)
2765 {
2766         struct nfs_server *server = NFS_SERVER(state->inode);
2767         nfs4_stateid *stateid = &state->open_stateid;
2768         struct rpc_cred *cred = state->owner->so_cred;
2769         int status;
2770
2771         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2772                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2773                         if (nfs4_have_delegation(state->inode, state->state))
2774                                 return NFS_OK;
2775                         return -NFS4ERR_OPENMODE;
2776                 }
2777                 return -NFS4ERR_BAD_STATEID;
2778         }
2779         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2780         trace_nfs4_test_open_stateid(state, NULL, status);
2781         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2782                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2783                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2784                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2785                 clear_bit(NFS_OPEN_STATE, &state->flags);
2786                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2787                 return status;
2788         }
2789         if (nfs_open_stateid_recover_openmode(state))
2790                 return -NFS4ERR_OPENMODE;
2791         return NFS_OK;
2792 }
2793
2794 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2795 {
2796         int status;
2797
2798         nfs41_check_delegation_stateid(state);
2799         status = nfs41_check_expired_locks(state);
2800         if (status != NFS_OK)
2801                 return status;
2802         status = nfs41_check_open_stateid(state);
2803         if (status != NFS_OK)
2804                 status = nfs4_open_expired(sp, state);
2805         return status;
2806 }
2807 #endif
2808
2809 /*
2810  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2811  * fields corresponding to attributes that were used to store the verifier.
2812  * Make sure we clobber those fields in the later setattr call
2813  */
2814 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2815                                 struct iattr *sattr, struct nfs4_label **label)
2816 {
2817         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2818         __u32 attrset[3];
2819         unsigned ret;
2820         unsigned i;
2821
2822         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2823                 attrset[i] = opendata->o_res.attrset[i];
2824                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2825                         attrset[i] &= ~bitmask[i];
2826         }
2827
2828         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2829                 sattr->ia_valid : 0;
2830
2831         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2832                 if (sattr->ia_valid & ATTR_ATIME_SET)
2833                         ret |= ATTR_ATIME_SET;
2834                 else
2835                         ret |= ATTR_ATIME;
2836         }
2837
2838         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2839                 if (sattr->ia_valid & ATTR_MTIME_SET)
2840                         ret |= ATTR_MTIME_SET;
2841                 else
2842                         ret |= ATTR_MTIME;
2843         }
2844
2845         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2846                 *label = NULL;
2847         return ret;
2848 }
2849
2850 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2851                 fmode_t fmode,
2852                 int flags,
2853                 struct nfs_open_context *ctx)
2854 {
2855         struct nfs4_state_owner *sp = opendata->owner;
2856         struct nfs_server *server = sp->so_server;
2857         struct dentry *dentry;
2858         struct nfs4_state *state;
2859         unsigned int seq;
2860         int ret;
2861
2862         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2863
2864         ret = _nfs4_proc_open(opendata, ctx);
2865         if (ret != 0)
2866                 goto out;
2867
2868         state = _nfs4_opendata_to_nfs4_state(opendata);
2869         ret = PTR_ERR(state);
2870         if (IS_ERR(state))
2871                 goto out;
2872         ctx->state = state;
2873         if (server->caps & NFS_CAP_POSIX_LOCK)
2874                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2875         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2876                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2877
2878         dentry = opendata->dentry;
2879         if (d_really_is_negative(dentry)) {
2880                 struct dentry *alias;
2881                 d_drop(dentry);
2882                 alias = d_exact_alias(dentry, state->inode);
2883                 if (!alias)
2884                         alias = d_splice_alias(igrab(state->inode), dentry);
2885                 /* d_splice_alias() can't fail here - it's a non-directory */
2886                 if (alias) {
2887                         dput(ctx->dentry);
2888                         ctx->dentry = dentry = alias;
2889                 }
2890                 nfs_set_verifier(dentry,
2891                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2892         }
2893
2894         /* Parse layoutget results before we check for access */
2895         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
2896
2897         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2898         if (ret != 0)
2899                 goto out;
2900
2901         if (d_inode(dentry) == state->inode) {
2902                 nfs_inode_attach_open_context(ctx);
2903                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2904                         nfs4_schedule_stateid_recovery(server, state);
2905         }
2906
2907 out:
2908         nfs4_sequence_free_slot(&opendata->o_res.seq_res);
2909         return ret;
2910 }
2911
2912 /*
2913  * Returns a referenced nfs4_state
2914  */
2915 static int _nfs4_do_open(struct inode *dir,
2916                         struct nfs_open_context *ctx,
2917                         int flags,
2918                         const struct nfs4_open_createattrs *c,
2919                         int *opened)
2920 {
2921         struct nfs4_state_owner  *sp;
2922         struct nfs4_state     *state = NULL;
2923         struct nfs_server       *server = NFS_SERVER(dir);
2924         struct nfs4_opendata *opendata;
2925         struct dentry *dentry = ctx->dentry;
2926         struct rpc_cred *cred = ctx->cred;
2927         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2928         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2929         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2930         struct iattr *sattr = c->sattr;
2931         struct nfs4_label *label = c->label;
2932         struct nfs4_label *olabel = NULL;
2933         int status;
2934
2935         /* Protect against reboot recovery conflicts */
2936         status = -ENOMEM;
2937         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2938         if (sp == NULL) {
2939                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2940                 goto out_err;
2941         }
2942         status = nfs4_client_recover_expired_lease(server->nfs_client);
2943         if (status != 0)
2944                 goto err_put_state_owner;
2945         if (d_really_is_positive(dentry))
2946                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2947         status = -ENOMEM;
2948         if (d_really_is_positive(dentry))
2949                 claim = NFS4_OPEN_CLAIM_FH;
2950         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
2951                         c, claim, GFP_KERNEL);
2952         if (opendata == NULL)
2953                 goto err_put_state_owner;
2954
2955         if (label) {
2956                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2957                 if (IS_ERR(olabel)) {
2958                         status = PTR_ERR(olabel);
2959                         goto err_opendata_put;
2960                 }
2961         }
2962
2963         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2964                 if (!opendata->f_attr.mdsthreshold) {
2965                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2966                         if (!opendata->f_attr.mdsthreshold)
2967                                 goto err_free_label;
2968                 }
2969                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2970         }
2971         if (d_really_is_positive(dentry))
2972                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2973
2974         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2975         if (status != 0)
2976                 goto err_free_label;
2977         state = ctx->state;
2978
2979         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2980             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2981                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
2982                 /*
2983                  * send create attributes which was not set by open
2984                  * with an extra setattr.
2985                  */
2986                 if (attrs || label) {
2987                         unsigned ia_old = sattr->ia_valid;
2988
2989                         sattr->ia_valid = attrs;
2990                         nfs_fattr_init(opendata->o_res.f_attr);
2991                         status = nfs4_do_setattr(state->inode, cred,
2992                                         opendata->o_res.f_attr, sattr,
2993                                         ctx, label, olabel);
2994                         if (status == 0) {
2995                                 nfs_setattr_update_inode(state->inode, sattr,
2996                                                 opendata->o_res.f_attr);
2997                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2998                         }
2999                         sattr->ia_valid = ia_old;
3000                 }
3001         }
3002         if (opened && opendata->file_created)
3003                 *opened = 1;
3004
3005         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3006                 *ctx_th = opendata->f_attr.mdsthreshold;
3007                 opendata->f_attr.mdsthreshold = NULL;
3008         }
3009
3010         nfs4_label_free(olabel);
3011
3012         nfs4_opendata_put(opendata);
3013         nfs4_put_state_owner(sp);
3014         return 0;
3015 err_free_label:
3016         nfs4_label_free(olabel);
3017 err_opendata_put:
3018         nfs4_opendata_put(opendata);
3019 err_put_state_owner:
3020         nfs4_put_state_owner(sp);
3021 out_err:
3022         return status;
3023 }
3024
3025
3026 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3027                                         struct nfs_open_context *ctx,
3028                                         int flags,
3029                                         struct iattr *sattr,
3030                                         struct nfs4_label *label,
3031                                         int *opened)
3032 {
3033         struct nfs_server *server = NFS_SERVER(dir);
3034         struct nfs4_exception exception = { };
3035         struct nfs4_state *res;
3036         struct nfs4_open_createattrs c = {
3037                 .label = label,
3038                 .sattr = sattr,
3039                 .verf = {
3040                         [0] = (__u32)jiffies,
3041                         [1] = (__u32)current->pid,
3042                 },
3043         };
3044         int status;
3045
3046         do {
3047                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3048                 res = ctx->state;
3049                 trace_nfs4_open_file(ctx, flags, status);
3050                 if (status == 0)
3051                         break;
3052                 /* NOTE: BAD_SEQID means the server and client disagree about the
3053                  * book-keeping w.r.t. state-changing operations
3054                  * (OPEN/CLOSE/LOCK/LOCKU...)
3055                  * It is actually a sign of a bug on the client or on the server.
3056                  *
3057                  * If we receive a BAD_SEQID error in the particular case of
3058                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3059                  * have unhashed the old state_owner for us, and that we can
3060                  * therefore safely retry using a new one. We should still warn
3061                  * the user though...
3062                  */
3063                 if (status == -NFS4ERR_BAD_SEQID) {
3064                         pr_warn_ratelimited("NFS: v4 server %s "
3065                                         " returned a bad sequence-id error!\n",
3066                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3067                         exception.retry = 1;
3068                         continue;
3069                 }
3070                 /*
3071                  * BAD_STATEID on OPEN means that the server cancelled our
3072                  * state before it received the OPEN_CONFIRM.
3073                  * Recover by retrying the request as per the discussion
3074                  * on Page 181 of RFC3530.
3075                  */
3076                 if (status == -NFS4ERR_BAD_STATEID) {
3077                         exception.retry = 1;
3078                         continue;
3079                 }
3080                 if (status == -EAGAIN) {
3081                         /* We must have found a delegation */
3082                         exception.retry = 1;
3083                         continue;
3084                 }
3085                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3086                         continue;
3087                 res = ERR_PTR(nfs4_handle_exception(server,
3088                                         status, &exception));
3089         } while (exception.retry);
3090         return res;
3091 }
3092
3093 static int _nfs4_do_setattr(struct inode *inode,
3094                             struct nfs_setattrargs *arg,
3095                             struct nfs_setattrres *res,
3096                             struct rpc_cred *cred,
3097                             struct nfs_open_context *ctx)
3098 {
3099         struct nfs_server *server = NFS_SERVER(inode);
3100         struct rpc_message msg = {
3101                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3102                 .rpc_argp       = arg,
3103                 .rpc_resp       = res,
3104                 .rpc_cred       = cred,
3105         };
3106         struct rpc_cred *delegation_cred = NULL;
3107         unsigned long timestamp = jiffies;
3108         bool truncate;
3109         int status;
3110
3111         nfs_fattr_init(res->fattr);
3112
3113         /* Servers should only apply open mode checks for file size changes */
3114         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3115         if (!truncate)
3116                 goto zero_stateid;
3117
3118         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3119                 /* Use that stateid */
3120         } else if (ctx != NULL) {
3121                 struct nfs_lock_context *l_ctx;
3122                 if (!nfs4_valid_open_stateid(ctx->state))
3123                         return -EBADF;
3124                 l_ctx = nfs_get_lock_context(ctx);
3125                 if (IS_ERR(l_ctx))
3126                         return PTR_ERR(l_ctx);
3127                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3128                                                 &arg->stateid, &delegation_cred);
3129                 nfs_put_lock_context(l_ctx);
3130                 if (status == -EIO)
3131                         return -EBADF;
3132         } else {
3133 zero_stateid:
3134                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3135         }
3136         if (delegation_cred)
3137                 msg.rpc_cred = delegation_cred;
3138
3139         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3140
3141         put_rpccred(delegation_cred);
3142         if (status == 0 && ctx != NULL)
3143                 renew_lease(server, timestamp);
3144         trace_nfs4_setattr(inode, &arg->stateid, status);
3145         return status;
3146 }
3147
3148 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
3149                            struct nfs_fattr *fattr, struct iattr *sattr,
3150                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3151                            struct nfs4_label *olabel)
3152 {
3153         struct nfs_server *server = NFS_SERVER(inode);
3154         __u32 bitmask[NFS4_BITMASK_SZ];
3155         struct nfs4_state *state = ctx ? ctx->state : NULL;
3156         struct nfs_setattrargs  arg = {
3157                 .fh             = NFS_FH(inode),
3158                 .iap            = sattr,
3159                 .server         = server,
3160                 .bitmask = bitmask,
3161                 .label          = ilabel,
3162         };
3163         struct nfs_setattrres  res = {
3164                 .fattr          = fattr,
3165                 .label          = olabel,
3166                 .server         = server,
3167         };
3168         struct nfs4_exception exception = {
3169                 .state = state,
3170                 .inode = inode,
3171                 .stateid = &arg.stateid,
3172         };
3173         int err;
3174
3175         do {
3176                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3177                                 nfs4_bitmask(server, olabel),
3178                                 inode);
3179
3180                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3181                 switch (err) {
3182                 case -NFS4ERR_OPENMODE:
3183                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3184                                 pr_warn_once("NFSv4: server %s is incorrectly "
3185                                                 "applying open mode checks to "
3186                                                 "a SETATTR that is not "
3187                                                 "changing file size.\n",
3188                                                 server->nfs_client->cl_hostname);
3189                         }
3190                         if (state && !(state->state & FMODE_WRITE)) {
3191                                 err = -EBADF;
3192                                 if (sattr->ia_valid & ATTR_OPEN)
3193                                         err = -EACCES;
3194                                 goto out;
3195                         }
3196                 }
3197                 err = nfs4_handle_exception(server, err, &exception);
3198         } while (exception.retry);
3199 out:
3200         return err;
3201 }
3202
3203 static bool
3204 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3205 {
3206         if (inode == NULL || !nfs_have_layout(inode))
3207                 return false;
3208
3209         return pnfs_wait_on_layoutreturn(inode, task);
3210 }
3211
3212 struct nfs4_closedata {
3213         struct inode *inode;
3214         struct nfs4_state *state;
3215         struct nfs_closeargs arg;
3216         struct nfs_closeres res;
3217         struct {
3218                 struct nfs4_layoutreturn_args arg;
3219                 struct nfs4_layoutreturn_res res;
3220                 struct nfs4_xdr_opaque_data ld_private;
3221                 u32 roc_barrier;
3222                 bool roc;
3223         } lr;
3224         struct nfs_fattr fattr;
3225         unsigned long timestamp;
3226 };
3227
3228 static void nfs4_free_closedata(void *data)
3229 {
3230         struct nfs4_closedata *calldata = data;
3231         struct nfs4_state_owner *sp = calldata->state->owner;
3232         struct super_block *sb = calldata->state->inode->i_sb;
3233
3234         if (calldata->lr.roc)
3235                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3236                                 calldata->res.lr_ret);
3237         nfs4_put_open_state(calldata->state);
3238         nfs_free_seqid(calldata->arg.seqid);
3239         nfs4_put_state_owner(sp);
3240         nfs_sb_deactive(sb);
3241         kfree(calldata);
3242 }
3243
3244 static void nfs4_close_done(struct rpc_task *task, void *data)
3245 {
3246         struct nfs4_closedata *calldata = data;
3247         struct nfs4_state *state = calldata->state;
3248         struct nfs_server *server = NFS_SERVER(calldata->inode);
3249         nfs4_stateid *res_stateid = NULL;
3250         struct nfs4_exception exception = {
3251                 .state = state,
3252                 .inode = calldata->inode,
3253                 .stateid = &calldata->arg.stateid,
3254         };
3255
3256         dprintk("%s: begin!\n", __func__);
3257         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3258                 return;
3259         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3260
3261         /* Handle Layoutreturn errors */
3262         if (calldata->arg.lr_args && task->tk_status != 0) {
3263                 switch (calldata->res.lr_ret) {
3264                 default:
3265                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3266                         break;
3267                 case 0:
3268                         calldata->arg.lr_args = NULL;
3269                         calldata->res.lr_res = NULL;
3270                         break;
3271                 case -NFS4ERR_OLD_STATEID:
3272                         if (nfs4_layoutreturn_refresh_stateid(&calldata->arg.lr_args->stateid,
3273                                                 &calldata->arg.lr_args->range,
3274                                                 calldata->inode))
3275                                 goto lr_restart;
3276                         /* Fallthrough */
3277                 case -NFS4ERR_ADMIN_REVOKED:
3278                 case -NFS4ERR_DELEG_REVOKED:
3279                 case -NFS4ERR_EXPIRED:
3280                 case -NFS4ERR_BAD_STATEID:
3281                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3282                 case -NFS4ERR_WRONG_CRED:
3283                         calldata->arg.lr_args = NULL;
3284                         calldata->res.lr_res = NULL;
3285                         goto lr_restart;
3286                 }
3287         }
3288
3289         /* hmm. we are done with the inode, and in the process of freeing
3290          * the state_owner. we keep this around to process errors
3291          */
3292         switch (task->tk_status) {
3293                 case 0:
3294                         res_stateid = &calldata->res.stateid;
3295                         renew_lease(server, calldata->timestamp);
3296                         break;
3297                 case -NFS4ERR_ACCESS:
3298                         if (calldata->arg.bitmask != NULL) {
3299                                 calldata->arg.bitmask = NULL;
3300                                 calldata->res.fattr = NULL;
3301                                 goto out_restart;
3302
3303                         }
3304                         break;
3305                 case -NFS4ERR_OLD_STATEID:
3306                         /* Did we race with OPEN? */
3307                         if (nfs4_refresh_open_stateid(&calldata->arg.stateid,
3308                                                 state))
3309                                 goto out_restart;
3310                         goto out_release;
3311                 case -NFS4ERR_ADMIN_REVOKED:
3312                 case -NFS4ERR_STALE_STATEID:
3313                 case -NFS4ERR_EXPIRED:
3314                         nfs4_free_revoked_stateid(server,
3315                                         &calldata->arg.stateid,
3316                                         task->tk_msg.rpc_cred);
3317                         /* Fallthrough */
3318                 case -NFS4ERR_BAD_STATEID:
3319                         break;
3320                 default:
3321                         task->tk_status = nfs4_async_handle_exception(task,
3322                                         server, task->tk_status, &exception);
3323                         if (exception.retry)
3324                                 goto out_restart;
3325         }
3326         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3327                         res_stateid, calldata->arg.fmode);
3328 out_release:
3329         task->tk_status = 0;
3330         nfs_release_seqid(calldata->arg.seqid);
3331         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3332         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3333         return;
3334 lr_restart:
3335         calldata->res.lr_ret = 0;
3336 out_restart:
3337         task->tk_status = 0;
3338         rpc_restart_call_prepare(task);
3339         goto out_release;
3340 }
3341
3342 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3343 {
3344         struct nfs4_closedata *calldata = data;
3345         struct nfs4_state *state = calldata->state;
3346         struct inode *inode = calldata->inode;
3347         struct pnfs_layout_hdr *lo;
3348         bool is_rdonly, is_wronly, is_rdwr;
3349         int call_close = 0;
3350
3351         dprintk("%s: begin!\n", __func__);
3352         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3353                 goto out_wait;
3354
3355         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3356         spin_lock(&state->owner->so_lock);
3357         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3358         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3359         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3360         /* Calculate the change in open mode */
3361         calldata->arg.fmode = 0;
3362         if (state->n_rdwr == 0) {
3363                 if (state->n_rdonly == 0)
3364                         call_close |= is_rdonly;
3365                 else if (is_rdonly)
3366                         calldata->arg.fmode |= FMODE_READ;
3367                 if (state->n_wronly == 0)
3368                         call_close |= is_wronly;
3369                 else if (is_wronly)
3370                         calldata->arg.fmode |= FMODE_WRITE;
3371                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3372                         call_close |= is_rdwr;
3373         } else if (is_rdwr)
3374                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3375
3376         if (!nfs4_valid_open_stateid(state) ||
3377             !nfs4_refresh_open_stateid(&calldata->arg.stateid, state))
3378                 call_close = 0;
3379         spin_unlock(&state->owner->so_lock);
3380
3381         if (!call_close) {
3382                 /* Note: exit _without_ calling nfs4_close_done */
3383                 goto out_no_action;
3384         }
3385
3386         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3387                 nfs_release_seqid(calldata->arg.seqid);
3388                 goto out_wait;
3389         }
3390
3391         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3392         if (lo && !pnfs_layout_is_valid(lo)) {
3393                 calldata->arg.lr_args = NULL;
3394                 calldata->res.lr_res = NULL;
3395         }
3396
3397         if (calldata->arg.fmode == 0)
3398                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3399
3400         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3401                 /* Close-to-open cache consistency revalidation */
3402                 if (!nfs4_have_delegation(inode, FMODE_READ))
3403                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3404                 else
3405                         calldata->arg.bitmask = NULL;
3406         }
3407
3408         calldata->arg.share_access =
3409                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3410                                 calldata->arg.fmode, 0);
3411
3412         if (calldata->res.fattr == NULL)
3413                 calldata->arg.bitmask = NULL;
3414         else if (calldata->arg.bitmask == NULL)
3415                 calldata->res.fattr = NULL;
3416         calldata->timestamp = jiffies;
3417         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3418                                 &calldata->arg.seq_args,
3419                                 &calldata->res.seq_res,
3420                                 task) != 0)
3421                 nfs_release_seqid(calldata->arg.seqid);
3422         dprintk("%s: done!\n", __func__);
3423         return;
3424 out_no_action:
3425         task->tk_action = NULL;
3426 out_wait:
3427         nfs4_sequence_done(task, &calldata->res.seq_res);
3428 }
3429
3430 static const struct rpc_call_ops nfs4_close_ops = {
3431         .rpc_call_prepare = nfs4_close_prepare,
3432         .rpc_call_done = nfs4_close_done,
3433         .rpc_release = nfs4_free_closedata,
3434 };
3435
3436 /*
3437  * It is possible for data to be read/written from a mem-mapped file
3438  * after the sys_close call (which hits the vfs layer as a flush).
3439  * This means that we can't safely call nfsv4 close on a file until
3440  * the inode is cleared. This in turn means that we are not good
3441  * NFSv4 citizens - we do not indicate to the server to update the file's
3442  * share state even when we are done with one of the three share
3443  * stateid's in the inode.
3444  *
3445  * NOTE: Caller must be holding the sp->so_owner semaphore!
3446  */
3447 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3448 {
3449         struct nfs_server *server = NFS_SERVER(state->inode);
3450         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3451         struct nfs4_closedata *calldata;
3452         struct nfs4_state_owner *sp = state->owner;
3453         struct rpc_task *task;
3454         struct rpc_message msg = {
3455                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3456                 .rpc_cred = state->owner->so_cred,
3457         };
3458         struct rpc_task_setup task_setup_data = {
3459                 .rpc_client = server->client,
3460                 .rpc_message = &msg,
3461                 .callback_ops = &nfs4_close_ops,
3462                 .workqueue = nfsiod_workqueue,
3463                 .flags = RPC_TASK_ASYNC,
3464         };
3465         int status = -ENOMEM;
3466
3467         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3468                 &task_setup_data.rpc_client, &msg);
3469
3470         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3471         if (calldata == NULL)
3472                 goto out;
3473         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3474         calldata->inode = state->inode;
3475         calldata->state = state;
3476         calldata->arg.fh = NFS_FH(state->inode);
3477         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3478                 goto out_free_calldata;
3479         /* Serialization for the sequence id */
3480         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3481         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3482         if (IS_ERR(calldata->arg.seqid))
3483                 goto out_free_calldata;
3484         nfs_fattr_init(&calldata->fattr);
3485         calldata->arg.fmode = 0;
3486         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3487         calldata->res.fattr = &calldata->fattr;
3488         calldata->res.seqid = calldata->arg.seqid;
3489         calldata->res.server = server;
3490         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3491         calldata->lr.roc = pnfs_roc(state->inode,
3492                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3493         if (calldata->lr.roc) {
3494                 calldata->arg.lr_args = &calldata->lr.arg;
3495                 calldata->res.lr_res = &calldata->lr.res;
3496         }
3497         nfs_sb_active(calldata->inode->i_sb);
3498
3499         msg.rpc_argp = &calldata->arg;
3500         msg.rpc_resp = &calldata->res;
3501         task_setup_data.callback_data = calldata;
3502         task = rpc_run_task(&task_setup_data);
3503         if (IS_ERR(task))
3504                 return PTR_ERR(task);
3505         status = 0;
3506         if (wait)
3507                 status = rpc_wait_for_completion_task(task);
3508         rpc_put_task(task);
3509         return status;
3510 out_free_calldata:
3511         kfree(calldata);
3512 out:
3513         nfs4_put_open_state(state);
3514         nfs4_put_state_owner(sp);
3515         return status;
3516 }
3517
3518 static struct inode *
3519 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3520                 int open_flags, struct iattr *attr, int *opened)
3521 {
3522         struct nfs4_state *state;
3523         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3524
3525         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3526
3527         /* Protect against concurrent sillydeletes */
3528         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3529
3530         nfs4_label_release_security(label);
3531
3532         if (IS_ERR(state))
3533                 return ERR_CAST(state);
3534         return state->inode;
3535 }
3536
3537 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3538 {
3539         if (ctx->state == NULL)
3540                 return;
3541         if (is_sync)
3542                 nfs4_close_sync(ctx->state, ctx->mode);
3543         else
3544                 nfs4_close_state(ctx->state, ctx->mode);
3545 }
3546
3547 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3548 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3549 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3550
3551 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3552 {
3553         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3554         struct nfs4_server_caps_arg args = {
3555                 .fhandle = fhandle,
3556                 .bitmask = bitmask,
3557         };
3558         struct nfs4_server_caps_res res = {};
3559         struct rpc_message msg = {
3560                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3561                 .rpc_argp = &args,
3562                 .rpc_resp = &res,
3563         };
3564         int status;
3565         int i;
3566
3567         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3568                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3569                      FATTR4_WORD0_LINK_SUPPORT |
3570                      FATTR4_WORD0_SYMLINK_SUPPORT |
3571                      FATTR4_WORD0_ACLSUPPORT;
3572         if (minorversion)
3573                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3574
3575         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3576         if (status == 0) {
3577                 /* Sanity check the server answers */
3578                 switch (minorversion) {
3579                 case 0:
3580                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3581                         res.attr_bitmask[2] = 0;
3582                         break;
3583                 case 1:
3584                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3585                         break;
3586                 case 2:
3587                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3588                 }
3589                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3590                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3591                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3592                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3593                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3594                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3595                                 NFS_CAP_SECURITY_LABEL);
3596                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3597                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3598                         server->caps |= NFS_CAP_ACLS;
3599                 if (res.has_links != 0)
3600                         server->caps |= NFS_CAP_HARDLINKS;
3601                 if (res.has_symlinks != 0)
3602                         server->caps |= NFS_CAP_SYMLINKS;
3603                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3604                         server->caps |= NFS_CAP_FILEID;
3605                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3606                         server->caps |= NFS_CAP_MODE;
3607                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3608                         server->caps |= NFS_CAP_NLINK;
3609                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3610                         server->caps |= NFS_CAP_OWNER;
3611                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3612                         server->caps |= NFS_CAP_OWNER_GROUP;
3613                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3614                         server->caps |= NFS_CAP_ATIME;
3615                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3616                         server->caps |= NFS_CAP_CTIME;
3617                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3618                         server->caps |= NFS_CAP_MTIME;
3619 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3620                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3621                         server->caps |= NFS_CAP_SECURITY_LABEL;
3622 #endif
3623                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3624                                 sizeof(server->attr_bitmask));
3625                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3626
3627                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3628                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3629                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3630                 server->cache_consistency_bitmask[2] = 0;
3631
3632                 /* Avoid a regression due to buggy server */
3633                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3634                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3635                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3636                         sizeof(server->exclcreat_bitmask));
3637
3638                 server->acl_bitmask = res.acl_bitmask;
3639                 server->fh_expire_type = res.fh_expire_type;
3640         }
3641
3642         return status;
3643 }
3644
3645 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3646 {
3647         struct nfs4_exception exception = { };
3648         int err;
3649         do {
3650                 err = nfs4_handle_exception(server,
3651                                 _nfs4_server_capabilities(server, fhandle),
3652                                 &exception);
3653         } while (exception.retry);
3654         return err;
3655 }
3656
3657 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3658                 struct nfs_fsinfo *info)
3659 {
3660         u32 bitmask[3];
3661         struct nfs4_lookup_root_arg args = {
3662                 .bitmask = bitmask,
3663         };
3664         struct nfs4_lookup_res res = {
3665                 .server = server,
3666                 .fattr = info->fattr,
3667                 .fh = fhandle,
3668         };
3669         struct rpc_message msg = {
3670                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3671                 .rpc_argp = &args,
3672                 .rpc_resp = &res,
3673         };
3674
3675         bitmask[0] = nfs4_fattr_bitmap[0];
3676         bitmask[1] = nfs4_fattr_bitmap[1];
3677         /*
3678          * Process the label in the upcoming getfattr
3679          */
3680         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3681
3682         nfs_fattr_init(info->fattr);
3683         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3684 }
3685
3686 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3687                 struct nfs_fsinfo *info)
3688 {
3689         struct nfs4_exception exception = { };
3690         int err;
3691         do {
3692                 err = _nfs4_lookup_root(server, fhandle, info);
3693                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3694                 switch (err) {
3695                 case 0:
3696                 case -NFS4ERR_WRONGSEC:
3697                         goto out;
3698                 default:
3699                         err = nfs4_handle_exception(server, err, &exception);
3700                 }
3701         } while (exception.retry);
3702 out:
3703         return err;
3704 }
3705
3706 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3707                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3708 {
3709         struct rpc_auth_create_args auth_args = {
3710                 .pseudoflavor = flavor,
3711         };
3712         struct rpc_auth *auth;
3713
3714         auth = rpcauth_create(&auth_args, server->client);
3715         if (IS_ERR(auth))
3716                 return -EACCES;
3717         return nfs4_lookup_root(server, fhandle, info);
3718 }
3719
3720 /*
3721  * Retry pseudoroot lookup with various security flavors.  We do this when:
3722  *
3723  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3724  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3725  *
3726  * Returns zero on success, or a negative NFS4ERR value, or a
3727  * negative errno value.
3728  */
3729 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3730                               struct nfs_fsinfo *info)
3731 {
3732         /* Per 3530bis 15.33.5 */
3733         static const rpc_authflavor_t flav_array[] = {
3734                 RPC_AUTH_GSS_KRB5P,
3735                 RPC_AUTH_GSS_KRB5I,
3736                 RPC_AUTH_GSS_KRB5,
3737                 RPC_AUTH_UNIX,                  /* courtesy */
3738                 RPC_AUTH_NULL,
3739         };
3740         int status = -EPERM;
3741         size_t i;
3742
3743         if (server->auth_info.flavor_len > 0) {
3744                 /* try each flavor specified by user */
3745                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3746                         status = nfs4_lookup_root_sec(server, fhandle, info,
3747                                                 server->auth_info.flavors[i]);
3748                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3749                                 continue;
3750                         break;
3751                 }
3752         } else {
3753                 /* no flavors specified by user, try default list */
3754                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3755                         status = nfs4_lookup_root_sec(server, fhandle, info,
3756                                                       flav_array[i]);
3757                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3758                                 continue;
3759                         break;
3760                 }
3761         }
3762
3763         /*
3764          * -EACCESS could mean that the user doesn't have correct permissions
3765          * to access the mount.  It could also mean that we tried to mount
3766          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3767          * existing mount programs don't handle -EACCES very well so it should
3768          * be mapped to -EPERM instead.
3769          */
3770         if (status == -EACCES)
3771                 status = -EPERM;
3772         return status;
3773 }
3774
3775 /**
3776  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3777  * @server: initialized nfs_server handle
3778  * @fhandle: we fill in the pseudo-fs root file handle
3779  * @info: we fill in an FSINFO struct
3780  * @auth_probe: probe the auth flavours
3781  *
3782  * Returns zero on success, or a negative errno.
3783  */
3784 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3785                          struct nfs_fsinfo *info,
3786                          bool auth_probe)
3787 {
3788         int status = 0;
3789
3790         if (!auth_probe)
3791                 status = nfs4_lookup_root(server, fhandle, info);
3792
3793         if (auth_probe || status == NFS4ERR_WRONGSEC)
3794                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3795                                 fhandle, info);
3796
3797         if (status == 0)
3798                 status = nfs4_server_capabilities(server, fhandle);
3799         if (status == 0)
3800                 status = nfs4_do_fsinfo(server, fhandle, info);
3801
3802         return nfs4_map_errors(status);
3803 }
3804
3805 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3806                               struct nfs_fsinfo *info)
3807 {
3808         int error;
3809         struct nfs_fattr *fattr = info->fattr;
3810         struct nfs4_label *label = NULL;
3811
3812         error = nfs4_server_capabilities(server, mntfh);
3813         if (error < 0) {
3814                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3815                 return error;
3816         }
3817
3818         label = nfs4_label_alloc(server, GFP_KERNEL);
3819         if (IS_ERR(label))
3820                 return PTR_ERR(label);
3821
3822         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
3823         if (error < 0) {
3824                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3825                 goto err_free_label;
3826         }
3827
3828         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3829             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3830                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3831
3832 err_free_label:
3833         nfs4_label_free(label);
3834
3835         return error;
3836 }
3837
3838 /*
3839  * Get locations and (maybe) other attributes of a referral.
3840  * Note that we'll actually follow the referral later when
3841  * we detect fsid mismatch in inode revalidation
3842  */
3843 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3844                              const struct qstr *name, struct nfs_fattr *fattr,
3845                              struct nfs_fh *fhandle)
3846 {
3847         int status = -ENOMEM;
3848         struct page *page = NULL;
3849         struct nfs4_fs_locations *locations = NULL;
3850
3851         page = alloc_page(GFP_KERNEL);
3852         if (page == NULL)
3853                 goto out;
3854         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3855         if (locations == NULL)
3856                 goto out;
3857
3858         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3859         if (status != 0)
3860                 goto out;
3861
3862         /*
3863          * If the fsid didn't change, this is a migration event, not a
3864          * referral.  Cause us to drop into the exception handler, which
3865          * will kick off migration recovery.
3866          */
3867         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3868                 dprintk("%s: server did not return a different fsid for"
3869                         " a referral at %s\n", __func__, name->name);
3870                 status = -NFS4ERR_MOVED;
3871                 goto out;
3872         }
3873         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3874         nfs_fixup_referral_attributes(&locations->fattr);
3875
3876         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3877         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3878         memset(fhandle, 0, sizeof(struct nfs_fh));
3879 out:
3880         if (page)
3881                 __free_page(page);
3882         kfree(locations);
3883         return status;
3884 }
3885
3886 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3887                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3888                                 struct inode *inode)
3889 {
3890         __u32 bitmask[NFS4_BITMASK_SZ];
3891         struct nfs4_getattr_arg args = {
3892                 .fh = fhandle,
3893                 .bitmask = bitmask,
3894         };
3895         struct nfs4_getattr_res res = {
3896                 .fattr = fattr,
3897                 .label = label,
3898                 .server = server,
3899         };
3900         struct rpc_message msg = {
3901                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3902                 .rpc_argp = &args,
3903                 .rpc_resp = &res,
3904         };
3905
3906         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
3907
3908         nfs_fattr_init(fattr);
3909         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3910 }
3911
3912 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3913                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3914                                 struct inode *inode)
3915 {
3916         struct nfs4_exception exception = { };
3917         int err;
3918         do {
3919                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
3920                 trace_nfs4_getattr(server, fhandle, fattr, err);
3921                 err = nfs4_handle_exception(server, err,
3922                                 &exception);
3923         } while (exception.retry);
3924         return err;
3925 }
3926
3927 /*
3928  * The file is not closed if it is opened due to the a request to change
3929  * the size of the file. The open call will not be needed once the
3930  * VFS layer lookup-intents are implemented.
3931  *
3932  * Close is called when the inode is destroyed.
3933  * If we haven't opened the file for O_WRONLY, we
3934  * need to in the size_change case to obtain a stateid.
3935  *
3936  * Got race?
3937  * Because OPEN is always done by name in nfsv4, it is
3938  * possible that we opened a different file by the same
3939  * name.  We can recognize this race condition, but we
3940  * can't do anything about it besides returning an error.
3941  *
3942  * This will be fixed with VFS changes (lookup-intent).
3943  */
3944 static int
3945 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3946                   struct iattr *sattr)
3947 {
3948         struct inode *inode = d_inode(dentry);
3949         struct rpc_cred *cred = NULL;
3950         struct nfs_open_context *ctx = NULL;
3951         struct nfs4_label *label = NULL;
3952         int status;
3953
3954         if (pnfs_ld_layoutret_on_setattr(inode) &&
3955             sattr->ia_valid & ATTR_SIZE &&
3956             sattr->ia_size < i_size_read(inode))
3957                 pnfs_commit_and_return_layout(inode);
3958
3959         nfs_fattr_init(fattr);
3960
3961         /* Deal with open(O_TRUNC) */
3962         if (sattr->ia_valid & ATTR_OPEN)
3963                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3964
3965         /* Optimization: if the end result is no change, don't RPC */
3966         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3967                 return 0;
3968
3969         /* Search for an existing open(O_WRITE) file */
3970         if (sattr->ia_valid & ATTR_FILE) {
3971
3972                 ctx = nfs_file_open_context(sattr->ia_file);
3973                 if (ctx)
3974                         cred = ctx->cred;
3975         }
3976
3977         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3978         if (IS_ERR(label))
3979                 return PTR_ERR(label);
3980
3981         /* Return any delegations if we're going to change ACLs */
3982         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
3983                 nfs4_inode_make_writeable(inode);
3984
3985         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3986         if (status == 0) {
3987                 nfs_setattr_update_inode(inode, sattr, fattr);
3988                 nfs_setsecurity(inode, fattr, label);
3989         }
3990         nfs4_label_free(label);
3991         return status;
3992 }
3993
3994 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3995                 const struct qstr *name, struct nfs_fh *fhandle,
3996                 struct nfs_fattr *fattr, struct nfs4_label *label)
3997 {
3998         struct nfs_server *server = NFS_SERVER(dir);
3999         int                    status;
4000         struct nfs4_lookup_arg args = {
4001                 .bitmask = server->attr_bitmask,
4002                 .dir_fh = NFS_FH(dir),
4003                 .name = name,
4004         };
4005         struct nfs4_lookup_res res = {
4006                 .server = server,
4007                 .fattr = fattr,
4008                 .label = label,
4009                 .fh = fhandle,
4010         };
4011         struct rpc_message msg = {
4012                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4013                 .rpc_argp = &args,
4014                 .rpc_resp = &res,
4015         };
4016
4017         args.bitmask = nfs4_bitmask(server, label);
4018
4019         nfs_fattr_init(fattr);
4020
4021         dprintk("NFS call  lookup %s\n", name->name);
4022         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
4023         dprintk("NFS reply lookup: %d\n", status);
4024         return status;
4025 }
4026
4027 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4028 {
4029         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4030                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4031         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4032         fattr->nlink = 2;
4033 }
4034
4035 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4036                                    const struct qstr *name, struct nfs_fh *fhandle,
4037                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4038 {
4039         struct nfs4_exception exception = { };
4040         struct rpc_clnt *client = *clnt;
4041         int err;
4042         do {
4043                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
4044                 trace_nfs4_lookup(dir, name, err);
4045                 switch (err) {
4046                 case -NFS4ERR_BADNAME:
4047                         err = -ENOENT;
4048                         goto out;
4049                 case -NFS4ERR_MOVED:
4050                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4051                         if (err == -NFS4ERR_MOVED)
4052                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4053                         goto out;
4054                 case -NFS4ERR_WRONGSEC:
4055                         err = -EPERM;
4056                         if (client != *clnt)
4057                                 goto out;
4058                         client = nfs4_negotiate_security(client, dir, name);
4059                         if (IS_ERR(client))
4060                                 return PTR_ERR(client);
4061
4062                         exception.retry = 1;
4063                         break;
4064                 default:
4065                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4066                 }
4067         } while (exception.retry);
4068
4069 out:
4070         if (err == 0)
4071                 *clnt = client;
4072         else if (client != *clnt)
4073                 rpc_shutdown_client(client);
4074
4075         return err;
4076 }
4077
4078 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
4079                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4080                             struct nfs4_label *label)
4081 {
4082         int status;
4083         struct rpc_clnt *client = NFS_CLIENT(dir);
4084
4085         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
4086         if (client != NFS_CLIENT(dir)) {
4087                 rpc_shutdown_client(client);
4088                 nfs_fixup_secinfo_attributes(fattr);
4089         }
4090         return status;
4091 }
4092
4093 struct rpc_clnt *
4094 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
4095                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4096 {
4097         struct rpc_clnt *client = NFS_CLIENT(dir);
4098         int status;
4099
4100         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
4101         if (status < 0)
4102                 return ERR_PTR(status);
4103         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4104 }
4105
4106 static int _nfs4_proc_lookupp(struct inode *inode,
4107                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4108                 struct nfs4_label *label)
4109 {
4110         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4111         struct nfs_server *server = NFS_SERVER(inode);
4112         int                    status;
4113         struct nfs4_lookupp_arg args = {
4114                 .bitmask = server->attr_bitmask,
4115                 .fh = NFS_FH(inode),
4116         };
4117         struct nfs4_lookupp_res res = {
4118                 .server = server,
4119                 .fattr = fattr,
4120                 .label = label,
4121                 .fh = fhandle,
4122         };
4123         struct rpc_message msg = {
4124                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4125                 .rpc_argp = &args,
4126                 .rpc_resp = &res,
4127         };
4128
4129         args.bitmask = nfs4_bitmask(server, label);
4130
4131         nfs_fattr_init(fattr);
4132
4133         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4134         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4135                                 &res.seq_res, 0);
4136         dprintk("NFS reply lookupp: %d\n", status);
4137         return status;
4138 }
4139
4140 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4141                              struct nfs_fattr *fattr, struct nfs4_label *label)
4142 {
4143         struct nfs4_exception exception = { };
4144         int err;
4145         do {
4146                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4147                 trace_nfs4_lookupp(inode, err);
4148                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4149                                 &exception);
4150         } while (exception.retry);
4151         return err;
4152 }
4153
4154 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4155 {
4156         struct nfs_server *server = NFS_SERVER(inode);
4157         struct nfs4_accessargs args = {
4158                 .fh = NFS_FH(inode),
4159                 .access = entry->mask,
4160         };
4161         struct nfs4_accessres res = {
4162                 .server = server,
4163         };
4164         struct rpc_message msg = {
4165                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4166                 .rpc_argp = &args,
4167                 .rpc_resp = &res,
4168                 .rpc_cred = entry->cred,
4169         };
4170         int status = 0;
4171
4172         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4173                 res.fattr = nfs_alloc_fattr();
4174                 if (res.fattr == NULL)
4175                         return -ENOMEM;
4176                 args.bitmask = server->cache_consistency_bitmask;
4177         }
4178
4179         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4180         if (!status) {
4181                 nfs_access_set_mask(entry, res.access);
4182                 if (res.fattr)
4183                         nfs_refresh_inode(inode, res.fattr);
4184         }
4185         nfs_free_fattr(res.fattr);
4186         return status;
4187 }
4188
4189 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4190 {
4191         struct nfs4_exception exception = { };
4192         int err;
4193         do {
4194                 err = _nfs4_proc_access(inode, entry);
4195                 trace_nfs4_access(inode, err);
4196                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4197                                 &exception);
4198         } while (exception.retry);
4199         return err;
4200 }
4201
4202 /*
4203  * TODO: For the time being, we don't try to get any attributes
4204  * along with any of the zero-copy operations READ, READDIR,
4205  * READLINK, WRITE.
4206  *
4207  * In the case of the first three, we want to put the GETATTR
4208  * after the read-type operation -- this is because it is hard
4209  * to predict the length of a GETATTR response in v4, and thus
4210  * align the READ data correctly.  This means that the GETATTR
4211  * may end up partially falling into the page cache, and we should
4212  * shift it into the 'tail' of the xdr_buf before processing.
4213  * To do this efficiently, we need to know the total length
4214  * of data received, which doesn't seem to be available outside
4215  * of the RPC layer.
4216  *
4217  * In the case of WRITE, we also want to put the GETATTR after
4218  * the operation -- in this case because we want to make sure
4219  * we get the post-operation mtime and size.
4220  *
4221  * Both of these changes to the XDR layer would in fact be quite
4222  * minor, but I decided to leave them for a subsequent patch.
4223  */
4224 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4225                 unsigned int pgbase, unsigned int pglen)
4226 {
4227         struct nfs4_readlink args = {
4228                 .fh       = NFS_FH(inode),
4229                 .pgbase   = pgbase,
4230                 .pglen    = pglen,
4231                 .pages    = &page,
4232         };
4233         struct nfs4_readlink_res res;
4234         struct rpc_message msg = {
4235                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4236                 .rpc_argp = &args,
4237                 .rpc_resp = &res,
4238         };
4239
4240         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4241 }
4242
4243 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4244                 unsigned int pgbase, unsigned int pglen)
4245 {
4246         struct nfs4_exception exception = { };
4247         int err;
4248         do {
4249                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4250                 trace_nfs4_readlink(inode, err);
4251                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4252                                 &exception);
4253         } while (exception.retry);
4254         return err;
4255 }
4256
4257 /*
4258  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4259  */
4260 static int
4261 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4262                  int flags)
4263 {
4264         struct nfs_server *server = NFS_SERVER(dir);
4265         struct nfs4_label l, *ilabel = NULL;
4266         struct nfs_open_context *ctx;
4267         struct nfs4_state *state;
4268         int status = 0;
4269
4270         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4271         if (IS_ERR(ctx))
4272                 return PTR_ERR(ctx);
4273
4274         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4275
4276         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4277                 sattr->ia_mode &= ~current_umask();
4278         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4279         if (IS_ERR(state)) {
4280                 status = PTR_ERR(state);
4281                 goto out;
4282         }
4283 out:
4284         nfs4_label_release_security(ilabel);
4285         put_nfs_open_context(ctx);
4286         return status;
4287 }
4288
4289 static int
4290 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4291 {
4292         struct nfs_server *server = NFS_SERVER(dir);
4293         struct nfs_removeargs args = {
4294                 .fh = NFS_FH(dir),
4295                 .name = *name,
4296         };
4297         struct nfs_removeres res = {
4298                 .server = server,
4299         };
4300         struct rpc_message msg = {
4301                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4302                 .rpc_argp = &args,
4303                 .rpc_resp = &res,
4304         };
4305         unsigned long timestamp = jiffies;
4306         int status;
4307
4308         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4309         if (status == 0) {
4310                 spin_lock(&dir->i_lock);
4311                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4312                 /* Removing a directory decrements nlink in the parent */
4313                 if (ftype == NF4DIR && dir->i_nlink > 2)
4314                         nfs4_dec_nlink_locked(dir);
4315                 spin_unlock(&dir->i_lock);
4316         }
4317         return status;
4318 }
4319
4320 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4321 {
4322         struct nfs4_exception exception = { };
4323         struct inode *inode = d_inode(dentry);
4324         int err;
4325
4326         if (inode) {
4327                 if (inode->i_nlink == 1)
4328                         nfs4_inode_return_delegation(inode);
4329                 else
4330                         nfs4_inode_make_writeable(inode);
4331         }
4332         do {
4333                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4334                 trace_nfs4_remove(dir, &dentry->d_name, err);
4335                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4336                                 &exception);
4337         } while (exception.retry);
4338         return err;
4339 }
4340
4341 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4342 {
4343         struct nfs4_exception exception = { };
4344         int err;
4345
4346         do {
4347                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4348                 trace_nfs4_remove(dir, name, err);
4349                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4350                                 &exception);
4351         } while (exception.retry);
4352         return err;
4353 }
4354
4355 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4356                 struct dentry *dentry,
4357                 struct inode *inode)
4358 {
4359         struct nfs_removeargs *args = msg->rpc_argp;
4360         struct nfs_removeres *res = msg->rpc_resp;
4361
4362         res->server = NFS_SB(dentry->d_sb);
4363         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4364         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4365
4366         nfs_fattr_init(res->dir_attr);
4367
4368         if (inode)
4369                 nfs4_inode_return_delegation(inode);
4370 }
4371
4372 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4373 {
4374         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4375                         &data->args.seq_args,
4376                         &data->res.seq_res,
4377                         task);
4378 }
4379
4380 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4381 {
4382         struct nfs_unlinkdata *data = task->tk_calldata;
4383         struct nfs_removeres *res = &data->res;
4384
4385         if (!nfs4_sequence_done(task, &res->seq_res))
4386                 return 0;
4387         if (nfs4_async_handle_error(task, res->server, NULL,
4388                                     &data->timeout) == -EAGAIN)
4389                 return 0;
4390         if (task->tk_status == 0)
4391                 update_changeattr(dir, &res->cinfo,
4392                                 res->dir_attr->time_start, 0);
4393         return 1;
4394 }
4395
4396 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4397                 struct dentry *old_dentry,
4398                 struct dentry *new_dentry)
4399 {
4400         struct nfs_renameargs *arg = msg->rpc_argp;
4401         struct nfs_renameres *res = msg->rpc_resp;
4402         struct inode *old_inode = d_inode(old_dentry);
4403         struct inode *new_inode = d_inode(new_dentry);
4404
4405         if (old_inode)
4406                 nfs4_inode_make_writeable(old_inode);
4407         if (new_inode)
4408                 nfs4_inode_return_delegation(new_inode);
4409         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4410         res->server = NFS_SB(old_dentry->d_sb);
4411         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4412 }
4413
4414 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4415 {
4416         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4417                         &data->args.seq_args,
4418                         &data->res.seq_res,
4419                         task);
4420 }
4421
4422 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4423                                  struct inode *new_dir)
4424 {
4425         struct nfs_renamedata *data = task->tk_calldata;
4426         struct nfs_renameres *res = &data->res;
4427
4428         if (!nfs4_sequence_done(task, &res->seq_res))
4429                 return 0;
4430         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4431                 return 0;
4432
4433         if (task->tk_status == 0) {
4434                 if (new_dir != old_dir) {
4435                         /* Note: If we moved a directory, nlink will change */
4436                         update_changeattr(old_dir, &res->old_cinfo,
4437                                         res->old_fattr->time_start,
4438                                         NFS_INO_INVALID_OTHER);
4439                         update_changeattr(new_dir, &res->new_cinfo,
4440                                         res->new_fattr->time_start,
4441                                         NFS_INO_INVALID_OTHER);
4442                 } else
4443                         update_changeattr(old_dir, &res->old_cinfo,
4444                                         res->old_fattr->time_start,
4445                                         0);
4446         }
4447         return 1;
4448 }
4449
4450 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4451 {
4452         struct nfs_server *server = NFS_SERVER(inode);
4453         __u32 bitmask[NFS4_BITMASK_SZ];
4454         struct nfs4_link_arg arg = {
4455                 .fh     = NFS_FH(inode),
4456                 .dir_fh = NFS_FH(dir),
4457                 .name   = name,
4458                 .bitmask = bitmask,
4459         };
4460         struct nfs4_link_res res = {
4461                 .server = server,
4462                 .label = NULL,
4463         };
4464         struct rpc_message msg = {
4465                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4466                 .rpc_argp = &arg,
4467                 .rpc_resp = &res,
4468         };
4469         int status = -ENOMEM;
4470
4471         res.fattr = nfs_alloc_fattr();
4472         if (res.fattr == NULL)
4473                 goto out;
4474
4475         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4476         if (IS_ERR(res.label)) {
4477                 status = PTR_ERR(res.label);
4478                 goto out;
4479         }
4480
4481         nfs4_inode_make_writeable(inode);
4482         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4483
4484         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4485         if (!status) {
4486                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4487                 status = nfs_post_op_update_inode(inode, res.fattr);
4488                 if (!status)
4489                         nfs_setsecurity(inode, res.fattr, res.label);
4490         }
4491
4492
4493         nfs4_label_free(res.label);
4494
4495 out:
4496         nfs_free_fattr(res.fattr);
4497         return status;
4498 }
4499
4500 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4501 {
4502         struct nfs4_exception exception = { };
4503         int err;
4504         do {
4505                 err = nfs4_handle_exception(NFS_SERVER(inode),
4506                                 _nfs4_proc_link(inode, dir, name),
4507                                 &exception);
4508         } while (exception.retry);
4509         return err;
4510 }
4511
4512 struct nfs4_createdata {
4513         struct rpc_message msg;
4514         struct nfs4_create_arg arg;
4515         struct nfs4_create_res res;
4516         struct nfs_fh fh;
4517         struct nfs_fattr fattr;
4518         struct nfs4_label *label;
4519 };
4520
4521 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4522                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4523 {
4524         struct nfs4_createdata *data;
4525
4526         data = kzalloc(sizeof(*data), GFP_KERNEL);
4527         if (data != NULL) {
4528                 struct nfs_server *server = NFS_SERVER(dir);
4529
4530                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4531                 if (IS_ERR(data->label))
4532                         goto out_free;
4533
4534                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4535                 data->msg.rpc_argp = &data->arg;
4536                 data->msg.rpc_resp = &data->res;
4537                 data->arg.dir_fh = NFS_FH(dir);
4538                 data->arg.server = server;
4539                 data->arg.name = name;
4540                 data->arg.attrs = sattr;
4541                 data->arg.ftype = ftype;
4542                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4543                 data->arg.umask = current_umask();
4544                 data->res.server = server;
4545                 data->res.fh = &data->fh;
4546                 data->res.fattr = &data->fattr;
4547                 data->res.label = data->label;
4548                 nfs_fattr_init(data->res.fattr);
4549         }
4550         return data;
4551 out_free:
4552         kfree(data);
4553         return NULL;
4554 }
4555
4556 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4557 {
4558         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4559                                     &data->arg.seq_args, &data->res.seq_res, 1);
4560         if (status == 0) {
4561                 spin_lock(&dir->i_lock);
4562                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4563                                 data->res.fattr->time_start, 0);
4564                 /* Creating a directory bumps nlink in the parent */
4565                 if (data->arg.ftype == NF4DIR)
4566                         nfs4_inc_nlink_locked(dir);
4567                 spin_unlock(&dir->i_lock);
4568                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4569         }
4570         return status;
4571 }
4572
4573 static void nfs4_free_createdata(struct nfs4_createdata *data)
4574 {
4575         nfs4_label_free(data->label);
4576         kfree(data);
4577 }
4578
4579 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4580                 struct page *page, unsigned int len, struct iattr *sattr,
4581                 struct nfs4_label *label)
4582 {
4583         struct nfs4_createdata *data;
4584         int status = -ENAMETOOLONG;
4585
4586         if (len > NFS4_MAXPATHLEN)
4587                 goto out;
4588
4589         status = -ENOMEM;
4590         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4591         if (data == NULL)
4592                 goto out;
4593
4594         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4595         data->arg.u.symlink.pages = &page;
4596         data->arg.u.symlink.len = len;
4597         data->arg.label = label;
4598
4599         status = nfs4_do_create(dir, dentry, data);
4600
4601         nfs4_free_createdata(data);
4602 out:
4603         return status;
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 {
4609         struct nfs4_exception exception = { };
4610         struct nfs4_label l, *label = NULL;
4611         int err;
4612
4613         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4614
4615         do {
4616                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4617                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4618                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4619                                 &exception);
4620         } while (exception.retry);
4621
4622         nfs4_label_release_security(label);
4623         return err;
4624 }
4625
4626 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4627                 struct iattr *sattr, struct nfs4_label *label)
4628 {
4629         struct nfs4_createdata *data;
4630         int status = -ENOMEM;
4631
4632         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4633         if (data == NULL)
4634                 goto out;
4635
4636         data->arg.label = label;
4637         status = nfs4_do_create(dir, dentry, data);
4638
4639         nfs4_free_createdata(data);
4640 out:
4641         return status;
4642 }
4643
4644 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4645                 struct iattr *sattr)
4646 {
4647         struct nfs_server *server = NFS_SERVER(dir);
4648         struct nfs4_exception exception = { };
4649         struct nfs4_label l, *label = NULL;
4650         int err;
4651
4652         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4653
4654         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4655                 sattr->ia_mode &= ~current_umask();
4656         do {
4657                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4658                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4659                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4660                                 &exception);
4661         } while (exception.retry);
4662         nfs4_label_release_security(label);
4663
4664         return err;
4665 }
4666
4667 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4668                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4669 {
4670         struct inode            *dir = d_inode(dentry);
4671         struct nfs4_readdir_arg args = {
4672                 .fh = NFS_FH(dir),
4673                 .pages = pages,
4674                 .pgbase = 0,
4675                 .count = count,
4676                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4677                 .plus = plus,
4678         };
4679         struct nfs4_readdir_res res;
4680         struct rpc_message msg = {
4681                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4682                 .rpc_argp = &args,
4683                 .rpc_resp = &res,
4684                 .rpc_cred = cred,
4685         };
4686         int                     status;
4687
4688         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4689                         dentry,
4690                         (unsigned long long)cookie);
4691         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4692         res.pgbase = args.pgbase;
4693         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4694         if (status >= 0) {
4695                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4696                 status += args.pgbase;
4697         }
4698
4699         nfs_invalidate_atime(dir);
4700
4701         dprintk("%s: returns %d\n", __func__, status);
4702         return status;
4703 }
4704
4705 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4706                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4707 {
4708         struct nfs4_exception exception = { };
4709         int err;
4710         do {
4711                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4712                                 pages, count, plus);
4713                 trace_nfs4_readdir(d_inode(dentry), err);
4714                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4715                                 &exception);
4716         } while (exception.retry);
4717         return err;
4718 }
4719
4720 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4721                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4722 {
4723         struct nfs4_createdata *data;
4724         int mode = sattr->ia_mode;
4725         int status = -ENOMEM;
4726
4727         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4728         if (data == NULL)
4729                 goto out;
4730
4731         if (S_ISFIFO(mode))
4732                 data->arg.ftype = NF4FIFO;
4733         else if (S_ISBLK(mode)) {
4734                 data->arg.ftype = NF4BLK;
4735                 data->arg.u.device.specdata1 = MAJOR(rdev);
4736                 data->arg.u.device.specdata2 = MINOR(rdev);
4737         }
4738         else if (S_ISCHR(mode)) {
4739                 data->arg.ftype = NF4CHR;
4740                 data->arg.u.device.specdata1 = MAJOR(rdev);
4741                 data->arg.u.device.specdata2 = MINOR(rdev);
4742         } else if (!S_ISSOCK(mode)) {
4743                 status = -EINVAL;
4744                 goto out_free;
4745         }
4746
4747         data->arg.label = label;
4748         status = nfs4_do_create(dir, dentry, data);
4749 out_free:
4750         nfs4_free_createdata(data);
4751 out:
4752         return status;
4753 }
4754
4755 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4756                 struct iattr *sattr, dev_t rdev)
4757 {
4758         struct nfs_server *server = NFS_SERVER(dir);
4759         struct nfs4_exception exception = { };
4760         struct nfs4_label l, *label = NULL;
4761         int err;
4762
4763         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4764
4765         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4766                 sattr->ia_mode &= ~current_umask();
4767         do {
4768                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4769                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4770                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4771                                 &exception);
4772         } while (exception.retry);
4773
4774         nfs4_label_release_security(label);
4775
4776         return err;
4777 }
4778
4779 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4780                  struct nfs_fsstat *fsstat)
4781 {
4782         struct nfs4_statfs_arg args = {
4783                 .fh = fhandle,
4784                 .bitmask = server->attr_bitmask,
4785         };
4786         struct nfs4_statfs_res res = {
4787                 .fsstat = fsstat,
4788         };
4789         struct rpc_message msg = {
4790                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4791                 .rpc_argp = &args,
4792                 .rpc_resp = &res,
4793         };
4794
4795         nfs_fattr_init(fsstat->fattr);
4796         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4797 }
4798
4799 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4800 {
4801         struct nfs4_exception exception = { };
4802         int err;
4803         do {
4804                 err = nfs4_handle_exception(server,
4805                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4806                                 &exception);
4807         } while (exception.retry);
4808         return err;
4809 }
4810
4811 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4812                 struct nfs_fsinfo *fsinfo)
4813 {
4814         struct nfs4_fsinfo_arg args = {
4815                 .fh = fhandle,
4816                 .bitmask = server->attr_bitmask,
4817         };
4818         struct nfs4_fsinfo_res res = {
4819                 .fsinfo = fsinfo,
4820         };
4821         struct rpc_message msg = {
4822                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4823                 .rpc_argp = &args,
4824                 .rpc_resp = &res,
4825         };
4826
4827         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4828 }
4829
4830 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4831 {
4832         struct nfs4_exception exception = { };
4833         unsigned long now = jiffies;
4834         int err;
4835
4836         do {
4837                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4838                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4839                 if (err == 0) {
4840                         nfs4_set_lease_period(server->nfs_client,
4841                                         fsinfo->lease_time * HZ,
4842                                         now);
4843                         break;
4844                 }
4845                 err = nfs4_handle_exception(server, err, &exception);
4846         } while (exception.retry);
4847         return err;
4848 }
4849
4850 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4851 {
4852         int error;
4853
4854         nfs_fattr_init(fsinfo->fattr);
4855         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4856         if (error == 0) {
4857                 /* block layout checks this! */
4858                 server->pnfs_blksize = fsinfo->blksize;
4859                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4860         }
4861
4862         return error;
4863 }
4864
4865 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4866                 struct nfs_pathconf *pathconf)
4867 {
4868         struct nfs4_pathconf_arg args = {
4869                 .fh = fhandle,
4870                 .bitmask = server->attr_bitmask,
4871         };
4872         struct nfs4_pathconf_res res = {
4873                 .pathconf = pathconf,
4874         };
4875         struct rpc_message msg = {
4876                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4877                 .rpc_argp = &args,
4878                 .rpc_resp = &res,
4879         };
4880
4881         /* None of the pathconf attributes are mandatory to implement */
4882         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4883                 memset(pathconf, 0, sizeof(*pathconf));
4884                 return 0;
4885         }
4886
4887         nfs_fattr_init(pathconf->fattr);
4888         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4889 }
4890
4891 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4892                 struct nfs_pathconf *pathconf)
4893 {
4894         struct nfs4_exception exception = { };
4895         int err;
4896
4897         do {
4898                 err = nfs4_handle_exception(server,
4899                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4900                                 &exception);
4901         } while (exception.retry);
4902         return err;
4903 }
4904
4905 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4906                 const struct nfs_open_context *ctx,
4907                 const struct nfs_lock_context *l_ctx,
4908                 fmode_t fmode)
4909 {
4910         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4911 }
4912 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4913
4914 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4915                 const struct nfs_open_context *ctx,
4916                 const struct nfs_lock_context *l_ctx,
4917                 fmode_t fmode)
4918 {
4919         nfs4_stateid current_stateid;
4920
4921         /* If the current stateid represents a lost lock, then exit */
4922         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4923                 return true;
4924         return nfs4_stateid_match(stateid, &current_stateid);
4925 }
4926
4927 static bool nfs4_error_stateid_expired(int err)
4928 {
4929         switch (err) {
4930         case -NFS4ERR_DELEG_REVOKED:
4931         case -NFS4ERR_ADMIN_REVOKED:
4932         case -NFS4ERR_BAD_STATEID:
4933         case -NFS4ERR_STALE_STATEID:
4934         case -NFS4ERR_OLD_STATEID:
4935         case -NFS4ERR_OPENMODE:
4936         case -NFS4ERR_EXPIRED:
4937                 return true;
4938         }
4939         return false;
4940 }
4941
4942 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4943 {
4944         struct nfs_server *server = NFS_SERVER(hdr->inode);
4945
4946         trace_nfs4_read(hdr, task->tk_status);
4947         if (task->tk_status < 0) {
4948                 struct nfs4_exception exception = {
4949                         .inode = hdr->inode,
4950                         .state = hdr->args.context->state,
4951                         .stateid = &hdr->args.stateid,
4952                 };
4953                 task->tk_status = nfs4_async_handle_exception(task,
4954                                 server, task->tk_status, &exception);
4955                 if (exception.retry) {
4956                         rpc_restart_call_prepare(task);
4957                         return -EAGAIN;
4958                 }
4959         }
4960
4961         if (task->tk_status > 0)
4962                 renew_lease(server, hdr->timestamp);
4963         return 0;
4964 }
4965
4966 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4967                 struct nfs_pgio_args *args)
4968 {
4969
4970         if (!nfs4_error_stateid_expired(task->tk_status) ||
4971                 nfs4_stateid_is_current(&args->stateid,
4972                                 args->context,
4973                                 args->lock_context,
4974                                 FMODE_READ))
4975                 return false;
4976         rpc_restart_call_prepare(task);
4977         return true;
4978 }
4979
4980 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4981 {
4982
4983         dprintk("--> %s\n", __func__);
4984
4985         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4986                 return -EAGAIN;
4987         if (nfs4_read_stateid_changed(task, &hdr->args))
4988                 return -EAGAIN;
4989         if (task->tk_status > 0)
4990                 nfs_invalidate_atime(hdr->inode);
4991         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4992                                     nfs4_read_done_cb(task, hdr);
4993 }
4994
4995 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4996                                  struct rpc_message *msg)
4997 {
4998         hdr->timestamp   = jiffies;
4999         if (!hdr->pgio_done_cb)
5000                 hdr->pgio_done_cb = nfs4_read_done_cb;
5001         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5002         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5003 }
5004
5005 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5006                                       struct nfs_pgio_header *hdr)
5007 {
5008         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5009                         &hdr->args.seq_args,
5010                         &hdr->res.seq_res,
5011                         task))
5012                 return 0;
5013         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5014                                 hdr->args.lock_context,
5015                                 hdr->rw_mode) == -EIO)
5016                 return -EIO;
5017         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5018                 return -EIO;
5019         return 0;
5020 }
5021
5022 static int nfs4_write_done_cb(struct rpc_task *task,
5023                               struct nfs_pgio_header *hdr)
5024 {
5025         struct inode *inode = hdr->inode;
5026
5027         trace_nfs4_write(hdr, task->tk_status);
5028         if (task->tk_status < 0) {
5029                 struct nfs4_exception exception = {
5030                         .inode = hdr->inode,
5031                         .state = hdr->args.context->state,
5032                         .stateid = &hdr->args.stateid,
5033                 };
5034                 task->tk_status = nfs4_async_handle_exception(task,
5035                                 NFS_SERVER(inode), task->tk_status,
5036                                 &exception);
5037                 if (exception.retry) {
5038                         rpc_restart_call_prepare(task);
5039                         return -EAGAIN;
5040                 }
5041         }
5042         if (task->tk_status >= 0) {
5043                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5044                 nfs_writeback_update_inode(hdr);
5045         }
5046         return 0;
5047 }
5048
5049 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5050                 struct nfs_pgio_args *args)
5051 {
5052
5053         if (!nfs4_error_stateid_expired(task->tk_status) ||
5054                 nfs4_stateid_is_current(&args->stateid,
5055                                 args->context,
5056                                 args->lock_context,
5057                                 FMODE_WRITE))
5058                 return false;
5059         rpc_restart_call_prepare(task);
5060         return true;
5061 }
5062
5063 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5064 {
5065         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5066                 return -EAGAIN;
5067         if (nfs4_write_stateid_changed(task, &hdr->args))
5068                 return -EAGAIN;
5069         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5070                 nfs4_write_done_cb(task, hdr);
5071 }
5072
5073 static
5074 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5075 {
5076         /* Don't request attributes for pNFS or O_DIRECT writes */
5077         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5078                 return false;
5079         /* Otherwise, request attributes if and only if we don't hold
5080          * a delegation
5081          */
5082         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5083 }
5084
5085 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5086                                   struct rpc_message *msg,
5087                                   struct rpc_clnt **clnt)
5088 {
5089         struct nfs_server *server = NFS_SERVER(hdr->inode);
5090
5091         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5092                 hdr->args.bitmask = NULL;
5093                 hdr->res.fattr = NULL;
5094         } else
5095                 hdr->args.bitmask = server->cache_consistency_bitmask;
5096
5097         if (!hdr->pgio_done_cb)
5098                 hdr->pgio_done_cb = nfs4_write_done_cb;
5099         hdr->res.server = server;
5100         hdr->timestamp   = jiffies;
5101
5102         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5103         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1, 0);
5104         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5105 }
5106
5107 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5108 {
5109         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5110                         &data->args.seq_args,
5111                         &data->res.seq_res,
5112                         task);
5113 }
5114
5115 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5116 {
5117         struct inode *inode = data->inode;
5118
5119         trace_nfs4_commit(data, task->tk_status);
5120         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5121                                     NULL, NULL) == -EAGAIN) {
5122                 rpc_restart_call_prepare(task);
5123                 return -EAGAIN;
5124         }
5125         return 0;
5126 }
5127
5128 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5129 {
5130         if (!nfs4_sequence_done(task, &data->res.seq_res))
5131                 return -EAGAIN;
5132         return data->commit_done_cb(task, data);
5133 }
5134
5135 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5136                                    struct rpc_clnt **clnt)
5137 {
5138         struct nfs_server *server = NFS_SERVER(data->inode);
5139
5140         if (data->commit_done_cb == NULL)
5141                 data->commit_done_cb = nfs4_commit_done_cb;
5142         data->res.server = server;
5143         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5144         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5145         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5146 }
5147
5148 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5149                                 struct nfs_commitres *res)
5150 {
5151         struct inode *dst_inode = file_inode(dst);
5152         struct nfs_server *server = NFS_SERVER(dst_inode);
5153         struct rpc_message msg = {
5154                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5155                 .rpc_argp = args,
5156                 .rpc_resp = res,
5157         };
5158
5159         args->fh = NFS_FH(dst_inode);
5160         return nfs4_call_sync(server->client, server, &msg,
5161                         &args->seq_args, &res->seq_res, 1);
5162 }
5163
5164 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5165 {
5166         struct nfs_commitargs args = {
5167                 .offset = offset,
5168                 .count = count,
5169         };
5170         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5171         struct nfs4_exception exception = { };
5172         int status;
5173
5174         do {
5175                 status = _nfs4_proc_commit(dst, &args, res);
5176                 status = nfs4_handle_exception(dst_server, status, &exception);
5177         } while (exception.retry);
5178
5179         return status;
5180 }
5181
5182 struct nfs4_renewdata {
5183         struct nfs_client       *client;
5184         unsigned long           timestamp;
5185 };
5186
5187 /*
5188  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5189  * standalone procedure for queueing an asynchronous RENEW.
5190  */
5191 static void nfs4_renew_release(void *calldata)
5192 {
5193         struct nfs4_renewdata *data = calldata;
5194         struct nfs_client *clp = data->client;
5195
5196         if (refcount_read(&clp->cl_count) > 1)
5197                 nfs4_schedule_state_renewal(clp);
5198         nfs_put_client(clp);
5199         kfree(data);
5200 }
5201
5202 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5203 {
5204         struct nfs4_renewdata *data = calldata;
5205         struct nfs_client *clp = data->client;
5206         unsigned long timestamp = data->timestamp;
5207
5208         trace_nfs4_renew_async(clp, task->tk_status);
5209         switch (task->tk_status) {
5210         case 0:
5211                 break;
5212         case -NFS4ERR_LEASE_MOVED:
5213                 nfs4_schedule_lease_moved_recovery(clp);
5214                 break;
5215         default:
5216                 /* Unless we're shutting down, schedule state recovery! */
5217                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5218                         return;
5219                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5220                         nfs4_schedule_lease_recovery(clp);
5221                         return;
5222                 }
5223                 nfs4_schedule_path_down_recovery(clp);
5224         }
5225         do_renew_lease(clp, timestamp);
5226 }
5227
5228 static const struct rpc_call_ops nfs4_renew_ops = {
5229         .rpc_call_done = nfs4_renew_done,
5230         .rpc_release = nfs4_renew_release,
5231 };
5232
5233 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5234 {
5235         struct rpc_message msg = {
5236                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5237                 .rpc_argp       = clp,
5238                 .rpc_cred       = cred,
5239         };
5240         struct nfs4_renewdata *data;
5241
5242         if (renew_flags == 0)
5243                 return 0;
5244         if (!refcount_inc_not_zero(&clp->cl_count))
5245                 return -EIO;
5246         data = kmalloc(sizeof(*data), GFP_NOFS);
5247         if (data == NULL) {
5248                 nfs_put_client(clp);
5249                 return -ENOMEM;
5250         }
5251         data->client = clp;
5252         data->timestamp = jiffies;
5253         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5254                         &nfs4_renew_ops, data);
5255 }
5256
5257 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
5258 {
5259         struct rpc_message msg = {
5260                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5261                 .rpc_argp       = clp,
5262                 .rpc_cred       = cred,
5263         };
5264         unsigned long now = jiffies;
5265         int status;
5266
5267         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5268         if (status < 0)
5269                 return status;
5270         do_renew_lease(clp, now);
5271         return 0;
5272 }
5273
5274 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5275 {
5276         return server->caps & NFS_CAP_ACLS;
5277 }
5278
5279 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5280  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5281  * the stack.
5282  */
5283 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5284
5285 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5286                 struct page **pages)
5287 {
5288         struct page *newpage, **spages;
5289         int rc = 0;
5290         size_t len;
5291         spages = pages;
5292
5293         do {
5294                 len = min_t(size_t, PAGE_SIZE, buflen);
5295                 newpage = alloc_page(GFP_KERNEL);
5296
5297                 if (newpage == NULL)
5298                         goto unwind;
5299                 memcpy(page_address(newpage), buf, len);
5300                 buf += len;
5301                 buflen -= len;
5302                 *pages++ = newpage;
5303                 rc++;
5304         } while (buflen != 0);
5305
5306         return rc;
5307
5308 unwind:
5309         for(; rc > 0; rc--)
5310                 __free_page(spages[rc-1]);
5311         return -ENOMEM;
5312 }
5313
5314 struct nfs4_cached_acl {
5315         int cached;
5316         size_t len;
5317         char data[0];
5318 };
5319
5320 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5321 {
5322         struct nfs_inode *nfsi = NFS_I(inode);
5323
5324         spin_lock(&inode->i_lock);
5325         kfree(nfsi->nfs4_acl);
5326         nfsi->nfs4_acl = acl;
5327         spin_unlock(&inode->i_lock);
5328 }
5329
5330 static void nfs4_zap_acl_attr(struct inode *inode)
5331 {
5332         nfs4_set_cached_acl(inode, NULL);
5333 }
5334
5335 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5336 {
5337         struct nfs_inode *nfsi = NFS_I(inode);
5338         struct nfs4_cached_acl *acl;
5339         int ret = -ENOENT;
5340
5341         spin_lock(&inode->i_lock);
5342         acl = nfsi->nfs4_acl;
5343         if (acl == NULL)
5344                 goto out;
5345         if (buf == NULL) /* user is just asking for length */
5346                 goto out_len;
5347         if (acl->cached == 0)
5348                 goto out;
5349         ret = -ERANGE; /* see getxattr(2) man page */
5350         if (acl->len > buflen)
5351                 goto out;
5352         memcpy(buf, acl->data, acl->len);
5353 out_len:
5354         ret = acl->len;
5355 out:
5356         spin_unlock(&inode->i_lock);
5357         return ret;
5358 }
5359
5360 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5361 {
5362         struct nfs4_cached_acl *acl;
5363         size_t buflen = sizeof(*acl) + acl_len;
5364
5365         if (buflen <= PAGE_SIZE) {
5366                 acl = kmalloc(buflen, GFP_KERNEL);
5367                 if (acl == NULL)
5368                         goto out;
5369                 acl->cached = 1;
5370                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5371         } else {
5372                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5373                 if (acl == NULL)
5374                         goto out;
5375                 acl->cached = 0;
5376         }
5377         acl->len = acl_len;
5378 out:
5379         nfs4_set_cached_acl(inode, acl);
5380 }
5381
5382 /*
5383  * The getxattr API returns the required buffer length when called with a
5384  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5385  * the required buf.  On a NULL buf, we send a page of data to the server
5386  * guessing that the ACL request can be serviced by a page. If so, we cache
5387  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5388  * the cache. If not so, we throw away the page, and cache the required
5389  * length. The next getxattr call will then produce another round trip to
5390  * the server, this time with the input buf of the required size.
5391  */
5392 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5393 {
5394         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5395         struct nfs_getaclargs args = {
5396                 .fh = NFS_FH(inode),
5397                 .acl_pages = pages,
5398                 .acl_len = buflen,
5399         };
5400         struct nfs_getaclres res = {
5401                 .acl_len = buflen,
5402         };
5403         struct rpc_message msg = {
5404                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5405                 .rpc_argp = &args,
5406                 .rpc_resp = &res,
5407         };
5408         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5409         int ret = -ENOMEM, i;
5410
5411         if (npages > ARRAY_SIZE(pages))
5412                 return -ERANGE;
5413
5414         for (i = 0; i < npages; i++) {
5415                 pages[i] = alloc_page(GFP_KERNEL);
5416                 if (!pages[i])
5417                         goto out_free;
5418         }
5419
5420         /* for decoding across pages */
5421         res.acl_scratch = alloc_page(GFP_KERNEL);
5422         if (!res.acl_scratch)
5423                 goto out_free;
5424
5425         args.acl_len = npages * PAGE_SIZE;
5426
5427         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5428                 __func__, buf, buflen, npages, args.acl_len);
5429         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5430                              &msg, &args.seq_args, &res.seq_res, 0);
5431         if (ret)
5432                 goto out_free;
5433
5434         /* Handle the case where the passed-in buffer is too short */
5435         if (res.acl_flags & NFS4_ACL_TRUNC) {
5436                 /* Did the user only issue a request for the acl length? */
5437                 if (buf == NULL)
5438                         goto out_ok;
5439                 ret = -ERANGE;
5440                 goto out_free;
5441         }
5442         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5443         if (buf) {
5444                 if (res.acl_len > buflen) {
5445                         ret = -ERANGE;
5446                         goto out_free;
5447                 }
5448                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5449         }
5450 out_ok:
5451         ret = res.acl_len;
5452 out_free:
5453         for (i = 0; i < npages; i++)
5454                 if (pages[i])
5455                         __free_page(pages[i]);
5456         if (res.acl_scratch)
5457                 __free_page(res.acl_scratch);
5458         return ret;
5459 }
5460
5461 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5462 {
5463         struct nfs4_exception exception = { };
5464         ssize_t ret;
5465         do {
5466                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5467                 trace_nfs4_get_acl(inode, ret);
5468                 if (ret >= 0)
5469                         break;
5470                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5471         } while (exception.retry);
5472         return ret;
5473 }
5474
5475 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5476 {
5477         struct nfs_server *server = NFS_SERVER(inode);
5478         int ret;
5479
5480         if (!nfs4_server_supports_acls(server))
5481                 return -EOPNOTSUPP;
5482         ret = nfs_revalidate_inode(server, inode);
5483         if (ret < 0)
5484                 return ret;
5485         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5486                 nfs_zap_acl_cache(inode);
5487         ret = nfs4_read_cached_acl(inode, buf, buflen);
5488         if (ret != -ENOENT)
5489                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5490                  * but no cached acl data, just the acl length */
5491                 return ret;
5492         return nfs4_get_acl_uncached(inode, buf, buflen);
5493 }
5494
5495 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5496 {
5497         struct nfs_server *server = NFS_SERVER(inode);
5498         struct page *pages[NFS4ACL_MAXPAGES];
5499         struct nfs_setaclargs arg = {
5500                 .fh             = NFS_FH(inode),
5501                 .acl_pages      = pages,
5502                 .acl_len        = buflen,
5503         };
5504         struct nfs_setaclres res;
5505         struct rpc_message msg = {
5506                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5507                 .rpc_argp       = &arg,
5508                 .rpc_resp       = &res,
5509         };
5510         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5511         int ret, i;
5512
5513         if (!nfs4_server_supports_acls(server))
5514                 return -EOPNOTSUPP;
5515         if (npages > ARRAY_SIZE(pages))
5516                 return -ERANGE;
5517         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5518         if (i < 0)
5519                 return i;
5520         nfs4_inode_make_writeable(inode);
5521         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5522
5523         /*
5524          * Free each page after tx, so the only ref left is
5525          * held by the network stack
5526          */
5527         for (; i > 0; i--)
5528                 put_page(pages[i-1]);
5529
5530         /*
5531          * Acl update can result in inode attribute update.
5532          * so mark the attribute cache invalid.
5533          */
5534         spin_lock(&inode->i_lock);
5535         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5536                 | NFS_INO_INVALID_CTIME
5537                 | NFS_INO_REVAL_FORCED;
5538         spin_unlock(&inode->i_lock);
5539         nfs_access_zap_cache(inode);
5540         nfs_zap_acl_cache(inode);
5541         return ret;
5542 }
5543
5544 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5545 {
5546         struct nfs4_exception exception = { };
5547         int err;
5548         do {
5549                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5550                 trace_nfs4_set_acl(inode, err);
5551                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5552                                 &exception);
5553         } while (exception.retry);
5554         return err;
5555 }
5556
5557 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5558 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5559                                         size_t buflen)
5560 {
5561         struct nfs_server *server = NFS_SERVER(inode);
5562         struct nfs_fattr fattr;
5563         struct nfs4_label label = {0, 0, buflen, buf};
5564
5565         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5566         struct nfs4_getattr_arg arg = {
5567                 .fh             = NFS_FH(inode),
5568                 .bitmask        = bitmask,
5569         };
5570         struct nfs4_getattr_res res = {
5571                 .fattr          = &fattr,
5572                 .label          = &label,
5573                 .server         = server,
5574         };
5575         struct rpc_message msg = {
5576                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5577                 .rpc_argp       = &arg,
5578                 .rpc_resp       = &res,
5579         };
5580         int ret;
5581
5582         nfs_fattr_init(&fattr);
5583
5584         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5585         if (ret)
5586                 return ret;
5587         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5588                 return -ENOENT;
5589         if (buflen < label.len)
5590                 return -ERANGE;
5591         return 0;
5592 }
5593
5594 static int nfs4_get_security_label(struct inode *inode, void *buf,
5595                                         size_t buflen)
5596 {
5597         struct nfs4_exception exception = { };
5598         int err;
5599
5600         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5601                 return -EOPNOTSUPP;
5602
5603         do {
5604                 err = _nfs4_get_security_label(inode, buf, buflen);
5605                 trace_nfs4_get_security_label(inode, err);
5606                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5607                                 &exception);
5608         } while (exception.retry);
5609         return err;
5610 }
5611
5612 static int _nfs4_do_set_security_label(struct inode *inode,
5613                 struct nfs4_label *ilabel,
5614                 struct nfs_fattr *fattr,
5615                 struct nfs4_label *olabel)
5616 {
5617
5618         struct iattr sattr = {0};
5619         struct nfs_server *server = NFS_SERVER(inode);
5620         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5621         struct nfs_setattrargs arg = {
5622                 .fh             = NFS_FH(inode),
5623                 .iap            = &sattr,
5624                 .server         = server,
5625                 .bitmask        = bitmask,
5626                 .label          = ilabel,
5627         };
5628         struct nfs_setattrres res = {
5629                 .fattr          = fattr,
5630                 .label          = olabel,
5631                 .server         = server,
5632         };
5633         struct rpc_message msg = {
5634                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5635                 .rpc_argp       = &arg,
5636                 .rpc_resp       = &res,
5637         };
5638         int status;
5639
5640         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5641
5642         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5643         if (status)
5644                 dprintk("%s failed: %d\n", __func__, status);
5645
5646         return status;
5647 }
5648
5649 static int nfs4_do_set_security_label(struct inode *inode,
5650                 struct nfs4_label *ilabel,
5651                 struct nfs_fattr *fattr,
5652                 struct nfs4_label *olabel)
5653 {
5654         struct nfs4_exception exception = { };
5655         int err;
5656
5657         do {
5658                 err = _nfs4_do_set_security_label(inode, ilabel,
5659                                 fattr, olabel);
5660                 trace_nfs4_set_security_label(inode, err);
5661                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5662                                 &exception);
5663         } while (exception.retry);
5664         return err;
5665 }
5666
5667 static int
5668 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5669 {
5670         struct nfs4_label ilabel, *olabel = NULL;
5671         struct nfs_fattr fattr;
5672         struct rpc_cred *cred;
5673         int status;
5674
5675         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5676                 return -EOPNOTSUPP;
5677
5678         nfs_fattr_init(&fattr);
5679
5680         ilabel.pi = 0;
5681         ilabel.lfs = 0;
5682         ilabel.label = (char *)buf;
5683         ilabel.len = buflen;
5684
5685         cred = rpc_lookup_cred();
5686         if (IS_ERR(cred))
5687                 return PTR_ERR(cred);
5688
5689         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5690         if (IS_ERR(olabel)) {
5691                 status = -PTR_ERR(olabel);
5692                 goto out;
5693         }
5694
5695         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5696         if (status == 0)
5697                 nfs_setsecurity(inode, &fattr, olabel);
5698
5699         nfs4_label_free(olabel);
5700 out:
5701         put_rpccred(cred);
5702         return status;
5703 }
5704 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5705
5706
5707 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5708                                     nfs4_verifier *bootverf)
5709 {
5710         __be32 verf[2];
5711
5712         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5713                 /* An impossible timestamp guarantees this value
5714                  * will never match a generated boot time. */
5715                 verf[0] = cpu_to_be32(U32_MAX);
5716                 verf[1] = cpu_to_be32(U32_MAX);
5717         } else {
5718                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5719                 u64 ns = ktime_to_ns(nn->boot_time);
5720
5721                 verf[0] = cpu_to_be32(ns >> 32);
5722                 verf[1] = cpu_to_be32(ns);
5723         }
5724         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5725 }
5726
5727 static int
5728 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5729 {
5730         size_t len;
5731         char *str;
5732
5733         if (clp->cl_owner_id != NULL)
5734                 return 0;
5735
5736         rcu_read_lock();
5737         len = 14 +
5738                 strlen(clp->cl_rpcclient->cl_nodename) +
5739                 1 +
5740                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5741                 1;
5742         rcu_read_unlock();
5743         if (nfs4_client_id_uniquifier[0] != '\0')
5744                 len += strlen(nfs4_client_id_uniquifier) + 1;
5745         if (len > NFS4_OPAQUE_LIMIT + 1)
5746                 return -EINVAL;
5747
5748         /*
5749          * Since this string is allocated at mount time, and held until the
5750          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5751          * about a memory-reclaim deadlock.
5752          */
5753         str = kmalloc(len, GFP_KERNEL);
5754         if (!str)
5755                 return -ENOMEM;
5756
5757         rcu_read_lock();
5758         if (nfs4_client_id_uniquifier[0] != '\0')
5759                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5760                           clp->cl_rpcclient->cl_nodename,
5761                           nfs4_client_id_uniquifier,
5762                           rpc_peeraddr2str(clp->cl_rpcclient,
5763                                            RPC_DISPLAY_ADDR));
5764         else
5765                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
5766                           clp->cl_rpcclient->cl_nodename,
5767                           rpc_peeraddr2str(clp->cl_rpcclient,
5768                                            RPC_DISPLAY_ADDR));
5769         rcu_read_unlock();
5770
5771         clp->cl_owner_id = str;
5772         return 0;
5773 }
5774
5775 static int
5776 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5777 {
5778         size_t len;
5779         char *str;
5780
5781         len = 10 + 10 + 1 + 10 + 1 +
5782                 strlen(nfs4_client_id_uniquifier) + 1 +
5783                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5784
5785         if (len > NFS4_OPAQUE_LIMIT + 1)
5786                 return -EINVAL;
5787
5788         /*
5789          * Since this string is allocated at mount time, and held until the
5790          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5791          * about a memory-reclaim deadlock.
5792          */
5793         str = kmalloc(len, GFP_KERNEL);
5794         if (!str)
5795                 return -ENOMEM;
5796
5797         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5798                         clp->rpc_ops->version, clp->cl_minorversion,
5799                         nfs4_client_id_uniquifier,
5800                         clp->cl_rpcclient->cl_nodename);
5801         clp->cl_owner_id = str;
5802         return 0;
5803 }
5804
5805 static int
5806 nfs4_init_uniform_client_string(struct nfs_client *clp)
5807 {
5808         size_t len;
5809         char *str;
5810
5811         if (clp->cl_owner_id != NULL)
5812                 return 0;
5813
5814         if (nfs4_client_id_uniquifier[0] != '\0')
5815                 return nfs4_init_uniquifier_client_string(clp);
5816
5817         len = 10 + 10 + 1 + 10 + 1 +
5818                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5819
5820         if (len > NFS4_OPAQUE_LIMIT + 1)
5821                 return -EINVAL;
5822
5823         /*
5824          * Since this string is allocated at mount time, and held until the
5825          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5826          * about a memory-reclaim deadlock.
5827          */
5828         str = kmalloc(len, GFP_KERNEL);
5829         if (!str)
5830                 return -ENOMEM;
5831
5832         scnprintf(str, len, "Linux NFSv%u.%u %s",
5833                         clp->rpc_ops->version, clp->cl_minorversion,
5834                         clp->cl_rpcclient->cl_nodename);
5835         clp->cl_owner_id = str;
5836         return 0;
5837 }
5838
5839 /*
5840  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5841  * services.  Advertise one based on the address family of the
5842  * clientaddr.
5843  */
5844 static unsigned int
5845 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5846 {
5847         if (strchr(clp->cl_ipaddr, ':') != NULL)
5848                 return scnprintf(buf, len, "tcp6");
5849         else
5850                 return scnprintf(buf, len, "tcp");
5851 }
5852
5853 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5854 {
5855         struct nfs4_setclientid *sc = calldata;
5856
5857         if (task->tk_status == 0)
5858                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5859 }
5860
5861 static const struct rpc_call_ops nfs4_setclientid_ops = {
5862         .rpc_call_done = nfs4_setclientid_done,
5863 };
5864
5865 /**
5866  * nfs4_proc_setclientid - Negotiate client ID
5867  * @clp: state data structure
5868  * @program: RPC program for NFSv4 callback service
5869  * @port: IP port number for NFS4 callback service
5870  * @cred: RPC credential to use for this call
5871  * @res: where to place the result
5872  *
5873  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5874  */
5875 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5876                 unsigned short port, struct rpc_cred *cred,
5877                 struct nfs4_setclientid_res *res)
5878 {
5879         nfs4_verifier sc_verifier;
5880         struct nfs4_setclientid setclientid = {
5881                 .sc_verifier = &sc_verifier,
5882                 .sc_prog = program,
5883                 .sc_clnt = clp,
5884         };
5885         struct rpc_message msg = {
5886                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5887                 .rpc_argp = &setclientid,
5888                 .rpc_resp = res,
5889                 .rpc_cred = cred,
5890         };
5891         struct rpc_task *task;
5892         struct rpc_task_setup task_setup_data = {
5893                 .rpc_client = clp->cl_rpcclient,
5894                 .rpc_message = &msg,
5895                 .callback_ops = &nfs4_setclientid_ops,
5896                 .callback_data = &setclientid,
5897                 .flags = RPC_TASK_TIMEOUT,
5898         };
5899         int status;
5900
5901         /* nfs_client_id4 */
5902         nfs4_init_boot_verifier(clp, &sc_verifier);
5903
5904         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5905                 status = nfs4_init_uniform_client_string(clp);
5906         else
5907                 status = nfs4_init_nonuniform_client_string(clp);
5908
5909         if (status)
5910                 goto out;
5911
5912         /* cb_client4 */
5913         setclientid.sc_netid_len =
5914                                 nfs4_init_callback_netid(clp,
5915                                                 setclientid.sc_netid,
5916                                                 sizeof(setclientid.sc_netid));
5917         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5918                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5919                                 clp->cl_ipaddr, port >> 8, port & 255);
5920
5921         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5922                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5923                 clp->cl_owner_id);
5924         task = rpc_run_task(&task_setup_data);
5925         if (IS_ERR(task)) {
5926                 status = PTR_ERR(task);
5927                 goto out;
5928         }
5929         status = task->tk_status;
5930         if (setclientid.sc_cred) {
5931                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5932                 put_rpccred(setclientid.sc_cred);
5933         }
5934         rpc_put_task(task);
5935 out:
5936         trace_nfs4_setclientid(clp, status);
5937         dprintk("NFS reply setclientid: %d\n", status);
5938         return status;
5939 }
5940
5941 /**
5942  * nfs4_proc_setclientid_confirm - Confirm client ID
5943  * @clp: state data structure
5944  * @res: result of a previous SETCLIENTID
5945  * @cred: RPC credential to use for this call
5946  *
5947  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5948  */
5949 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5950                 struct nfs4_setclientid_res *arg,
5951                 struct rpc_cred *cred)
5952 {
5953         struct rpc_message msg = {
5954                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5955                 .rpc_argp = arg,
5956                 .rpc_cred = cred,
5957         };
5958         int status;
5959
5960         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5961                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5962                 clp->cl_clientid);
5963         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5964         trace_nfs4_setclientid_confirm(clp, status);
5965         dprintk("NFS reply setclientid_confirm: %d\n", status);
5966         return status;
5967 }
5968
5969 struct nfs4_delegreturndata {
5970         struct nfs4_delegreturnargs args;
5971         struct nfs4_delegreturnres res;
5972         struct nfs_fh fh;
5973         nfs4_stateid stateid;
5974         unsigned long timestamp;
5975         struct {
5976                 struct nfs4_layoutreturn_args arg;
5977                 struct nfs4_layoutreturn_res res;
5978                 struct nfs4_xdr_opaque_data ld_private;
5979                 u32 roc_barrier;
5980                 bool roc;
5981         } lr;
5982         struct nfs_fattr fattr;
5983         int rpc_status;
5984         struct inode *inode;
5985 };
5986
5987 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5988 {
5989         struct nfs4_delegreturndata *data = calldata;
5990         struct nfs4_exception exception = {
5991                 .inode = data->inode,
5992                 .stateid = &data->stateid,
5993         };
5994
5995         if (!nfs4_sequence_done(task, &data->res.seq_res))
5996                 return;
5997
5998         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5999
6000         /* Handle Layoutreturn errors */
6001         if (data->args.lr_args && task->tk_status != 0) {
6002                 switch(data->res.lr_ret) {
6003                 default:
6004                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6005                         break;
6006                 case 0:
6007                         data->args.lr_args = NULL;
6008                         data->res.lr_res = NULL;
6009                         break;
6010                 case -NFS4ERR_OLD_STATEID:
6011                         if (nfs4_layoutreturn_refresh_stateid(&data->args.lr_args->stateid,
6012                                                 &data->args.lr_args->range,
6013                                                 data->inode))
6014                                 goto lr_restart;
6015                         /* Fallthrough */
6016                 case -NFS4ERR_ADMIN_REVOKED:
6017                 case -NFS4ERR_DELEG_REVOKED:
6018                 case -NFS4ERR_EXPIRED:
6019                 case -NFS4ERR_BAD_STATEID:
6020                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
6021                 case -NFS4ERR_WRONG_CRED:
6022                         data->args.lr_args = NULL;
6023                         data->res.lr_res = NULL;
6024                         goto lr_restart;
6025                 }
6026         }
6027
6028         switch (task->tk_status) {
6029         case 0:
6030                 renew_lease(data->res.server, data->timestamp);
6031                 break;
6032         case -NFS4ERR_ADMIN_REVOKED:
6033         case -NFS4ERR_DELEG_REVOKED:
6034         case -NFS4ERR_EXPIRED:
6035                 nfs4_free_revoked_stateid(data->res.server,
6036                                 data->args.stateid,
6037                                 task->tk_msg.rpc_cred);
6038                 /* Fallthrough */
6039         case -NFS4ERR_BAD_STATEID:
6040         case -NFS4ERR_STALE_STATEID:
6041                 task->tk_status = 0;
6042                 break;
6043         case -NFS4ERR_OLD_STATEID:
6044                 if (nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6045                         goto out_restart;
6046                 task->tk_status = 0;
6047                 break;
6048         case -NFS4ERR_ACCESS:
6049                 if (data->args.bitmask) {
6050                         data->args.bitmask = NULL;
6051                         data->res.fattr = NULL;
6052                         goto out_restart;
6053                 }
6054                 /* Fallthrough */
6055         default:
6056                 task->tk_status = nfs4_async_handle_exception(task,
6057                                 data->res.server, task->tk_status,
6058                                 &exception);
6059                 if (exception.retry)
6060                         goto out_restart;
6061         }
6062         data->rpc_status = task->tk_status;
6063         return;
6064 lr_restart:
6065         data->res.lr_ret = 0;
6066 out_restart:
6067         task->tk_status = 0;
6068         rpc_restart_call_prepare(task);
6069 }
6070
6071 static void nfs4_delegreturn_release(void *calldata)
6072 {
6073         struct nfs4_delegreturndata *data = calldata;
6074         struct inode *inode = data->inode;
6075
6076         if (inode) {
6077                 if (data->lr.roc)
6078                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6079                                         data->res.lr_ret);
6080                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6081                 nfs_iput_and_deactive(inode);
6082         }
6083         kfree(calldata);
6084 }
6085
6086 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6087 {
6088         struct nfs4_delegreturndata *d_data;
6089         struct pnfs_layout_hdr *lo;
6090
6091         d_data = (struct nfs4_delegreturndata *)data;
6092
6093         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
6094                 return;
6095
6096         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6097         if (lo && !pnfs_layout_is_valid(lo)) {
6098                 d_data->args.lr_args = NULL;
6099                 d_data->res.lr_res = NULL;
6100         }
6101
6102         nfs4_setup_sequence(d_data->res.server->nfs_client,
6103                         &d_data->args.seq_args,
6104                         &d_data->res.seq_res,
6105                         task);
6106 }
6107
6108 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6109         .rpc_call_prepare = nfs4_delegreturn_prepare,
6110         .rpc_call_done = nfs4_delegreturn_done,
6111         .rpc_release = nfs4_delegreturn_release,
6112 };
6113
6114 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
6115 {
6116         struct nfs4_delegreturndata *data;
6117         struct nfs_server *server = NFS_SERVER(inode);
6118         struct rpc_task *task;
6119         struct rpc_message msg = {
6120                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6121                 .rpc_cred = cred,
6122         };
6123         struct rpc_task_setup task_setup_data = {
6124                 .rpc_client = server->client,
6125                 .rpc_message = &msg,
6126                 .callback_ops = &nfs4_delegreturn_ops,
6127                 .flags = RPC_TASK_ASYNC,
6128         };
6129         int status = 0;
6130
6131         data = kzalloc(sizeof(*data), GFP_NOFS);
6132         if (data == NULL)
6133                 return -ENOMEM;
6134         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6135
6136         nfs4_state_protect(server->nfs_client,
6137                         NFS_SP4_MACH_CRED_CLEANUP,
6138                         &task_setup_data.rpc_client, &msg);
6139
6140         data->args.fhandle = &data->fh;
6141         data->args.stateid = &data->stateid;
6142         data->args.bitmask = server->cache_consistency_bitmask;
6143         nfs_copy_fh(&data->fh, NFS_FH(inode));
6144         nfs4_stateid_copy(&data->stateid, stateid);
6145         data->res.fattr = &data->fattr;
6146         data->res.server = server;
6147         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6148         data->lr.arg.ld_private = &data->lr.ld_private;
6149         nfs_fattr_init(data->res.fattr);
6150         data->timestamp = jiffies;
6151         data->rpc_status = 0;
6152         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6153         data->inode = nfs_igrab_and_active(inode);
6154         if (data->inode) {
6155                 if (data->lr.roc) {
6156                         data->args.lr_args = &data->lr.arg;
6157                         data->res.lr_res = &data->lr.res;
6158                 }
6159         } else if (data->lr.roc) {
6160                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6161                 data->lr.roc = false;
6162         }
6163
6164         task_setup_data.callback_data = data;
6165         msg.rpc_argp = &data->args;
6166         msg.rpc_resp = &data->res;
6167         task = rpc_run_task(&task_setup_data);
6168         if (IS_ERR(task))
6169                 return PTR_ERR(task);
6170         if (!issync)
6171                 goto out;
6172         status = rpc_wait_for_completion_task(task);
6173         if (status != 0)
6174                 goto out;
6175         status = data->rpc_status;
6176 out:
6177         rpc_put_task(task);
6178         return status;
6179 }
6180
6181 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
6182 {
6183         struct nfs_server *server = NFS_SERVER(inode);
6184         struct nfs4_exception exception = { };
6185         int err;
6186         do {
6187                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6188                 trace_nfs4_delegreturn(inode, stateid, err);
6189                 switch (err) {
6190                         case -NFS4ERR_STALE_STATEID:
6191                         case -NFS4ERR_EXPIRED:
6192                         case 0:
6193                                 return 0;
6194                 }
6195                 err = nfs4_handle_exception(server, err, &exception);
6196         } while (exception.retry);
6197         return err;
6198 }
6199
6200 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6201 {
6202         struct inode *inode = state->inode;
6203         struct nfs_server *server = NFS_SERVER(inode);
6204         struct nfs_client *clp = server->nfs_client;
6205         struct nfs_lockt_args arg = {
6206                 .fh = NFS_FH(inode),
6207                 .fl = request,
6208         };
6209         struct nfs_lockt_res res = {
6210                 .denied = request,
6211         };
6212         struct rpc_message msg = {
6213                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6214                 .rpc_argp       = &arg,
6215                 .rpc_resp       = &res,
6216                 .rpc_cred       = state->owner->so_cred,
6217         };
6218         struct nfs4_lock_state *lsp;
6219         int status;
6220
6221         arg.lock_owner.clientid = clp->cl_clientid;
6222         status = nfs4_set_lock_state(state, request);
6223         if (status != 0)
6224                 goto out;
6225         lsp = request->fl_u.nfs4_fl.owner;
6226         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6227         arg.lock_owner.s_dev = server->s_dev;
6228         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6229         switch (status) {
6230                 case 0:
6231                         request->fl_type = F_UNLCK;
6232                         break;
6233                 case -NFS4ERR_DENIED:
6234                         status = 0;
6235         }
6236         request->fl_ops->fl_release_private(request);
6237         request->fl_ops = NULL;
6238 out:
6239         return status;
6240 }
6241
6242 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6243 {
6244         struct nfs4_exception exception = { };
6245         int err;
6246
6247         do {
6248                 err = _nfs4_proc_getlk(state, cmd, request);
6249                 trace_nfs4_get_lock(request, state, cmd, err);
6250                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6251                                 &exception);
6252         } while (exception.retry);
6253         return err;
6254 }
6255
6256 struct nfs4_unlockdata {
6257         struct nfs_locku_args arg;
6258         struct nfs_locku_res res;
6259         struct nfs4_lock_state *lsp;
6260         struct nfs_open_context *ctx;
6261         struct nfs_lock_context *l_ctx;
6262         struct file_lock fl;
6263         struct nfs_server *server;
6264         unsigned long timestamp;
6265 };
6266
6267 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6268                 struct nfs_open_context *ctx,
6269                 struct nfs4_lock_state *lsp,
6270                 struct nfs_seqid *seqid)
6271 {
6272         struct nfs4_unlockdata *p;
6273         struct inode *inode = lsp->ls_state->inode;
6274
6275         p = kzalloc(sizeof(*p), GFP_NOFS);
6276         if (p == NULL)
6277                 return NULL;
6278         p->arg.fh = NFS_FH(inode);
6279         p->arg.fl = &p->fl;
6280         p->arg.seqid = seqid;
6281         p->res.seqid = seqid;
6282         p->lsp = lsp;
6283         refcount_inc(&lsp->ls_count);
6284         /* Ensure we don't close file until we're done freeing locks! */
6285         p->ctx = get_nfs_open_context(ctx);
6286         p->l_ctx = nfs_get_lock_context(ctx);
6287         memcpy(&p->fl, fl, sizeof(p->fl));
6288         p->server = NFS_SERVER(inode);
6289         return p;
6290 }
6291
6292 static void nfs4_locku_release_calldata(void *data)
6293 {
6294         struct nfs4_unlockdata *calldata = data;
6295         nfs_free_seqid(calldata->arg.seqid);
6296         nfs4_put_lock_state(calldata->lsp);
6297         nfs_put_lock_context(calldata->l_ctx);
6298         put_nfs_open_context(calldata->ctx);
6299         kfree(calldata);
6300 }
6301
6302 static void nfs4_locku_done(struct rpc_task *task, void *data)
6303 {
6304         struct nfs4_unlockdata *calldata = data;
6305         struct nfs4_exception exception = {
6306                 .inode = calldata->lsp->ls_state->inode,
6307                 .stateid = &calldata->arg.stateid,
6308         };
6309
6310         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6311                 return;
6312         switch (task->tk_status) {
6313                 case 0:
6314                         renew_lease(calldata->server, calldata->timestamp);
6315                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6316                         if (nfs4_update_lock_stateid(calldata->lsp,
6317                                         &calldata->res.stateid))
6318                                 break;
6319                         /* Fall through */
6320                 case -NFS4ERR_ADMIN_REVOKED:
6321                 case -NFS4ERR_EXPIRED:
6322                         nfs4_free_revoked_stateid(calldata->server,
6323                                         &calldata->arg.stateid,
6324                                         task->tk_msg.rpc_cred);
6325                         /* Fall through */
6326                 case -NFS4ERR_BAD_STATEID:
6327                 case -NFS4ERR_OLD_STATEID:
6328                 case -NFS4ERR_STALE_STATEID:
6329                         if (!nfs4_stateid_match(&calldata->arg.stateid,
6330                                                 &calldata->lsp->ls_stateid))
6331                                 rpc_restart_call_prepare(task);
6332                         break;
6333                 default:
6334                         task->tk_status = nfs4_async_handle_exception(task,
6335                                         calldata->server, task->tk_status,
6336                                         &exception);
6337                         if (exception.retry)
6338                                 rpc_restart_call_prepare(task);
6339         }
6340         nfs_release_seqid(calldata->arg.seqid);
6341 }
6342
6343 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6344 {
6345         struct nfs4_unlockdata *calldata = data;
6346
6347         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6348                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6349                 return;
6350
6351         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6352                 goto out_wait;
6353         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
6354         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6355                 /* Note: exit _without_ running nfs4_locku_done */
6356                 goto out_no_action;
6357         }
6358         calldata->timestamp = jiffies;
6359         if (nfs4_setup_sequence(calldata->server->nfs_client,
6360                                 &calldata->arg.seq_args,
6361                                 &calldata->res.seq_res,
6362                                 task) != 0)
6363                 nfs_release_seqid(calldata->arg.seqid);
6364         return;
6365 out_no_action:
6366         task->tk_action = NULL;
6367 out_wait:
6368         nfs4_sequence_done(task, &calldata->res.seq_res);
6369 }
6370
6371 static const struct rpc_call_ops nfs4_locku_ops = {
6372         .rpc_call_prepare = nfs4_locku_prepare,
6373         .rpc_call_done = nfs4_locku_done,
6374         .rpc_release = nfs4_locku_release_calldata,
6375 };
6376
6377 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6378                 struct nfs_open_context *ctx,
6379                 struct nfs4_lock_state *lsp,
6380                 struct nfs_seqid *seqid)
6381 {
6382         struct nfs4_unlockdata *data;
6383         struct rpc_message msg = {
6384                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6385                 .rpc_cred = ctx->cred,
6386         };
6387         struct rpc_task_setup task_setup_data = {
6388                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6389                 .rpc_message = &msg,
6390                 .callback_ops = &nfs4_locku_ops,
6391                 .workqueue = nfsiod_workqueue,
6392                 .flags = RPC_TASK_ASYNC,
6393         };
6394
6395         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6396                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6397
6398         /* Ensure this is an unlock - when canceling a lock, the
6399          * canceled lock is passed in, and it won't be an unlock.
6400          */
6401         fl->fl_type = F_UNLCK;
6402         if (fl->fl_flags & FL_CLOSE)
6403                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6404
6405         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6406         if (data == NULL) {
6407                 nfs_free_seqid(seqid);
6408                 return ERR_PTR(-ENOMEM);
6409         }
6410
6411         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6412         msg.rpc_argp = &data->arg;
6413         msg.rpc_resp = &data->res;
6414         task_setup_data.callback_data = data;
6415         return rpc_run_task(&task_setup_data);
6416 }
6417
6418 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6419 {
6420         struct inode *inode = state->inode;
6421         struct nfs4_state_owner *sp = state->owner;
6422         struct nfs_inode *nfsi = NFS_I(inode);
6423         struct nfs_seqid *seqid;
6424         struct nfs4_lock_state *lsp;
6425         struct rpc_task *task;
6426         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6427         int status = 0;
6428         unsigned char fl_flags = request->fl_flags;
6429
6430         status = nfs4_set_lock_state(state, request);
6431         /* Unlock _before_ we do the RPC call */
6432         request->fl_flags |= FL_EXISTS;
6433         /* Exclude nfs_delegation_claim_locks() */
6434         mutex_lock(&sp->so_delegreturn_mutex);
6435         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6436         down_read(&nfsi->rwsem);
6437         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6438                 up_read(&nfsi->rwsem);
6439                 mutex_unlock(&sp->so_delegreturn_mutex);
6440                 goto out;
6441         }
6442         up_read(&nfsi->rwsem);
6443         mutex_unlock(&sp->so_delegreturn_mutex);
6444         if (status != 0)
6445                 goto out;
6446         /* Is this a delegated lock? */
6447         lsp = request->fl_u.nfs4_fl.owner;
6448         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6449                 goto out;
6450         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6451         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6452         status = -ENOMEM;
6453         if (IS_ERR(seqid))
6454                 goto out;
6455         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6456         status = PTR_ERR(task);
6457         if (IS_ERR(task))
6458                 goto out;
6459         status = rpc_wait_for_completion_task(task);
6460         rpc_put_task(task);
6461 out:
6462         request->fl_flags = fl_flags;
6463         trace_nfs4_unlock(request, state, F_SETLK, status);
6464         return status;
6465 }
6466
6467 struct nfs4_lockdata {
6468         struct nfs_lock_args arg;
6469         struct nfs_lock_res res;
6470         struct nfs4_lock_state *lsp;
6471         struct nfs_open_context *ctx;
6472         struct file_lock fl;
6473         unsigned long timestamp;
6474         int rpc_status;
6475         int cancelled;
6476         struct nfs_server *server;
6477 };
6478
6479 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6480                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6481                 gfp_t gfp_mask)
6482 {
6483         struct nfs4_lockdata *p;
6484         struct inode *inode = lsp->ls_state->inode;
6485         struct nfs_server *server = NFS_SERVER(inode);
6486         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6487
6488         p = kzalloc(sizeof(*p), gfp_mask);
6489         if (p == NULL)
6490                 return NULL;
6491
6492         p->arg.fh = NFS_FH(inode);
6493         p->arg.fl = &p->fl;
6494         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6495         if (IS_ERR(p->arg.open_seqid))
6496                 goto out_free;
6497         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6498         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6499         if (IS_ERR(p->arg.lock_seqid))
6500                 goto out_free_seqid;
6501         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6502         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6503         p->arg.lock_owner.s_dev = server->s_dev;
6504         p->res.lock_seqid = p->arg.lock_seqid;
6505         p->lsp = lsp;
6506         p->server = server;
6507         refcount_inc(&lsp->ls_count);
6508         p->ctx = get_nfs_open_context(ctx);
6509         memcpy(&p->fl, fl, sizeof(p->fl));
6510         return p;
6511 out_free_seqid:
6512         nfs_free_seqid(p->arg.open_seqid);
6513 out_free:
6514         kfree(p);
6515         return NULL;
6516 }
6517
6518 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6519 {
6520         struct nfs4_lockdata *data = calldata;
6521         struct nfs4_state *state = data->lsp->ls_state;
6522
6523         dprintk("%s: begin!\n", __func__);
6524         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6525                 goto out_wait;
6526         /* Do we need to do an open_to_lock_owner? */
6527         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6528                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6529                         goto out_release_lock_seqid;
6530                 }
6531                 nfs4_stateid_copy(&data->arg.open_stateid,
6532                                 &state->open_stateid);
6533                 data->arg.new_lock_owner = 1;
6534                 data->res.open_seqid = data->arg.open_seqid;
6535         } else {
6536                 data->arg.new_lock_owner = 0;
6537                 nfs4_stateid_copy(&data->arg.lock_stateid,
6538                                 &data->lsp->ls_stateid);
6539         }
6540         if (!nfs4_valid_open_stateid(state)) {
6541                 data->rpc_status = -EBADF;
6542                 task->tk_action = NULL;
6543                 goto out_release_open_seqid;
6544         }
6545         data->timestamp = jiffies;
6546         if (nfs4_setup_sequence(data->server->nfs_client,
6547                                 &data->arg.seq_args,
6548                                 &data->res.seq_res,
6549                                 task) == 0)
6550                 return;
6551 out_release_open_seqid:
6552         nfs_release_seqid(data->arg.open_seqid);
6553 out_release_lock_seqid:
6554         nfs_release_seqid(data->arg.lock_seqid);
6555 out_wait:
6556         nfs4_sequence_done(task, &data->res.seq_res);
6557         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6558 }
6559
6560 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6561 {
6562         struct nfs4_lockdata *data = calldata;
6563         struct nfs4_lock_state *lsp = data->lsp;
6564
6565         dprintk("%s: begin!\n", __func__);
6566
6567         if (!nfs4_sequence_done(task, &data->res.seq_res))
6568                 return;
6569
6570         data->rpc_status = task->tk_status;
6571         switch (task->tk_status) {
6572         case 0:
6573                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6574                                 data->timestamp);
6575                 if (data->arg.new_lock && !data->cancelled) {
6576                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6577                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6578                                 goto out_restart;
6579                 }
6580                 if (data->arg.new_lock_owner != 0) {
6581                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6582                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6583                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6584                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6585                         goto out_restart;
6586                 break;
6587         case -NFS4ERR_BAD_STATEID:
6588         case -NFS4ERR_OLD_STATEID:
6589         case -NFS4ERR_STALE_STATEID:
6590         case -NFS4ERR_EXPIRED:
6591                 if (data->arg.new_lock_owner != 0) {
6592                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6593                                                 &lsp->ls_state->open_stateid))
6594                                 goto out_restart;
6595                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6596                                                 &lsp->ls_stateid))
6597                                 goto out_restart;
6598         }
6599 out_done:
6600         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6601         return;
6602 out_restart:
6603         if (!data->cancelled)
6604                 rpc_restart_call_prepare(task);
6605         goto out_done;
6606 }
6607
6608 static void nfs4_lock_release(void *calldata)
6609 {
6610         struct nfs4_lockdata *data = calldata;
6611
6612         dprintk("%s: begin!\n", __func__);
6613         nfs_free_seqid(data->arg.open_seqid);
6614         if (data->cancelled && data->rpc_status == 0) {
6615                 struct rpc_task *task;
6616                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6617                                 data->arg.lock_seqid);
6618                 if (!IS_ERR(task))
6619                         rpc_put_task_async(task);
6620                 dprintk("%s: cancelling lock!\n", __func__);
6621         } else
6622                 nfs_free_seqid(data->arg.lock_seqid);
6623         nfs4_put_lock_state(data->lsp);
6624         put_nfs_open_context(data->ctx);
6625         kfree(data);
6626         dprintk("%s: done!\n", __func__);
6627 }
6628
6629 static const struct rpc_call_ops nfs4_lock_ops = {
6630         .rpc_call_prepare = nfs4_lock_prepare,
6631         .rpc_call_done = nfs4_lock_done,
6632         .rpc_release = nfs4_lock_release,
6633 };
6634
6635 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6636 {
6637         switch (error) {
6638         case -NFS4ERR_ADMIN_REVOKED:
6639         case -NFS4ERR_EXPIRED:
6640         case -NFS4ERR_BAD_STATEID:
6641                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6642                 if (new_lock_owner != 0 ||
6643                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6644                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6645                 break;
6646         case -NFS4ERR_STALE_STATEID:
6647                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6648                 nfs4_schedule_lease_recovery(server->nfs_client);
6649         };
6650 }
6651
6652 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6653 {
6654         struct nfs4_lockdata *data;
6655         struct rpc_task *task;
6656         struct rpc_message msg = {
6657                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6658                 .rpc_cred = state->owner->so_cred,
6659         };
6660         struct rpc_task_setup task_setup_data = {
6661                 .rpc_client = NFS_CLIENT(state->inode),
6662                 .rpc_message = &msg,
6663                 .callback_ops = &nfs4_lock_ops,
6664                 .workqueue = nfsiod_workqueue,
6665                 .flags = RPC_TASK_ASYNC,
6666         };
6667         int ret;
6668
6669         dprintk("%s: begin!\n", __func__);
6670         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6671                         fl->fl_u.nfs4_fl.owner,
6672                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6673         if (data == NULL)
6674                 return -ENOMEM;
6675         if (IS_SETLKW(cmd))
6676                 data->arg.block = 1;
6677         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6678                                 recovery_type > NFS_LOCK_NEW);
6679         msg.rpc_argp = &data->arg;
6680         msg.rpc_resp = &data->res;
6681         task_setup_data.callback_data = data;
6682         if (recovery_type > NFS_LOCK_NEW) {
6683                 if (recovery_type == NFS_LOCK_RECLAIM)
6684                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6685         } else
6686                 data->arg.new_lock = 1;
6687         task = rpc_run_task(&task_setup_data);
6688         if (IS_ERR(task))
6689                 return PTR_ERR(task);
6690         ret = rpc_wait_for_completion_task(task);
6691         if (ret == 0) {
6692                 ret = data->rpc_status;
6693                 if (ret)
6694                         nfs4_handle_setlk_error(data->server, data->lsp,
6695                                         data->arg.new_lock_owner, ret);
6696         } else
6697                 data->cancelled = true;
6698         rpc_put_task(task);
6699         dprintk("%s: done, ret = %d!\n", __func__, ret);
6700         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6701         return ret;
6702 }
6703
6704 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6705 {
6706         struct nfs_server *server = NFS_SERVER(state->inode);
6707         struct nfs4_exception exception = {
6708                 .inode = state->inode,
6709         };
6710         int err;
6711
6712         do {
6713                 /* Cache the lock if possible... */
6714                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6715                         return 0;
6716                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6717                 if (err != -NFS4ERR_DELAY)
6718                         break;
6719                 nfs4_handle_exception(server, err, &exception);
6720         } while (exception.retry);
6721         return err;
6722 }
6723
6724 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6725 {
6726         struct nfs_server *server = NFS_SERVER(state->inode);
6727         struct nfs4_exception exception = {
6728                 .inode = state->inode,
6729         };
6730         int err;
6731
6732         err = nfs4_set_lock_state(state, request);
6733         if (err != 0)
6734                 return err;
6735         if (!recover_lost_locks) {
6736                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6737                 return 0;
6738         }
6739         do {
6740                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6741                         return 0;
6742                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6743                 switch (err) {
6744                 default:
6745                         goto out;
6746                 case -NFS4ERR_GRACE:
6747                 case -NFS4ERR_DELAY:
6748                         nfs4_handle_exception(server, err, &exception);
6749                         err = 0;
6750                 }
6751         } while (exception.retry);
6752 out:
6753         return err;
6754 }
6755
6756 #if defined(CONFIG_NFS_V4_1)
6757 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6758 {
6759         struct nfs4_lock_state *lsp;
6760         int status;
6761
6762         status = nfs4_set_lock_state(state, request);
6763         if (status != 0)
6764                 return status;
6765         lsp = request->fl_u.nfs4_fl.owner;
6766         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6767             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6768                 return 0;
6769         return nfs4_lock_expired(state, request);
6770 }
6771 #endif
6772
6773 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6774 {
6775         struct nfs_inode *nfsi = NFS_I(state->inode);
6776         struct nfs4_state_owner *sp = state->owner;
6777         unsigned char fl_flags = request->fl_flags;
6778         int status;
6779
6780         request->fl_flags |= FL_ACCESS;
6781         status = locks_lock_inode_wait(state->inode, request);
6782         if (status < 0)
6783                 goto out;
6784         mutex_lock(&sp->so_delegreturn_mutex);
6785         down_read(&nfsi->rwsem);
6786         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6787                 /* Yes: cache locks! */
6788                 /* ...but avoid races with delegation recall... */
6789                 request->fl_flags = fl_flags & ~FL_SLEEP;
6790                 status = locks_lock_inode_wait(state->inode, request);
6791                 up_read(&nfsi->rwsem);
6792                 mutex_unlock(&sp->so_delegreturn_mutex);
6793                 goto out;
6794         }
6795         up_read(&nfsi->rwsem);
6796         mutex_unlock(&sp->so_delegreturn_mutex);
6797         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6798 out:
6799         request->fl_flags = fl_flags;
6800         return status;
6801 }
6802
6803 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6804 {
6805         struct nfs4_exception exception = {
6806                 .state = state,
6807                 .inode = state->inode,
6808         };
6809         int err;
6810
6811         do {
6812                 err = _nfs4_proc_setlk(state, cmd, request);
6813                 if (err == -NFS4ERR_DENIED)
6814                         err = -EAGAIN;
6815                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6816                                 err, &exception);
6817         } while (exception.retry);
6818         return err;
6819 }
6820
6821 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6822 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6823
6824 static int
6825 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6826                         struct file_lock *request)
6827 {
6828         int             status = -ERESTARTSYS;
6829         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6830
6831         while(!signalled()) {
6832                 status = nfs4_proc_setlk(state, cmd, request);
6833                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6834                         break;
6835                 freezable_schedule_timeout_interruptible(timeout);
6836                 timeout *= 2;
6837                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6838                 status = -ERESTARTSYS;
6839         }
6840         return status;
6841 }
6842
6843 #ifdef CONFIG_NFS_V4_1
6844 struct nfs4_lock_waiter {
6845         struct task_struct      *task;
6846         struct inode            *inode;
6847         struct nfs_lowner       *owner;
6848         bool                    notified;
6849 };
6850
6851 static int
6852 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6853 {
6854         int ret;
6855         struct nfs4_lock_waiter *waiter = wait->private;
6856
6857         /* NULL key means to wake up everyone */
6858         if (key) {
6859                 struct cb_notify_lock_args      *cbnl = key;
6860                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
6861                                                 *wowner = waiter->owner;
6862
6863                 /* Only wake if the callback was for the same owner. */
6864                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
6865                         return 0;
6866
6867                 /* Make sure it's for the right inode */
6868                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6869                         return 0;
6870
6871                 waiter->notified = true;
6872         }
6873
6874         /* override "private" so we can use default_wake_function */
6875         wait->private = waiter->task;
6876         ret = autoremove_wake_function(wait, mode, flags, key);
6877         wait->private = waiter;
6878         return ret;
6879 }
6880
6881 static int
6882 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6883 {
6884         int status = -ERESTARTSYS;
6885         unsigned long flags;
6886         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6887         struct nfs_server *server = NFS_SERVER(state->inode);
6888         struct nfs_client *clp = server->nfs_client;
6889         wait_queue_head_t *q = &clp->cl_lock_waitq;
6890         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6891                                     .id = lsp->ls_seqid.owner_id,
6892                                     .s_dev = server->s_dev };
6893         struct nfs4_lock_waiter waiter = { .task  = current,
6894                                            .inode = state->inode,
6895                                            .owner = &owner,
6896                                            .notified = false };
6897         wait_queue_entry_t wait;
6898
6899         /* Don't bother with waitqueue if we don't expect a callback */
6900         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6901                 return nfs4_retry_setlk_simple(state, cmd, request);
6902
6903         init_wait(&wait);
6904         wait.private = &waiter;
6905         wait.func = nfs4_wake_lock_waiter;
6906         add_wait_queue(q, &wait);
6907
6908         while(!signalled()) {
6909                 waiter.notified = false;
6910                 status = nfs4_proc_setlk(state, cmd, request);
6911                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6912                         break;
6913
6914                 status = -ERESTARTSYS;
6915                 spin_lock_irqsave(&q->lock, flags);
6916                 if (waiter.notified) {
6917                         spin_unlock_irqrestore(&q->lock, flags);
6918                         continue;
6919                 }
6920                 set_current_state(TASK_INTERRUPTIBLE);
6921                 spin_unlock_irqrestore(&q->lock, flags);
6922
6923                 freezable_schedule_timeout(NFS4_LOCK_MAXTIMEOUT);
6924         }
6925
6926         finish_wait(q, &wait);
6927         return status;
6928 }
6929 #else /* !CONFIG_NFS_V4_1 */
6930 static inline int
6931 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6932 {
6933         return nfs4_retry_setlk_simple(state, cmd, request);
6934 }
6935 #endif
6936
6937 static int
6938 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6939 {
6940         struct nfs_open_context *ctx;
6941         struct nfs4_state *state;
6942         int status;
6943
6944         /* verify open state */
6945         ctx = nfs_file_open_context(filp);
6946         state = ctx->state;
6947
6948         if (IS_GETLK(cmd)) {
6949                 if (state != NULL)
6950                         return nfs4_proc_getlk(state, F_GETLK, request);
6951                 return 0;
6952         }
6953
6954         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6955                 return -EINVAL;
6956
6957         if (request->fl_type == F_UNLCK) {
6958                 if (state != NULL)
6959                         return nfs4_proc_unlck(state, cmd, request);
6960                 return 0;
6961         }
6962
6963         if (state == NULL)
6964                 return -ENOLCK;
6965
6966         if ((request->fl_flags & FL_POSIX) &&
6967             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6968                 return -ENOLCK;
6969
6970         /*
6971          * Don't rely on the VFS having checked the file open mode,
6972          * since it won't do this for flock() locks.
6973          */
6974         switch (request->fl_type) {
6975         case F_RDLCK:
6976                 if (!(filp->f_mode & FMODE_READ))
6977                         return -EBADF;
6978                 break;
6979         case F_WRLCK:
6980                 if (!(filp->f_mode & FMODE_WRITE))
6981                         return -EBADF;
6982         }
6983
6984         status = nfs4_set_lock_state(state, request);
6985         if (status != 0)
6986                 return status;
6987
6988         return nfs4_retry_setlk(state, cmd, request);
6989 }
6990
6991 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6992 {
6993         struct nfs_server *server = NFS_SERVER(state->inode);
6994         int err;
6995
6996         err = nfs4_set_lock_state(state, fl);
6997         if (err != 0)
6998                 return err;
6999         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7000         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7001 }
7002
7003 struct nfs_release_lockowner_data {
7004         struct nfs4_lock_state *lsp;
7005         struct nfs_server *server;
7006         struct nfs_release_lockowner_args args;
7007         struct nfs_release_lockowner_res res;
7008         unsigned long timestamp;
7009 };
7010
7011 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7012 {
7013         struct nfs_release_lockowner_data *data = calldata;
7014         struct nfs_server *server = data->server;
7015         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7016                            &data->res.seq_res, task);
7017         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7018         data->timestamp = jiffies;
7019 }
7020
7021 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7022 {
7023         struct nfs_release_lockowner_data *data = calldata;
7024         struct nfs_server *server = data->server;
7025
7026         nfs40_sequence_done(task, &data->res.seq_res);
7027
7028         switch (task->tk_status) {
7029         case 0:
7030                 renew_lease(server, data->timestamp);
7031                 break;
7032         case -NFS4ERR_STALE_CLIENTID:
7033         case -NFS4ERR_EXPIRED:
7034                 nfs4_schedule_lease_recovery(server->nfs_client);
7035                 break;
7036         case -NFS4ERR_LEASE_MOVED:
7037         case -NFS4ERR_DELAY:
7038                 if (nfs4_async_handle_error(task, server,
7039                                             NULL, NULL) == -EAGAIN)
7040                         rpc_restart_call_prepare(task);
7041         }
7042 }
7043
7044 static void nfs4_release_lockowner_release(void *calldata)
7045 {
7046         struct nfs_release_lockowner_data *data = calldata;
7047         nfs4_free_lock_state(data->server, data->lsp);
7048         kfree(calldata);
7049 }
7050
7051 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7052         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7053         .rpc_call_done = nfs4_release_lockowner_done,
7054         .rpc_release = nfs4_release_lockowner_release,
7055 };
7056
7057 static void
7058 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7059 {
7060         struct nfs_release_lockowner_data *data;
7061         struct rpc_message msg = {
7062                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7063         };
7064
7065         if (server->nfs_client->cl_mvops->minor_version != 0)
7066                 return;
7067
7068         data = kmalloc(sizeof(*data), GFP_NOFS);
7069         if (!data)
7070                 return;
7071         data->lsp = lsp;
7072         data->server = server;
7073         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7074         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7075         data->args.lock_owner.s_dev = server->s_dev;
7076
7077         msg.rpc_argp = &data->args;
7078         msg.rpc_resp = &data->res;
7079         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7080         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7081 }
7082
7083 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7084
7085 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7086                                    struct dentry *unused, struct inode *inode,
7087                                    const char *key, const void *buf,
7088                                    size_t buflen, int flags)
7089 {
7090         return nfs4_proc_set_acl(inode, buf, buflen);
7091 }
7092
7093 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7094                                    struct dentry *unused, struct inode *inode,
7095                                    const char *key, void *buf, size_t buflen)
7096 {
7097         return nfs4_proc_get_acl(inode, buf, buflen);
7098 }
7099
7100 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7101 {
7102         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7103 }
7104
7105 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7106
7107 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7108                                      struct dentry *unused, struct inode *inode,
7109                                      const char *key, const void *buf,
7110                                      size_t buflen, int flags)
7111 {
7112         if (security_ismaclabel(key))
7113                 return nfs4_set_security_label(inode, buf, buflen);
7114
7115         return -EOPNOTSUPP;
7116 }
7117
7118 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7119                                      struct dentry *unused, struct inode *inode,
7120                                      const char *key, void *buf, size_t buflen)
7121 {
7122         if (security_ismaclabel(key))
7123                 return nfs4_get_security_label(inode, buf, buflen);
7124         return -EOPNOTSUPP;
7125 }
7126
7127 static ssize_t
7128 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7129 {
7130         int len = 0;
7131
7132         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7133                 len = security_inode_listsecurity(inode, list, list_len);
7134                 if (list_len && len > list_len)
7135                         return -ERANGE;
7136         }
7137         return len;
7138 }
7139
7140 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7141         .prefix = XATTR_SECURITY_PREFIX,
7142         .get    = nfs4_xattr_get_nfs4_label,
7143         .set    = nfs4_xattr_set_nfs4_label,
7144 };
7145
7146 #else
7147
7148 static ssize_t
7149 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7150 {
7151         return 0;
7152 }
7153
7154 #endif
7155
7156 /*
7157  * nfs_fhget will use either the mounted_on_fileid or the fileid
7158  */
7159 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7160 {
7161         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7162                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7163               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7164               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7165                 return;
7166
7167         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7168                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7169         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7170         fattr->nlink = 2;
7171 }
7172
7173 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7174                                    const struct qstr *name,
7175                                    struct nfs4_fs_locations *fs_locations,
7176                                    struct page *page)
7177 {
7178         struct nfs_server *server = NFS_SERVER(dir);
7179         u32 bitmask[3];
7180         struct nfs4_fs_locations_arg args = {
7181                 .dir_fh = NFS_FH(dir),
7182                 .name = name,
7183                 .page = page,
7184                 .bitmask = bitmask,
7185         };
7186         struct nfs4_fs_locations_res res = {
7187                 .fs_locations = fs_locations,
7188         };
7189         struct rpc_message msg = {
7190                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7191                 .rpc_argp = &args,
7192                 .rpc_resp = &res,
7193         };
7194         int status;
7195
7196         dprintk("%s: start\n", __func__);
7197
7198         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7199         bitmask[1] = nfs4_fattr_bitmap[1];
7200
7201         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7202          * is not supported */
7203         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7204                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7205         else
7206                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7207
7208         nfs_fattr_init(&fs_locations->fattr);
7209         fs_locations->server = server;
7210         fs_locations->nlocations = 0;
7211         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7212         dprintk("%s: returned status = %d\n", __func__, status);
7213         return status;
7214 }
7215
7216 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7217                            const struct qstr *name,
7218                            struct nfs4_fs_locations *fs_locations,
7219                            struct page *page)
7220 {
7221         struct nfs4_exception exception = { };
7222         int err;
7223         do {
7224                 err = _nfs4_proc_fs_locations(client, dir, name,
7225                                 fs_locations, page);
7226                 trace_nfs4_get_fs_locations(dir, name, err);
7227                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7228                                 &exception);
7229         } while (exception.retry);
7230         return err;
7231 }
7232
7233 /*
7234  * This operation also signals the server that this client is
7235  * performing migration recovery.  The server can stop returning
7236  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7237  * appended to this compound to identify the client ID which is
7238  * performing recovery.
7239  */
7240 static int _nfs40_proc_get_locations(struct inode *inode,
7241                                      struct nfs4_fs_locations *locations,
7242                                      struct page *page, struct rpc_cred *cred)
7243 {
7244         struct nfs_server *server = NFS_SERVER(inode);
7245         struct rpc_clnt *clnt = server->client;
7246         u32 bitmask[2] = {
7247                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7248         };
7249         struct nfs4_fs_locations_arg args = {
7250                 .clientid       = server->nfs_client->cl_clientid,
7251                 .fh             = NFS_FH(inode),
7252                 .page           = page,
7253                 .bitmask        = bitmask,
7254                 .migration      = 1,            /* skip LOOKUP */
7255                 .renew          = 1,            /* append RENEW */
7256         };
7257         struct nfs4_fs_locations_res res = {
7258                 .fs_locations   = locations,
7259                 .migration      = 1,
7260                 .renew          = 1,
7261         };
7262         struct rpc_message msg = {
7263                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7264                 .rpc_argp       = &args,
7265                 .rpc_resp       = &res,
7266                 .rpc_cred       = cred,
7267         };
7268         unsigned long now = jiffies;
7269         int status;
7270
7271         nfs_fattr_init(&locations->fattr);
7272         locations->server = server;
7273         locations->nlocations = 0;
7274
7275         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7276         status = nfs4_call_sync_sequence(clnt, server, &msg,
7277                                         &args.seq_args, &res.seq_res);
7278         if (status)
7279                 return status;
7280
7281         renew_lease(server, now);
7282         return 0;
7283 }
7284
7285 #ifdef CONFIG_NFS_V4_1
7286
7287 /*
7288  * This operation also signals the server that this client is
7289  * performing migration recovery.  The server can stop asserting
7290  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7291  * performing this operation is identified in the SEQUENCE
7292  * operation in this compound.
7293  *
7294  * When the client supports GETATTR(fs_locations_info), it can
7295  * be plumbed in here.
7296  */
7297 static int _nfs41_proc_get_locations(struct inode *inode,
7298                                      struct nfs4_fs_locations *locations,
7299                                      struct page *page, struct rpc_cred *cred)
7300 {
7301         struct nfs_server *server = NFS_SERVER(inode);
7302         struct rpc_clnt *clnt = server->client;
7303         u32 bitmask[2] = {
7304                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7305         };
7306         struct nfs4_fs_locations_arg args = {
7307                 .fh             = NFS_FH(inode),
7308                 .page           = page,
7309                 .bitmask        = bitmask,
7310                 .migration      = 1,            /* skip LOOKUP */
7311         };
7312         struct nfs4_fs_locations_res res = {
7313                 .fs_locations   = locations,
7314                 .migration      = 1,
7315         };
7316         struct rpc_message msg = {
7317                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7318                 .rpc_argp       = &args,
7319                 .rpc_resp       = &res,
7320                 .rpc_cred       = cred,
7321         };
7322         int status;
7323
7324         nfs_fattr_init(&locations->fattr);
7325         locations->server = server;
7326         locations->nlocations = 0;
7327
7328         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7329         status = nfs4_call_sync_sequence(clnt, server, &msg,
7330                                         &args.seq_args, &res.seq_res);
7331         if (status == NFS4_OK &&
7332             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7333                 status = -NFS4ERR_LEASE_MOVED;
7334         return status;
7335 }
7336
7337 #endif  /* CONFIG_NFS_V4_1 */
7338
7339 /**
7340  * nfs4_proc_get_locations - discover locations for a migrated FSID
7341  * @inode: inode on FSID that is migrating
7342  * @locations: result of query
7343  * @page: buffer
7344  * @cred: credential to use for this operation
7345  *
7346  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7347  * operation failed, or a negative errno if a local error occurred.
7348  *
7349  * On success, "locations" is filled in, but if the server has
7350  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7351  * asserted.
7352  *
7353  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7354  * from this client that require migration recovery.
7355  */
7356 int nfs4_proc_get_locations(struct inode *inode,
7357                             struct nfs4_fs_locations *locations,
7358                             struct page *page, struct rpc_cred *cred)
7359 {
7360         struct nfs_server *server = NFS_SERVER(inode);
7361         struct nfs_client *clp = server->nfs_client;
7362         const struct nfs4_mig_recovery_ops *ops =
7363                                         clp->cl_mvops->mig_recovery_ops;
7364         struct nfs4_exception exception = { };
7365         int status;
7366
7367         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7368                 (unsigned long long)server->fsid.major,
7369                 (unsigned long long)server->fsid.minor,
7370                 clp->cl_hostname);
7371         nfs_display_fhandle(NFS_FH(inode), __func__);
7372
7373         do {
7374                 status = ops->get_locations(inode, locations, page, cred);
7375                 if (status != -NFS4ERR_DELAY)
7376                         break;
7377                 nfs4_handle_exception(server, status, &exception);
7378         } while (exception.retry);
7379         return status;
7380 }
7381
7382 /*
7383  * This operation also signals the server that this client is
7384  * performing "lease moved" recovery.  The server can stop
7385  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7386  * is appended to this compound to identify the client ID which is
7387  * performing recovery.
7388  */
7389 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7390 {
7391         struct nfs_server *server = NFS_SERVER(inode);
7392         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7393         struct rpc_clnt *clnt = server->client;
7394         struct nfs4_fsid_present_arg args = {
7395                 .fh             = NFS_FH(inode),
7396                 .clientid       = clp->cl_clientid,
7397                 .renew          = 1,            /* append RENEW */
7398         };
7399         struct nfs4_fsid_present_res res = {
7400                 .renew          = 1,
7401         };
7402         struct rpc_message msg = {
7403                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7404                 .rpc_argp       = &args,
7405                 .rpc_resp       = &res,
7406                 .rpc_cred       = cred,
7407         };
7408         unsigned long now = jiffies;
7409         int status;
7410
7411         res.fh = nfs_alloc_fhandle();
7412         if (res.fh == NULL)
7413                 return -ENOMEM;
7414
7415         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7416         status = nfs4_call_sync_sequence(clnt, server, &msg,
7417                                                 &args.seq_args, &res.seq_res);
7418         nfs_free_fhandle(res.fh);
7419         if (status)
7420                 return status;
7421
7422         do_renew_lease(clp, now);
7423         return 0;
7424 }
7425
7426 #ifdef CONFIG_NFS_V4_1
7427
7428 /*
7429  * This operation also signals the server that this client is
7430  * performing "lease moved" recovery.  The server can stop asserting
7431  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7432  * this operation is identified in the SEQUENCE operation in this
7433  * compound.
7434  */
7435 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7436 {
7437         struct nfs_server *server = NFS_SERVER(inode);
7438         struct rpc_clnt *clnt = server->client;
7439         struct nfs4_fsid_present_arg args = {
7440                 .fh             = NFS_FH(inode),
7441         };
7442         struct nfs4_fsid_present_res res = {
7443         };
7444         struct rpc_message msg = {
7445                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7446                 .rpc_argp       = &args,
7447                 .rpc_resp       = &res,
7448                 .rpc_cred       = cred,
7449         };
7450         int status;
7451
7452         res.fh = nfs_alloc_fhandle();
7453         if (res.fh == NULL)
7454                 return -ENOMEM;
7455
7456         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7457         status = nfs4_call_sync_sequence(clnt, server, &msg,
7458                                                 &args.seq_args, &res.seq_res);
7459         nfs_free_fhandle(res.fh);
7460         if (status == NFS4_OK &&
7461             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7462                 status = -NFS4ERR_LEASE_MOVED;
7463         return status;
7464 }
7465
7466 #endif  /* CONFIG_NFS_V4_1 */
7467
7468 /**
7469  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7470  * @inode: inode on FSID to check
7471  * @cred: credential to use for this operation
7472  *
7473  * Server indicates whether the FSID is present, moved, or not
7474  * recognized.  This operation is necessary to clear a LEASE_MOVED
7475  * condition for this client ID.
7476  *
7477  * Returns NFS4_OK if the FSID is present on this server,
7478  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7479  *  NFS4ERR code if some error occurred on the server, or a
7480  *  negative errno if a local failure occurred.
7481  */
7482 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7483 {
7484         struct nfs_server *server = NFS_SERVER(inode);
7485         struct nfs_client *clp = server->nfs_client;
7486         const struct nfs4_mig_recovery_ops *ops =
7487                                         clp->cl_mvops->mig_recovery_ops;
7488         struct nfs4_exception exception = { };
7489         int status;
7490
7491         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7492                 (unsigned long long)server->fsid.major,
7493                 (unsigned long long)server->fsid.minor,
7494                 clp->cl_hostname);
7495         nfs_display_fhandle(NFS_FH(inode), __func__);
7496
7497         do {
7498                 status = ops->fsid_present(inode, cred);
7499                 if (status != -NFS4ERR_DELAY)
7500                         break;
7501                 nfs4_handle_exception(server, status, &exception);
7502         } while (exception.retry);
7503         return status;
7504 }
7505
7506 /**
7507  * If 'use_integrity' is true and the state managment nfs_client
7508  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7509  * and the machine credential as per RFC3530bis and RFC5661 Security
7510  * Considerations sections. Otherwise, just use the user cred with the
7511  * filesystem's rpc_client.
7512  */
7513 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7514 {
7515         int status;
7516         struct nfs4_secinfo_arg args = {
7517                 .dir_fh = NFS_FH(dir),
7518                 .name   = name,
7519         };
7520         struct nfs4_secinfo_res res = {
7521                 .flavors     = flavors,
7522         };
7523         struct rpc_message msg = {
7524                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7525                 .rpc_argp = &args,
7526                 .rpc_resp = &res,
7527         };
7528         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7529         struct rpc_cred *cred = NULL;
7530
7531         if (use_integrity) {
7532                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7533                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7534                 msg.rpc_cred = cred;
7535         }
7536
7537         dprintk("NFS call  secinfo %s\n", name->name);
7538
7539         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7540                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7541
7542         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7543                                 &res.seq_res, 0);
7544         dprintk("NFS reply  secinfo: %d\n", status);
7545
7546         if (cred)
7547                 put_rpccred(cred);
7548
7549         return status;
7550 }
7551
7552 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7553                       struct nfs4_secinfo_flavors *flavors)
7554 {
7555         struct nfs4_exception exception = { };
7556         int err;
7557         do {
7558                 err = -NFS4ERR_WRONGSEC;
7559
7560                 /* try to use integrity protection with machine cred */
7561                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7562                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7563
7564                 /*
7565                  * if unable to use integrity protection, or SECINFO with
7566                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7567                  * disallowed by spec, but exists in deployed servers) use
7568                  * the current filesystem's rpc_client and the user cred.
7569                  */
7570                 if (err == -NFS4ERR_WRONGSEC)
7571                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7572
7573                 trace_nfs4_secinfo(dir, name, err);
7574                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7575                                 &exception);
7576         } while (exception.retry);
7577         return err;
7578 }
7579
7580 #ifdef CONFIG_NFS_V4_1
7581 /*
7582  * Check the exchange flags returned by the server for invalid flags, having
7583  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7584  * DS flags set.
7585  */
7586 static int nfs4_check_cl_exchange_flags(u32 flags)
7587 {
7588         if (flags & ~EXCHGID4_FLAG_MASK_R)
7589                 goto out_inval;
7590         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7591             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7592                 goto out_inval;
7593         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7594                 goto out_inval;
7595         return NFS_OK;
7596 out_inval:
7597         return -NFS4ERR_INVAL;
7598 }
7599
7600 static bool
7601 nfs41_same_server_scope(struct nfs41_server_scope *a,
7602                         struct nfs41_server_scope *b)
7603 {
7604         if (a->server_scope_sz != b->server_scope_sz)
7605                 return false;
7606         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7607 }
7608
7609 static void
7610 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7611 {
7612 }
7613
7614 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7615         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7616 };
7617
7618 /*
7619  * nfs4_proc_bind_one_conn_to_session()
7620  *
7621  * The 4.1 client currently uses the same TCP connection for the
7622  * fore and backchannel.
7623  */
7624 static
7625 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7626                 struct rpc_xprt *xprt,
7627                 struct nfs_client *clp,
7628                 struct rpc_cred *cred)
7629 {
7630         int status;
7631         struct nfs41_bind_conn_to_session_args args = {
7632                 .client = clp,
7633                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7634         };
7635         struct nfs41_bind_conn_to_session_res res;
7636         struct rpc_message msg = {
7637                 .rpc_proc =
7638                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7639                 .rpc_argp = &args,
7640                 .rpc_resp = &res,
7641                 .rpc_cred = cred,
7642         };
7643         struct rpc_task_setup task_setup_data = {
7644                 .rpc_client = clnt,
7645                 .rpc_xprt = xprt,
7646                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7647                 .rpc_message = &msg,
7648                 .flags = RPC_TASK_TIMEOUT,
7649         };
7650         struct rpc_task *task;
7651
7652         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7653         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7654                 args.dir = NFS4_CDFC4_FORE;
7655
7656         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7657         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7658                 args.dir = NFS4_CDFC4_FORE;
7659
7660         task = rpc_run_task(&task_setup_data);
7661         if (!IS_ERR(task)) {
7662                 status = task->tk_status;
7663                 rpc_put_task(task);
7664         } else
7665                 status = PTR_ERR(task);
7666         trace_nfs4_bind_conn_to_session(clp, status);
7667         if (status == 0) {
7668                 if (memcmp(res.sessionid.data,
7669                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7670                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7671                         return -EIO;
7672                 }
7673                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7674                         dprintk("NFS: %s: Unexpected direction from server\n",
7675                                 __func__);
7676                         return -EIO;
7677                 }
7678                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7679                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7680                                 __func__);
7681                         return -EIO;
7682                 }
7683         }
7684
7685         return status;
7686 }
7687
7688 struct rpc_bind_conn_calldata {
7689         struct nfs_client *clp;
7690         struct rpc_cred *cred;
7691 };
7692
7693 static int
7694 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7695                 struct rpc_xprt *xprt,
7696                 void *calldata)
7697 {
7698         struct rpc_bind_conn_calldata *p = calldata;
7699
7700         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7701 }
7702
7703 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7704 {
7705         struct rpc_bind_conn_calldata data = {
7706                 .clp = clp,
7707                 .cred = cred,
7708         };
7709         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7710                         nfs4_proc_bind_conn_to_session_callback, &data);
7711 }
7712
7713 /*
7714  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7715  * and operations we'd like to see to enable certain features in the allow map
7716  */
7717 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7718         .how = SP4_MACH_CRED,
7719         .enforce.u.words = {
7720                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7721                       1 << (OP_EXCHANGE_ID - 32) |
7722                       1 << (OP_CREATE_SESSION - 32) |
7723                       1 << (OP_DESTROY_SESSION - 32) |
7724                       1 << (OP_DESTROY_CLIENTID - 32)
7725         },
7726         .allow.u.words = {
7727                 [0] = 1 << (OP_CLOSE) |
7728                       1 << (OP_OPEN_DOWNGRADE) |
7729                       1 << (OP_LOCKU) |
7730                       1 << (OP_DELEGRETURN) |
7731                       1 << (OP_COMMIT),
7732                 [1] = 1 << (OP_SECINFO - 32) |
7733                       1 << (OP_SECINFO_NO_NAME - 32) |
7734                       1 << (OP_LAYOUTRETURN - 32) |
7735                       1 << (OP_TEST_STATEID - 32) |
7736                       1 << (OP_FREE_STATEID - 32) |
7737                       1 << (OP_WRITE - 32)
7738         }
7739 };
7740
7741 /*
7742  * Select the state protection mode for client `clp' given the server results
7743  * from exchange_id in `sp'.
7744  *
7745  * Returns 0 on success, negative errno otherwise.
7746  */
7747 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7748                                  struct nfs41_state_protection *sp)
7749 {
7750         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7751                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7752                       1 << (OP_EXCHANGE_ID - 32) |
7753                       1 << (OP_CREATE_SESSION - 32) |
7754                       1 << (OP_DESTROY_SESSION - 32) |
7755                       1 << (OP_DESTROY_CLIENTID - 32)
7756         };
7757         unsigned long flags = 0;
7758         unsigned int i;
7759         int ret = 0;
7760
7761         if (sp->how == SP4_MACH_CRED) {
7762                 /* Print state protect result */
7763                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7764                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7765                         if (test_bit(i, sp->enforce.u.longs))
7766                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7767                         if (test_bit(i, sp->allow.u.longs))
7768                                 dfprintk(MOUNT, "  allow op %d\n", i);
7769                 }
7770
7771                 /* make sure nothing is on enforce list that isn't supported */
7772                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7773                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7774                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7775                                 ret = -EINVAL;
7776                                 goto out;
7777                         }
7778                 }
7779
7780                 /*
7781                  * Minimal mode - state operations are allowed to use machine
7782                  * credential.  Note this already happens by default, so the
7783                  * client doesn't have to do anything more than the negotiation.
7784                  *
7785                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7786                  *       we're already using the machine cred for exchange_id
7787                  *       and will never use a different cred.
7788                  */
7789                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7790                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7791                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7792                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7793                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7794                         dfprintk(MOUNT, "  minimal mode enabled\n");
7795                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
7796                 } else {
7797                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7798                         ret = -EINVAL;
7799                         goto out;
7800                 }
7801
7802                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7803                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7804                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7805                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7806                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7807                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
7808                 }
7809
7810                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7811                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7812                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
7813                 }
7814
7815                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7816                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7817                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7818                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
7819                 }
7820
7821                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7822                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7823                         dfprintk(MOUNT, "  stateid mode enabled\n");
7824                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
7825                 }
7826
7827                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7828                         dfprintk(MOUNT, "  write mode enabled\n");
7829                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
7830                 }
7831
7832                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7833                         dfprintk(MOUNT, "  commit mode enabled\n");
7834                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
7835                 }
7836         }
7837 out:
7838         clp->cl_sp4_flags = flags;
7839         return ret;
7840 }
7841
7842 struct nfs41_exchange_id_data {
7843         struct nfs41_exchange_id_res res;
7844         struct nfs41_exchange_id_args args;
7845 };
7846
7847 static void nfs4_exchange_id_release(void *data)
7848 {
7849         struct nfs41_exchange_id_data *cdata =
7850                                         (struct nfs41_exchange_id_data *)data;
7851
7852         nfs_put_client(cdata->args.client);
7853         kfree(cdata->res.impl_id);
7854         kfree(cdata->res.server_scope);
7855         kfree(cdata->res.server_owner);
7856         kfree(cdata);
7857 }
7858
7859 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7860         .rpc_release = nfs4_exchange_id_release,
7861 };
7862
7863 /*
7864  * _nfs4_proc_exchange_id()
7865  *
7866  * Wrapper for EXCHANGE_ID operation.
7867  */
7868 static struct rpc_task *
7869 nfs4_run_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7870                         u32 sp4_how, struct rpc_xprt *xprt)
7871 {
7872         struct rpc_message msg = {
7873                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7874                 .rpc_cred = cred,
7875         };
7876         struct rpc_task_setup task_setup_data = {
7877                 .rpc_client = clp->cl_rpcclient,
7878                 .callback_ops = &nfs4_exchange_id_call_ops,
7879                 .rpc_message = &msg,
7880                 .flags = RPC_TASK_TIMEOUT,
7881         };
7882         struct nfs41_exchange_id_data *calldata;
7883         int status;
7884
7885         if (!refcount_inc_not_zero(&clp->cl_count))
7886                 return ERR_PTR(-EIO);
7887
7888         status = -ENOMEM;
7889         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7890         if (!calldata)
7891                 goto out;
7892
7893         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
7894
7895         status = nfs4_init_uniform_client_string(clp);
7896         if (status)
7897                 goto out_calldata;
7898
7899         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7900                                                 GFP_NOFS);
7901         status = -ENOMEM;
7902         if (unlikely(calldata->res.server_owner == NULL))
7903                 goto out_calldata;
7904
7905         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7906                                         GFP_NOFS);
7907         if (unlikely(calldata->res.server_scope == NULL))
7908                 goto out_server_owner;
7909
7910         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7911         if (unlikely(calldata->res.impl_id == NULL))
7912                 goto out_server_scope;
7913
7914         switch (sp4_how) {
7915         case SP4_NONE:
7916                 calldata->args.state_protect.how = SP4_NONE;
7917                 break;
7918
7919         case SP4_MACH_CRED:
7920                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7921                 break;
7922
7923         default:
7924                 /* unsupported! */
7925                 WARN_ON_ONCE(1);
7926                 status = -EINVAL;
7927                 goto out_impl_id;
7928         }
7929         if (xprt) {
7930                 task_setup_data.rpc_xprt = xprt;
7931                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
7932                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
7933                                 sizeof(calldata->args.verifier.data));
7934         }
7935         calldata->args.client = clp;
7936         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7937         EXCHGID4_FLAG_BIND_PRINC_STATEID;
7938 #ifdef CONFIG_NFS_V4_1_MIGRATION
7939         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
7940 #endif
7941         msg.rpc_argp = &calldata->args;
7942         msg.rpc_resp = &calldata->res;
7943         task_setup_data.callback_data = calldata;
7944
7945         return rpc_run_task(&task_setup_data);
7946
7947 out_impl_id:
7948         kfree(calldata->res.impl_id);
7949 out_server_scope:
7950         kfree(calldata->res.server_scope);
7951 out_server_owner:
7952         kfree(calldata->res.server_owner);
7953 out_calldata:
7954         kfree(calldata);
7955 out:
7956         nfs_put_client(clp);
7957         return ERR_PTR(status);
7958 }
7959
7960 /*
7961  * _nfs4_proc_exchange_id()
7962  *
7963  * Wrapper for EXCHANGE_ID operation.
7964  */
7965 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7966                         u32 sp4_how)
7967 {
7968         struct rpc_task *task;
7969         struct nfs41_exchange_id_args *argp;
7970         struct nfs41_exchange_id_res *resp;
7971         int status;
7972
7973         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
7974         if (IS_ERR(task))
7975                 return PTR_ERR(task);
7976
7977         argp = task->tk_msg.rpc_argp;
7978         resp = task->tk_msg.rpc_resp;
7979         status = task->tk_status;
7980         if (status  != 0)
7981                 goto out;
7982
7983         status = nfs4_check_cl_exchange_flags(resp->flags);
7984         if (status  != 0)
7985                 goto out;
7986
7987         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
7988         if (status != 0)
7989                 goto out;
7990
7991         clp->cl_clientid = resp->clientid;
7992         clp->cl_exchange_flags = resp->flags;
7993         clp->cl_seqid = resp->seqid;
7994         /* Client ID is not confirmed */
7995         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
7996                 clear_bit(NFS4_SESSION_ESTABLISHED,
7997                           &clp->cl_session->session_state);
7998
7999         if (clp->cl_serverscope != NULL &&
8000             !nfs41_same_server_scope(clp->cl_serverscope,
8001                                 resp->server_scope)) {
8002                 dprintk("%s: server_scope mismatch detected\n",
8003                         __func__);
8004                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8005         }
8006
8007         swap(clp->cl_serverowner, resp->server_owner);
8008         swap(clp->cl_serverscope, resp->server_scope);
8009         swap(clp->cl_implid, resp->impl_id);
8010
8011         /* Save the EXCHANGE_ID verifier session trunk tests */
8012         memcpy(clp->cl_confirm.data, argp->verifier.data,
8013                sizeof(clp->cl_confirm.data));
8014 out:
8015         trace_nfs4_exchange_id(clp, status);
8016         rpc_put_task(task);
8017         return status;
8018 }
8019
8020 /*
8021  * nfs4_proc_exchange_id()
8022  *
8023  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8024  *
8025  * Since the clientid has expired, all compounds using sessions
8026  * associated with the stale clientid will be returning
8027  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8028  * be in some phase of session reset.
8029  *
8030  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8031  */
8032 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
8033 {
8034         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8035         int status;
8036
8037         /* try SP4_MACH_CRED if krb5i/p */
8038         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8039             authflavor == RPC_AUTH_GSS_KRB5P) {
8040                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8041                 if (!status)
8042                         return 0;
8043         }
8044
8045         /* try SP4_NONE */
8046         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8047 }
8048
8049 /**
8050  * nfs4_test_session_trunk
8051  *
8052  * This is an add_xprt_test() test function called from
8053  * rpc_clnt_setup_test_and_add_xprt.
8054  *
8055  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8056  * and is dereferrenced in nfs4_exchange_id_release
8057  *
8058  * Upon success, add the new transport to the rpc_clnt
8059  *
8060  * @clnt: struct rpc_clnt to get new transport
8061  * @xprt: the rpc_xprt to test
8062  * @data: call data for _nfs4_proc_exchange_id.
8063  */
8064 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8065                             void *data)
8066 {
8067         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8068         struct rpc_task *task;
8069         int status;
8070
8071         u32 sp4_how;
8072
8073         dprintk("--> %s try %s\n", __func__,
8074                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8075
8076         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8077
8078         /* Test connection for session trunking. Async exchange_id call */
8079         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8080         if (IS_ERR(task))
8081                 return PTR_ERR(task);
8082
8083         status = task->tk_status;
8084         if (status == 0)
8085                 status = nfs4_detect_session_trunking(adata->clp,
8086                                 task->tk_msg.rpc_resp, xprt);
8087
8088         rpc_put_task(task);
8089         return status;
8090 }
8091 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8092
8093 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8094                 struct rpc_cred *cred)
8095 {
8096         struct rpc_message msg = {
8097                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8098                 .rpc_argp = clp,
8099                 .rpc_cred = cred,
8100         };
8101         int status;
8102
8103         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8104         trace_nfs4_destroy_clientid(clp, status);
8105         if (status)
8106                 dprintk("NFS: Got error %d from the server %s on "
8107                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8108         return status;
8109 }
8110
8111 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8112                 struct rpc_cred *cred)
8113 {
8114         unsigned int loop;
8115         int ret;
8116
8117         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8118                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8119                 switch (ret) {
8120                 case -NFS4ERR_DELAY:
8121                 case -NFS4ERR_CLIENTID_BUSY:
8122                         ssleep(1);
8123                         break;
8124                 default:
8125                         return ret;
8126                 }
8127         }
8128         return 0;
8129 }
8130
8131 int nfs4_destroy_clientid(struct nfs_client *clp)
8132 {
8133         struct rpc_cred *cred;
8134         int ret = 0;
8135
8136         if (clp->cl_mvops->minor_version < 1)
8137                 goto out;
8138         if (clp->cl_exchange_flags == 0)
8139                 goto out;
8140         if (clp->cl_preserve_clid)
8141                 goto out;
8142         cred = nfs4_get_clid_cred(clp);
8143         ret = nfs4_proc_destroy_clientid(clp, cred);
8144         if (cred)
8145                 put_rpccred(cred);
8146         switch (ret) {
8147         case 0:
8148         case -NFS4ERR_STALE_CLIENTID:
8149                 clp->cl_exchange_flags = 0;
8150         }
8151 out:
8152         return ret;
8153 }
8154
8155 struct nfs4_get_lease_time_data {
8156         struct nfs4_get_lease_time_args *args;
8157         struct nfs4_get_lease_time_res *res;
8158         struct nfs_client *clp;
8159 };
8160
8161 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8162                                         void *calldata)
8163 {
8164         struct nfs4_get_lease_time_data *data =
8165                         (struct nfs4_get_lease_time_data *)calldata;
8166
8167         dprintk("--> %s\n", __func__);
8168         /* just setup sequence, do not trigger session recovery
8169            since we're invoked within one */
8170         nfs4_setup_sequence(data->clp,
8171                         &data->args->la_seq_args,
8172                         &data->res->lr_seq_res,
8173                         task);
8174         dprintk("<-- %s\n", __func__);
8175 }
8176
8177 /*
8178  * Called from nfs4_state_manager thread for session setup, so don't recover
8179  * from sequence operation or clientid errors.
8180  */
8181 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8182 {
8183         struct nfs4_get_lease_time_data *data =
8184                         (struct nfs4_get_lease_time_data *)calldata;
8185
8186         dprintk("--> %s\n", __func__);
8187         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
8188                 return;
8189         switch (task->tk_status) {
8190         case -NFS4ERR_DELAY:
8191         case -NFS4ERR_GRACE:
8192                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8193                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8194                 task->tk_status = 0;
8195                 /* fall through */
8196         case -NFS4ERR_RETRY_UNCACHED_REP:
8197                 rpc_restart_call_prepare(task);
8198                 return;
8199         }
8200         dprintk("<-- %s\n", __func__);
8201 }
8202
8203 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8204         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8205         .rpc_call_done = nfs4_get_lease_time_done,
8206 };
8207
8208 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8209 {
8210         struct rpc_task *task;
8211         struct nfs4_get_lease_time_args args;
8212         struct nfs4_get_lease_time_res res = {
8213                 .lr_fsinfo = fsinfo,
8214         };
8215         struct nfs4_get_lease_time_data data = {
8216                 .args = &args,
8217                 .res = &res,
8218                 .clp = clp,
8219         };
8220         struct rpc_message msg = {
8221                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8222                 .rpc_argp = &args,
8223                 .rpc_resp = &res,
8224         };
8225         struct rpc_task_setup task_setup = {
8226                 .rpc_client = clp->cl_rpcclient,
8227                 .rpc_message = &msg,
8228                 .callback_ops = &nfs4_get_lease_time_ops,
8229                 .callback_data = &data,
8230                 .flags = RPC_TASK_TIMEOUT,
8231         };
8232         int status;
8233
8234         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8235         task = rpc_run_task(&task_setup);
8236
8237         if (IS_ERR(task))
8238                 return PTR_ERR(task);
8239
8240         status = task->tk_status;
8241         rpc_put_task(task);
8242         return status;
8243 }
8244
8245 /*
8246  * Initialize the values to be used by the client in CREATE_SESSION
8247  * If nfs4_init_session set the fore channel request and response sizes,
8248  * use them.
8249  *
8250  * Set the back channel max_resp_sz_cached to zero to force the client to
8251  * always set csa_cachethis to FALSE because the current implementation
8252  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8253  */
8254 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8255                                     struct rpc_clnt *clnt)
8256 {
8257         unsigned int max_rqst_sz, max_resp_sz;
8258         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8259
8260         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8261         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8262
8263         /* Fore channel attributes */
8264         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8265         args->fc_attrs.max_resp_sz = max_resp_sz;
8266         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8267         args->fc_attrs.max_reqs = max_session_slots;
8268
8269         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8270                 "max_ops=%u max_reqs=%u\n",
8271                 __func__,
8272                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8273                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8274
8275         /* Back channel attributes */
8276         args->bc_attrs.max_rqst_sz = max_bc_payload;
8277         args->bc_attrs.max_resp_sz = max_bc_payload;
8278         args->bc_attrs.max_resp_sz_cached = 0;
8279         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8280         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8281
8282         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8283                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8284                 __func__,
8285                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8286                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8287                 args->bc_attrs.max_reqs);
8288 }
8289
8290 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8291                 struct nfs41_create_session_res *res)
8292 {
8293         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8294         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8295
8296         if (rcvd->max_resp_sz > sent->max_resp_sz)
8297                 return -EINVAL;
8298         /*
8299          * Our requested max_ops is the minimum we need; we're not
8300          * prepared to break up compounds into smaller pieces than that.
8301          * So, no point even trying to continue if the server won't
8302          * cooperate:
8303          */
8304         if (rcvd->max_ops < sent->max_ops)
8305                 return -EINVAL;
8306         if (rcvd->max_reqs == 0)
8307                 return -EINVAL;
8308         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8309                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8310         return 0;
8311 }
8312
8313 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8314                 struct nfs41_create_session_res *res)
8315 {
8316         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8317         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8318
8319         if (!(res->flags & SESSION4_BACK_CHAN))
8320                 goto out;
8321         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8322                 return -EINVAL;
8323         if (rcvd->max_resp_sz < sent->max_resp_sz)
8324                 return -EINVAL;
8325         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8326                 return -EINVAL;
8327         if (rcvd->max_ops > sent->max_ops)
8328                 return -EINVAL;
8329         if (rcvd->max_reqs > sent->max_reqs)
8330                 return -EINVAL;
8331 out:
8332         return 0;
8333 }
8334
8335 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8336                                      struct nfs41_create_session_res *res)
8337 {
8338         int ret;
8339
8340         ret = nfs4_verify_fore_channel_attrs(args, res);
8341         if (ret)
8342                 return ret;
8343         return nfs4_verify_back_channel_attrs(args, res);
8344 }
8345
8346 static void nfs4_update_session(struct nfs4_session *session,
8347                 struct nfs41_create_session_res *res)
8348 {
8349         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8350         /* Mark client id and session as being confirmed */
8351         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8352         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8353         session->flags = res->flags;
8354         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8355         if (res->flags & SESSION4_BACK_CHAN)
8356                 memcpy(&session->bc_attrs, &res->bc_attrs,
8357                                 sizeof(session->bc_attrs));
8358 }
8359
8360 static int _nfs4_proc_create_session(struct nfs_client *clp,
8361                 struct rpc_cred *cred)
8362 {
8363         struct nfs4_session *session = clp->cl_session;
8364         struct nfs41_create_session_args args = {
8365                 .client = clp,
8366                 .clientid = clp->cl_clientid,
8367                 .seqid = clp->cl_seqid,
8368                 .cb_program = NFS4_CALLBACK,
8369         };
8370         struct nfs41_create_session_res res;
8371
8372         struct rpc_message msg = {
8373                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8374                 .rpc_argp = &args,
8375                 .rpc_resp = &res,
8376                 .rpc_cred = cred,
8377         };
8378         int status;
8379
8380         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8381         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8382
8383         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8384         trace_nfs4_create_session(clp, status);
8385
8386         switch (status) {
8387         case -NFS4ERR_STALE_CLIENTID:
8388         case -NFS4ERR_DELAY:
8389         case -ETIMEDOUT:
8390         case -EACCES:
8391         case -EAGAIN:
8392                 goto out;
8393         };
8394
8395         clp->cl_seqid++;
8396         if (!status) {
8397                 /* Verify the session's negotiated channel_attrs values */
8398                 status = nfs4_verify_channel_attrs(&args, &res);
8399                 /* Increment the clientid slot sequence id */
8400                 if (status)
8401                         goto out;
8402                 nfs4_update_session(session, &res);
8403         }
8404 out:
8405         return status;
8406 }
8407
8408 /*
8409  * Issues a CREATE_SESSION operation to the server.
8410  * It is the responsibility of the caller to verify the session is
8411  * expired before calling this routine.
8412  */
8413 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
8414 {
8415         int status;
8416         unsigned *ptr;
8417         struct nfs4_session *session = clp->cl_session;
8418
8419         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8420
8421         status = _nfs4_proc_create_session(clp, cred);
8422         if (status)
8423                 goto out;
8424
8425         /* Init or reset the session slot tables */
8426         status = nfs4_setup_session_slot_tables(session);
8427         dprintk("slot table setup returned %d\n", status);
8428         if (status)
8429                 goto out;
8430
8431         ptr = (unsigned *)&session->sess_id.data[0];
8432         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8433                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8434 out:
8435         dprintk("<-- %s\n", __func__);
8436         return status;
8437 }
8438
8439 /*
8440  * Issue the over-the-wire RPC DESTROY_SESSION.
8441  * The caller must serialize access to this routine.
8442  */
8443 int nfs4_proc_destroy_session(struct nfs4_session *session,
8444                 struct rpc_cred *cred)
8445 {
8446         struct rpc_message msg = {
8447                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8448                 .rpc_argp = session,
8449                 .rpc_cred = cred,
8450         };
8451         int status = 0;
8452
8453         dprintk("--> nfs4_proc_destroy_session\n");
8454
8455         /* session is still being setup */
8456         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8457                 return 0;
8458
8459         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8460         trace_nfs4_destroy_session(session->clp, status);
8461
8462         if (status)
8463                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8464                         "Session has been destroyed regardless...\n", status);
8465
8466         dprintk("<-- nfs4_proc_destroy_session\n");
8467         return status;
8468 }
8469
8470 /*
8471  * Renew the cl_session lease.
8472  */
8473 struct nfs4_sequence_data {
8474         struct nfs_client *clp;
8475         struct nfs4_sequence_args args;
8476         struct nfs4_sequence_res res;
8477 };
8478
8479 static void nfs41_sequence_release(void *data)
8480 {
8481         struct nfs4_sequence_data *calldata = data;
8482         struct nfs_client *clp = calldata->clp;
8483
8484         if (refcount_read(&clp->cl_count) > 1)
8485                 nfs4_schedule_state_renewal(clp);
8486         nfs_put_client(clp);
8487         kfree(calldata);
8488 }
8489
8490 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8491 {
8492         switch(task->tk_status) {
8493         case -NFS4ERR_DELAY:
8494                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8495                 return -EAGAIN;
8496         default:
8497                 nfs4_schedule_lease_recovery(clp);
8498         }
8499         return 0;
8500 }
8501
8502 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8503 {
8504         struct nfs4_sequence_data *calldata = data;
8505         struct nfs_client *clp = calldata->clp;
8506
8507         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8508                 return;
8509
8510         trace_nfs4_sequence(clp, task->tk_status);
8511         if (task->tk_status < 0) {
8512                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8513                 if (refcount_read(&clp->cl_count) == 1)
8514                         goto out;
8515
8516                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8517                         rpc_restart_call_prepare(task);
8518                         return;
8519                 }
8520         }
8521         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8522 out:
8523         dprintk("<-- %s\n", __func__);
8524 }
8525
8526 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8527 {
8528         struct nfs4_sequence_data *calldata = data;
8529         struct nfs_client *clp = calldata->clp;
8530         struct nfs4_sequence_args *args;
8531         struct nfs4_sequence_res *res;
8532
8533         args = task->tk_msg.rpc_argp;
8534         res = task->tk_msg.rpc_resp;
8535
8536         nfs4_setup_sequence(clp, args, res, task);
8537 }
8538
8539 static const struct rpc_call_ops nfs41_sequence_ops = {
8540         .rpc_call_done = nfs41_sequence_call_done,
8541         .rpc_call_prepare = nfs41_sequence_prepare,
8542         .rpc_release = nfs41_sequence_release,
8543 };
8544
8545 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8546                 struct rpc_cred *cred,
8547                 struct nfs4_slot *slot,
8548                 bool is_privileged)
8549 {
8550         struct nfs4_sequence_data *calldata;
8551         struct rpc_message msg = {
8552                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8553                 .rpc_cred = cred,
8554         };
8555         struct rpc_task_setup task_setup_data = {
8556                 .rpc_client = clp->cl_rpcclient,
8557                 .rpc_message = &msg,
8558                 .callback_ops = &nfs41_sequence_ops,
8559                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8560         };
8561         struct rpc_task *ret;
8562
8563         ret = ERR_PTR(-EIO);
8564         if (!refcount_inc_not_zero(&clp->cl_count))
8565                 goto out_err;
8566
8567         ret = ERR_PTR(-ENOMEM);
8568         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8569         if (calldata == NULL)
8570                 goto out_put_clp;
8571         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8572         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8573         msg.rpc_argp = &calldata->args;
8574         msg.rpc_resp = &calldata->res;
8575         calldata->clp = clp;
8576         task_setup_data.callback_data = calldata;
8577
8578         ret = rpc_run_task(&task_setup_data);
8579         if (IS_ERR(ret))
8580                 goto out_err;
8581         return ret;
8582 out_put_clp:
8583         nfs_put_client(clp);
8584 out_err:
8585         nfs41_release_slot(slot);
8586         return ret;
8587 }
8588
8589 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8590 {
8591         struct rpc_task *task;
8592         int ret = 0;
8593
8594         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8595                 return -EAGAIN;
8596         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8597         if (IS_ERR(task))
8598                 ret = PTR_ERR(task);
8599         else
8600                 rpc_put_task_async(task);
8601         dprintk("<-- %s status=%d\n", __func__, ret);
8602         return ret;
8603 }
8604
8605 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8606 {
8607         struct rpc_task *task;
8608         int ret;
8609
8610         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8611         if (IS_ERR(task)) {
8612                 ret = PTR_ERR(task);
8613                 goto out;
8614         }
8615         ret = rpc_wait_for_completion_task(task);
8616         if (!ret)
8617                 ret = task->tk_status;
8618         rpc_put_task(task);
8619 out:
8620         dprintk("<-- %s status=%d\n", __func__, ret);
8621         return ret;
8622 }
8623
8624 struct nfs4_reclaim_complete_data {
8625         struct nfs_client *clp;
8626         struct nfs41_reclaim_complete_args arg;
8627         struct nfs41_reclaim_complete_res res;
8628 };
8629
8630 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8631 {
8632         struct nfs4_reclaim_complete_data *calldata = data;
8633
8634         nfs4_setup_sequence(calldata->clp,
8635                         &calldata->arg.seq_args,
8636                         &calldata->res.seq_res,
8637                         task);
8638 }
8639
8640 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8641 {
8642         switch(task->tk_status) {
8643         case 0:
8644                 wake_up_all(&clp->cl_lock_waitq);
8645                 /* Fallthrough */
8646         case -NFS4ERR_COMPLETE_ALREADY:
8647         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8648                 break;
8649         case -NFS4ERR_DELAY:
8650                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8651                 /* fall through */
8652         case -NFS4ERR_RETRY_UNCACHED_REP:
8653                 return -EAGAIN;
8654         case -NFS4ERR_BADSESSION:
8655         case -NFS4ERR_DEADSESSION:
8656         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8657                 nfs4_schedule_session_recovery(clp->cl_session,
8658                                 task->tk_status);
8659                 break;
8660         default:
8661                 nfs4_schedule_lease_recovery(clp);
8662         }
8663         return 0;
8664 }
8665
8666 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8667 {
8668         struct nfs4_reclaim_complete_data *calldata = data;
8669         struct nfs_client *clp = calldata->clp;
8670         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8671
8672         dprintk("--> %s\n", __func__);
8673         if (!nfs41_sequence_done(task, res))
8674                 return;
8675
8676         trace_nfs4_reclaim_complete(clp, task->tk_status);
8677         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8678                 rpc_restart_call_prepare(task);
8679                 return;
8680         }
8681         dprintk("<-- %s\n", __func__);
8682 }
8683
8684 static void nfs4_free_reclaim_complete_data(void *data)
8685 {
8686         struct nfs4_reclaim_complete_data *calldata = data;
8687
8688         kfree(calldata);
8689 }
8690
8691 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8692         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8693         .rpc_call_done = nfs4_reclaim_complete_done,
8694         .rpc_release = nfs4_free_reclaim_complete_data,
8695 };
8696
8697 /*
8698  * Issue a global reclaim complete.
8699  */
8700 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8701                 struct rpc_cred *cred)
8702 {
8703         struct nfs4_reclaim_complete_data *calldata;
8704         struct rpc_task *task;
8705         struct rpc_message msg = {
8706                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8707                 .rpc_cred = cred,
8708         };
8709         struct rpc_task_setup task_setup_data = {
8710                 .rpc_client = clp->cl_rpcclient,
8711                 .rpc_message = &msg,
8712                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8713                 .flags = RPC_TASK_ASYNC,
8714         };
8715         int status = -ENOMEM;
8716
8717         dprintk("--> %s\n", __func__);
8718         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8719         if (calldata == NULL)
8720                 goto out;
8721         calldata->clp = clp;
8722         calldata->arg.one_fs = 0;
8723
8724         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
8725         msg.rpc_argp = &calldata->arg;
8726         msg.rpc_resp = &calldata->res;
8727         task_setup_data.callback_data = calldata;
8728         task = rpc_run_task(&task_setup_data);
8729         if (IS_ERR(task)) {
8730                 status = PTR_ERR(task);
8731                 goto out;
8732         }
8733         status = rpc_wait_for_completion_task(task);
8734         if (status == 0)
8735                 status = task->tk_status;
8736         rpc_put_task(task);
8737 out:
8738         dprintk("<-- %s status=%d\n", __func__, status);
8739         return status;
8740 }
8741
8742 static void
8743 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8744 {
8745         struct nfs4_layoutget *lgp = calldata;
8746         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8747
8748         dprintk("--> %s\n", __func__);
8749         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8750                                 &lgp->res.seq_res, task);
8751         dprintk("<-- %s\n", __func__);
8752 }
8753
8754 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8755 {
8756         struct nfs4_layoutget *lgp = calldata;
8757
8758         dprintk("--> %s\n", __func__);
8759         nfs41_sequence_process(task, &lgp->res.seq_res);
8760         dprintk("<-- %s\n", __func__);
8761 }
8762
8763 static int
8764 nfs4_layoutget_handle_exception(struct rpc_task *task,
8765                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8766 {
8767         struct inode *inode = lgp->args.inode;
8768         struct nfs_server *server = NFS_SERVER(inode);
8769         struct pnfs_layout_hdr *lo;
8770         int nfs4err = task->tk_status;
8771         int err, status = 0;
8772         LIST_HEAD(head);
8773
8774         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8775
8776         nfs4_sequence_free_slot(&lgp->res.seq_res);
8777
8778         switch (nfs4err) {
8779         case 0:
8780                 goto out;
8781
8782         /*
8783          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8784          * on the file. set tk_status to -ENODATA to tell upper layer to
8785          * retry go inband.
8786          */
8787         case -NFS4ERR_LAYOUTUNAVAILABLE:
8788                 status = -ENODATA;
8789                 goto out;
8790         /*
8791          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8792          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8793          */
8794         case -NFS4ERR_BADLAYOUT:
8795                 status = -EOVERFLOW;
8796                 goto out;
8797         /*
8798          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8799          * (or clients) writing to the same RAID stripe except when
8800          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8801          *
8802          * Treat it like we would RECALLCONFLICT -- we retry for a little
8803          * while, and then eventually give up.
8804          */
8805         case -NFS4ERR_LAYOUTTRYLATER:
8806                 if (lgp->args.minlength == 0) {
8807                         status = -EOVERFLOW;
8808                         goto out;
8809                 }
8810                 status = -EBUSY;
8811                 break;
8812         case -NFS4ERR_RECALLCONFLICT:
8813                 status = -ERECALLCONFLICT;
8814                 break;
8815         case -NFS4ERR_DELEG_REVOKED:
8816         case -NFS4ERR_ADMIN_REVOKED:
8817         case -NFS4ERR_EXPIRED:
8818         case -NFS4ERR_BAD_STATEID:
8819                 exception->timeout = 0;
8820                 spin_lock(&inode->i_lock);
8821                 lo = NFS_I(inode)->layout;
8822                 /* If the open stateid was bad, then recover it. */
8823                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8824                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
8825                         spin_unlock(&inode->i_lock);
8826                         exception->state = lgp->args.ctx->state;
8827                         exception->stateid = &lgp->args.stateid;
8828                         break;
8829                 }
8830
8831                 /*
8832                  * Mark the bad layout state as invalid, then retry
8833                  */
8834                 pnfs_mark_layout_stateid_invalid(lo, &head);
8835                 spin_unlock(&inode->i_lock);
8836                 nfs_commit_inode(inode, 0);
8837                 pnfs_free_lseg_list(&head);
8838                 status = -EAGAIN;
8839                 goto out;
8840         }
8841
8842         err = nfs4_handle_exception(server, nfs4err, exception);
8843         if (!status) {
8844                 if (exception->retry)
8845                         status = -EAGAIN;
8846                 else
8847                         status = err;
8848         }
8849 out:
8850         dprintk("<-- %s\n", __func__);
8851         return status;
8852 }
8853
8854 size_t max_response_pages(struct nfs_server *server)
8855 {
8856         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8857         return nfs_page_array_len(0, max_resp_sz);
8858 }
8859
8860 static void nfs4_layoutget_release(void *calldata)
8861 {
8862         struct nfs4_layoutget *lgp = calldata;
8863
8864         dprintk("--> %s\n", __func__);
8865         nfs4_sequence_free_slot(&lgp->res.seq_res);
8866         pnfs_layoutget_free(lgp);
8867         dprintk("<-- %s\n", __func__);
8868 }
8869
8870 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8871         .rpc_call_prepare = nfs4_layoutget_prepare,
8872         .rpc_call_done = nfs4_layoutget_done,
8873         .rpc_release = nfs4_layoutget_release,
8874 };
8875
8876 struct pnfs_layout_segment *
8877 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
8878 {
8879         struct inode *inode = lgp->args.inode;
8880         struct nfs_server *server = NFS_SERVER(inode);
8881         struct rpc_task *task;
8882         struct rpc_message msg = {
8883                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8884                 .rpc_argp = &lgp->args,
8885                 .rpc_resp = &lgp->res,
8886                 .rpc_cred = lgp->cred,
8887         };
8888         struct rpc_task_setup task_setup_data = {
8889                 .rpc_client = server->client,
8890                 .rpc_message = &msg,
8891                 .callback_ops = &nfs4_layoutget_call_ops,
8892                 .callback_data = lgp,
8893                 .flags = RPC_TASK_ASYNC,
8894         };
8895         struct pnfs_layout_segment *lseg = NULL;
8896         struct nfs4_exception exception = {
8897                 .inode = inode,
8898                 .timeout = *timeout,
8899         };
8900         int status = 0;
8901
8902         dprintk("--> %s\n", __func__);
8903
8904         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8905         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8906
8907         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
8908
8909         task = rpc_run_task(&task_setup_data);
8910         if (IS_ERR(task))
8911                 return ERR_CAST(task);
8912         status = rpc_wait_for_completion_task(task);
8913         if (status != 0)
8914                 goto out;
8915
8916         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8917         if (task->tk_status < 0 || lgp->res.layoutp->len == 0) {
8918                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8919                 *timeout = exception.timeout;
8920         } else
8921                 lseg = pnfs_layout_process(lgp);
8922 out:
8923         trace_nfs4_layoutget(lgp->args.ctx,
8924                         &lgp->args.range,
8925                         &lgp->res.range,
8926                         &lgp->res.stateid,
8927                         status);
8928
8929         rpc_put_task(task);
8930         dprintk("<-- %s status=%d\n", __func__, status);
8931         if (status)
8932                 return ERR_PTR(status);
8933         return lseg;
8934 }
8935
8936 static void
8937 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8938 {
8939         struct nfs4_layoutreturn *lrp = calldata;
8940
8941         dprintk("--> %s\n", __func__);
8942         nfs4_setup_sequence(lrp->clp,
8943                         &lrp->args.seq_args,
8944                         &lrp->res.seq_res,
8945                         task);
8946         if (!pnfs_layout_is_valid(lrp->args.layout))
8947                 rpc_exit(task, 0);
8948 }
8949
8950 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8951 {
8952         struct nfs4_layoutreturn *lrp = calldata;
8953         struct nfs_server *server;
8954
8955         dprintk("--> %s\n", __func__);
8956
8957         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8958                 return;
8959
8960         server = NFS_SERVER(lrp->args.inode);
8961         switch (task->tk_status) {
8962         case -NFS4ERR_OLD_STATEID:
8963                 if (nfs4_layoutreturn_refresh_stateid(&lrp->args.stateid,
8964                                         &lrp->args.range,
8965                                         lrp->args.inode))
8966                         goto out_restart;
8967                 /* Fallthrough */
8968         default:
8969                 task->tk_status = 0;
8970                 /* Fallthrough */
8971         case 0:
8972                 break;
8973         case -NFS4ERR_DELAY:
8974                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8975                         break;
8976                 goto out_restart;
8977         }
8978         dprintk("<-- %s\n", __func__);
8979         return;
8980 out_restart:
8981         task->tk_status = 0;
8982         nfs4_sequence_free_slot(&lrp->res.seq_res);
8983         rpc_restart_call_prepare(task);
8984 }
8985
8986 static void nfs4_layoutreturn_release(void *calldata)
8987 {
8988         struct nfs4_layoutreturn *lrp = calldata;
8989         struct pnfs_layout_hdr *lo = lrp->args.layout;
8990
8991         dprintk("--> %s\n", __func__);
8992         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8993                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8994         nfs4_sequence_free_slot(&lrp->res.seq_res);
8995         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8996                 lrp->ld_private.ops->free(&lrp->ld_private);
8997         pnfs_put_layout_hdr(lrp->args.layout);
8998         nfs_iput_and_deactive(lrp->inode);
8999         kfree(calldata);
9000         dprintk("<-- %s\n", __func__);
9001 }
9002
9003 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9004         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9005         .rpc_call_done = nfs4_layoutreturn_done,
9006         .rpc_release = nfs4_layoutreturn_release,
9007 };
9008
9009 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9010 {
9011         struct rpc_task *task;
9012         struct rpc_message msg = {
9013                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9014                 .rpc_argp = &lrp->args,
9015                 .rpc_resp = &lrp->res,
9016                 .rpc_cred = lrp->cred,
9017         };
9018         struct rpc_task_setup task_setup_data = {
9019                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9020                 .rpc_message = &msg,
9021                 .callback_ops = &nfs4_layoutreturn_call_ops,
9022                 .callback_data = lrp,
9023         };
9024         int status = 0;
9025
9026         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9027                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9028                         &task_setup_data.rpc_client, &msg);
9029
9030         dprintk("--> %s\n", __func__);
9031         if (!sync) {
9032                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9033                 if (!lrp->inode) {
9034                         nfs4_layoutreturn_release(lrp);
9035                         return -EAGAIN;
9036                 }
9037                 task_setup_data.flags |= RPC_TASK_ASYNC;
9038         }
9039         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9040         task = rpc_run_task(&task_setup_data);
9041         if (IS_ERR(task))
9042                 return PTR_ERR(task);
9043         if (sync)
9044                 status = task->tk_status;
9045         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9046         dprintk("<-- %s status=%d\n", __func__, status);
9047         rpc_put_task(task);
9048         return status;
9049 }
9050
9051 static int
9052 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9053                 struct pnfs_device *pdev,
9054                 struct rpc_cred *cred)
9055 {
9056         struct nfs4_getdeviceinfo_args args = {
9057                 .pdev = pdev,
9058                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9059                         NOTIFY_DEVICEID4_DELETE,
9060         };
9061         struct nfs4_getdeviceinfo_res res = {
9062                 .pdev = pdev,
9063         };
9064         struct rpc_message msg = {
9065                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9066                 .rpc_argp = &args,
9067                 .rpc_resp = &res,
9068                 .rpc_cred = cred,
9069         };
9070         int status;
9071
9072         dprintk("--> %s\n", __func__);
9073         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9074         if (res.notification & ~args.notify_types)
9075                 dprintk("%s: unsupported notification\n", __func__);
9076         if (res.notification != args.notify_types)
9077                 pdev->nocache = 1;
9078
9079         dprintk("<-- %s status=%d\n", __func__, status);
9080
9081         return status;
9082 }
9083
9084 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9085                 struct pnfs_device *pdev,
9086                 struct rpc_cred *cred)
9087 {
9088         struct nfs4_exception exception = { };
9089         int err;
9090
9091         do {
9092                 err = nfs4_handle_exception(server,
9093                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9094                                         &exception);
9095         } while (exception.retry);
9096         return err;
9097 }
9098 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9099
9100 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9101 {
9102         struct nfs4_layoutcommit_data *data = calldata;
9103         struct nfs_server *server = NFS_SERVER(data->args.inode);
9104
9105         nfs4_setup_sequence(server->nfs_client,
9106                         &data->args.seq_args,
9107                         &data->res.seq_res,
9108                         task);
9109 }
9110
9111 static void
9112 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9113 {
9114         struct nfs4_layoutcommit_data *data = calldata;
9115         struct nfs_server *server = NFS_SERVER(data->args.inode);
9116
9117         if (!nfs41_sequence_done(task, &data->res.seq_res))
9118                 return;
9119
9120         switch (task->tk_status) { /* Just ignore these failures */
9121         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9122         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9123         case -NFS4ERR_BADLAYOUT:     /* no layout */
9124         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9125                 task->tk_status = 0;
9126         case 0:
9127                 break;
9128         default:
9129                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9130                         rpc_restart_call_prepare(task);
9131                         return;
9132                 }
9133         }
9134 }
9135
9136 static void nfs4_layoutcommit_release(void *calldata)
9137 {
9138         struct nfs4_layoutcommit_data *data = calldata;
9139
9140         pnfs_cleanup_layoutcommit(data);
9141         nfs_post_op_update_inode_force_wcc(data->args.inode,
9142                                            data->res.fattr);
9143         put_rpccred(data->cred);
9144         nfs_iput_and_deactive(data->inode);
9145         kfree(data);
9146 }
9147
9148 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9149         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9150         .rpc_call_done = nfs4_layoutcommit_done,
9151         .rpc_release = nfs4_layoutcommit_release,
9152 };
9153
9154 int
9155 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9156 {
9157         struct rpc_message msg = {
9158                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9159                 .rpc_argp = &data->args,
9160                 .rpc_resp = &data->res,
9161                 .rpc_cred = data->cred,
9162         };
9163         struct rpc_task_setup task_setup_data = {
9164                 .task = &data->task,
9165                 .rpc_client = NFS_CLIENT(data->args.inode),
9166                 .rpc_message = &msg,
9167                 .callback_ops = &nfs4_layoutcommit_ops,
9168                 .callback_data = data,
9169         };
9170         struct rpc_task *task;
9171         int status = 0;
9172
9173         dprintk("NFS: initiating layoutcommit call. sync %d "
9174                 "lbw: %llu inode %lu\n", sync,
9175                 data->args.lastbytewritten,
9176                 data->args.inode->i_ino);
9177
9178         if (!sync) {
9179                 data->inode = nfs_igrab_and_active(data->args.inode);
9180                 if (data->inode == NULL) {
9181                         nfs4_layoutcommit_release(data);
9182                         return -EAGAIN;
9183                 }
9184                 task_setup_data.flags = RPC_TASK_ASYNC;
9185         }
9186         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9187         task = rpc_run_task(&task_setup_data);
9188         if (IS_ERR(task))
9189                 return PTR_ERR(task);
9190         if (sync)
9191                 status = task->tk_status;
9192         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9193         dprintk("%s: status %d\n", __func__, status);
9194         rpc_put_task(task);
9195         return status;
9196 }
9197
9198 /**
9199  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9200  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9201  */
9202 static int
9203 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9204                     struct nfs_fsinfo *info,
9205                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9206 {
9207         struct nfs41_secinfo_no_name_args args = {
9208                 .style = SECINFO_STYLE_CURRENT_FH,
9209         };
9210         struct nfs4_secinfo_res res = {
9211                 .flavors = flavors,
9212         };
9213         struct rpc_message msg = {
9214                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9215                 .rpc_argp = &args,
9216                 .rpc_resp = &res,
9217         };
9218         struct rpc_clnt *clnt = server->client;
9219         struct rpc_cred *cred = NULL;
9220         int status;
9221
9222         if (use_integrity) {
9223                 clnt = server->nfs_client->cl_rpcclient;
9224                 cred = nfs4_get_clid_cred(server->nfs_client);
9225                 msg.rpc_cred = cred;
9226         }
9227
9228         dprintk("--> %s\n", __func__);
9229         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
9230                                 &res.seq_res, 0);
9231         dprintk("<-- %s status=%d\n", __func__, status);
9232
9233         if (cred)
9234                 put_rpccred(cred);
9235
9236         return status;
9237 }
9238
9239 static int
9240 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9241                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9242 {
9243         struct nfs4_exception exception = { };
9244         int err;
9245         do {
9246                 /* first try using integrity protection */
9247                 err = -NFS4ERR_WRONGSEC;
9248
9249                 /* try to use integrity protection with machine cred */
9250                 if (_nfs4_is_integrity_protected(server->nfs_client))
9251                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9252                                                           flavors, true);
9253
9254                 /*
9255                  * if unable to use integrity protection, or SECINFO with
9256                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9257                  * disallowed by spec, but exists in deployed servers) use
9258                  * the current filesystem's rpc_client and the user cred.
9259                  */
9260                 if (err == -NFS4ERR_WRONGSEC)
9261                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9262                                                           flavors, false);
9263
9264                 switch (err) {
9265                 case 0:
9266                 case -NFS4ERR_WRONGSEC:
9267                 case -ENOTSUPP:
9268                         goto out;
9269                 default:
9270                         err = nfs4_handle_exception(server, err, &exception);
9271                 }
9272         } while (exception.retry);
9273 out:
9274         return err;
9275 }
9276
9277 static int
9278 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9279                     struct nfs_fsinfo *info)
9280 {
9281         int err;
9282         struct page *page;
9283         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9284         struct nfs4_secinfo_flavors *flavors;
9285         struct nfs4_secinfo4 *secinfo;
9286         int i;
9287
9288         page = alloc_page(GFP_KERNEL);
9289         if (!page) {
9290                 err = -ENOMEM;
9291                 goto out;
9292         }
9293
9294         flavors = page_address(page);
9295         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9296
9297         /*
9298          * Fall back on "guess and check" method if
9299          * the server doesn't support SECINFO_NO_NAME
9300          */
9301         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9302                 err = nfs4_find_root_sec(server, fhandle, info);
9303                 goto out_freepage;
9304         }
9305         if (err)
9306                 goto out_freepage;
9307
9308         for (i = 0; i < flavors->num_flavors; i++) {
9309                 secinfo = &flavors->flavors[i];
9310
9311                 switch (secinfo->flavor) {
9312                 case RPC_AUTH_NULL:
9313                 case RPC_AUTH_UNIX:
9314                 case RPC_AUTH_GSS:
9315                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9316                                         &secinfo->flavor_info);
9317                         break;
9318                 default:
9319                         flavor = RPC_AUTH_MAXFLAVOR;
9320                         break;
9321                 }
9322
9323                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9324                         flavor = RPC_AUTH_MAXFLAVOR;
9325
9326                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9327                         err = nfs4_lookup_root_sec(server, fhandle,
9328                                                    info, flavor);
9329                         if (!err)
9330                                 break;
9331                 }
9332         }
9333
9334         if (flavor == RPC_AUTH_MAXFLAVOR)
9335                 err = -EPERM;
9336
9337 out_freepage:
9338         put_page(page);
9339         if (err == -EACCES)
9340                 return -EPERM;
9341 out:
9342         return err;
9343 }
9344
9345 static int _nfs41_test_stateid(struct nfs_server *server,
9346                 nfs4_stateid *stateid,
9347                 struct rpc_cred *cred)
9348 {
9349         int status;
9350         struct nfs41_test_stateid_args args = {
9351                 .stateid = stateid,
9352         };
9353         struct nfs41_test_stateid_res res;
9354         struct rpc_message msg = {
9355                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9356                 .rpc_argp = &args,
9357                 .rpc_resp = &res,
9358                 .rpc_cred = cred,
9359         };
9360         struct rpc_clnt *rpc_client = server->client;
9361
9362         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9363                 &rpc_client, &msg);
9364
9365         dprintk("NFS call  test_stateid %p\n", stateid);
9366         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9367         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9368                         &args.seq_args, &res.seq_res);
9369         if (status != NFS_OK) {
9370                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9371                 return status;
9372         }
9373         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9374         return -res.status;
9375 }
9376
9377 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9378                 int err, struct nfs4_exception *exception)
9379 {
9380         exception->retry = 0;
9381         switch(err) {
9382         case -NFS4ERR_DELAY:
9383         case -NFS4ERR_RETRY_UNCACHED_REP:
9384                 nfs4_handle_exception(server, err, exception);
9385                 break;
9386         case -NFS4ERR_BADSESSION:
9387         case -NFS4ERR_BADSLOT:
9388         case -NFS4ERR_BAD_HIGH_SLOT:
9389         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9390         case -NFS4ERR_DEADSESSION:
9391                 nfs4_do_handle_exception(server, err, exception);
9392         }
9393 }
9394
9395 /**
9396  * nfs41_test_stateid - perform a TEST_STATEID operation
9397  *
9398  * @server: server / transport on which to perform the operation
9399  * @stateid: state ID to test
9400  * @cred: credential
9401  *
9402  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9403  * Otherwise a negative NFS4ERR value is returned if the operation
9404  * failed or the state ID is not currently valid.
9405  */
9406 static int nfs41_test_stateid(struct nfs_server *server,
9407                 nfs4_stateid *stateid,
9408                 struct rpc_cred *cred)
9409 {
9410         struct nfs4_exception exception = { };
9411         int err;
9412         do {
9413                 err = _nfs41_test_stateid(server, stateid, cred);
9414                 nfs4_handle_delay_or_session_error(server, err, &exception);
9415         } while (exception.retry);
9416         return err;
9417 }
9418
9419 struct nfs_free_stateid_data {
9420         struct nfs_server *server;
9421         struct nfs41_free_stateid_args args;
9422         struct nfs41_free_stateid_res res;
9423 };
9424
9425 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9426 {
9427         struct nfs_free_stateid_data *data = calldata;
9428         nfs4_setup_sequence(data->server->nfs_client,
9429                         &data->args.seq_args,
9430                         &data->res.seq_res,
9431                         task);
9432 }
9433
9434 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9435 {
9436         struct nfs_free_stateid_data *data = calldata;
9437
9438         nfs41_sequence_done(task, &data->res.seq_res);
9439
9440         switch (task->tk_status) {
9441         case -NFS4ERR_DELAY:
9442                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9443                         rpc_restart_call_prepare(task);
9444         }
9445 }
9446
9447 static void nfs41_free_stateid_release(void *calldata)
9448 {
9449         kfree(calldata);
9450 }
9451
9452 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9453         .rpc_call_prepare = nfs41_free_stateid_prepare,
9454         .rpc_call_done = nfs41_free_stateid_done,
9455         .rpc_release = nfs41_free_stateid_release,
9456 };
9457
9458 /**
9459  * nfs41_free_stateid - perform a FREE_STATEID operation
9460  *
9461  * @server: server / transport on which to perform the operation
9462  * @stateid: state ID to release
9463  * @cred: credential
9464  * @is_recovery: set to true if this call needs to be privileged
9465  *
9466  * Note: this function is always asynchronous.
9467  */
9468 static int nfs41_free_stateid(struct nfs_server *server,
9469                 const nfs4_stateid *stateid,
9470                 struct rpc_cred *cred,
9471                 bool privileged)
9472 {
9473         struct rpc_message msg = {
9474                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9475                 .rpc_cred = cred,
9476         };
9477         struct rpc_task_setup task_setup = {
9478                 .rpc_client = server->client,
9479                 .rpc_message = &msg,
9480                 .callback_ops = &nfs41_free_stateid_ops,
9481                 .flags = RPC_TASK_ASYNC,
9482         };
9483         struct nfs_free_stateid_data *data;
9484         struct rpc_task *task;
9485
9486         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9487                 &task_setup.rpc_client, &msg);
9488
9489         dprintk("NFS call  free_stateid %p\n", stateid);
9490         data = kmalloc(sizeof(*data), GFP_NOFS);
9491         if (!data)
9492                 return -ENOMEM;
9493         data->server = server;
9494         nfs4_stateid_copy(&data->args.stateid, stateid);
9495
9496         task_setup.callback_data = data;
9497
9498         msg.rpc_argp = &data->args;
9499         msg.rpc_resp = &data->res;
9500         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9501         task = rpc_run_task(&task_setup);
9502         if (IS_ERR(task))
9503                 return PTR_ERR(task);
9504         rpc_put_task(task);
9505         return 0;
9506 }
9507
9508 static void
9509 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9510 {
9511         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9512
9513         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9514         nfs4_free_lock_state(server, lsp);
9515 }
9516
9517 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9518                 const nfs4_stateid *s2)
9519 {
9520         if (s1->type != s2->type)
9521                 return false;
9522
9523         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9524                 return false;
9525
9526         if (s1->seqid == s2->seqid)
9527                 return true;
9528
9529         return s1->seqid == 0 || s2->seqid == 0;
9530 }
9531
9532 #endif /* CONFIG_NFS_V4_1 */
9533
9534 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9535                 const nfs4_stateid *s2)
9536 {
9537         return nfs4_stateid_match(s1, s2);
9538 }
9539
9540
9541 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9542         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9543         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9544         .recover_open   = nfs4_open_reclaim,
9545         .recover_lock   = nfs4_lock_reclaim,
9546         .establish_clid = nfs4_init_clientid,
9547         .detect_trunking = nfs40_discover_server_trunking,
9548 };
9549
9550 #if defined(CONFIG_NFS_V4_1)
9551 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9552         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9553         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9554         .recover_open   = nfs4_open_reclaim,
9555         .recover_lock   = nfs4_lock_reclaim,
9556         .establish_clid = nfs41_init_clientid,
9557         .reclaim_complete = nfs41_proc_reclaim_complete,
9558         .detect_trunking = nfs41_discover_server_trunking,
9559 };
9560 #endif /* CONFIG_NFS_V4_1 */
9561
9562 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9563         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9564         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9565         .recover_open   = nfs40_open_expired,
9566         .recover_lock   = nfs4_lock_expired,
9567         .establish_clid = nfs4_init_clientid,
9568 };
9569
9570 #if defined(CONFIG_NFS_V4_1)
9571 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9572         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9573         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9574         .recover_open   = nfs41_open_expired,
9575         .recover_lock   = nfs41_lock_expired,
9576         .establish_clid = nfs41_init_clientid,
9577 };
9578 #endif /* CONFIG_NFS_V4_1 */
9579
9580 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9581         .sched_state_renewal = nfs4_proc_async_renew,
9582         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9583         .renew_lease = nfs4_proc_renew,
9584 };
9585
9586 #if defined(CONFIG_NFS_V4_1)
9587 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9588         .sched_state_renewal = nfs41_proc_async_sequence,
9589         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9590         .renew_lease = nfs4_proc_sequence,
9591 };
9592 #endif
9593
9594 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9595         .get_locations = _nfs40_proc_get_locations,
9596         .fsid_present = _nfs40_proc_fsid_present,
9597 };
9598
9599 #if defined(CONFIG_NFS_V4_1)
9600 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9601         .get_locations = _nfs41_proc_get_locations,
9602         .fsid_present = _nfs41_proc_fsid_present,
9603 };
9604 #endif  /* CONFIG_NFS_V4_1 */
9605
9606 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9607         .minor_version = 0,
9608         .init_caps = NFS_CAP_READDIRPLUS
9609                 | NFS_CAP_ATOMIC_OPEN
9610                 | NFS_CAP_POSIX_LOCK,
9611         .init_client = nfs40_init_client,
9612         .shutdown_client = nfs40_shutdown_client,
9613         .match_stateid = nfs4_match_stateid,
9614         .find_root_sec = nfs4_find_root_sec,
9615         .free_lock_state = nfs4_release_lockowner,
9616         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9617         .alloc_seqid = nfs_alloc_seqid,
9618         .call_sync_ops = &nfs40_call_sync_ops,
9619         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9620         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9621         .state_renewal_ops = &nfs40_state_renewal_ops,
9622         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9623 };
9624
9625 #if defined(CONFIG_NFS_V4_1)
9626 static struct nfs_seqid *
9627 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9628 {
9629         return NULL;
9630 }
9631
9632 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9633         .minor_version = 1,
9634         .init_caps = NFS_CAP_READDIRPLUS
9635                 | NFS_CAP_ATOMIC_OPEN
9636                 | NFS_CAP_POSIX_LOCK
9637                 | NFS_CAP_STATEID_NFSV41
9638                 | NFS_CAP_ATOMIC_OPEN_V1
9639                 | NFS_CAP_LGOPEN,
9640         .init_client = nfs41_init_client,
9641         .shutdown_client = nfs41_shutdown_client,
9642         .match_stateid = nfs41_match_stateid,
9643         .find_root_sec = nfs41_find_root_sec,
9644         .free_lock_state = nfs41_free_lock_state,
9645         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9646         .alloc_seqid = nfs_alloc_no_seqid,
9647         .session_trunk = nfs4_test_session_trunk,
9648         .call_sync_ops = &nfs41_call_sync_ops,
9649         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9650         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9651         .state_renewal_ops = &nfs41_state_renewal_ops,
9652         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9653 };
9654 #endif
9655
9656 #if defined(CONFIG_NFS_V4_2)
9657 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9658         .minor_version = 2,
9659         .init_caps = NFS_CAP_READDIRPLUS
9660                 | NFS_CAP_ATOMIC_OPEN
9661                 | NFS_CAP_POSIX_LOCK
9662                 | NFS_CAP_STATEID_NFSV41
9663                 | NFS_CAP_ATOMIC_OPEN_V1
9664                 | NFS_CAP_LGOPEN
9665                 | NFS_CAP_ALLOCATE
9666                 | NFS_CAP_COPY
9667                 | NFS_CAP_OFFLOAD_CANCEL
9668                 | NFS_CAP_DEALLOCATE
9669                 | NFS_CAP_SEEK
9670                 | NFS_CAP_LAYOUTSTATS
9671                 | NFS_CAP_CLONE,
9672         .init_client = nfs41_init_client,
9673         .shutdown_client = nfs41_shutdown_client,
9674         .match_stateid = nfs41_match_stateid,
9675         .find_root_sec = nfs41_find_root_sec,
9676         .free_lock_state = nfs41_free_lock_state,
9677         .call_sync_ops = &nfs41_call_sync_ops,
9678         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9679         .alloc_seqid = nfs_alloc_no_seqid,
9680         .session_trunk = nfs4_test_session_trunk,
9681         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9682         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9683         .state_renewal_ops = &nfs41_state_renewal_ops,
9684         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9685 };
9686 #endif
9687
9688 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9689         [0] = &nfs_v4_0_minor_ops,
9690 #if defined(CONFIG_NFS_V4_1)
9691         [1] = &nfs_v4_1_minor_ops,
9692 #endif
9693 #if defined(CONFIG_NFS_V4_2)
9694         [2] = &nfs_v4_2_minor_ops,
9695 #endif
9696 };
9697
9698 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9699 {
9700         ssize_t error, error2;
9701
9702         error = generic_listxattr(dentry, list, size);
9703         if (error < 0)
9704                 return error;
9705         if (list) {
9706                 list += error;
9707                 size -= error;
9708         }
9709
9710         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9711         if (error2 < 0)
9712                 return error2;
9713         return error + error2;
9714 }
9715
9716 static const struct inode_operations nfs4_dir_inode_operations = {
9717         .create         = nfs_create,
9718         .lookup         = nfs_lookup,
9719         .atomic_open    = nfs_atomic_open,
9720         .link           = nfs_link,
9721         .unlink         = nfs_unlink,
9722         .symlink        = nfs_symlink,
9723         .mkdir          = nfs_mkdir,
9724         .rmdir          = nfs_rmdir,
9725         .mknod          = nfs_mknod,
9726         .rename         = nfs_rename,
9727         .permission     = nfs_permission,
9728         .getattr        = nfs_getattr,
9729         .setattr        = nfs_setattr,
9730         .listxattr      = nfs4_listxattr,
9731 };
9732
9733 static const struct inode_operations nfs4_file_inode_operations = {
9734         .permission     = nfs_permission,
9735         .getattr        = nfs_getattr,
9736         .setattr        = nfs_setattr,
9737         .listxattr      = nfs4_listxattr,
9738 };
9739
9740 const struct nfs_rpc_ops nfs_v4_clientops = {
9741         .version        = 4,                    /* protocol version */
9742         .dentry_ops     = &nfs4_dentry_operations,
9743         .dir_inode_ops  = &nfs4_dir_inode_operations,
9744         .file_inode_ops = &nfs4_file_inode_operations,
9745         .file_ops       = &nfs4_file_operations,
9746         .getroot        = nfs4_proc_get_root,
9747         .submount       = nfs4_submount,
9748         .try_mount      = nfs4_try_mount,
9749         .getattr        = nfs4_proc_getattr,
9750         .setattr        = nfs4_proc_setattr,
9751         .lookup         = nfs4_proc_lookup,
9752         .lookupp        = nfs4_proc_lookupp,
9753         .access         = nfs4_proc_access,
9754         .readlink       = nfs4_proc_readlink,
9755         .create         = nfs4_proc_create,
9756         .remove         = nfs4_proc_remove,
9757         .unlink_setup   = nfs4_proc_unlink_setup,
9758         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9759         .unlink_done    = nfs4_proc_unlink_done,
9760         .rename_setup   = nfs4_proc_rename_setup,
9761         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9762         .rename_done    = nfs4_proc_rename_done,
9763         .link           = nfs4_proc_link,
9764         .symlink        = nfs4_proc_symlink,
9765         .mkdir          = nfs4_proc_mkdir,
9766         .rmdir          = nfs4_proc_rmdir,
9767         .readdir        = nfs4_proc_readdir,
9768         .mknod          = nfs4_proc_mknod,
9769         .statfs         = nfs4_proc_statfs,
9770         .fsinfo         = nfs4_proc_fsinfo,
9771         .pathconf       = nfs4_proc_pathconf,
9772         .set_capabilities = nfs4_server_capabilities,
9773         .decode_dirent  = nfs4_decode_dirent,
9774         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9775         .read_setup     = nfs4_proc_read_setup,
9776         .read_done      = nfs4_read_done,
9777         .write_setup    = nfs4_proc_write_setup,
9778         .write_done     = nfs4_write_done,
9779         .commit_setup   = nfs4_proc_commit_setup,
9780         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9781         .commit_done    = nfs4_commit_done,
9782         .lock           = nfs4_proc_lock,
9783         .clear_acl_cache = nfs4_zap_acl_attr,
9784         .close_context  = nfs4_close_context,
9785         .open_context   = nfs4_atomic_open,
9786         .have_delegation = nfs4_have_delegation,
9787         .alloc_client   = nfs4_alloc_client,
9788         .init_client    = nfs4_init_client,
9789         .free_client    = nfs4_free_client,
9790         .create_server  = nfs4_create_server,
9791         .clone_server   = nfs_clone_server,
9792 };
9793
9794 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9795         .name   = XATTR_NAME_NFSV4_ACL,
9796         .list   = nfs4_xattr_list_nfs4_acl,
9797         .get    = nfs4_xattr_get_nfs4_acl,
9798         .set    = nfs4_xattr_set_nfs4_acl,
9799 };
9800
9801 const struct xattr_handler *nfs4_xattr_handlers[] = {
9802         &nfs4_xattr_nfs4_acl_handler,
9803 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9804         &nfs4_xattr_nfs4_label_handler,
9805 #endif
9806         NULL
9807 };
9808
9809 /*
9810  * Local variables:
9811  *  c-basic-offset: 8
9812  * End:
9813  */