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 *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  95 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  96                             struct nfs_fattr *fattr, struct iattr *sattr,
  97                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  98                             struct nfs4_label *olabel);
  99 #ifdef CONFIG_NFS_V4_1
 100 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 101                 const struct cred *cred,
 102                 struct nfs4_slot *slot,
 103                 bool is_privileged);
 104 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 105                 const struct cred *);
 106 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 107                 const struct cred *, bool);
 108 #endif
 109
 110 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 111 static inline struct nfs4_label *
 112 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 113         struct iattr *sattr, struct nfs4_label *label)
 114 {
 115         int err;
 116
 117         if (label == NULL)
 118                 return NULL;
 119
 120         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 121                 return NULL;
 122
 123         err = security_dentry_init_security(dentry, sattr->ia_mode,
 124                                 &dentry->d_name, (void **)&label->label, &label->len);
 125         if (err == 0)
 126                 return label;
 127
 128         return NULL;
 129 }
 130 static inline void
 131 nfs4_label_release_security(struct nfs4_label *label)
 132 {
 133         if (label)
 134                 security_release_secctx(label->label, label->len);
 135 }
 136 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 137 {
 138         if (label)
 139                 return server->attr_bitmask;
 140
 141         return server->attr_bitmask_nl;
 142 }
 143 #else
 144 static inline struct nfs4_label *
 145 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 146         struct iattr *sattr, struct nfs4_label *l)
 147 { return NULL; }
 148 static inline void
 149 nfs4_label_release_security(struct nfs4_label *label)
 150 { return; }
 151 static inline u32 *
 152 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 153 { return server->attr_bitmask; }
 154 #endif
 155
 156 /* Prevent leaks of NFSv4 errors into userland */
 157 static int nfs4_map_errors(int err)
 158 {
 159         if (err >= -1000)
 160                 return err;
 161         switch (err) {
 162         case -NFS4ERR_RESOURCE:
 163         case -NFS4ERR_LAYOUTTRYLATER:
 164         case -NFS4ERR_RECALLCONFLICT:
 165                 return -EREMOTEIO;
 166         case -NFS4ERR_WRONGSEC:
 167         case -NFS4ERR_WRONG_CRED:
 168                 return -EPERM;
 169         case -NFS4ERR_BADOWNER:
 170         case -NFS4ERR_BADNAME:
 171                 return -EINVAL;
 172         case -NFS4ERR_SHARE_DENIED:
 173                 return -EACCES;
 174         case -NFS4ERR_MINOR_VERS_MISMATCH:
 175                 return -EPROTONOSUPPORT;
 176         case -NFS4ERR_FILE_OPEN:
 177                 return -EBUSY;
 178         default:
 179                 dprintk("%s could not handle NFSv4 error %d\n",
 180                                 __func__, -err);
 181                 break;
 182         }
 183         return -EIO;
 184 }
 185
 186 /*
 187  * This is our standard bitmap for GETATTR requests.
 188  */
 189 const u32 nfs4_fattr_bitmap[3] = {
 190         FATTR4_WORD0_TYPE
 191         | FATTR4_WORD0_CHANGE
 192         | FATTR4_WORD0_SIZE
 193         | FATTR4_WORD0_FSID
 194         | FATTR4_WORD0_FILEID,
 195         FATTR4_WORD1_MODE
 196         | FATTR4_WORD1_NUMLINKS
 197         | FATTR4_WORD1_OWNER
 198         | FATTR4_WORD1_OWNER_GROUP
 199         | FATTR4_WORD1_RAWDEV
 200         | FATTR4_WORD1_SPACE_USED
 201         | FATTR4_WORD1_TIME_ACCESS
 202         | FATTR4_WORD1_TIME_METADATA
 203         | FATTR4_WORD1_TIME_MODIFY
 204         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 205 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 206         FATTR4_WORD2_SECURITY_LABEL
 207 #endif
 208 };
 209
 210 static const u32 nfs4_pnfs_open_bitmap[3] = {
 211         FATTR4_WORD0_TYPE
 212         | FATTR4_WORD0_CHANGE
 213         | FATTR4_WORD0_SIZE
 214         | FATTR4_WORD0_FSID
 215         | FATTR4_WORD0_FILEID,
 216         FATTR4_WORD1_MODE
 217         | FATTR4_WORD1_NUMLINKS
 218         | FATTR4_WORD1_OWNER
 219         | FATTR4_WORD1_OWNER_GROUP
 220         | FATTR4_WORD1_RAWDEV
 221         | FATTR4_WORD1_SPACE_USED
 222         | FATTR4_WORD1_TIME_ACCESS
 223         | FATTR4_WORD1_TIME_METADATA
 224         | FATTR4_WORD1_TIME_MODIFY,
 225         FATTR4_WORD2_MDSTHRESHOLD
 226 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 227         | FATTR4_WORD2_SECURITY_LABEL
 228 #endif
 229 };
 230
 231 static const u32 nfs4_open_noattr_bitmap[3] = {
 232         FATTR4_WORD0_TYPE
 233         | FATTR4_WORD0_FILEID,
 234 };
 235
 236 const u32 nfs4_statfs_bitmap[3] = {
 237         FATTR4_WORD0_FILES_AVAIL
 238         | FATTR4_WORD0_FILES_FREE
 239         | FATTR4_WORD0_FILES_TOTAL,
 240         FATTR4_WORD1_SPACE_AVAIL
 241         | FATTR4_WORD1_SPACE_FREE
 242         | FATTR4_WORD1_SPACE_TOTAL
 243 };
 244
 245 const u32 nfs4_pathconf_bitmap[3] = {
 246         FATTR4_WORD0_MAXLINK
 247         | FATTR4_WORD0_MAXNAME,
 248         0
 249 };
 250
 251 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 252                         | FATTR4_WORD0_MAXREAD
 253                         | FATTR4_WORD0_MAXWRITE
 254                         | FATTR4_WORD0_LEASE_TIME,
 255                         FATTR4_WORD1_TIME_DELTA
 256                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 257                         FATTR4_WORD2_LAYOUT_BLKSIZE
 258                         | FATTR4_WORD2_CLONE_BLKSIZE
 259 };
 260
 261 const u32 nfs4_fs_locations_bitmap[3] = {
 262         FATTR4_WORD0_CHANGE
 263         | FATTR4_WORD0_SIZE
 264         | FATTR4_WORD0_FSID
 265         | FATTR4_WORD0_FILEID
 266         | FATTR4_WORD0_FS_LOCATIONS,
 267         FATTR4_WORD1_OWNER
 268         | FATTR4_WORD1_OWNER_GROUP
 269         | FATTR4_WORD1_RAWDEV
 270         | FATTR4_WORD1_SPACE_USED
 271         | FATTR4_WORD1_TIME_ACCESS
 272         | FATTR4_WORD1_TIME_METADATA
 273         | FATTR4_WORD1_TIME_MODIFY
 274         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 275 };
 276
 277 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 278                 struct inode *inode)
 279 {
 280         unsigned long cache_validity;
 281
 282         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 283         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 284                 return;
 285
 286         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 287         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 288                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 289                                 | NFS_INO_INVALID_SIZE);
 290
 291         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 292                 dst[0] &= ~FATTR4_WORD0_SIZE;
 293
 294         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 295                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 296 }
 297
 298 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 299                 const __u32 *src, struct inode *inode)
 300 {
 301         nfs4_bitmap_copy_adjust(dst, src, inode);
 302 }
 303
 304 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 305                 struct nfs4_readdir_arg *readdir)
 306 {
 307         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 308         __be32 *start, *p;
 309
 310         if (cookie > 2) {
 311                 readdir->cookie = cookie;
 312                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 313                 return;
 314         }
 315
 316         readdir->cookie = 0;
 317         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 318         if (cookie == 2)
 319                 return;
 320
 321         /*
 322          * NFSv4 servers do not return entries for '.' and '..'
 323          * Therefore, we fake these entries here.  We let '.'
 324          * have cookie 0 and '..' have cookie 1.  Note that
 325          * when talking to the server, we always send cookie 0
 326          * instead of 1 or 2.
 327          */
 328         start = p = kmap_atomic(*readdir->pages);
 329
 330         if (cookie == 0) {
 331                 *p++ = xdr_one;                                  /* next */
 332                 *p++ = xdr_zero;                   /* cookie, first word */
 333                 *p++ = xdr_one;                   /* cookie, second word */
 334                 *p++ = xdr_one;                             /* entry len */
 335                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 336                 p++;
 337                 *p++ = xdr_one;                         /* bitmap length */
 338                 *p++ = htonl(attrs);                           /* bitmap */
 339                 *p++ = htonl(12);             /* attribute buffer length */
 340                 *p++ = htonl(NF4DIR);
 341                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 342         }
 343
 344         *p++ = xdr_one;                                  /* next */
 345         *p++ = xdr_zero;                   /* cookie, first word */
 346         *p++ = xdr_two;                   /* cookie, second word */
 347         *p++ = xdr_two;                             /* entry len */
 348         memcpy(p, "..\0\0", 4);                         /* entry */
 349         p++;
 350         *p++ = xdr_one;                         /* bitmap length */
 351         *p++ = htonl(attrs);                           /* bitmap */
 352         *p++ = htonl(12);             /* attribute buffer length */
 353         *p++ = htonl(NF4DIR);
 354         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 355
 356         readdir->pgbase = (char *)p - (char *)start;
 357         readdir->count -= readdir->pgbase;
 358         kunmap_atomic(start);
 359 }
 360
 361 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 362                 nfs4_stateid *stateid,
 363                 const struct cred *cred)
 364 {
 365         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 366
 367         ops->test_and_free_expired(server, stateid, cred);
 368 }
 369
 370 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 371                 nfs4_stateid *stateid,
 372                 const struct cred *cred)
 373 {
 374         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 375         nfs4_test_and_free_stateid(server, stateid, cred);
 376 }
 377
 378 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 379                 const nfs4_stateid *stateid,
 380                 const struct cred *cred)
 381 {
 382         nfs4_stateid tmp;
 383
 384         nfs4_stateid_copy(&tmp, stateid);
 385         __nfs4_free_revoked_stateid(server, &tmp, cred);
 386 }
 387
 388 static long nfs4_update_delay(long *timeout)
 389 {
 390         long ret;
 391         if (!timeout)
 392                 return NFS4_POLL_RETRY_MAX;
 393         if (*timeout <= 0)
 394                 *timeout = NFS4_POLL_RETRY_MIN;
 395         if (*timeout > NFS4_POLL_RETRY_MAX)
 396                 *timeout = NFS4_POLL_RETRY_MAX;
 397         ret = *timeout;
 398         *timeout <<= 1;
 399         return ret;
 400 }
 401
 402 static int nfs4_delay_killable(long *timeout)
 403 {
 404         might_sleep();
 405
 406         freezable_schedule_timeout_killable_unsafe(
 407                 nfs4_update_delay(timeout));
 408         if (!__fatal_signal_pending(current))
 409                 return 0;
 410         return -EINTR;
 411 }
 412
 413 static int nfs4_delay_interruptible(long *timeout)
 414 {
 415         might_sleep();
 416
 417         freezable_schedule_timeout_interruptible(nfs4_update_delay(timeout));
 418         if (!signal_pending(current))
 419                 return 0;
 420         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 421 }
 422
 423 static int nfs4_delay(long *timeout, bool interruptible)
 424 {
 425         if (interruptible)
 426                 return nfs4_delay_interruptible(timeout);
 427         return nfs4_delay_killable(timeout);
 428 }
 429
 430 static const nfs4_stateid *
 431 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 432 {
 433         if (!stateid)
 434                 return NULL;
 435         switch (stateid->type) {
 436         case NFS4_OPEN_STATEID_TYPE:
 437         case NFS4_LOCK_STATEID_TYPE:
 438         case NFS4_DELEGATION_STATEID_TYPE:
 439                 return stateid;
 440         default:
 441                 break;
 442         }
 443         return NULL;
 444 }
 445
 446 /* This is the error handling routine for processes that are allowed
 447  * to sleep.
 448  */
 449 static int nfs4_do_handle_exception(struct nfs_server *server,
 450                 int errorcode, struct nfs4_exception *exception)
 451 {
 452         struct nfs_client *clp = server->nfs_client;
 453         struct nfs4_state *state = exception->state;
 454         const nfs4_stateid *stateid;
 455         struct inode *inode = exception->inode;
 456         int ret = errorcode;
 457
 458         exception->delay = 0;
 459         exception->recovering = 0;
 460         exception->retry = 0;
 461
 462         stateid = nfs4_recoverable_stateid(exception->stateid);
 463         if (stateid == NULL && state != NULL)
 464                 stateid = nfs4_recoverable_stateid(&state->stateid);
 465
 466         switch(errorcode) {
 467                 case 0:
 468                         return 0;
 469                 case -NFS4ERR_BADHANDLE:
 470                 case -ESTALE:
 471                         if (inode != NULL && S_ISREG(inode->i_mode))
 472                                 pnfs_destroy_layout(NFS_I(inode));
 473                         break;
 474                 case -NFS4ERR_DELEG_REVOKED:
 475                 case -NFS4ERR_ADMIN_REVOKED:
 476                 case -NFS4ERR_EXPIRED:
 477                 case -NFS4ERR_BAD_STATEID:
 478                 case -NFS4ERR_PARTNER_NO_AUTH:
 479                         if (inode != NULL && stateid != NULL) {
 480                                 nfs_inode_find_state_and_recover(inode,
 481                                                 stateid);
 482                                 goto wait_on_recovery;
 483                         }
 484                         /* Fall through */
 485                 case -NFS4ERR_OPENMODE:
 486                         if (inode) {
 487                                 int err;
 488
 489                                 err = nfs_async_inode_return_delegation(inode,
 490                                                 stateid);
 491                                 if (err == 0)
 492                                         goto wait_on_recovery;
 493                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 494                                         exception->retry = 1;
 495                                         break;
 496                                 }
 497                         }
 498                         if (state == NULL)
 499                                 break;
 500                         ret = nfs4_schedule_stateid_recovery(server, state);
 501                         if (ret < 0)
 502                                 break;
 503                         goto wait_on_recovery;
 504                 case -NFS4ERR_STALE_STATEID:
 505                 case -NFS4ERR_STALE_CLIENTID:
 506                         nfs4_schedule_lease_recovery(clp);
 507                         goto wait_on_recovery;
 508                 case -NFS4ERR_MOVED:
 509                         ret = nfs4_schedule_migration_recovery(server);
 510                         if (ret < 0)
 511                                 break;
 512                         goto wait_on_recovery;
 513                 case -NFS4ERR_LEASE_MOVED:
 514                         nfs4_schedule_lease_moved_recovery(clp);
 515                         goto wait_on_recovery;
 516 #if defined(CONFIG_NFS_V4_1)
 517                 case -NFS4ERR_BADSESSION:
 518                 case -NFS4ERR_BADSLOT:
 519                 case -NFS4ERR_BAD_HIGH_SLOT:
 520                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 521                 case -NFS4ERR_DEADSESSION:
 522                 case -NFS4ERR_SEQ_FALSE_RETRY:
 523                 case -NFS4ERR_SEQ_MISORDERED:
 524                         /* Handled in nfs41_sequence_process() */
 525                         goto wait_on_recovery;
 526 #endif /* defined(CONFIG_NFS_V4_1) */
 527                 case -NFS4ERR_FILE_OPEN:
 528                         if (exception->timeout > HZ) {
 529                                 /* We have retried a decent amount, time to
 530                                  * fail
 531                                  */
 532                                 ret = -EBUSY;
 533                                 break;
 534                         }
 535                         /* Fall through */
 536                 case -NFS4ERR_DELAY:
 537                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 538                         /* Fall through */
 539                 case -NFS4ERR_GRACE:
 540                 case -NFS4ERR_LAYOUTTRYLATER:
 541                 case -NFS4ERR_RECALLCONFLICT:
 542                         exception->delay = 1;
 543                         return 0;
 544
 545                 case -NFS4ERR_RETRY_UNCACHED_REP:
 546                 case -NFS4ERR_OLD_STATEID:
 547                         exception->retry = 1;
 548                         break;
 549                 case -NFS4ERR_BADOWNER:
 550                         /* The following works around a Linux server bug! */
 551                 case -NFS4ERR_BADNAME:
 552                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 553                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 554                                 exception->retry = 1;
 555                                 printk(KERN_WARNING "NFS: v4 server %s "
 556                                                 "does not accept raw "
 557                                                 "uid/gids. "
 558                                                 "Reenabling the idmapper.\n",
 559                                                 server->nfs_client->cl_hostname);
 560                         }
 561         }
 562         /* We failed to handle the error */
 563         return nfs4_map_errors(ret);
 564 wait_on_recovery:
 565         exception->recovering = 1;
 566         return 0;
 567 }
 568
 569 /* This is the error handling routine for processes that are allowed
 570  * to sleep.
 571  */
 572 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 573 {
 574         struct nfs_client *clp = server->nfs_client;
 575         int ret;
 576
 577         ret = nfs4_do_handle_exception(server, errorcode, exception);
 578         if (exception->delay) {
 579                 ret = nfs4_delay(&exception->timeout,
 580                                 exception->interruptible);
 581                 goto out_retry;
 582         }
 583         if (exception->recovering) {
 584                 ret = nfs4_wait_clnt_recover(clp);
 585                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 586                         return -EIO;
 587                 goto out_retry;
 588         }
 589         return ret;
 590 out_retry:
 591         if (ret == 0)
 592                 exception->retry = 1;
 593         return ret;
 594 }
 595
 596 static int
 597 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 598                 int errorcode, struct nfs4_exception *exception)
 599 {
 600         struct nfs_client *clp = server->nfs_client;
 601         int ret;
 602
 603         ret = nfs4_do_handle_exception(server, errorcode, exception);
 604         if (exception->delay) {
 605                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 606                 goto out_retry;
 607         }
 608         if (exception->recovering) {
 609                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 610                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 611                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 612                 goto out_retry;
 613         }
 614         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 615                 ret = -EIO;
 616         return ret;
 617 out_retry:
 618         if (ret == 0) {
 619                 exception->retry = 1;
 620                 /*
 621                  * For NFS4ERR_MOVED, the client transport will need to
 622                  * be recomputed after migration recovery has completed.
 623                  */
 624                 if (errorcode == -NFS4ERR_MOVED)
 625                         rpc_task_release_transport(task);
 626         }
 627         return ret;
 628 }
 629
 630 int
 631 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 632                         struct nfs4_state *state, long *timeout)
 633 {
 634         struct nfs4_exception exception = {
 635                 .state = state,
 636         };
 637
 638         if (task->tk_status >= 0)
 639                 return 0;
 640         if (timeout)
 641                 exception.timeout = *timeout;
 642         task->tk_status = nfs4_async_handle_exception(task, server,
 643                         task->tk_status,
 644                         &exception);
 645         if (exception.delay && timeout)
 646                 *timeout = exception.timeout;
 647         if (exception.retry)
 648                 return -EAGAIN;
 649         return 0;
 650 }
 651
 652 /*
 653  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 654  * or 'false' otherwise.
 655  */
 656 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 657 {
 658         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 659         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 660 }
 661
 662 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 663 {
 664         spin_lock(&clp->cl_lock);
 665         if (time_before(clp->cl_last_renewal,timestamp))
 666                 clp->cl_last_renewal = timestamp;
 667         spin_unlock(&clp->cl_lock);
 668 }
 669
 670 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 671 {
 672         struct nfs_client *clp = server->nfs_client;
 673
 674         if (!nfs4_has_session(clp))
 675                 do_renew_lease(clp, timestamp);
 676 }
 677
 678 struct nfs4_call_sync_data {
 679         const struct nfs_server *seq_server;
 680         struct nfs4_sequence_args *seq_args;
 681         struct nfs4_sequence_res *seq_res;
 682 };
 683
 684 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 685                         struct nfs4_sequence_res *res, int cache_reply,
 686                         int privileged)
 687 {
 688         args->sa_slot = NULL;
 689         args->sa_cache_this = cache_reply;
 690         args->sa_privileged = privileged;
 691
 692         res->sr_slot = NULL;
 693 }
 694
 695 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 696 {
 697         struct nfs4_slot *slot = res->sr_slot;
 698         struct nfs4_slot_table *tbl;
 699
 700         tbl = slot->table;
 701         spin_lock(&tbl->slot_tbl_lock);
 702         if (!nfs41_wake_and_assign_slot(tbl, slot))
 703                 nfs4_free_slot(tbl, slot);
 704         spin_unlock(&tbl->slot_tbl_lock);
 705
 706         res->sr_slot = NULL;
 707 }
 708
 709 static int nfs40_sequence_done(struct rpc_task *task,
 710                                struct nfs4_sequence_res *res)
 711 {
 712         if (res->sr_slot != NULL)
 713                 nfs40_sequence_free_slot(res);
 714         return 1;
 715 }
 716
 717 #if defined(CONFIG_NFS_V4_1)
 718
 719 static void nfs41_release_slot(struct nfs4_slot *slot)
 720 {
 721         struct nfs4_session *session;
 722         struct nfs4_slot_table *tbl;
 723         bool send_new_highest_used_slotid = false;
 724
 725         if (!slot)
 726                 return;
 727         tbl = slot->table;
 728         session = tbl->session;
 729
 730         /* Bump the slot sequence number */
 731         if (slot->seq_done)
 732                 slot->seq_nr++;
 733         slot->seq_done = 0;
 734
 735         spin_lock(&tbl->slot_tbl_lock);
 736         /* Be nice to the server: try to ensure that the last transmitted
 737          * value for highest_user_slotid <= target_highest_slotid
 738          */
 739         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 740                 send_new_highest_used_slotid = true;
 741
 742         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 743                 send_new_highest_used_slotid = false;
 744                 goto out_unlock;
 745         }
 746         nfs4_free_slot(tbl, slot);
 747
 748         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 749                 send_new_highest_used_slotid = false;
 750 out_unlock:
 751         spin_unlock(&tbl->slot_tbl_lock);
 752         if (send_new_highest_used_slotid)
 753                 nfs41_notify_server(session->clp);
 754         if (waitqueue_active(&tbl->slot_waitq))
 755                 wake_up_all(&tbl->slot_waitq);
 756 }
 757
 758 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 759 {
 760         nfs41_release_slot(res->sr_slot);
 761         res->sr_slot = NULL;
 762 }
 763
 764 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 765                 u32 seqnr)
 766 {
 767         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 768                 slot->seq_nr_highest_sent = seqnr;
 769 }
 770 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 771                 u32 seqnr)
 772 {
 773         slot->seq_nr_highest_sent = seqnr;
 774         slot->seq_nr_last_acked = seqnr;
 775 }
 776
 777 static int nfs41_sequence_process(struct rpc_task *task,
 778                 struct nfs4_sequence_res *res)
 779 {
 780         struct nfs4_session *session;
 781         struct nfs4_slot *slot = res->sr_slot;
 782         struct nfs_client *clp;
 783         int status;
 784         int ret = 1;
 785
 786         if (slot == NULL)
 787                 goto out_noaction;
 788         /* don't increment the sequence number if the task wasn't sent */
 789         if (!RPC_WAS_SENT(task) || slot->seq_done)
 790                 goto out;
 791
 792         session = slot->table->session;
 793
 794         trace_nfs4_sequence_done(session, res);
 795
 796         status = res->sr_status;
 797         if (task->tk_status == -NFS4ERR_DEADSESSION)
 798                 status = -NFS4ERR_DEADSESSION;
 799
 800         /* Check the SEQUENCE operation status */
 801         switch (status) {
 802         case 0:
 803                 /* Mark this sequence number as having been acked */
 804                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 805                 /* Update the slot's sequence and clientid lease timer */
 806                 slot->seq_done = 1;
 807                 clp = session->clp;
 808                 do_renew_lease(clp, res->sr_timestamp);
 809                 /* Check sequence flags */
 810                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 811                                 !!slot->privileged);
 812                 nfs41_update_target_slotid(slot->table, slot, res);
 813                 break;
 814         case 1:
 815                 /*
 816                  * sr_status remains 1 if an RPC level error occurred.
 817                  * The server may or may not have processed the sequence
 818                  * operation..
 819                  */
 820                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 821                 slot->seq_done = 1;
 822                 goto out;
 823         case -NFS4ERR_DELAY:
 824                 /* The server detected a resend of the RPC call and
 825                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 826                  * of RFC5661.
 827                  */
 828                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 829                         __func__,
 830                         slot->slot_nr,
 831                         slot->seq_nr);
 832                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 833                 goto out_retry;
 834         case -NFS4ERR_RETRY_UNCACHED_REP:
 835         case -NFS4ERR_SEQ_FALSE_RETRY:
 836                 /*
 837                  * The server thinks we tried to replay a request.
 838                  * Retry the call after bumping the sequence ID.
 839                  */
 840                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 841                 goto retry_new_seq;
 842         case -NFS4ERR_BADSLOT:
 843                 /*
 844                  * The slot id we used was probably retired. Try again
 845                  * using a different slot id.
 846                  */
 847                 if (slot->slot_nr < slot->table->target_highest_slotid)
 848                         goto session_recover;
 849                 goto retry_nowait;
 850         case -NFS4ERR_SEQ_MISORDERED:
 851                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 852                 /*
 853                  * Were one or more calls using this slot interrupted?
 854                  * If the server never received the request, then our
 855                  * transmitted slot sequence number may be too high.
 856                  */
 857                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 858                         slot->seq_nr--;
 859                         goto retry_nowait;
 860                 }
 861                 /*
 862                  * RFC5661:
 863                  * A retry might be sent while the original request is
 864                  * still in progress on the replier. The replier SHOULD
 865                  * deal with the issue by returning NFS4ERR_DELAY as the
 866                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 867                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 868                  *
 869                  * Restart the search after a delay.
 870                  */
 871                 slot->seq_nr = slot->seq_nr_highest_sent;
 872                 goto out_retry;
 873         case -NFS4ERR_BADSESSION:
 874         case -NFS4ERR_DEADSESSION:
 875         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 876                 goto session_recover;
 877         default:
 878                 /* Just update the slot sequence no. */
 879                 slot->seq_done = 1;
 880         }
 881 out:
 882         /* The session may be reset by one of the error handlers. */
 883         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 884 out_noaction:
 885         return ret;
 886 session_recover:
 887         nfs4_schedule_session_recovery(session, status);
 888         dprintk("%s ERROR: %d Reset session\n", __func__, status);
 889         nfs41_sequence_free_slot(res);
 890         goto out;
 891 retry_new_seq:
 892         ++slot->seq_nr;
 893 retry_nowait:
 894         if (rpc_restart_call_prepare(task)) {
 895                 nfs41_sequence_free_slot(res);
 896                 task->tk_status = 0;
 897                 ret = 0;
 898         }
 899         goto out;
 900 out_retry:
 901         if (!rpc_restart_call(task))
 902                 goto out;
 903         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 904         return 0;
 905 }
 906
 907 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 908 {
 909         if (!nfs41_sequence_process(task, res))
 910                 return 0;
 911         if (res->sr_slot != NULL)
 912                 nfs41_sequence_free_slot(res);
 913         return 1;
 914
 915 }
 916 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 917
 918 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 919 {
 920         if (res->sr_slot == NULL)
 921                 return 1;
 922         if (res->sr_slot->table->session != NULL)
 923                 return nfs41_sequence_process(task, res);
 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                 if (res->sr_slot->table->session != NULL)
 931                         nfs41_sequence_free_slot(res);
 932                 else
 933                         nfs40_sequence_free_slot(res);
 934         }
 935 }
 936
 937 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 938 {
 939         if (res->sr_slot == NULL)
 940                 return 1;
 941         if (!res->sr_slot->table->session)
 942                 return nfs40_sequence_done(task, res);
 943         return nfs41_sequence_done(task, res);
 944 }
 945 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 946
 947 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 948 {
 949         struct nfs4_call_sync_data *data = calldata;
 950
 951         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 952
 953         nfs4_setup_sequence(data->seq_server->nfs_client,
 954                             data->seq_args, data->seq_res, task);
 955 }
 956
 957 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 958 {
 959         struct nfs4_call_sync_data *data = calldata;
 960
 961         nfs41_sequence_done(task, data->seq_res);
 962 }
 963
 964 static const struct rpc_call_ops nfs41_call_sync_ops = {
 965         .rpc_call_prepare = nfs41_call_sync_prepare,
 966         .rpc_call_done = nfs41_call_sync_done,
 967 };
 968
 969 #else   /* !CONFIG_NFS_V4_1 */
 970
 971 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 972 {
 973         return nfs40_sequence_done(task, res);
 974 }
 975
 976 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 977 {
 978         if (res->sr_slot != NULL)
 979                 nfs40_sequence_free_slot(res);
 980 }
 981
 982 int nfs4_sequence_done(struct rpc_task *task,
 983                        struct nfs4_sequence_res *res)
 984 {
 985         return nfs40_sequence_done(task, res);
 986 }
 987 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 988
 989 #endif  /* !CONFIG_NFS_V4_1 */
 990
 991 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
 992 {
 993         res->sr_timestamp = jiffies;
 994         res->sr_status_flags = 0;
 995         res->sr_status = 1;
 996 }
 997
 998 static
 999 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1000                 struct nfs4_sequence_res *res,
1001                 struct nfs4_slot *slot)
1002 {
1003         if (!slot)
1004                 return;
1005         slot->privileged = args->sa_privileged ? 1 : 0;
1006         args->sa_slot = slot;
1007
1008         res->sr_slot = slot;
1009 }
1010
1011 int nfs4_setup_sequence(struct nfs_client *client,
1012                         struct nfs4_sequence_args *args,
1013                         struct nfs4_sequence_res *res,
1014                         struct rpc_task *task)
1015 {
1016         struct nfs4_session *session = nfs4_get_session(client);
1017         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1018         struct nfs4_slot *slot;
1019
1020         /* slot already allocated? */
1021         if (res->sr_slot != NULL)
1022                 goto out_start;
1023
1024         if (session)
1025                 tbl = &session->fc_slot_table;
1026
1027         spin_lock(&tbl->slot_tbl_lock);
1028         /* The state manager will wait until the slot table is empty */
1029         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1030                 goto out_sleep;
1031
1032         slot = nfs4_alloc_slot(tbl);
1033         if (IS_ERR(slot)) {
1034                 if (slot == ERR_PTR(-ENOMEM))
1035                         goto out_sleep_timeout;
1036                 goto out_sleep;
1037         }
1038         spin_unlock(&tbl->slot_tbl_lock);
1039
1040         nfs4_sequence_attach_slot(args, res, slot);
1041
1042         trace_nfs4_setup_sequence(session, args);
1043 out_start:
1044         nfs41_sequence_res_init(res);
1045         rpc_call_start(task);
1046         return 0;
1047 out_sleep_timeout:
1048         /* Try again in 1/4 second */
1049         if (args->sa_privileged)
1050                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1051                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1052         else
1053                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1054                                 NULL, jiffies + (HZ >> 2));
1055         spin_unlock(&tbl->slot_tbl_lock);
1056         return -EAGAIN;
1057 out_sleep:
1058         if (args->sa_privileged)
1059                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1060                                 RPC_PRIORITY_PRIVILEGED);
1061         else
1062                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1063         spin_unlock(&tbl->slot_tbl_lock);
1064         return -EAGAIN;
1065 }
1066 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1067
1068 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1069 {
1070         struct nfs4_call_sync_data *data = calldata;
1071         nfs4_setup_sequence(data->seq_server->nfs_client,
1072                                 data->seq_args, data->seq_res, task);
1073 }
1074
1075 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1076 {
1077         struct nfs4_call_sync_data *data = calldata;
1078         nfs4_sequence_done(task, data->seq_res);
1079 }
1080
1081 static const struct rpc_call_ops nfs40_call_sync_ops = {
1082         .rpc_call_prepare = nfs40_call_sync_prepare,
1083         .rpc_call_done = nfs40_call_sync_done,
1084 };
1085
1086 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1087 {
1088         int ret;
1089         struct rpc_task *task;
1090
1091         task = rpc_run_task(task_setup);
1092         if (IS_ERR(task))
1093                 return PTR_ERR(task);
1094
1095         ret = task->tk_status;
1096         rpc_put_task(task);
1097         return ret;
1098 }
1099
1100 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1101                                    struct nfs_server *server,
1102                                    struct rpc_message *msg,
1103                                    struct nfs4_sequence_args *args,
1104                                    struct nfs4_sequence_res *res)
1105 {
1106         struct nfs_client *clp = server->nfs_client;
1107         struct nfs4_call_sync_data data = {
1108                 .seq_server = server,
1109                 .seq_args = args,
1110                 .seq_res = res,
1111         };
1112         struct rpc_task_setup task_setup = {
1113                 .rpc_client = clnt,
1114                 .rpc_message = msg,
1115                 .callback_ops = clp->cl_mvops->call_sync_ops,
1116                 .callback_data = &data
1117         };
1118
1119         return nfs4_call_sync_custom(&task_setup);
1120 }
1121
1122 int nfs4_call_sync(struct rpc_clnt *clnt,
1123                    struct nfs_server *server,
1124                    struct rpc_message *msg,
1125                    struct nfs4_sequence_args *args,
1126                    struct nfs4_sequence_res *res,
1127                    int cache_reply)
1128 {
1129         nfs4_init_sequence(args, res, cache_reply, 0);
1130         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1131 }
1132
1133 static void
1134 nfs4_inc_nlink_locked(struct inode *inode)
1135 {
1136         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1137         inc_nlink(inode);
1138 }
1139
1140 static void
1141 nfs4_dec_nlink_locked(struct inode *inode)
1142 {
1143         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1144         drop_nlink(inode);
1145 }
1146
1147 static void
1148 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1149                 unsigned long timestamp, unsigned long cache_validity)
1150 {
1151         struct nfs_inode *nfsi = NFS_I(dir);
1152
1153         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1154                 | NFS_INO_INVALID_MTIME
1155                 | NFS_INO_INVALID_DATA
1156                 | cache_validity;
1157         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1158                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1159                 nfsi->attrtimeo_timestamp = jiffies;
1160         } else {
1161                 nfs_force_lookup_revalidate(dir);
1162                 if (cinfo->before != inode_peek_iversion_raw(dir))
1163                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1164                                 NFS_INO_INVALID_ACL;
1165         }
1166         inode_set_iversion_raw(dir, cinfo->after);
1167         nfsi->read_cache_jiffies = timestamp;
1168         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1169         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1170         nfs_fscache_invalidate(dir);
1171 }
1172
1173 static void
1174 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1175                 unsigned long timestamp, unsigned long cache_validity)
1176 {
1177         spin_lock(&dir->i_lock);
1178         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1179         spin_unlock(&dir->i_lock);
1180 }
1181
1182 struct nfs4_open_createattrs {
1183         struct nfs4_label *label;
1184         struct iattr *sattr;
1185         const __u32 verf[2];
1186 };
1187
1188 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1189                 int err, struct nfs4_exception *exception)
1190 {
1191         if (err != -EINVAL)
1192                 return false;
1193         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1194                 return false;
1195         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1196         exception->retry = 1;
1197         return true;
1198 }
1199
1200 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1201 {
1202          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1203 }
1204
1205 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1206 {
1207         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1208
1209         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1210 }
1211
1212 static u32
1213 nfs4_map_atomic_open_share(struct nfs_server *server,
1214                 fmode_t fmode, int openflags)
1215 {
1216         u32 res = 0;
1217
1218         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1219         case FMODE_READ:
1220                 res = NFS4_SHARE_ACCESS_READ;
1221                 break;
1222         case FMODE_WRITE:
1223                 res = NFS4_SHARE_ACCESS_WRITE;
1224                 break;
1225         case FMODE_READ|FMODE_WRITE:
1226                 res = NFS4_SHARE_ACCESS_BOTH;
1227         }
1228         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1229                 goto out;
1230         /* Want no delegation if we're using O_DIRECT */
1231         if (openflags & O_DIRECT)
1232                 res |= NFS4_SHARE_WANT_NO_DELEG;
1233 out:
1234         return res;
1235 }
1236
1237 static enum open_claim_type4
1238 nfs4_map_atomic_open_claim(struct nfs_server *server,
1239                 enum open_claim_type4 claim)
1240 {
1241         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1242                 return claim;
1243         switch (claim) {
1244         default:
1245                 return claim;
1246         case NFS4_OPEN_CLAIM_FH:
1247                 return NFS4_OPEN_CLAIM_NULL;
1248         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1249                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1250         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1251                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1252         }
1253 }
1254
1255 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1256 {
1257         p->o_res.f_attr = &p->f_attr;
1258         p->o_res.f_label = p->f_label;
1259         p->o_res.seqid = p->o_arg.seqid;
1260         p->c_res.seqid = p->c_arg.seqid;
1261         p->o_res.server = p->o_arg.server;
1262         p->o_res.access_request = p->o_arg.access;
1263         nfs_fattr_init(&p->f_attr);
1264         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1265 }
1266
1267 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1268                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1269                 const struct nfs4_open_createattrs *c,
1270                 enum open_claim_type4 claim,
1271                 gfp_t gfp_mask)
1272 {
1273         struct dentry *parent = dget_parent(dentry);
1274         struct inode *dir = d_inode(parent);
1275         struct nfs_server *server = NFS_SERVER(dir);
1276         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1277         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1278         struct nfs4_opendata *p;
1279
1280         p = kzalloc(sizeof(*p), gfp_mask);
1281         if (p == NULL)
1282                 goto err;
1283
1284         p->f_label = nfs4_label_alloc(server, gfp_mask);
1285         if (IS_ERR(p->f_label))
1286                 goto err_free_p;
1287
1288         p->a_label = nfs4_label_alloc(server, gfp_mask);
1289         if (IS_ERR(p->a_label))
1290                 goto err_free_f;
1291
1292         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1293         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1294         if (IS_ERR(p->o_arg.seqid))
1295                 goto err_free_label;
1296         nfs_sb_active(dentry->d_sb);
1297         p->dentry = dget(dentry);
1298         p->dir = parent;
1299         p->owner = sp;
1300         atomic_inc(&sp->so_count);
1301         p->o_arg.open_flags = flags;
1302         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1303         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1304         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1305                         fmode, flags);
1306         if (flags & O_CREAT) {
1307                 p->o_arg.umask = current_umask();
1308                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1309                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1310                         p->o_arg.u.attrs = &p->attrs;
1311                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1312
1313                         memcpy(p->o_arg.u.verifier.data, c->verf,
1314                                         sizeof(p->o_arg.u.verifier.data));
1315                 }
1316         }
1317         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1318          * will return permission denied for all bits until close */
1319         if (!(flags & O_EXCL)) {
1320                 /* ask server to check for all possible rights as results
1321                  * are cached */
1322                 switch (p->o_arg.claim) {
1323                 default:
1324                         break;
1325                 case NFS4_OPEN_CLAIM_NULL:
1326                 case NFS4_OPEN_CLAIM_FH:
1327                         p->o_arg.access = NFS4_ACCESS_READ |
1328                                 NFS4_ACCESS_MODIFY |
1329                                 NFS4_ACCESS_EXTEND |
1330                                 NFS4_ACCESS_EXECUTE;
1331                 }
1332         }
1333         p->o_arg.clientid = server->nfs_client->cl_clientid;
1334         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1335         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1336         p->o_arg.name = &dentry->d_name;
1337         p->o_arg.server = server;
1338         p->o_arg.bitmask = nfs4_bitmask(server, label);
1339         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1340         switch (p->o_arg.claim) {
1341         case NFS4_OPEN_CLAIM_NULL:
1342         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1343         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1344                 p->o_arg.fh = NFS_FH(dir);
1345                 break;
1346         case NFS4_OPEN_CLAIM_PREVIOUS:
1347         case NFS4_OPEN_CLAIM_FH:
1348         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1349         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1350                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1351         }
1352         p->c_arg.fh = &p->o_res.fh;
1353         p->c_arg.stateid = &p->o_res.stateid;
1354         p->c_arg.seqid = p->o_arg.seqid;
1355         nfs4_init_opendata_res(p);
1356         kref_init(&p->kref);
1357         return p;
1358
1359 err_free_label:
1360         nfs4_label_free(p->a_label);
1361 err_free_f:
1362         nfs4_label_free(p->f_label);
1363 err_free_p:
1364         kfree(p);
1365 err:
1366         dput(parent);
1367         return NULL;
1368 }
1369
1370 static void nfs4_opendata_free(struct kref *kref)
1371 {
1372         struct nfs4_opendata *p = container_of(kref,
1373                         struct nfs4_opendata, kref);
1374         struct super_block *sb = p->dentry->d_sb;
1375
1376         nfs4_lgopen_release(p->lgp);
1377         nfs_free_seqid(p->o_arg.seqid);
1378         nfs4_sequence_free_slot(&p->o_res.seq_res);
1379         if (p->state != NULL)
1380                 nfs4_put_open_state(p->state);
1381         nfs4_put_state_owner(p->owner);
1382
1383         nfs4_label_free(p->a_label);
1384         nfs4_label_free(p->f_label);
1385
1386         dput(p->dir);
1387         dput(p->dentry);
1388         nfs_sb_deactive(sb);
1389         nfs_fattr_free_names(&p->f_attr);
1390         kfree(p->f_attr.mdsthreshold);
1391         kfree(p);
1392 }
1393
1394 static void nfs4_opendata_put(struct nfs4_opendata *p)
1395 {
1396         if (p != NULL)
1397                 kref_put(&p->kref, nfs4_opendata_free);
1398 }
1399
1400 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1401                 fmode_t fmode)
1402 {
1403         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1404         case FMODE_READ|FMODE_WRITE:
1405                 return state->n_rdwr != 0;
1406         case FMODE_WRITE:
1407                 return state->n_wronly != 0;
1408         case FMODE_READ:
1409                 return state->n_rdonly != 0;
1410         }
1411         WARN_ON_ONCE(1);
1412         return false;
1413 }
1414
1415 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1416                 int open_mode, enum open_claim_type4 claim)
1417 {
1418         int ret = 0;
1419
1420         if (open_mode & (O_EXCL|O_TRUNC))
1421                 goto out;
1422         switch (claim) {
1423         case NFS4_OPEN_CLAIM_NULL:
1424         case NFS4_OPEN_CLAIM_FH:
1425                 goto out;
1426         default:
1427                 break;
1428         }
1429         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1430                 case FMODE_READ:
1431                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1432                                 && state->n_rdonly != 0;
1433                         break;
1434                 case FMODE_WRITE:
1435                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1436                                 && state->n_wronly != 0;
1437                         break;
1438                 case FMODE_READ|FMODE_WRITE:
1439                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1440                                 && state->n_rdwr != 0;
1441         }
1442 out:
1443         return ret;
1444 }
1445
1446 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1447                 enum open_claim_type4 claim)
1448 {
1449         if (delegation == NULL)
1450                 return 0;
1451         if ((delegation->type & fmode) != fmode)
1452                 return 0;
1453         switch (claim) {
1454         case NFS4_OPEN_CLAIM_NULL:
1455         case NFS4_OPEN_CLAIM_FH:
1456                 break;
1457         case NFS4_OPEN_CLAIM_PREVIOUS:
1458                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1459                         break;
1460                 /* Fall through */
1461         default:
1462                 return 0;
1463         }
1464         nfs_mark_delegation_referenced(delegation);
1465         return 1;
1466 }
1467
1468 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1469 {
1470         switch (fmode) {
1471                 case FMODE_WRITE:
1472                         state->n_wronly++;
1473                         break;
1474                 case FMODE_READ:
1475                         state->n_rdonly++;
1476                         break;
1477                 case FMODE_READ|FMODE_WRITE:
1478                         state->n_rdwr++;
1479         }
1480         nfs4_state_set_mode_locked(state, state->state | fmode);
1481 }
1482
1483 #ifdef CONFIG_NFS_V4_1
1484 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1485 {
1486         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1487                 return true;
1488         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1489                 return true;
1490         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1491                 return true;
1492         return false;
1493 }
1494 #endif /* CONFIG_NFS_V4_1 */
1495
1496 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1497 {
1498         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1499                 wake_up_all(&state->waitq);
1500 }
1501
1502 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1503                 const nfs4_stateid *stateid)
1504 {
1505         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1506         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1507
1508         if (stateid_seqid == state_seqid + 1U ||
1509             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1510                 nfs_state_log_update_open_stateid(state);
1511         else
1512                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1513 }
1514
1515 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1516 {
1517         struct nfs_client *clp = state->owner->so_server->nfs_client;
1518         bool need_recover = false;
1519
1520         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1521                 need_recover = true;
1522         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1523                 need_recover = true;
1524         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1525                 need_recover = true;
1526         if (need_recover)
1527                 nfs4_state_mark_reclaim_nograce(clp, state);
1528 }
1529
1530 /*
1531  * Check for whether or not the caller may update the open stateid
1532  * to the value passed in by stateid.
1533  *
1534  * Note: This function relies heavily on the server implementing
1535  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1536  * correctly.
1537  * i.e. The stateid seqids have to be initialised to 1, and
1538  * are then incremented on every state transition.
1539  */
1540 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1541                 const nfs4_stateid *stateid)
1542 {
1543         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1544             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1545                 if (stateid->seqid == cpu_to_be32(1))
1546                         nfs_state_log_update_open_stateid(state);
1547                 else
1548                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1549                 return true;
1550         }
1551
1552         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1553                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1554                 return true;
1555         }
1556         return false;
1557 }
1558
1559 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1560 {
1561         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1562                 return;
1563         if (state->n_wronly)
1564                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1565         if (state->n_rdonly)
1566                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1567         if (state->n_rdwr)
1568                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1569         set_bit(NFS_OPEN_STATE, &state->flags);
1570 }
1571
1572 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1573                 nfs4_stateid *stateid, fmode_t fmode)
1574 {
1575         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1576         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1577         case FMODE_WRITE:
1578                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1579                 break;
1580         case FMODE_READ:
1581                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1582                 break;
1583         case 0:
1584                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1585                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1586                 clear_bit(NFS_OPEN_STATE, &state->flags);
1587         }
1588         if (stateid == NULL)
1589                 return;
1590         /* Handle OPEN+OPEN_DOWNGRADE races */
1591         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1592             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1593                 nfs_resync_open_stateid_locked(state);
1594                 goto out;
1595         }
1596         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1597                 nfs4_stateid_copy(&state->stateid, stateid);
1598         nfs4_stateid_copy(&state->open_stateid, stateid);
1599         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1600 out:
1601         nfs_state_log_update_open_stateid(state);
1602 }
1603
1604 static void nfs_clear_open_stateid(struct nfs4_state *state,
1605         nfs4_stateid *arg_stateid,
1606         nfs4_stateid *stateid, fmode_t fmode)
1607 {
1608         write_seqlock(&state->seqlock);
1609         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1610         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1611                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1612         write_sequnlock(&state->seqlock);
1613         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1614                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1615 }
1616
1617 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1618                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1619         __must_hold(&state->owner->so_lock)
1620         __must_hold(&state->seqlock)
1621         __must_hold(RCU)
1622
1623 {
1624         DEFINE_WAIT(wait);
1625         int status = 0;
1626         for (;;) {
1627
1628                 if (!nfs_need_update_open_stateid(state, stateid))
1629                         return;
1630                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1631                         break;
1632                 if (status)
1633                         break;
1634                 /* Rely on seqids for serialisation with NFSv4.0 */
1635                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1636                         break;
1637
1638                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1639                 /*
1640                  * Ensure we process the state changes in the same order
1641                  * in which the server processed them by delaying the
1642                  * update of the stateid until we are in sequence.
1643                  */
1644                 write_sequnlock(&state->seqlock);
1645                 spin_unlock(&state->owner->so_lock);
1646                 rcu_read_unlock();
1647                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1648                 if (!signal_pending(current)) {
1649                         if (schedule_timeout(5*HZ) == 0)
1650                                 status = -EAGAIN;
1651                         else
1652                                 status = 0;
1653                 } else
1654                         status = -EINTR;
1655                 finish_wait(&state->waitq, &wait);
1656                 rcu_read_lock();
1657                 spin_lock(&state->owner->so_lock);
1658                 write_seqlock(&state->seqlock);
1659         }
1660
1661         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1662             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1663                 nfs4_stateid_copy(freeme, &state->open_stateid);
1664                 nfs_test_and_clear_all_open_stateid(state);
1665         }
1666
1667         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1668                 nfs4_stateid_copy(&state->stateid, stateid);
1669         nfs4_stateid_copy(&state->open_stateid, stateid);
1670         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1671         nfs_state_log_update_open_stateid(state);
1672 }
1673
1674 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1675                 const nfs4_stateid *open_stateid,
1676                 fmode_t fmode,
1677                 nfs4_stateid *freeme)
1678 {
1679         /*
1680          * Protect the call to nfs4_state_set_mode_locked and
1681          * serialise the stateid update
1682          */
1683         write_seqlock(&state->seqlock);
1684         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1685         switch (fmode) {
1686         case FMODE_READ:
1687                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1688                 break;
1689         case FMODE_WRITE:
1690                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1691                 break;
1692         case FMODE_READ|FMODE_WRITE:
1693                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1694         }
1695         set_bit(NFS_OPEN_STATE, &state->flags);
1696         write_sequnlock(&state->seqlock);
1697 }
1698
1699 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1700 {
1701         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1702         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1703         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1704         clear_bit(NFS_OPEN_STATE, &state->flags);
1705 }
1706
1707 static void nfs_state_set_delegation(struct nfs4_state *state,
1708                 const nfs4_stateid *deleg_stateid,
1709                 fmode_t fmode)
1710 {
1711         /*
1712          * Protect the call to nfs4_state_set_mode_locked and
1713          * serialise the stateid update
1714          */
1715         write_seqlock(&state->seqlock);
1716         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1717         set_bit(NFS_DELEGATED_STATE, &state->flags);
1718         write_sequnlock(&state->seqlock);
1719 }
1720
1721 static void nfs_state_clear_delegation(struct nfs4_state *state)
1722 {
1723         write_seqlock(&state->seqlock);
1724         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1725         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1726         write_sequnlock(&state->seqlock);
1727 }
1728
1729 int update_open_stateid(struct nfs4_state *state,
1730                 const nfs4_stateid *open_stateid,
1731                 const nfs4_stateid *delegation,
1732                 fmode_t fmode)
1733 {
1734         struct nfs_server *server = NFS_SERVER(state->inode);
1735         struct nfs_client *clp = server->nfs_client;
1736         struct nfs_inode *nfsi = NFS_I(state->inode);
1737         struct nfs_delegation *deleg_cur;
1738         nfs4_stateid freeme = { };
1739         int ret = 0;
1740
1741         fmode &= (FMODE_READ|FMODE_WRITE);
1742
1743         rcu_read_lock();
1744         spin_lock(&state->owner->so_lock);
1745         if (open_stateid != NULL) {
1746                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1747                 ret = 1;
1748         }
1749
1750         deleg_cur = nfs4_get_valid_delegation(state->inode);
1751         if (deleg_cur == NULL)
1752                 goto no_delegation;
1753
1754         spin_lock(&deleg_cur->lock);
1755         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1756            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1757             (deleg_cur->type & fmode) != fmode)
1758                 goto no_delegation_unlock;
1759
1760         if (delegation == NULL)
1761                 delegation = &deleg_cur->stateid;
1762         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1763                 goto no_delegation_unlock;
1764
1765         nfs_mark_delegation_referenced(deleg_cur);
1766         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1767         ret = 1;
1768 no_delegation_unlock:
1769         spin_unlock(&deleg_cur->lock);
1770 no_delegation:
1771         if (ret)
1772                 update_open_stateflags(state, fmode);
1773         spin_unlock(&state->owner->so_lock);
1774         rcu_read_unlock();
1775
1776         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1777                 nfs4_schedule_state_manager(clp);
1778         if (freeme.type != 0)
1779                 nfs4_test_and_free_stateid(server, &freeme,
1780                                 state->owner->so_cred);
1781
1782         return ret;
1783 }
1784
1785 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1786                 const nfs4_stateid *stateid)
1787 {
1788         struct nfs4_state *state = lsp->ls_state;
1789         bool ret = false;
1790
1791         spin_lock(&state->state_lock);
1792         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1793                 goto out_noupdate;
1794         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1795                 goto out_noupdate;
1796         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1797         ret = true;
1798 out_noupdate:
1799         spin_unlock(&state->state_lock);
1800         return ret;
1801 }
1802
1803 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1804 {
1805         struct nfs_delegation *delegation;
1806
1807         fmode &= FMODE_READ|FMODE_WRITE;
1808         rcu_read_lock();
1809         delegation = nfs4_get_valid_delegation(inode);
1810         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1811                 rcu_read_unlock();
1812                 return;
1813         }
1814         rcu_read_unlock();
1815         nfs4_inode_return_delegation(inode);
1816 }
1817
1818 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1819 {
1820         struct nfs4_state *state = opendata->state;
1821         struct nfs_delegation *delegation;
1822         int open_mode = opendata->o_arg.open_flags;
1823         fmode_t fmode = opendata->o_arg.fmode;
1824         enum open_claim_type4 claim = opendata->o_arg.claim;
1825         nfs4_stateid stateid;
1826         int ret = -EAGAIN;
1827
1828         for (;;) {
1829                 spin_lock(&state->owner->so_lock);
1830                 if (can_open_cached(state, fmode, open_mode, claim)) {
1831                         update_open_stateflags(state, fmode);
1832                         spin_unlock(&state->owner->so_lock);
1833                         goto out_return_state;
1834                 }
1835                 spin_unlock(&state->owner->so_lock);
1836                 rcu_read_lock();
1837                 delegation = nfs4_get_valid_delegation(state->inode);
1838                 if (!can_open_delegated(delegation, fmode, claim)) {
1839                         rcu_read_unlock();
1840                         break;
1841                 }
1842                 /* Save the delegation */
1843                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1844                 rcu_read_unlock();
1845                 nfs_release_seqid(opendata->o_arg.seqid);
1846                 if (!opendata->is_recover) {
1847                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1848                         if (ret != 0)
1849                                 goto out;
1850                 }
1851                 ret = -EAGAIN;
1852
1853                 /* Try to update the stateid using the delegation */
1854                 if (update_open_stateid(state, NULL, &stateid, fmode))
1855                         goto out_return_state;
1856         }
1857 out:
1858         return ERR_PTR(ret);
1859 out_return_state:
1860         refcount_inc(&state->count);
1861         return state;
1862 }
1863
1864 static void
1865 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1866 {
1867         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1868         struct nfs_delegation *delegation;
1869         int delegation_flags = 0;
1870
1871         rcu_read_lock();
1872         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1873         if (delegation)
1874                 delegation_flags = delegation->flags;
1875         rcu_read_unlock();
1876         switch (data->o_arg.claim) {
1877         default:
1878                 break;
1879         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1880         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1881                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1882                                    "returning a delegation for "
1883                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1884                                    clp->cl_hostname);
1885                 return;
1886         }
1887         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1888                 nfs_inode_set_delegation(state->inode,
1889                                 data->owner->so_cred,
1890                                 data->o_res.delegation_type,
1891                                 &data->o_res.delegation,
1892                                 data->o_res.pagemod_limit);
1893         else
1894                 nfs_inode_reclaim_delegation(state->inode,
1895                                 data->owner->so_cred,
1896                                 data->o_res.delegation_type,
1897                                 &data->o_res.delegation,
1898                                 data->o_res.pagemod_limit);
1899
1900         if (data->o_res.do_recall)
1901                 nfs_async_inode_return_delegation(state->inode,
1902                                                   &data->o_res.delegation);
1903 }
1904
1905 /*
1906  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1907  * and update the nfs4_state.
1908  */
1909 static struct nfs4_state *
1910 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1911 {
1912         struct inode *inode = data->state->inode;
1913         struct nfs4_state *state = data->state;
1914         int ret;
1915
1916         if (!data->rpc_done) {
1917                 if (data->rpc_status)
1918                         return ERR_PTR(data->rpc_status);
1919                 /* cached opens have already been processed */
1920                 goto update;
1921         }
1922
1923         ret = nfs_refresh_inode(inode, &data->f_attr);
1924         if (ret)
1925                 return ERR_PTR(ret);
1926
1927         if (data->o_res.delegation_type != 0)
1928                 nfs4_opendata_check_deleg(data, state);
1929 update:
1930         if (!update_open_stateid(state, &data->o_res.stateid,
1931                                 NULL, data->o_arg.fmode))
1932                 return ERR_PTR(-EAGAIN);
1933         refcount_inc(&state->count);
1934
1935         return state;
1936 }
1937
1938 static struct inode *
1939 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1940 {
1941         struct inode *inode;
1942
1943         switch (data->o_arg.claim) {
1944         case NFS4_OPEN_CLAIM_NULL:
1945         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1946         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1947                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1948                         return ERR_PTR(-EAGAIN);
1949                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1950                                 &data->f_attr, data->f_label);
1951                 break;
1952         default:
1953                 inode = d_inode(data->dentry);
1954                 ihold(inode);
1955                 nfs_refresh_inode(inode, &data->f_attr);
1956         }
1957         return inode;
1958 }
1959
1960 static struct nfs4_state *
1961 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1962 {
1963         struct nfs4_state *state;
1964         struct inode *inode;
1965
1966         inode = nfs4_opendata_get_inode(data);
1967         if (IS_ERR(inode))
1968                 return ERR_CAST(inode);
1969         if (data->state != NULL && data->state->inode == inode) {
1970                 state = data->state;
1971                 refcount_inc(&state->count);
1972         } else
1973                 state = nfs4_get_open_state(inode, data->owner);
1974         iput(inode);
1975         if (state == NULL)
1976                 state = ERR_PTR(-ENOMEM);
1977         return state;
1978 }
1979
1980 static struct nfs4_state *
1981 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1982 {
1983         struct nfs4_state *state;
1984
1985         if (!data->rpc_done) {
1986                 state = nfs4_try_open_cached(data);
1987                 trace_nfs4_cached_open(data->state);
1988                 goto out;
1989         }
1990
1991         state = nfs4_opendata_find_nfs4_state(data);
1992         if (IS_ERR(state))
1993                 goto out;
1994
1995         if (data->o_res.delegation_type != 0)
1996                 nfs4_opendata_check_deleg(data, state);
1997         if (!update_open_stateid(state, &data->o_res.stateid,
1998                                 NULL, data->o_arg.fmode)) {
1999                 nfs4_put_open_state(state);
2000                 state = ERR_PTR(-EAGAIN);
2001         }
2002 out:
2003         nfs_release_seqid(data->o_arg.seqid);
2004         return state;
2005 }
2006
2007 static struct nfs4_state *
2008 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2009 {
2010         struct nfs4_state *ret;
2011
2012         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2013                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2014         else
2015                 ret = _nfs4_opendata_to_nfs4_state(data);
2016         nfs4_sequence_free_slot(&data->o_res.seq_res);
2017         return ret;
2018 }
2019
2020 static struct nfs_open_context *
2021 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2022 {
2023         struct nfs_inode *nfsi = NFS_I(state->inode);
2024         struct nfs_open_context *ctx;
2025
2026         rcu_read_lock();
2027         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2028                 if (ctx->state != state)
2029                         continue;
2030                 if ((ctx->mode & mode) != mode)
2031                         continue;
2032                 if (!get_nfs_open_context(ctx))
2033                         continue;
2034                 rcu_read_unlock();
2035                 return ctx;
2036         }
2037         rcu_read_unlock();
2038         return ERR_PTR(-ENOENT);
2039 }
2040
2041 static struct nfs_open_context *
2042 nfs4_state_find_open_context(struct nfs4_state *state)
2043 {
2044         struct nfs_open_context *ctx;
2045
2046         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2047         if (!IS_ERR(ctx))
2048                 return ctx;
2049         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2050         if (!IS_ERR(ctx))
2051                 return ctx;
2052         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2053 }
2054
2055 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2056                 struct nfs4_state *state, enum open_claim_type4 claim)
2057 {
2058         struct nfs4_opendata *opendata;
2059
2060         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2061                         NULL, claim, GFP_NOFS);
2062         if (opendata == NULL)
2063                 return ERR_PTR(-ENOMEM);
2064         opendata->state = state;
2065         refcount_inc(&state->count);
2066         return opendata;
2067 }
2068
2069 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2070                 fmode_t fmode)
2071 {
2072         struct nfs4_state *newstate;
2073         int ret;
2074
2075         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2076                 return 0;
2077         opendata->o_arg.open_flags = 0;
2078         opendata->o_arg.fmode = fmode;
2079         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2080                         NFS_SB(opendata->dentry->d_sb),
2081                         fmode, 0);
2082         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2083         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2084         nfs4_init_opendata_res(opendata);
2085         ret = _nfs4_recover_proc_open(opendata);
2086         if (ret != 0)
2087                 return ret;
2088         newstate = nfs4_opendata_to_nfs4_state(opendata);
2089         if (IS_ERR(newstate))
2090                 return PTR_ERR(newstate);
2091         if (newstate != opendata->state)
2092                 ret = -ESTALE;
2093         nfs4_close_state(newstate, fmode);
2094         return ret;
2095 }
2096
2097 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2098 {
2099         int ret;
2100
2101         /* memory barrier prior to reading state->n_* */
2102         smp_rmb();
2103         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2104         if (ret != 0)
2105                 return ret;
2106         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2107         if (ret != 0)
2108                 return ret;
2109         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2110         if (ret != 0)
2111                 return ret;
2112         /*
2113          * We may have performed cached opens for all three recoveries.
2114          * Check if we need to update the current stateid.
2115          */
2116         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2117             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2118                 write_seqlock(&state->seqlock);
2119                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2120                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2121                 write_sequnlock(&state->seqlock);
2122         }
2123         return 0;
2124 }
2125
2126 /*
2127  * OPEN_RECLAIM:
2128  *      reclaim state on the server after a reboot.
2129  */
2130 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2131 {
2132         struct nfs_delegation *delegation;
2133         struct nfs4_opendata *opendata;
2134         fmode_t delegation_type = 0;
2135         int status;
2136
2137         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2138                         NFS4_OPEN_CLAIM_PREVIOUS);
2139         if (IS_ERR(opendata))
2140                 return PTR_ERR(opendata);
2141         rcu_read_lock();
2142         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2143         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2144                 delegation_type = delegation->type;
2145         rcu_read_unlock();
2146         opendata->o_arg.u.delegation_type = delegation_type;
2147         status = nfs4_open_recover(opendata, state);
2148         nfs4_opendata_put(opendata);
2149         return status;
2150 }
2151
2152 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2153 {
2154         struct nfs_server *server = NFS_SERVER(state->inode);
2155         struct nfs4_exception exception = { };
2156         int err;
2157         do {
2158                 err = _nfs4_do_open_reclaim(ctx, state);
2159                 trace_nfs4_open_reclaim(ctx, 0, err);
2160                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2161                         continue;
2162                 if (err != -NFS4ERR_DELAY)
2163                         break;
2164                 nfs4_handle_exception(server, err, &exception);
2165         } while (exception.retry);
2166         return err;
2167 }
2168
2169 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2170 {
2171         struct nfs_open_context *ctx;
2172         int ret;
2173
2174         ctx = nfs4_state_find_open_context(state);
2175         if (IS_ERR(ctx))
2176                 return -EAGAIN;
2177         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2178         nfs_state_clear_open_state_flags(state);
2179         ret = nfs4_do_open_reclaim(ctx, state);
2180         put_nfs_open_context(ctx);
2181         return ret;
2182 }
2183
2184 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)
2185 {
2186         switch (err) {
2187                 default:
2188                         printk(KERN_ERR "NFS: %s: unhandled error "
2189                                         "%d.\n", __func__, err);
2190                 case 0:
2191                 case -ENOENT:
2192                 case -EAGAIN:
2193                 case -ESTALE:
2194                 case -ETIMEDOUT:
2195                         break;
2196                 case -NFS4ERR_BADSESSION:
2197                 case -NFS4ERR_BADSLOT:
2198                 case -NFS4ERR_BAD_HIGH_SLOT:
2199                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2200                 case -NFS4ERR_DEADSESSION:
2201                         return -EAGAIN;
2202                 case -NFS4ERR_STALE_CLIENTID:
2203                 case -NFS4ERR_STALE_STATEID:
2204                         /* Don't recall a delegation if it was lost */
2205                         nfs4_schedule_lease_recovery(server->nfs_client);
2206                         return -EAGAIN;
2207                 case -NFS4ERR_MOVED:
2208                         nfs4_schedule_migration_recovery(server);
2209                         return -EAGAIN;
2210                 case -NFS4ERR_LEASE_MOVED:
2211                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2212                         return -EAGAIN;
2213                 case -NFS4ERR_DELEG_REVOKED:
2214                 case -NFS4ERR_ADMIN_REVOKED:
2215                 case -NFS4ERR_EXPIRED:
2216                 case -NFS4ERR_BAD_STATEID:
2217                 case -NFS4ERR_OPENMODE:
2218                         nfs_inode_find_state_and_recover(state->inode,
2219                                         stateid);
2220                         nfs4_schedule_stateid_recovery(server, state);
2221                         return -EAGAIN;
2222                 case -NFS4ERR_DELAY:
2223                 case -NFS4ERR_GRACE:
2224                         ssleep(1);
2225                         return -EAGAIN;
2226                 case -ENOMEM:
2227                 case -NFS4ERR_DENIED:
2228                         if (fl) {
2229                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2230                                 if (lsp)
2231                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2232                         }
2233                         return 0;
2234         }
2235         return err;
2236 }
2237
2238 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2239                 struct nfs4_state *state, const nfs4_stateid *stateid)
2240 {
2241         struct nfs_server *server = NFS_SERVER(state->inode);
2242         struct nfs4_opendata *opendata;
2243         int err = 0;
2244
2245         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2246                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2247         if (IS_ERR(opendata))
2248                 return PTR_ERR(opendata);
2249         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2250         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2251                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2252                 if (err)
2253                         goto out;
2254         }
2255         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2256                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2257                 if (err)
2258                         goto out;
2259         }
2260         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2261                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2262                 if (err)
2263                         goto out;
2264         }
2265         nfs_state_clear_delegation(state);
2266 out:
2267         nfs4_opendata_put(opendata);
2268         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2269 }
2270
2271 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2272 {
2273         struct nfs4_opendata *data = calldata;
2274
2275         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2276                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2277 }
2278
2279 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2280 {
2281         struct nfs4_opendata *data = calldata;
2282
2283         nfs40_sequence_done(task, &data->c_res.seq_res);
2284
2285         data->rpc_status = task->tk_status;
2286         if (data->rpc_status == 0) {
2287                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2288                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2289                 renew_lease(data->o_res.server, data->timestamp);
2290                 data->rpc_done = true;
2291         }
2292 }
2293
2294 static void nfs4_open_confirm_release(void *calldata)
2295 {
2296         struct nfs4_opendata *data = calldata;
2297         struct nfs4_state *state = NULL;
2298
2299         /* If this request hasn't been cancelled, do nothing */
2300         if (!data->cancelled)
2301                 goto out_free;
2302         /* In case of error, no cleanup! */
2303         if (!data->rpc_done)
2304                 goto out_free;
2305         state = nfs4_opendata_to_nfs4_state(data);
2306         if (!IS_ERR(state))
2307                 nfs4_close_state(state, data->o_arg.fmode);
2308 out_free:
2309         nfs4_opendata_put(data);
2310 }
2311
2312 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2313         .rpc_call_prepare = nfs4_open_confirm_prepare,
2314         .rpc_call_done = nfs4_open_confirm_done,
2315         .rpc_release = nfs4_open_confirm_release,
2316 };
2317
2318 /*
2319  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2320  */
2321 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2322 {
2323         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2324         struct rpc_task *task;
2325         struct  rpc_message msg = {
2326                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2327                 .rpc_argp = &data->c_arg,
2328                 .rpc_resp = &data->c_res,
2329                 .rpc_cred = data->owner->so_cred,
2330         };
2331         struct rpc_task_setup task_setup_data = {
2332                 .rpc_client = server->client,
2333                 .rpc_message = &msg,
2334                 .callback_ops = &nfs4_open_confirm_ops,
2335                 .callback_data = data,
2336                 .workqueue = nfsiod_workqueue,
2337                 .flags = RPC_TASK_ASYNC,
2338         };
2339         int status;
2340
2341         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2342                                 data->is_recover);
2343         kref_get(&data->kref);
2344         data->rpc_done = false;
2345         data->rpc_status = 0;
2346         data->timestamp = jiffies;
2347         task = rpc_run_task(&task_setup_data);
2348         if (IS_ERR(task))
2349                 return PTR_ERR(task);
2350         status = rpc_wait_for_completion_task(task);
2351         if (status != 0) {
2352                 data->cancelled = true;
2353                 smp_wmb();
2354         } else
2355                 status = data->rpc_status;
2356         rpc_put_task(task);
2357         return status;
2358 }
2359
2360 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2361 {
2362         struct nfs4_opendata *data = calldata;
2363         struct nfs4_state_owner *sp = data->owner;
2364         struct nfs_client *clp = sp->so_server->nfs_client;
2365         enum open_claim_type4 claim = data->o_arg.claim;
2366
2367         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2368                 goto out_wait;
2369         /*
2370          * Check if we still need to send an OPEN call, or if we can use
2371          * a delegation instead.
2372          */
2373         if (data->state != NULL) {
2374                 struct nfs_delegation *delegation;
2375
2376                 if (can_open_cached(data->state, data->o_arg.fmode,
2377                                         data->o_arg.open_flags, claim))
2378                         goto out_no_action;
2379                 rcu_read_lock();
2380                 delegation = nfs4_get_valid_delegation(data->state->inode);
2381                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2382                         goto unlock_no_action;
2383                 rcu_read_unlock();
2384         }
2385         /* Update client id. */
2386         data->o_arg.clientid = clp->cl_clientid;
2387         switch (claim) {
2388         default:
2389                 break;
2390         case NFS4_OPEN_CLAIM_PREVIOUS:
2391         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2392         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2393                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2394                 /* Fall through */
2395         case NFS4_OPEN_CLAIM_FH:
2396                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2397         }
2398         data->timestamp = jiffies;
2399         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2400                                 &data->o_arg.seq_args,
2401                                 &data->o_res.seq_res,
2402                                 task) != 0)
2403                 nfs_release_seqid(data->o_arg.seqid);
2404
2405         /* Set the create mode (note dependency on the session type) */
2406         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2407         if (data->o_arg.open_flags & O_EXCL) {
2408                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2409                 if (nfs4_has_persistent_session(clp))
2410                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2411                 else if (clp->cl_mvops->minor_version > 0)
2412                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2413         }
2414         return;
2415 unlock_no_action:
2416         trace_nfs4_cached_open(data->state);
2417         rcu_read_unlock();
2418 out_no_action:
2419         task->tk_action = NULL;
2420 out_wait:
2421         nfs4_sequence_done(task, &data->o_res.seq_res);
2422 }
2423
2424 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2425 {
2426         struct nfs4_opendata *data = calldata;
2427
2428         data->rpc_status = task->tk_status;
2429
2430         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2431                 return;
2432
2433         if (task->tk_status == 0) {
2434                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2435                         switch (data->o_res.f_attr->mode & S_IFMT) {
2436                         case S_IFREG:
2437                                 break;
2438                         case S_IFLNK:
2439                                 data->rpc_status = -ELOOP;
2440                                 break;
2441                         case S_IFDIR:
2442                                 data->rpc_status = -EISDIR;
2443                                 break;
2444                         default:
2445                                 data->rpc_status = -ENOTDIR;
2446                         }
2447                 }
2448                 renew_lease(data->o_res.server, data->timestamp);
2449                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2450                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2451         }
2452         data->rpc_done = true;
2453 }
2454
2455 static void nfs4_open_release(void *calldata)
2456 {
2457         struct nfs4_opendata *data = calldata;
2458         struct nfs4_state *state = NULL;
2459
2460         /* If this request hasn't been cancelled, do nothing */
2461         if (!data->cancelled)
2462                 goto out_free;
2463         /* In case of error, no cleanup! */
2464         if (data->rpc_status != 0 || !data->rpc_done)
2465                 goto out_free;
2466         /* In case we need an open_confirm, no cleanup! */
2467         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2468                 goto out_free;
2469         state = nfs4_opendata_to_nfs4_state(data);
2470         if (!IS_ERR(state))
2471                 nfs4_close_state(state, data->o_arg.fmode);
2472 out_free:
2473         nfs4_opendata_put(data);
2474 }
2475
2476 static const struct rpc_call_ops nfs4_open_ops = {
2477         .rpc_call_prepare = nfs4_open_prepare,
2478         .rpc_call_done = nfs4_open_done,
2479         .rpc_release = nfs4_open_release,
2480 };
2481
2482 static int nfs4_run_open_task(struct nfs4_opendata *data,
2483                               struct nfs_open_context *ctx)
2484 {
2485         struct inode *dir = d_inode(data->dir);
2486         struct nfs_server *server = NFS_SERVER(dir);
2487         struct nfs_openargs *o_arg = &data->o_arg;
2488         struct nfs_openres *o_res = &data->o_res;
2489         struct rpc_task *task;
2490         struct rpc_message msg = {
2491                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2492                 .rpc_argp = o_arg,
2493                 .rpc_resp = o_res,
2494                 .rpc_cred = data->owner->so_cred,
2495         };
2496         struct rpc_task_setup task_setup_data = {
2497                 .rpc_client = server->client,
2498                 .rpc_message = &msg,
2499                 .callback_ops = &nfs4_open_ops,
2500                 .callback_data = data,
2501                 .workqueue = nfsiod_workqueue,
2502                 .flags = RPC_TASK_ASYNC,
2503         };
2504         int status;
2505
2506         kref_get(&data->kref);
2507         data->rpc_done = false;
2508         data->rpc_status = 0;
2509         data->cancelled = false;
2510         data->is_recover = false;
2511         if (!ctx) {
2512                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2513                 data->is_recover = true;
2514                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2515         } else {
2516                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2517                 pnfs_lgopen_prepare(data, ctx);
2518         }
2519         task = rpc_run_task(&task_setup_data);
2520         if (IS_ERR(task))
2521                 return PTR_ERR(task);
2522         status = rpc_wait_for_completion_task(task);
2523         if (status != 0) {
2524                 data->cancelled = true;
2525                 smp_wmb();
2526         } else
2527                 status = data->rpc_status;
2528         rpc_put_task(task);
2529
2530         return status;
2531 }
2532
2533 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2534 {
2535         struct inode *dir = d_inode(data->dir);
2536         struct nfs_openres *o_res = &data->o_res;
2537         int status;
2538
2539         status = nfs4_run_open_task(data, NULL);
2540         if (status != 0 || !data->rpc_done)
2541                 return status;
2542
2543         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2544
2545         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2546                 status = _nfs4_proc_open_confirm(data);
2547
2548         return status;
2549 }
2550
2551 /*
2552  * Additional permission checks in order to distinguish between an
2553  * open for read, and an open for execute. This works around the
2554  * fact that NFSv4 OPEN treats read and execute permissions as being
2555  * the same.
2556  * Note that in the non-execute case, we want to turn off permission
2557  * checking if we just created a new file (POSIX open() semantics).
2558  */
2559 static int nfs4_opendata_access(const struct cred *cred,
2560                                 struct nfs4_opendata *opendata,
2561                                 struct nfs4_state *state, fmode_t fmode,
2562                                 int openflags)
2563 {
2564         struct nfs_access_entry cache;
2565         u32 mask, flags;
2566
2567         /* access call failed or for some reason the server doesn't
2568          * support any access modes -- defer access call until later */
2569         if (opendata->o_res.access_supported == 0)
2570                 return 0;
2571
2572         mask = 0;
2573         /*
2574          * Use openflags to check for exec, because fmode won't
2575          * always have FMODE_EXEC set when file open for exec.
2576          */
2577         if (openflags & __FMODE_EXEC) {
2578                 /* ONLY check for exec rights */
2579                 if (S_ISDIR(state->inode->i_mode))
2580                         mask = NFS4_ACCESS_LOOKUP;
2581                 else
2582                         mask = NFS4_ACCESS_EXECUTE;
2583         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2584                 mask = NFS4_ACCESS_READ;
2585
2586         cache.cred = cred;
2587         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2588         nfs_access_add_cache(state->inode, &cache);
2589
2590         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2591         if ((mask & ~cache.mask & flags) == 0)
2592                 return 0;
2593
2594         return -EACCES;
2595 }
2596
2597 /*
2598  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2599  */
2600 static int _nfs4_proc_open(struct nfs4_opendata *data,
2601                            struct nfs_open_context *ctx)
2602 {
2603         struct inode *dir = d_inode(data->dir);
2604         struct nfs_server *server = NFS_SERVER(dir);
2605         struct nfs_openargs *o_arg = &data->o_arg;
2606         struct nfs_openres *o_res = &data->o_res;
2607         int status;
2608
2609         status = nfs4_run_open_task(data, ctx);
2610         if (!data->rpc_done)
2611                 return status;
2612         if (status != 0) {
2613                 if (status == -NFS4ERR_BADNAME &&
2614                                 !(o_arg->open_flags & O_CREAT))
2615                         return -ENOENT;
2616                 return status;
2617         }
2618
2619         nfs_fattr_map_and_free_names(server, &data->f_attr);
2620
2621         if (o_arg->open_flags & O_CREAT) {
2622                 if (o_arg->open_flags & O_EXCL)
2623                         data->file_created = true;
2624                 else if (o_res->cinfo.before != o_res->cinfo.after)
2625                         data->file_created = true;
2626                 if (data->file_created ||
2627                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2628                         update_changeattr(dir, &o_res->cinfo,
2629                                         o_res->f_attr->time_start, 0);
2630         }
2631         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2632                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2633         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2634                 status = _nfs4_proc_open_confirm(data);
2635                 if (status != 0)
2636                         return status;
2637         }
2638         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2639                 nfs4_sequence_free_slot(&o_res->seq_res);
2640                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2641                                 o_res->f_label, NULL);
2642         }
2643         return 0;
2644 }
2645
2646 /*
2647  * OPEN_EXPIRED:
2648  *      reclaim state on the server after a network partition.
2649  *      Assumes caller holds the appropriate lock
2650  */
2651 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2652 {
2653         struct nfs4_opendata *opendata;
2654         int ret;
2655
2656         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2657                         NFS4_OPEN_CLAIM_FH);
2658         if (IS_ERR(opendata))
2659                 return PTR_ERR(opendata);
2660         ret = nfs4_open_recover(opendata, state);
2661         if (ret == -ESTALE)
2662                 d_drop(ctx->dentry);
2663         nfs4_opendata_put(opendata);
2664         return ret;
2665 }
2666
2667 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2668 {
2669         struct nfs_server *server = NFS_SERVER(state->inode);
2670         struct nfs4_exception exception = { };
2671         int err;
2672
2673         do {
2674                 err = _nfs4_open_expired(ctx, state);
2675                 trace_nfs4_open_expired(ctx, 0, err);
2676                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2677                         continue;
2678                 switch (err) {
2679                 default:
2680                         goto out;
2681                 case -NFS4ERR_GRACE:
2682                 case -NFS4ERR_DELAY:
2683                         nfs4_handle_exception(server, err, &exception);
2684                         err = 0;
2685                 }
2686         } while (exception.retry);
2687 out:
2688         return err;
2689 }
2690
2691 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2692 {
2693         struct nfs_open_context *ctx;
2694         int ret;
2695
2696         ctx = nfs4_state_find_open_context(state);
2697         if (IS_ERR(ctx))
2698                 return -EAGAIN;
2699         ret = nfs4_do_open_expired(ctx, state);
2700         put_nfs_open_context(ctx);
2701         return ret;
2702 }
2703
2704 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2705                 const nfs4_stateid *stateid)
2706 {
2707         nfs_remove_bad_delegation(state->inode, stateid);
2708         nfs_state_clear_delegation(state);
2709 }
2710
2711 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2712 {
2713         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2714                 nfs_finish_clear_delegation_stateid(state, NULL);
2715 }
2716
2717 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2718 {
2719         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2720         nfs40_clear_delegation_stateid(state);
2721         nfs_state_clear_open_state_flags(state);
2722         return nfs4_open_expired(sp, state);
2723 }
2724
2725 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2726                 nfs4_stateid *stateid,
2727                 const struct cred *cred)
2728 {
2729         return -NFS4ERR_BAD_STATEID;
2730 }
2731
2732 #if defined(CONFIG_NFS_V4_1)
2733 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2734                 nfs4_stateid *stateid,
2735                 const struct cred *cred)
2736 {
2737         int status;
2738
2739         switch (stateid->type) {
2740         default:
2741                 break;
2742         case NFS4_INVALID_STATEID_TYPE:
2743         case NFS4_SPECIAL_STATEID_TYPE:
2744                 return -NFS4ERR_BAD_STATEID;
2745         case NFS4_REVOKED_STATEID_TYPE:
2746                 goto out_free;
2747         }
2748
2749         status = nfs41_test_stateid(server, stateid, cred);
2750         switch (status) {
2751         case -NFS4ERR_EXPIRED:
2752         case -NFS4ERR_ADMIN_REVOKED:
2753         case -NFS4ERR_DELEG_REVOKED:
2754                 break;
2755         default:
2756                 return status;
2757         }
2758 out_free:
2759         /* Ack the revoked state to the server */
2760         nfs41_free_stateid(server, stateid, cred, true);
2761         return -NFS4ERR_EXPIRED;
2762 }
2763
2764 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2765 {
2766         struct nfs_server *server = NFS_SERVER(state->inode);
2767         nfs4_stateid stateid;
2768         struct nfs_delegation *delegation;
2769         const struct cred *cred = NULL;
2770         int status, ret = NFS_OK;
2771
2772         /* Get the delegation credential for use by test/free_stateid */
2773         rcu_read_lock();
2774         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2775         if (delegation == NULL) {
2776                 rcu_read_unlock();
2777                 nfs_state_clear_delegation(state);
2778                 return NFS_OK;
2779         }
2780
2781         nfs4_stateid_copy(&stateid, &delegation->stateid);
2782
2783         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2784                                 &delegation->flags)) {
2785                 rcu_read_unlock();
2786                 return NFS_OK;
2787         }
2788
2789         if (delegation->cred)
2790                 cred = get_cred(delegation->cred);
2791         rcu_read_unlock();
2792         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2793         trace_nfs4_test_delegation_stateid(state, NULL, status);
2794         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2795                 nfs_finish_clear_delegation_stateid(state, &stateid);
2796         else
2797                 ret = status;
2798
2799         put_cred(cred);
2800         return ret;
2801 }
2802
2803 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2804 {
2805         nfs4_stateid tmp;
2806
2807         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2808             nfs4_copy_delegation_stateid(state->inode, state->state,
2809                                 &tmp, NULL) &&
2810             nfs4_stateid_match_other(&state->stateid, &tmp))
2811                 nfs_state_set_delegation(state, &tmp, state->state);
2812         else
2813                 nfs_state_clear_delegation(state);
2814 }
2815
2816 /**
2817  * nfs41_check_expired_locks - possibly free a lock stateid
2818  *
2819  * @state: NFSv4 state for an inode
2820  *
2821  * Returns NFS_OK if recovery for this stateid is now finished.
2822  * Otherwise a negative NFS4ERR value is returned.
2823  */
2824 static int nfs41_check_expired_locks(struct nfs4_state *state)
2825 {
2826         int status, ret = NFS_OK;
2827         struct nfs4_lock_state *lsp, *prev = NULL;
2828         struct nfs_server *server = NFS_SERVER(state->inode);
2829
2830         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2831                 goto out;
2832
2833         spin_lock(&state->state_lock);
2834         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2835                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2836                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2837
2838                         refcount_inc(&lsp->ls_count);
2839                         spin_unlock(&state->state_lock);
2840
2841                         nfs4_put_lock_state(prev);
2842                         prev = lsp;
2843
2844                         status = nfs41_test_and_free_expired_stateid(server,
2845                                         &lsp->ls_stateid,
2846                                         cred);
2847                         trace_nfs4_test_lock_stateid(state, lsp, status);
2848                         if (status == -NFS4ERR_EXPIRED ||
2849                             status == -NFS4ERR_BAD_STATEID) {
2850                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2851                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2852                                 if (!recover_lost_locks)
2853                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2854                         } else if (status != NFS_OK) {
2855                                 ret = status;
2856                                 nfs4_put_lock_state(prev);
2857                                 goto out;
2858                         }
2859                         spin_lock(&state->state_lock);
2860                 }
2861         }
2862         spin_unlock(&state->state_lock);
2863         nfs4_put_lock_state(prev);
2864 out:
2865         return ret;
2866 }
2867
2868 /**
2869  * nfs41_check_open_stateid - possibly free an open stateid
2870  *
2871  * @state: NFSv4 state for an inode
2872  *
2873  * Returns NFS_OK if recovery for this stateid is now finished.
2874  * Otherwise a negative NFS4ERR value is returned.
2875  */
2876 static int nfs41_check_open_stateid(struct nfs4_state *state)
2877 {
2878         struct nfs_server *server = NFS_SERVER(state->inode);
2879         nfs4_stateid *stateid = &state->open_stateid;
2880         const struct cred *cred = state->owner->so_cred;
2881         int status;
2882
2883         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2884                 return -NFS4ERR_BAD_STATEID;
2885         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2886         trace_nfs4_test_open_stateid(state, NULL, status);
2887         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2888                 nfs_state_clear_open_state_flags(state);
2889                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2890                 return status;
2891         }
2892         if (nfs_open_stateid_recover_openmode(state))
2893                 return -NFS4ERR_OPENMODE;
2894         return NFS_OK;
2895 }
2896
2897 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2898 {
2899         int status;
2900
2901         status = nfs41_check_delegation_stateid(state);
2902         if (status != NFS_OK)
2903                 return status;
2904         nfs41_delegation_recover_stateid(state);
2905
2906         status = nfs41_check_expired_locks(state);
2907         if (status != NFS_OK)
2908                 return status;
2909         status = nfs41_check_open_stateid(state);
2910         if (status != NFS_OK)
2911                 status = nfs4_open_expired(sp, state);
2912         return status;
2913 }
2914 #endif
2915
2916 /*
2917  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2918  * fields corresponding to attributes that were used to store the verifier.
2919  * Make sure we clobber those fields in the later setattr call
2920  */
2921 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2922                                 struct iattr *sattr, struct nfs4_label **label)
2923 {
2924         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2925         __u32 attrset[3];
2926         unsigned ret;
2927         unsigned i;
2928
2929         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2930                 attrset[i] = opendata->o_res.attrset[i];
2931                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2932                         attrset[i] &= ~bitmask[i];
2933         }
2934
2935         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2936                 sattr->ia_valid : 0;
2937
2938         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2939                 if (sattr->ia_valid & ATTR_ATIME_SET)
2940                         ret |= ATTR_ATIME_SET;
2941                 else
2942                         ret |= ATTR_ATIME;
2943         }
2944
2945         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2946                 if (sattr->ia_valid & ATTR_MTIME_SET)
2947                         ret |= ATTR_MTIME_SET;
2948                 else
2949                         ret |= ATTR_MTIME;
2950         }
2951
2952         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2953                 *label = NULL;
2954         return ret;
2955 }
2956
2957 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2958                 int flags, struct nfs_open_context *ctx)
2959 {
2960         struct nfs4_state_owner *sp = opendata->owner;
2961         struct nfs_server *server = sp->so_server;
2962         struct dentry *dentry;
2963         struct nfs4_state *state;
2964         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
2965         unsigned int seq;
2966         int ret;
2967
2968         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2969
2970         ret = _nfs4_proc_open(opendata, ctx);
2971         if (ret != 0)
2972                 goto out;
2973
2974         state = _nfs4_opendata_to_nfs4_state(opendata);
2975         ret = PTR_ERR(state);
2976         if (IS_ERR(state))
2977                 goto out;
2978         ctx->state = state;
2979         if (server->caps & NFS_CAP_POSIX_LOCK)
2980                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2981         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2982                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2983
2984         dentry = opendata->dentry;
2985         if (d_really_is_negative(dentry)) {
2986                 struct dentry *alias;
2987                 d_drop(dentry);
2988                 alias = d_exact_alias(dentry, state->inode);
2989                 if (!alias)
2990                         alias = d_splice_alias(igrab(state->inode), dentry);
2991                 /* d_splice_alias() can't fail here - it's a non-directory */
2992                 if (alias) {
2993                         dput(ctx->dentry);
2994                         ctx->dentry = dentry = alias;
2995                 }
2996                 nfs_set_verifier(dentry,
2997                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2998         }
2999
3000         /* Parse layoutget results before we check for access */
3001         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3002
3003         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3004                         acc_mode, flags);
3005         if (ret != 0)
3006                 goto out;
3007
3008         if (d_inode(dentry) == state->inode) {
3009                 nfs_inode_attach_open_context(ctx);
3010                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3011                         nfs4_schedule_stateid_recovery(server, state);
3012         }
3013
3014 out:
3015         if (!opendata->cancelled)
3016                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3017         return ret;
3018 }
3019
3020 /*
3021  * Returns a referenced nfs4_state
3022  */
3023 static int _nfs4_do_open(struct inode *dir,
3024                         struct nfs_open_context *ctx,
3025                         int flags,
3026                         const struct nfs4_open_createattrs *c,
3027                         int *opened)
3028 {
3029         struct nfs4_state_owner  *sp;
3030         struct nfs4_state     *state = NULL;
3031         struct nfs_server       *server = NFS_SERVER(dir);
3032         struct nfs4_opendata *opendata;
3033         struct dentry *dentry = ctx->dentry;
3034         const struct cred *cred = ctx->cred;
3035         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3036         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3037         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3038         struct iattr *sattr = c->sattr;
3039         struct nfs4_label *label = c->label;
3040         struct nfs4_label *olabel = NULL;
3041         int status;
3042
3043         /* Protect against reboot recovery conflicts */
3044         status = -ENOMEM;
3045         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3046         if (sp == NULL) {
3047                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3048                 goto out_err;
3049         }
3050         status = nfs4_client_recover_expired_lease(server->nfs_client);
3051         if (status != 0)
3052                 goto err_put_state_owner;
3053         if (d_really_is_positive(dentry))
3054                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3055         status = -ENOMEM;
3056         if (d_really_is_positive(dentry))
3057                 claim = NFS4_OPEN_CLAIM_FH;
3058         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3059                         c, claim, GFP_KERNEL);
3060         if (opendata == NULL)
3061                 goto err_put_state_owner;
3062
3063         if (label) {
3064                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3065                 if (IS_ERR(olabel)) {
3066                         status = PTR_ERR(olabel);
3067                         goto err_opendata_put;
3068                 }
3069         }
3070
3071         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3072                 if (!opendata->f_attr.mdsthreshold) {
3073                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3074                         if (!opendata->f_attr.mdsthreshold)
3075                                 goto err_free_label;
3076                 }
3077                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3078         }
3079         if (d_really_is_positive(dentry))
3080                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3081
3082         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3083         if (status != 0)
3084                 goto err_free_label;
3085         state = ctx->state;
3086
3087         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3088             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3089                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3090                 /*
3091                  * send create attributes which was not set by open
3092                  * with an extra setattr.
3093                  */
3094                 if (attrs || label) {
3095                         unsigned ia_old = sattr->ia_valid;
3096
3097                         sattr->ia_valid = attrs;
3098                         nfs_fattr_init(opendata->o_res.f_attr);
3099                         status = nfs4_do_setattr(state->inode, cred,
3100                                         opendata->o_res.f_attr, sattr,
3101                                         ctx, label, olabel);
3102                         if (status == 0) {
3103                                 nfs_setattr_update_inode(state->inode, sattr,
3104                                                 opendata->o_res.f_attr);
3105                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3106                         }
3107                         sattr->ia_valid = ia_old;
3108                 }
3109         }
3110         if (opened && opendata->file_created)
3111                 *opened = 1;
3112
3113         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3114                 *ctx_th = opendata->f_attr.mdsthreshold;
3115                 opendata->f_attr.mdsthreshold = NULL;
3116         }
3117
3118         nfs4_label_free(olabel);
3119
3120         nfs4_opendata_put(opendata);
3121         nfs4_put_state_owner(sp);
3122         return 0;
3123 err_free_label:
3124         nfs4_label_free(olabel);
3125 err_opendata_put:
3126         nfs4_opendata_put(opendata);
3127 err_put_state_owner:
3128         nfs4_put_state_owner(sp);
3129 out_err:
3130         return status;
3131 }
3132
3133
3134 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3135                                         struct nfs_open_context *ctx,
3136                                         int flags,
3137                                         struct iattr *sattr,
3138                                         struct nfs4_label *label,
3139                                         int *opened)
3140 {
3141         struct nfs_server *server = NFS_SERVER(dir);
3142         struct nfs4_exception exception = {
3143                 .interruptible = true,
3144         };
3145         struct nfs4_state *res;
3146         struct nfs4_open_createattrs c = {
3147                 .label = label,
3148                 .sattr = sattr,
3149                 .verf = {
3150                         [0] = (__u32)jiffies,
3151                         [1] = (__u32)current->pid,
3152                 },
3153         };
3154         int status;
3155
3156         do {
3157                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3158                 res = ctx->state;
3159                 trace_nfs4_open_file(ctx, flags, status);
3160                 if (status == 0)
3161                         break;
3162                 /* NOTE: BAD_SEQID means the server and client disagree about the
3163                  * book-keeping w.r.t. state-changing operations
3164                  * (OPEN/CLOSE/LOCK/LOCKU...)
3165                  * It is actually a sign of a bug on the client or on the server.
3166                  *
3167                  * If we receive a BAD_SEQID error in the particular case of
3168                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3169                  * have unhashed the old state_owner for us, and that we can
3170                  * therefore safely retry using a new one. We should still warn
3171                  * the user though...
3172                  */
3173                 if (status == -NFS4ERR_BAD_SEQID) {
3174                         pr_warn_ratelimited("NFS: v4 server %s "
3175                                         " returned a bad sequence-id error!\n",
3176                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3177                         exception.retry = 1;
3178                         continue;
3179                 }
3180                 /*
3181                  * BAD_STATEID on OPEN means that the server cancelled our
3182                  * state before it received the OPEN_CONFIRM.
3183                  * Recover by retrying the request as per the discussion
3184                  * on Page 181 of RFC3530.
3185                  */
3186                 if (status == -NFS4ERR_BAD_STATEID) {
3187                         exception.retry = 1;
3188                         continue;
3189                 }
3190                 if (status == -EAGAIN) {
3191                         /* We must have found a delegation */
3192                         exception.retry = 1;
3193                         continue;
3194                 }
3195                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3196                         continue;
3197                 res = ERR_PTR(nfs4_handle_exception(server,
3198                                         status, &exception));
3199         } while (exception.retry);
3200         return res;
3201 }
3202
3203 static int _nfs4_do_setattr(struct inode *inode,
3204                             struct nfs_setattrargs *arg,
3205                             struct nfs_setattrres *res,
3206                             const struct cred *cred,
3207                             struct nfs_open_context *ctx)
3208 {
3209         struct nfs_server *server = NFS_SERVER(inode);
3210         struct rpc_message msg = {
3211                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3212                 .rpc_argp       = arg,
3213                 .rpc_resp       = res,
3214                 .rpc_cred       = cred,
3215         };
3216         const struct cred *delegation_cred = NULL;
3217         unsigned long timestamp = jiffies;
3218         bool truncate;
3219         int status;
3220
3221         nfs_fattr_init(res->fattr);
3222
3223         /* Servers should only apply open mode checks for file size changes */
3224         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3225         if (!truncate)
3226                 goto zero_stateid;
3227
3228         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3229                 /* Use that stateid */
3230         } else if (ctx != NULL && ctx->state) {
3231                 struct nfs_lock_context *l_ctx;
3232                 if (!nfs4_valid_open_stateid(ctx->state))
3233                         return -EBADF;
3234                 l_ctx = nfs_get_lock_context(ctx);
3235                 if (IS_ERR(l_ctx))
3236                         return PTR_ERR(l_ctx);
3237                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3238                                                 &arg->stateid, &delegation_cred);
3239                 nfs_put_lock_context(l_ctx);
3240                 if (status == -EIO)
3241                         return -EBADF;
3242         } else {
3243 zero_stateid:
3244                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3245         }
3246         if (delegation_cred)
3247                 msg.rpc_cred = delegation_cred;
3248
3249         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3250
3251         put_cred(delegation_cred);
3252         if (status == 0 && ctx != NULL)
3253                 renew_lease(server, timestamp);
3254         trace_nfs4_setattr(inode, &arg->stateid, status);
3255         return status;
3256 }
3257
3258 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3259                            struct nfs_fattr *fattr, struct iattr *sattr,
3260                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3261                            struct nfs4_label *olabel)
3262 {
3263         struct nfs_server *server = NFS_SERVER(inode);
3264         __u32 bitmask[NFS4_BITMASK_SZ];
3265         struct nfs4_state *state = ctx ? ctx->state : NULL;
3266         struct nfs_setattrargs  arg = {
3267                 .fh             = NFS_FH(inode),
3268                 .iap            = sattr,
3269                 .server         = server,
3270                 .bitmask = bitmask,
3271                 .label          = ilabel,
3272         };
3273         struct nfs_setattrres  res = {
3274                 .fattr          = fattr,
3275                 .label          = olabel,
3276                 .server         = server,
3277         };
3278         struct nfs4_exception exception = {
3279                 .state = state,
3280                 .inode = inode,
3281                 .stateid = &arg.stateid,
3282         };
3283         int err;
3284
3285         do {
3286                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3287                                 nfs4_bitmask(server, olabel),
3288                                 inode);
3289
3290                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3291                 switch (err) {
3292                 case -NFS4ERR_OPENMODE:
3293                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3294                                 pr_warn_once("NFSv4: server %s is incorrectly "
3295                                                 "applying open mode checks to "
3296                                                 "a SETATTR that is not "
3297                                                 "changing file size.\n",
3298                                                 server->nfs_client->cl_hostname);
3299                         }
3300                         if (state && !(state->state & FMODE_WRITE)) {
3301                                 err = -EBADF;
3302                                 if (sattr->ia_valid & ATTR_OPEN)
3303                                         err = -EACCES;
3304                                 goto out;
3305                         }
3306                 }
3307                 err = nfs4_handle_exception(server, err, &exception);
3308         } while (exception.retry);
3309 out:
3310         return err;
3311 }
3312
3313 static bool
3314 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3315 {
3316         if (inode == NULL || !nfs_have_layout(inode))
3317                 return false;
3318
3319         return pnfs_wait_on_layoutreturn(inode, task);
3320 }
3321
3322 /*
3323  * Update the seqid of an open stateid
3324  */
3325 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3326                 struct nfs4_state *state)
3327 {
3328         __be32 seqid_open;
3329         u32 dst_seqid;
3330         int seq;
3331
3332         for (;;) {
3333                 if (!nfs4_valid_open_stateid(state))
3334                         break;
3335                 seq = read_seqbegin(&state->seqlock);
3336                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3337                         nfs4_stateid_copy(dst, &state->open_stateid);
3338                         if (read_seqretry(&state->seqlock, seq))
3339                                 continue;
3340                         break;
3341                 }
3342                 seqid_open = state->open_stateid.seqid;
3343                 if (read_seqretry(&state->seqlock, seq))
3344                         continue;
3345
3346                 dst_seqid = be32_to_cpu(dst->seqid);
3347                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3348                         dst->seqid = seqid_open;
3349                 break;
3350         }
3351 }
3352
3353 /*
3354  * Update the seqid of an open stateid after receiving
3355  * NFS4ERR_OLD_STATEID
3356  */
3357 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3358                 struct nfs4_state *state)
3359 {
3360         __be32 seqid_open;
3361         u32 dst_seqid;
3362         bool ret;
3363         int seq;
3364
3365         for (;;) {
3366                 ret = false;
3367                 if (!nfs4_valid_open_stateid(state))
3368                         break;
3369                 seq = read_seqbegin(&state->seqlock);
3370                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3371                         if (read_seqretry(&state->seqlock, seq))
3372                                 continue;
3373                         break;
3374                 }
3375                 seqid_open = state->open_stateid.seqid;
3376                 if (read_seqretry(&state->seqlock, seq))
3377                         continue;
3378
3379                 dst_seqid = be32_to_cpu(dst->seqid);
3380                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) >= 0)
3381                         dst->seqid = cpu_to_be32(dst_seqid + 1);
3382                 else
3383                         dst->seqid = seqid_open;
3384                 ret = true;
3385                 break;
3386         }
3387
3388         return ret;
3389 }
3390
3391 struct nfs4_closedata {
3392         struct inode *inode;
3393         struct nfs4_state *state;
3394         struct nfs_closeargs arg;
3395         struct nfs_closeres res;
3396         struct {
3397                 struct nfs4_layoutreturn_args arg;
3398                 struct nfs4_layoutreturn_res res;
3399                 struct nfs4_xdr_opaque_data ld_private;
3400                 u32 roc_barrier;
3401                 bool roc;
3402         } lr;
3403         struct nfs_fattr fattr;
3404         unsigned long timestamp;
3405 };
3406
3407 static void nfs4_free_closedata(void *data)
3408 {
3409         struct nfs4_closedata *calldata = data;
3410         struct nfs4_state_owner *sp = calldata->state->owner;
3411         struct super_block *sb = calldata->state->inode->i_sb;
3412
3413         if (calldata->lr.roc)
3414                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3415                                 calldata->res.lr_ret);
3416         nfs4_put_open_state(calldata->state);
3417         nfs_free_seqid(calldata->arg.seqid);
3418         nfs4_put_state_owner(sp);
3419         nfs_sb_deactive(sb);
3420         kfree(calldata);
3421 }
3422
3423 static void nfs4_close_done(struct rpc_task *task, void *data)
3424 {
3425         struct nfs4_closedata *calldata = data;
3426         struct nfs4_state *state = calldata->state;
3427         struct nfs_server *server = NFS_SERVER(calldata->inode);
3428         nfs4_stateid *res_stateid = NULL;
3429         struct nfs4_exception exception = {
3430                 .state = state,
3431                 .inode = calldata->inode,
3432                 .stateid = &calldata->arg.stateid,
3433         };
3434
3435         dprintk("%s: begin!\n", __func__);
3436         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3437                 return;
3438         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3439
3440         /* Handle Layoutreturn errors */
3441         if (pnfs_roc_done(task, calldata->inode,
3442                                 &calldata->arg.lr_args,
3443                                 &calldata->res.lr_res,
3444                                 &calldata->res.lr_ret) == -EAGAIN)
3445                 goto out_restart;
3446
3447         /* hmm. we are done with the inode, and in the process of freeing
3448          * the state_owner. we keep this around to process errors
3449          */
3450         switch (task->tk_status) {
3451                 case 0:
3452                         res_stateid = &calldata->res.stateid;
3453                         renew_lease(server, calldata->timestamp);
3454                         break;
3455                 case -NFS4ERR_ACCESS:
3456                         if (calldata->arg.bitmask != NULL) {
3457                                 calldata->arg.bitmask = NULL;
3458                                 calldata->res.fattr = NULL;
3459                                 goto out_restart;
3460
3461                         }
3462                         break;
3463                 case -NFS4ERR_OLD_STATEID:
3464                         /* Did we race with OPEN? */
3465                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3466                                                 state))
3467                                 goto out_restart;
3468                         goto out_release;
3469                 case -NFS4ERR_ADMIN_REVOKED:
3470                 case -NFS4ERR_STALE_STATEID:
3471                 case -NFS4ERR_EXPIRED:
3472                         nfs4_free_revoked_stateid(server,
3473                                         &calldata->arg.stateid,
3474                                         task->tk_msg.rpc_cred);
3475                         /* Fallthrough */
3476                 case -NFS4ERR_BAD_STATEID:
3477                         if (calldata->arg.fmode == 0)
3478                                 break;
3479                         /* Fallthrough */
3480                 default:
3481                         task->tk_status = nfs4_async_handle_exception(task,
3482                                         server, task->tk_status, &exception);
3483                         if (exception.retry)
3484                                 goto out_restart;
3485         }
3486         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3487                         res_stateid, calldata->arg.fmode);
3488 out_release:
3489         task->tk_status = 0;
3490         nfs_release_seqid(calldata->arg.seqid);
3491         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3492         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3493         return;
3494 out_restart:
3495         task->tk_status = 0;
3496         rpc_restart_call_prepare(task);
3497         goto out_release;
3498 }
3499
3500 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3501 {
3502         struct nfs4_closedata *calldata = data;
3503         struct nfs4_state *state = calldata->state;
3504         struct inode *inode = calldata->inode;
3505         struct pnfs_layout_hdr *lo;
3506         bool is_rdonly, is_wronly, is_rdwr;
3507         int call_close = 0;
3508
3509         dprintk("%s: begin!\n", __func__);
3510         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3511                 goto out_wait;
3512
3513         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3514         spin_lock(&state->owner->so_lock);
3515         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3516         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3517         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3518         /* Calculate the change in open mode */
3519         calldata->arg.fmode = 0;
3520         if (state->n_rdwr == 0) {
3521                 if (state->n_rdonly == 0)
3522                         call_close |= is_rdonly;
3523                 else if (is_rdonly)
3524                         calldata->arg.fmode |= FMODE_READ;
3525                 if (state->n_wronly == 0)
3526                         call_close |= is_wronly;
3527                 else if (is_wronly)
3528                         calldata->arg.fmode |= FMODE_WRITE;
3529                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3530                         call_close |= is_rdwr;
3531         } else if (is_rdwr)
3532                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3533
3534         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3535         if (!nfs4_valid_open_stateid(state))
3536                 call_close = 0;
3537         spin_unlock(&state->owner->so_lock);
3538
3539         if (!call_close) {
3540                 /* Note: exit _without_ calling nfs4_close_done */
3541                 goto out_no_action;
3542         }
3543
3544         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3545                 nfs_release_seqid(calldata->arg.seqid);
3546                 goto out_wait;
3547         }
3548
3549         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3550         if (lo && !pnfs_layout_is_valid(lo)) {
3551                 calldata->arg.lr_args = NULL;
3552                 calldata->res.lr_res = NULL;
3553         }
3554
3555         if (calldata->arg.fmode == 0)
3556                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3557
3558         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3559                 /* Close-to-open cache consistency revalidation */
3560                 if (!nfs4_have_delegation(inode, FMODE_READ))
3561                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3562                 else
3563                         calldata->arg.bitmask = NULL;
3564         }
3565
3566         calldata->arg.share_access =
3567                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3568                                 calldata->arg.fmode, 0);
3569
3570         if (calldata->res.fattr == NULL)
3571                 calldata->arg.bitmask = NULL;
3572         else if (calldata->arg.bitmask == NULL)
3573                 calldata->res.fattr = NULL;
3574         calldata->timestamp = jiffies;
3575         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3576                                 &calldata->arg.seq_args,
3577                                 &calldata->res.seq_res,
3578                                 task) != 0)
3579                 nfs_release_seqid(calldata->arg.seqid);
3580         dprintk("%s: done!\n", __func__);
3581         return;
3582 out_no_action:
3583         task->tk_action = NULL;
3584 out_wait:
3585         nfs4_sequence_done(task, &calldata->res.seq_res);
3586 }
3587
3588 static const struct rpc_call_ops nfs4_close_ops = {
3589         .rpc_call_prepare = nfs4_close_prepare,
3590         .rpc_call_done = nfs4_close_done,
3591         .rpc_release = nfs4_free_closedata,
3592 };
3593
3594 /*
3595  * It is possible for data to be read/written from a mem-mapped file
3596  * after the sys_close call (which hits the vfs layer as a flush).
3597  * This means that we can't safely call nfsv4 close on a file until
3598  * the inode is cleared. This in turn means that we are not good
3599  * NFSv4 citizens - we do not indicate to the server to update the file's
3600  * share state even when we are done with one of the three share
3601  * stateid's in the inode.
3602  *
3603  * NOTE: Caller must be holding the sp->so_owner semaphore!
3604  */
3605 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3606 {
3607         struct nfs_server *server = NFS_SERVER(state->inode);
3608         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3609         struct nfs4_closedata *calldata;
3610         struct nfs4_state_owner *sp = state->owner;
3611         struct rpc_task *task;
3612         struct rpc_message msg = {
3613                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3614                 .rpc_cred = state->owner->so_cred,
3615         };
3616         struct rpc_task_setup task_setup_data = {
3617                 .rpc_client = server->client,
3618                 .rpc_message = &msg,
3619                 .callback_ops = &nfs4_close_ops,
3620                 .workqueue = nfsiod_workqueue,
3621                 .flags = RPC_TASK_ASYNC,
3622         };
3623         int status = -ENOMEM;
3624
3625         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3626                 &task_setup_data.rpc_client, &msg);
3627
3628         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3629         if (calldata == NULL)
3630                 goto out;
3631         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3632         calldata->inode = state->inode;
3633         calldata->state = state;
3634         calldata->arg.fh = NFS_FH(state->inode);
3635         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3636                 goto out_free_calldata;
3637         /* Serialization for the sequence id */
3638         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3639         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3640         if (IS_ERR(calldata->arg.seqid))
3641                 goto out_free_calldata;
3642         nfs_fattr_init(&calldata->fattr);
3643         calldata->arg.fmode = 0;
3644         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3645         calldata->res.fattr = &calldata->fattr;
3646         calldata->res.seqid = calldata->arg.seqid;
3647         calldata->res.server = server;
3648         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3649         calldata->lr.roc = pnfs_roc(state->inode,
3650                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3651         if (calldata->lr.roc) {
3652                 calldata->arg.lr_args = &calldata->lr.arg;
3653                 calldata->res.lr_res = &calldata->lr.res;
3654         }
3655         nfs_sb_active(calldata->inode->i_sb);
3656
3657         msg.rpc_argp = &calldata->arg;
3658         msg.rpc_resp = &calldata->res;
3659         task_setup_data.callback_data = calldata;
3660         task = rpc_run_task(&task_setup_data);
3661         if (IS_ERR(task))
3662                 return PTR_ERR(task);
3663         status = 0;
3664         if (wait)
3665                 status = rpc_wait_for_completion_task(task);
3666         rpc_put_task(task);
3667         return status;
3668 out_free_calldata:
3669         kfree(calldata);
3670 out:
3671         nfs4_put_open_state(state);
3672         nfs4_put_state_owner(sp);
3673         return status;
3674 }
3675
3676 static struct inode *
3677 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3678                 int open_flags, struct iattr *attr, int *opened)
3679 {
3680         struct nfs4_state *state;
3681         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3682
3683         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3684
3685         /* Protect against concurrent sillydeletes */
3686         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3687
3688         nfs4_label_release_security(label);
3689
3690         if (IS_ERR(state))
3691                 return ERR_CAST(state);
3692         return state->inode;
3693 }
3694
3695 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3696 {
3697         if (ctx->state == NULL)
3698                 return;
3699         if (is_sync)
3700                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3701         else
3702                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3703 }
3704
3705 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3706 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3707 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3708
3709 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3710 {
3711         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3712         struct nfs4_server_caps_arg args = {
3713                 .fhandle = fhandle,
3714                 .bitmask = bitmask,
3715         };
3716         struct nfs4_server_caps_res res = {};
3717         struct rpc_message msg = {
3718                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3719                 .rpc_argp = &args,
3720                 .rpc_resp = &res,
3721         };
3722         int status;
3723         int i;
3724
3725         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3726                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3727                      FATTR4_WORD0_LINK_SUPPORT |
3728                      FATTR4_WORD0_SYMLINK_SUPPORT |
3729                      FATTR4_WORD0_ACLSUPPORT;
3730         if (minorversion)
3731                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3732
3733         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3734         if (status == 0) {
3735                 /* Sanity check the server answers */
3736                 switch (minorversion) {
3737                 case 0:
3738                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3739                         res.attr_bitmask[2] = 0;
3740                         break;
3741                 case 1:
3742                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3743                         break;
3744                 case 2:
3745                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3746                 }
3747                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3748                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3749                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3750                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3751                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3752                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3753                                 NFS_CAP_SECURITY_LABEL);
3754                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3755                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3756                         server->caps |= NFS_CAP_ACLS;
3757                 if (res.has_links != 0)
3758                         server->caps |= NFS_CAP_HARDLINKS;
3759                 if (res.has_symlinks != 0)
3760                         server->caps |= NFS_CAP_SYMLINKS;
3761                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3762                         server->caps |= NFS_CAP_FILEID;
3763                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3764                         server->caps |= NFS_CAP_MODE;
3765                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3766                         server->caps |= NFS_CAP_NLINK;
3767                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3768                         server->caps |= NFS_CAP_OWNER;
3769                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3770                         server->caps |= NFS_CAP_OWNER_GROUP;
3771                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3772                         server->caps |= NFS_CAP_ATIME;
3773                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3774                         server->caps |= NFS_CAP_CTIME;
3775                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3776                         server->caps |= NFS_CAP_MTIME;
3777 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3778                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3779                         server->caps |= NFS_CAP_SECURITY_LABEL;
3780 #endif
3781                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3782                                 sizeof(server->attr_bitmask));
3783                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3784
3785                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3786                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3787                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3788                 server->cache_consistency_bitmask[2] = 0;
3789
3790                 /* Avoid a regression due to buggy server */
3791                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3792                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3793                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3794                         sizeof(server->exclcreat_bitmask));
3795
3796                 server->acl_bitmask = res.acl_bitmask;
3797                 server->fh_expire_type = res.fh_expire_type;
3798         }
3799
3800         return status;
3801 }
3802
3803 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3804 {
3805         struct nfs4_exception exception = {
3806                 .interruptible = true,
3807         };
3808         int err;
3809         do {
3810                 err = nfs4_handle_exception(server,
3811                                 _nfs4_server_capabilities(server, fhandle),
3812                                 &exception);
3813         } while (exception.retry);
3814         return err;
3815 }
3816
3817 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3818                 struct nfs_fsinfo *info)
3819 {
3820         u32 bitmask[3];
3821         struct nfs4_lookup_root_arg args = {
3822                 .bitmask = bitmask,
3823         };
3824         struct nfs4_lookup_res res = {
3825                 .server = server,
3826                 .fattr = info->fattr,
3827                 .fh = fhandle,
3828         };
3829         struct rpc_message msg = {
3830                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3831                 .rpc_argp = &args,
3832                 .rpc_resp = &res,
3833         };
3834
3835         bitmask[0] = nfs4_fattr_bitmap[0];
3836         bitmask[1] = nfs4_fattr_bitmap[1];
3837         /*
3838          * Process the label in the upcoming getfattr
3839          */
3840         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3841
3842         nfs_fattr_init(info->fattr);
3843         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3844 }
3845
3846 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3847                 struct nfs_fsinfo *info)
3848 {
3849         struct nfs4_exception exception = {
3850                 .interruptible = true,
3851         };
3852         int err;
3853         do {
3854                 err = _nfs4_lookup_root(server, fhandle, info);
3855                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3856                 switch (err) {
3857                 case 0:
3858                 case -NFS4ERR_WRONGSEC:
3859                         goto out;
3860                 default:
3861                         err = nfs4_handle_exception(server, err, &exception);
3862                 }
3863         } while (exception.retry);
3864 out:
3865         return err;
3866 }
3867
3868 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3869                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3870 {
3871         struct rpc_auth_create_args auth_args = {
3872                 .pseudoflavor = flavor,
3873         };
3874         struct rpc_auth *auth;
3875
3876         auth = rpcauth_create(&auth_args, server->client);
3877         if (IS_ERR(auth))
3878                 return -EACCES;
3879         return nfs4_lookup_root(server, fhandle, info);
3880 }
3881
3882 /*
3883  * Retry pseudoroot lookup with various security flavors.  We do this when:
3884  *
3885  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3886  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3887  *
3888  * Returns zero on success, or a negative NFS4ERR value, or a
3889  * negative errno value.
3890  */
3891 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3892                               struct nfs_fsinfo *info)
3893 {
3894         /* Per 3530bis 15.33.5 */
3895         static const rpc_authflavor_t flav_array[] = {
3896                 RPC_AUTH_GSS_KRB5P,
3897                 RPC_AUTH_GSS_KRB5I,
3898                 RPC_AUTH_GSS_KRB5,
3899                 RPC_AUTH_UNIX,                  /* courtesy */
3900                 RPC_AUTH_NULL,
3901         };
3902         int status = -EPERM;
3903         size_t i;
3904
3905         if (server->auth_info.flavor_len > 0) {
3906                 /* try each flavor specified by user */
3907                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3908                         status = nfs4_lookup_root_sec(server, fhandle, info,
3909                                                 server->auth_info.flavors[i]);
3910                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3911                                 continue;
3912                         break;
3913                 }
3914         } else {
3915                 /* no flavors specified by user, try default list */
3916                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3917                         status = nfs4_lookup_root_sec(server, fhandle, info,
3918                                                       flav_array[i]);
3919                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3920                                 continue;
3921                         break;
3922                 }
3923         }
3924
3925         /*
3926          * -EACCES could mean that the user doesn't have correct permissions
3927          * to access the mount.  It could also mean that we tried to mount
3928          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3929          * existing mount programs don't handle -EACCES very well so it should
3930          * be mapped to -EPERM instead.
3931          */
3932         if (status == -EACCES)
3933                 status = -EPERM;
3934         return status;
3935 }
3936
3937 /**
3938  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3939  * @server: initialized nfs_server handle
3940  * @fhandle: we fill in the pseudo-fs root file handle
3941  * @info: we fill in an FSINFO struct
3942  * @auth_probe: probe the auth flavours
3943  *
3944  * Returns zero on success, or a negative errno.
3945  */
3946 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3947                          struct nfs_fsinfo *info,
3948                          bool auth_probe)
3949 {
3950         int status = 0;
3951
3952         if (!auth_probe)
3953                 status = nfs4_lookup_root(server, fhandle, info);
3954
3955         if (auth_probe || status == NFS4ERR_WRONGSEC)
3956                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3957                                 fhandle, info);
3958
3959         if (status == 0)
3960                 status = nfs4_server_capabilities(server, fhandle);
3961         if (status == 0)
3962                 status = nfs4_do_fsinfo(server, fhandle, info);
3963
3964         return nfs4_map_errors(status);
3965 }
3966
3967 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3968                               struct nfs_fsinfo *info)
3969 {
3970         int error;
3971         struct nfs_fattr *fattr = info->fattr;
3972         struct nfs4_label *label = NULL;
3973
3974         error = nfs4_server_capabilities(server, mntfh);
3975         if (error < 0) {
3976                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3977                 return error;
3978         }
3979
3980         label = nfs4_label_alloc(server, GFP_KERNEL);
3981         if (IS_ERR(label))
3982                 return PTR_ERR(label);
3983
3984         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
3985         if (error < 0) {
3986                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3987                 goto err_free_label;
3988         }
3989
3990         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3991             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3992                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3993
3994 err_free_label:
3995         nfs4_label_free(label);
3996
3997         return error;
3998 }
3999
4000 /*
4001  * Get locations and (maybe) other attributes of a referral.
4002  * Note that we'll actually follow the referral later when
4003  * we detect fsid mismatch in inode revalidation
4004  */
4005 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4006                              const struct qstr *name, struct nfs_fattr *fattr,
4007                              struct nfs_fh *fhandle)
4008 {
4009         int status = -ENOMEM;
4010         struct page *page = NULL;
4011         struct nfs4_fs_locations *locations = NULL;
4012
4013         page = alloc_page(GFP_KERNEL);
4014         if (page == NULL)
4015                 goto out;
4016         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4017         if (locations == NULL)
4018                 goto out;
4019
4020         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4021         if (status != 0)
4022                 goto out;
4023
4024         /*
4025          * If the fsid didn't change, this is a migration event, not a
4026          * referral.  Cause us to drop into the exception handler, which
4027          * will kick off migration recovery.
4028          */
4029         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4030                 dprintk("%s: server did not return a different fsid for"
4031                         " a referral at %s\n", __func__, name->name);
4032                 status = -NFS4ERR_MOVED;
4033                 goto out;
4034         }
4035         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4036         nfs_fixup_referral_attributes(&locations->fattr);
4037
4038         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4039         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4040         memset(fhandle, 0, sizeof(struct nfs_fh));
4041 out:
4042         if (page)
4043                 __free_page(page);
4044         kfree(locations);
4045         return status;
4046 }
4047
4048 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4049                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4050                                 struct inode *inode)
4051 {
4052         __u32 bitmask[NFS4_BITMASK_SZ];
4053         struct nfs4_getattr_arg args = {
4054                 .fh = fhandle,
4055                 .bitmask = bitmask,
4056         };
4057         struct nfs4_getattr_res res = {
4058                 .fattr = fattr,
4059                 .label = label,
4060                 .server = server,
4061         };
4062         struct rpc_message msg = {
4063                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4064                 .rpc_argp = &args,
4065                 .rpc_resp = &res,
4066         };
4067
4068         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4069
4070         nfs_fattr_init(fattr);
4071         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4072 }
4073
4074 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4075                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4076                                 struct inode *inode)
4077 {
4078         struct nfs4_exception exception = {
4079                 .interruptible = true,
4080         };
4081         int err;
4082         do {
4083                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4084                 trace_nfs4_getattr(server, fhandle, fattr, err);
4085                 err = nfs4_handle_exception(server, err,
4086                                 &exception);
4087         } while (exception.retry);
4088         return err;
4089 }
4090
4091 /*
4092  * The file is not closed if it is opened due to the a request to change
4093  * the size of the file. The open call will not be needed once the
4094  * VFS layer lookup-intents are implemented.
4095  *
4096  * Close is called when the inode is destroyed.
4097  * If we haven't opened the file for O_WRONLY, we
4098  * need to in the size_change case to obtain a stateid.
4099  *
4100  * Got race?
4101  * Because OPEN is always done by name in nfsv4, it is
4102  * possible that we opened a different file by the same
4103  * name.  We can recognize this race condition, but we
4104  * can't do anything about it besides returning an error.
4105  *
4106  * This will be fixed with VFS changes (lookup-intent).
4107  */
4108 static int
4109 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4110                   struct iattr *sattr)
4111 {
4112         struct inode *inode = d_inode(dentry);
4113         const struct cred *cred = NULL;
4114         struct nfs_open_context *ctx = NULL;
4115         struct nfs4_label *label = NULL;
4116         int status;
4117
4118         if (pnfs_ld_layoutret_on_setattr(inode) &&
4119             sattr->ia_valid & ATTR_SIZE &&
4120             sattr->ia_size < i_size_read(inode))
4121                 pnfs_commit_and_return_layout(inode);
4122
4123         nfs_fattr_init(fattr);
4124
4125         /* Deal with open(O_TRUNC) */
4126         if (sattr->ia_valid & ATTR_OPEN)
4127                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4128
4129         /* Optimization: if the end result is no change, don't RPC */
4130         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4131                 return 0;
4132
4133         /* Search for an existing open(O_WRITE) file */
4134         if (sattr->ia_valid & ATTR_FILE) {
4135
4136                 ctx = nfs_file_open_context(sattr->ia_file);
4137                 if (ctx)
4138                         cred = ctx->cred;
4139         }
4140
4141         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4142         if (IS_ERR(label))
4143                 return PTR_ERR(label);
4144
4145         /* Return any delegations if we're going to change ACLs */
4146         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4147                 nfs4_inode_make_writeable(inode);
4148
4149         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4150         if (status == 0) {
4151                 nfs_setattr_update_inode(inode, sattr, fattr);
4152                 nfs_setsecurity(inode, fattr, label);
4153         }
4154         nfs4_label_free(label);
4155         return status;
4156 }
4157
4158 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4159                 const struct qstr *name, struct nfs_fh *fhandle,
4160                 struct nfs_fattr *fattr, struct nfs4_label *label)
4161 {
4162         struct nfs_server *server = NFS_SERVER(dir);
4163         int                    status;
4164         struct nfs4_lookup_arg args = {
4165                 .bitmask = server->attr_bitmask,
4166                 .dir_fh = NFS_FH(dir),
4167                 .name = name,
4168         };
4169         struct nfs4_lookup_res res = {
4170                 .server = server,
4171                 .fattr = fattr,
4172                 .label = label,
4173                 .fh = fhandle,
4174         };
4175         struct rpc_message msg = {
4176                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4177                 .rpc_argp = &args,
4178                 .rpc_resp = &res,
4179         };
4180
4181         args.bitmask = nfs4_bitmask(server, label);
4182
4183         nfs_fattr_init(fattr);
4184
4185         dprintk("NFS call  lookup %s\n", name->name);
4186         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
4187         dprintk("NFS reply lookup: %d\n", status);
4188         return status;
4189 }
4190
4191 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4192 {
4193         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4194                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4195         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4196         fattr->nlink = 2;
4197 }
4198
4199 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4200                                    const struct qstr *name, struct nfs_fh *fhandle,
4201                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4202 {
4203         struct nfs4_exception exception = {
4204                 .interruptible = true,
4205         };
4206         struct rpc_clnt *client = *clnt;
4207         int err;
4208         do {
4209                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
4210                 trace_nfs4_lookup(dir, name, err);
4211                 switch (err) {
4212                 case -NFS4ERR_BADNAME:
4213                         err = -ENOENT;
4214                         goto out;
4215                 case -NFS4ERR_MOVED:
4216                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4217                         if (err == -NFS4ERR_MOVED)
4218                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4219                         goto out;
4220                 case -NFS4ERR_WRONGSEC:
4221                         err = -EPERM;
4222                         if (client != *clnt)
4223                                 goto out;
4224                         client = nfs4_negotiate_security(client, dir, name);
4225                         if (IS_ERR(client))
4226                                 return PTR_ERR(client);
4227
4228                         exception.retry = 1;
4229                         break;
4230                 default:
4231                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4232                 }
4233         } while (exception.retry);
4234
4235 out:
4236         if (err == 0)
4237                 *clnt = client;
4238         else if (client != *clnt)
4239                 rpc_shutdown_client(client);
4240
4241         return err;
4242 }
4243
4244 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
4245                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4246                             struct nfs4_label *label)
4247 {
4248         int status;
4249         struct rpc_clnt *client = NFS_CLIENT(dir);
4250
4251         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
4252         if (client != NFS_CLIENT(dir)) {
4253                 rpc_shutdown_client(client);
4254                 nfs_fixup_secinfo_attributes(fattr);
4255         }
4256         return status;
4257 }
4258
4259 struct rpc_clnt *
4260 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
4261                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4262 {
4263         struct rpc_clnt *client = NFS_CLIENT(dir);
4264         int status;
4265
4266         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
4267         if (status < 0)
4268                 return ERR_PTR(status);
4269         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4270 }
4271
4272 static int _nfs4_proc_lookupp(struct inode *inode,
4273                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4274                 struct nfs4_label *label)
4275 {
4276         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4277         struct nfs_server *server = NFS_SERVER(inode);
4278         int                    status;
4279         struct nfs4_lookupp_arg args = {
4280                 .bitmask = server->attr_bitmask,
4281                 .fh = NFS_FH(inode),
4282         };
4283         struct nfs4_lookupp_res res = {
4284                 .server = server,
4285                 .fattr = fattr,
4286                 .label = label,
4287                 .fh = fhandle,
4288         };
4289         struct rpc_message msg = {
4290                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4291                 .rpc_argp = &args,
4292                 .rpc_resp = &res,
4293         };
4294
4295         args.bitmask = nfs4_bitmask(server, label);
4296
4297         nfs_fattr_init(fattr);
4298
4299         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4300         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4301                                 &res.seq_res, 0);
4302         dprintk("NFS reply lookupp: %d\n", status);
4303         return status;
4304 }
4305
4306 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4307                              struct nfs_fattr *fattr, struct nfs4_label *label)
4308 {
4309         struct nfs4_exception exception = {
4310                 .interruptible = true,
4311         };
4312         int err;
4313         do {
4314                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4315                 trace_nfs4_lookupp(inode, err);
4316                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4317                                 &exception);
4318         } while (exception.retry);
4319         return err;
4320 }
4321
4322 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4323 {
4324         struct nfs_server *server = NFS_SERVER(inode);
4325         struct nfs4_accessargs args = {
4326                 .fh = NFS_FH(inode),
4327                 .access = entry->mask,
4328         };
4329         struct nfs4_accessres res = {
4330                 .server = server,
4331         };
4332         struct rpc_message msg = {
4333                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4334                 .rpc_argp = &args,
4335                 .rpc_resp = &res,
4336                 .rpc_cred = entry->cred,
4337         };
4338         int status = 0;
4339
4340         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4341                 res.fattr = nfs_alloc_fattr();
4342                 if (res.fattr == NULL)
4343                         return -ENOMEM;
4344                 args.bitmask = server->cache_consistency_bitmask;
4345         }
4346         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4347         if (!status) {
4348                 nfs_access_set_mask(entry, res.access);
4349                 if (res.fattr)
4350                         nfs_refresh_inode(inode, res.fattr);
4351         }
4352         nfs_free_fattr(res.fattr);
4353         return status;
4354 }
4355
4356 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4357 {
4358         struct nfs4_exception exception = {
4359                 .interruptible = true,
4360         };
4361         int err;
4362         do {
4363                 err = _nfs4_proc_access(inode, entry);
4364                 trace_nfs4_access(inode, err);
4365                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4366                                 &exception);
4367         } while (exception.retry);
4368         return err;
4369 }
4370
4371 /*
4372  * TODO: For the time being, we don't try to get any attributes
4373  * along with any of the zero-copy operations READ, READDIR,
4374  * READLINK, WRITE.
4375  *
4376  * In the case of the first three, we want to put the GETATTR
4377  * after the read-type operation -- this is because it is hard
4378  * to predict the length of a GETATTR response in v4, and thus
4379  * align the READ data correctly.  This means that the GETATTR
4380  * may end up partially falling into the page cache, and we should
4381  * shift it into the 'tail' of the xdr_buf before processing.
4382  * To do this efficiently, we need to know the total length
4383  * of data received, which doesn't seem to be available outside
4384  * of the RPC layer.
4385  *
4386  * In the case of WRITE, we also want to put the GETATTR after
4387  * the operation -- in this case because we want to make sure
4388  * we get the post-operation mtime and size.
4389  *
4390  * Both of these changes to the XDR layer would in fact be quite
4391  * minor, but I decided to leave them for a subsequent patch.
4392  */
4393 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4394                 unsigned int pgbase, unsigned int pglen)
4395 {
4396         struct nfs4_readlink args = {
4397                 .fh       = NFS_FH(inode),
4398                 .pgbase   = pgbase,
4399                 .pglen    = pglen,
4400                 .pages    = &page,
4401         };
4402         struct nfs4_readlink_res res;
4403         struct rpc_message msg = {
4404                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4405                 .rpc_argp = &args,
4406                 .rpc_resp = &res,
4407         };
4408
4409         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4410 }
4411
4412 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4413                 unsigned int pgbase, unsigned int pglen)
4414 {
4415         struct nfs4_exception exception = {
4416                 .interruptible = true,
4417         };
4418         int err;
4419         do {
4420                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4421                 trace_nfs4_readlink(inode, err);
4422                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4423                                 &exception);
4424         } while (exception.retry);
4425         return err;
4426 }
4427
4428 /*
4429  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4430  */
4431 static int
4432 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4433                  int flags)
4434 {
4435         struct nfs_server *server = NFS_SERVER(dir);
4436         struct nfs4_label l, *ilabel = NULL;
4437         struct nfs_open_context *ctx;
4438         struct nfs4_state *state;
4439         int status = 0;
4440
4441         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4442         if (IS_ERR(ctx))
4443                 return PTR_ERR(ctx);
4444
4445         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4446
4447         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4448                 sattr->ia_mode &= ~current_umask();
4449         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4450         if (IS_ERR(state)) {
4451                 status = PTR_ERR(state);
4452                 goto out;
4453         }
4454 out:
4455         nfs4_label_release_security(ilabel);
4456         put_nfs_open_context(ctx);
4457         return status;
4458 }
4459
4460 static int
4461 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4462 {
4463         struct nfs_server *server = NFS_SERVER(dir);
4464         struct nfs_removeargs args = {
4465                 .fh = NFS_FH(dir),
4466                 .name = *name,
4467         };
4468         struct nfs_removeres res = {
4469                 .server = server,
4470         };
4471         struct rpc_message msg = {
4472                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4473                 .rpc_argp = &args,
4474                 .rpc_resp = &res,
4475         };
4476         unsigned long timestamp = jiffies;
4477         int status;
4478
4479         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4480         if (status == 0) {
4481                 spin_lock(&dir->i_lock);
4482                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4483                 /* Removing a directory decrements nlink in the parent */
4484                 if (ftype == NF4DIR && dir->i_nlink > 2)
4485                         nfs4_dec_nlink_locked(dir);
4486                 spin_unlock(&dir->i_lock);
4487         }
4488         return status;
4489 }
4490
4491 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4492 {
4493         struct nfs4_exception exception = {
4494                 .interruptible = true,
4495         };
4496         struct inode *inode = d_inode(dentry);
4497         int err;
4498
4499         if (inode) {
4500                 if (inode->i_nlink == 1)
4501                         nfs4_inode_return_delegation(inode);
4502                 else
4503                         nfs4_inode_make_writeable(inode);
4504         }
4505         do {
4506                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4507                 trace_nfs4_remove(dir, &dentry->d_name, err);
4508                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4509                                 &exception);
4510         } while (exception.retry);
4511         return err;
4512 }
4513
4514 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4515 {
4516         struct nfs4_exception exception = {
4517                 .interruptible = true,
4518         };
4519         int err;
4520
4521         do {
4522                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4523                 trace_nfs4_remove(dir, name, err);
4524                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4525                                 &exception);
4526         } while (exception.retry);
4527         return err;
4528 }
4529
4530 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4531                 struct dentry *dentry,
4532                 struct inode *inode)
4533 {
4534         struct nfs_removeargs *args = msg->rpc_argp;
4535         struct nfs_removeres *res = msg->rpc_resp;
4536
4537         res->server = NFS_SB(dentry->d_sb);
4538         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4539         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4540
4541         nfs_fattr_init(res->dir_attr);
4542
4543         if (inode)
4544                 nfs4_inode_return_delegation(inode);
4545 }
4546
4547 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4548 {
4549         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4550                         &data->args.seq_args,
4551                         &data->res.seq_res,
4552                         task);
4553 }
4554
4555 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4556 {
4557         struct nfs_unlinkdata *data = task->tk_calldata;
4558         struct nfs_removeres *res = &data->res;
4559
4560         if (!nfs4_sequence_done(task, &res->seq_res))
4561                 return 0;
4562         if (nfs4_async_handle_error(task, res->server, NULL,
4563                                     &data->timeout) == -EAGAIN)
4564                 return 0;
4565         if (task->tk_status == 0)
4566                 update_changeattr(dir, &res->cinfo,
4567                                 res->dir_attr->time_start, 0);
4568         return 1;
4569 }
4570
4571 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4572                 struct dentry *old_dentry,
4573                 struct dentry *new_dentry)
4574 {
4575         struct nfs_renameargs *arg = msg->rpc_argp;
4576         struct nfs_renameres *res = msg->rpc_resp;
4577         struct inode *old_inode = d_inode(old_dentry);
4578         struct inode *new_inode = d_inode(new_dentry);
4579
4580         if (old_inode)
4581                 nfs4_inode_make_writeable(old_inode);
4582         if (new_inode)
4583                 nfs4_inode_return_delegation(new_inode);
4584         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4585         res->server = NFS_SB(old_dentry->d_sb);
4586         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4587 }
4588
4589 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4590 {
4591         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4592                         &data->args.seq_args,
4593                         &data->res.seq_res,
4594                         task);
4595 }
4596
4597 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4598                                  struct inode *new_dir)
4599 {
4600         struct nfs_renamedata *data = task->tk_calldata;
4601         struct nfs_renameres *res = &data->res;
4602
4603         if (!nfs4_sequence_done(task, &res->seq_res))
4604                 return 0;
4605         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4606                 return 0;
4607
4608         if (task->tk_status == 0) {
4609                 if (new_dir != old_dir) {
4610                         /* Note: If we moved a directory, nlink will change */
4611                         update_changeattr(old_dir, &res->old_cinfo,
4612                                         res->old_fattr->time_start,
4613                                         NFS_INO_INVALID_OTHER);
4614                         update_changeattr(new_dir, &res->new_cinfo,
4615                                         res->new_fattr->time_start,
4616                                         NFS_INO_INVALID_OTHER);
4617                 } else
4618                         update_changeattr(old_dir, &res->old_cinfo,
4619                                         res->old_fattr->time_start,
4620                                         0);
4621         }
4622         return 1;
4623 }
4624
4625 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4626 {
4627         struct nfs_server *server = NFS_SERVER(inode);
4628         __u32 bitmask[NFS4_BITMASK_SZ];
4629         struct nfs4_link_arg arg = {
4630                 .fh     = NFS_FH(inode),
4631                 .dir_fh = NFS_FH(dir),
4632                 .name   = name,
4633                 .bitmask = bitmask,
4634         };
4635         struct nfs4_link_res res = {
4636                 .server = server,
4637                 .label = NULL,
4638         };
4639         struct rpc_message msg = {
4640                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4641                 .rpc_argp = &arg,
4642                 .rpc_resp = &res,
4643         };
4644         int status = -ENOMEM;
4645
4646         res.fattr = nfs_alloc_fattr();
4647         if (res.fattr == NULL)
4648                 goto out;
4649
4650         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4651         if (IS_ERR(res.label)) {
4652                 status = PTR_ERR(res.label);
4653                 goto out;
4654         }
4655
4656         nfs4_inode_make_writeable(inode);
4657         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4658
4659         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4660         if (!status) {
4661                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4662                 status = nfs_post_op_update_inode(inode, res.fattr);
4663                 if (!status)
4664                         nfs_setsecurity(inode, res.fattr, res.label);
4665         }
4666
4667
4668         nfs4_label_free(res.label);
4669
4670 out:
4671         nfs_free_fattr(res.fattr);
4672         return status;
4673 }
4674
4675 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4676 {
4677         struct nfs4_exception exception = {
4678                 .interruptible = true,
4679         };
4680         int err;
4681         do {
4682                 err = nfs4_handle_exception(NFS_SERVER(inode),
4683                                 _nfs4_proc_link(inode, dir, name),
4684                                 &exception);
4685         } while (exception.retry);
4686         return err;
4687 }
4688
4689 struct nfs4_createdata {
4690         struct rpc_message msg;
4691         struct nfs4_create_arg arg;
4692         struct nfs4_create_res res;
4693         struct nfs_fh fh;
4694         struct nfs_fattr fattr;
4695         struct nfs4_label *label;
4696 };
4697
4698 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4699                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4700 {
4701         struct nfs4_createdata *data;
4702
4703         data = kzalloc(sizeof(*data), GFP_KERNEL);
4704         if (data != NULL) {
4705                 struct nfs_server *server = NFS_SERVER(dir);
4706
4707                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4708                 if (IS_ERR(data->label))
4709                         goto out_free;
4710
4711                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4712                 data->msg.rpc_argp = &data->arg;
4713                 data->msg.rpc_resp = &data->res;
4714                 data->arg.dir_fh = NFS_FH(dir);
4715                 data->arg.server = server;
4716                 data->arg.name = name;
4717                 data->arg.attrs = sattr;
4718                 data->arg.ftype = ftype;
4719                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4720                 data->arg.umask = current_umask();
4721                 data->res.server = server;
4722                 data->res.fh = &data->fh;
4723                 data->res.fattr = &data->fattr;
4724                 data->res.label = data->label;
4725                 nfs_fattr_init(data->res.fattr);
4726         }
4727         return data;
4728 out_free:
4729         kfree(data);
4730         return NULL;
4731 }
4732
4733 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4734 {
4735         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4736                                     &data->arg.seq_args, &data->res.seq_res, 1);
4737         if (status == 0) {
4738                 spin_lock(&dir->i_lock);
4739                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4740                                 data->res.fattr->time_start, 0);
4741                 /* Creating a directory bumps nlink in the parent */
4742                 if (data->arg.ftype == NF4DIR)
4743                         nfs4_inc_nlink_locked(dir);
4744                 spin_unlock(&dir->i_lock);
4745                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4746         }
4747         return status;
4748 }
4749
4750 static void nfs4_free_createdata(struct nfs4_createdata *data)
4751 {
4752         nfs4_label_free(data->label);
4753         kfree(data);
4754 }
4755
4756 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4757                 struct page *page, unsigned int len, struct iattr *sattr,
4758                 struct nfs4_label *label)
4759 {
4760         struct nfs4_createdata *data;
4761         int status = -ENAMETOOLONG;
4762
4763         if (len > NFS4_MAXPATHLEN)
4764                 goto out;
4765
4766         status = -ENOMEM;
4767         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4768         if (data == NULL)
4769                 goto out;
4770
4771         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4772         data->arg.u.symlink.pages = &page;
4773         data->arg.u.symlink.len = len;
4774         data->arg.label = label;
4775
4776         status = nfs4_do_create(dir, dentry, data);
4777
4778         nfs4_free_createdata(data);
4779 out:
4780         return status;
4781 }
4782
4783 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4784                 struct page *page, unsigned int len, struct iattr *sattr)
4785 {
4786         struct nfs4_exception exception = {
4787                 .interruptible = true,
4788         };
4789         struct nfs4_label l, *label = NULL;
4790         int err;
4791
4792         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4793
4794         do {
4795                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4796                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4797                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4798                                 &exception);
4799         } while (exception.retry);
4800
4801         nfs4_label_release_security(label);
4802         return err;
4803 }
4804
4805 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4806                 struct iattr *sattr, struct nfs4_label *label)
4807 {
4808         struct nfs4_createdata *data;
4809         int status = -ENOMEM;
4810
4811         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4812         if (data == NULL)
4813                 goto out;
4814
4815         data->arg.label = label;
4816         status = nfs4_do_create(dir, dentry, data);
4817
4818         nfs4_free_createdata(data);
4819 out:
4820         return status;
4821 }
4822
4823 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4824                 struct iattr *sattr)
4825 {
4826         struct nfs_server *server = NFS_SERVER(dir);
4827         struct nfs4_exception exception = {
4828                 .interruptible = true,
4829         };
4830         struct nfs4_label l, *label = NULL;
4831         int err;
4832
4833         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4834
4835         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4836                 sattr->ia_mode &= ~current_umask();
4837         do {
4838                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4839                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4840                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4841                                 &exception);
4842         } while (exception.retry);
4843         nfs4_label_release_security(label);
4844
4845         return err;
4846 }
4847
4848 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4849                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4850 {
4851         struct inode            *dir = d_inode(dentry);
4852         struct nfs4_readdir_arg args = {
4853                 .fh = NFS_FH(dir),
4854                 .pages = pages,
4855                 .pgbase = 0,
4856                 .count = count,
4857                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4858                 .plus = plus,
4859         };
4860         struct nfs4_readdir_res res;
4861         struct rpc_message msg = {
4862                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4863                 .rpc_argp = &args,
4864                 .rpc_resp = &res,
4865                 .rpc_cred = cred,
4866         };
4867         int                     status;
4868
4869         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4870                         dentry,
4871                         (unsigned long long)cookie);
4872         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4873         res.pgbase = args.pgbase;
4874         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4875         if (status >= 0) {
4876                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4877                 status += args.pgbase;
4878         }
4879
4880         nfs_invalidate_atime(dir);
4881
4882         dprintk("%s: returns %d\n", __func__, status);
4883         return status;
4884 }
4885
4886 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4887                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4888 {
4889         struct nfs4_exception exception = {
4890                 .interruptible = true,
4891         };
4892         int err;
4893         do {
4894                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4895                                 pages, count, plus);
4896                 trace_nfs4_readdir(d_inode(dentry), err);
4897                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4898                                 &exception);
4899         } while (exception.retry);
4900         return err;
4901 }
4902
4903 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4904                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4905 {
4906         struct nfs4_createdata *data;
4907         int mode = sattr->ia_mode;
4908         int status = -ENOMEM;
4909
4910         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4911         if (data == NULL)
4912                 goto out;
4913
4914         if (S_ISFIFO(mode))
4915                 data->arg.ftype = NF4FIFO;
4916         else if (S_ISBLK(mode)) {
4917                 data->arg.ftype = NF4BLK;
4918                 data->arg.u.device.specdata1 = MAJOR(rdev);
4919                 data->arg.u.device.specdata2 = MINOR(rdev);
4920         }
4921         else if (S_ISCHR(mode)) {
4922                 data->arg.ftype = NF4CHR;
4923                 data->arg.u.device.specdata1 = MAJOR(rdev);
4924                 data->arg.u.device.specdata2 = MINOR(rdev);
4925         } else if (!S_ISSOCK(mode)) {
4926                 status = -EINVAL;
4927                 goto out_free;
4928         }
4929
4930         data->arg.label = label;
4931         status = nfs4_do_create(dir, dentry, data);
4932 out_free:
4933         nfs4_free_createdata(data);
4934 out:
4935         return status;
4936 }
4937
4938 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4939                 struct iattr *sattr, dev_t rdev)
4940 {
4941         struct nfs_server *server = NFS_SERVER(dir);
4942         struct nfs4_exception exception = {
4943                 .interruptible = true,
4944         };
4945         struct nfs4_label l, *label = NULL;
4946         int err;
4947
4948         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4949
4950         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4951                 sattr->ia_mode &= ~current_umask();
4952         do {
4953                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4954                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4955                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4956                                 &exception);
4957         } while (exception.retry);
4958
4959         nfs4_label_release_security(label);
4960
4961         return err;
4962 }
4963
4964 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4965                  struct nfs_fsstat *fsstat)
4966 {
4967         struct nfs4_statfs_arg args = {
4968                 .fh = fhandle,
4969                 .bitmask = server->attr_bitmask,
4970         };
4971         struct nfs4_statfs_res res = {
4972                 .fsstat = fsstat,
4973         };
4974         struct rpc_message msg = {
4975                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4976                 .rpc_argp = &args,
4977                 .rpc_resp = &res,
4978         };
4979
4980         nfs_fattr_init(fsstat->fattr);
4981         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4982 }
4983
4984 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4985 {
4986         struct nfs4_exception exception = {
4987                 .interruptible = true,
4988         };
4989         int err;
4990         do {
4991                 err = nfs4_handle_exception(server,
4992                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4993                                 &exception);
4994         } while (exception.retry);
4995         return err;
4996 }
4997
4998 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4999                 struct nfs_fsinfo *fsinfo)
5000 {
5001         struct nfs4_fsinfo_arg args = {
5002                 .fh = fhandle,
5003                 .bitmask = server->attr_bitmask,
5004         };
5005         struct nfs4_fsinfo_res res = {
5006                 .fsinfo = fsinfo,
5007         };
5008         struct rpc_message msg = {
5009                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5010                 .rpc_argp = &args,
5011                 .rpc_resp = &res,
5012         };
5013
5014         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5015 }
5016
5017 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5018 {
5019         struct nfs4_exception exception = {
5020                 .interruptible = true,
5021         };
5022         unsigned long now = jiffies;
5023         int err;
5024
5025         do {
5026                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5027                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5028                 if (err == 0) {
5029                         nfs4_set_lease_period(server->nfs_client,
5030                                         fsinfo->lease_time * HZ,
5031                                         now);
5032                         break;
5033                 }
5034                 err = nfs4_handle_exception(server, err, &exception);
5035         } while (exception.retry);
5036         return err;
5037 }
5038
5039 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5040 {
5041         int error;
5042
5043         nfs_fattr_init(fsinfo->fattr);
5044         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5045         if (error == 0) {
5046                 /* block layout checks this! */
5047                 server->pnfs_blksize = fsinfo->blksize;
5048                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5049         }
5050
5051         return error;
5052 }
5053
5054 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5055                 struct nfs_pathconf *pathconf)
5056 {
5057         struct nfs4_pathconf_arg args = {
5058                 .fh = fhandle,
5059                 .bitmask = server->attr_bitmask,
5060         };
5061         struct nfs4_pathconf_res res = {
5062                 .pathconf = pathconf,
5063         };
5064         struct rpc_message msg = {
5065                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5066                 .rpc_argp = &args,
5067                 .rpc_resp = &res,
5068         };
5069
5070         /* None of the pathconf attributes are mandatory to implement */
5071         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5072                 memset(pathconf, 0, sizeof(*pathconf));
5073                 return 0;
5074         }
5075
5076         nfs_fattr_init(pathconf->fattr);
5077         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5078 }
5079
5080 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5081                 struct nfs_pathconf *pathconf)
5082 {
5083         struct nfs4_exception exception = {
5084                 .interruptible = true,
5085         };
5086         int err;
5087
5088         do {
5089                 err = nfs4_handle_exception(server,
5090                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5091                                 &exception);
5092         } while (exception.retry);
5093         return err;
5094 }
5095
5096 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5097                 const struct nfs_open_context *ctx,
5098                 const struct nfs_lock_context *l_ctx,
5099                 fmode_t fmode)
5100 {
5101         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5102 }
5103 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5104
5105 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5106                 const struct nfs_open_context *ctx,
5107                 const struct nfs_lock_context *l_ctx,
5108                 fmode_t fmode)
5109 {
5110         nfs4_stateid _current_stateid;
5111
5112         /* If the current stateid represents a lost lock, then exit */
5113         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5114                 return true;
5115         return nfs4_stateid_match(stateid, &_current_stateid);
5116 }
5117
5118 static bool nfs4_error_stateid_expired(int err)
5119 {
5120         switch (err) {
5121         case -NFS4ERR_DELEG_REVOKED:
5122         case -NFS4ERR_ADMIN_REVOKED:
5123         case -NFS4ERR_BAD_STATEID:
5124         case -NFS4ERR_STALE_STATEID:
5125         case -NFS4ERR_OLD_STATEID:
5126         case -NFS4ERR_OPENMODE:
5127         case -NFS4ERR_EXPIRED:
5128                 return true;
5129         }
5130         return false;
5131 }
5132
5133 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5134 {
5135         struct nfs_server *server = NFS_SERVER(hdr->inode);
5136
5137         trace_nfs4_read(hdr, task->tk_status);
5138         if (task->tk_status < 0) {
5139                 struct nfs4_exception exception = {
5140                         .inode = hdr->inode,
5141                         .state = hdr->args.context->state,
5142                         .stateid = &hdr->args.stateid,
5143                 };
5144                 task->tk_status = nfs4_async_handle_exception(task,
5145                                 server, task->tk_status, &exception);
5146                 if (exception.retry) {
5147                         rpc_restart_call_prepare(task);
5148                         return -EAGAIN;
5149                 }
5150         }
5151
5152         if (task->tk_status > 0)
5153                 renew_lease(server, hdr->timestamp);
5154         return 0;
5155 }
5156
5157 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5158                 struct nfs_pgio_args *args)
5159 {
5160
5161         if (!nfs4_error_stateid_expired(task->tk_status) ||
5162                 nfs4_stateid_is_current(&args->stateid,
5163                                 args->context,
5164                                 args->lock_context,
5165                                 FMODE_READ))
5166                 return false;
5167         rpc_restart_call_prepare(task);
5168         return true;
5169 }
5170
5171 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5172 {
5173
5174         dprintk("--> %s\n", __func__);
5175
5176         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5177                 return -EAGAIN;
5178         if (nfs4_read_stateid_changed(task, &hdr->args))
5179                 return -EAGAIN;
5180         if (task->tk_status > 0)
5181                 nfs_invalidate_atime(hdr->inode);
5182         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5183                                     nfs4_read_done_cb(task, hdr);
5184 }
5185
5186 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5187                                  struct rpc_message *msg)
5188 {
5189         hdr->timestamp   = jiffies;
5190         if (!hdr->pgio_done_cb)
5191                 hdr->pgio_done_cb = nfs4_read_done_cb;
5192         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5193         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5194 }
5195
5196 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5197                                       struct nfs_pgio_header *hdr)
5198 {
5199         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5200                         &hdr->args.seq_args,
5201                         &hdr->res.seq_res,
5202                         task))
5203                 return 0;
5204         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5205                                 hdr->args.lock_context,
5206                                 hdr->rw_mode) == -EIO)
5207                 return -EIO;
5208         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5209                 return -EIO;
5210         return 0;
5211 }
5212
5213 static int nfs4_write_done_cb(struct rpc_task *task,
5214                               struct nfs_pgio_header *hdr)
5215 {
5216         struct inode *inode = hdr->inode;
5217
5218         trace_nfs4_write(hdr, task->tk_status);
5219         if (task->tk_status < 0) {
5220                 struct nfs4_exception exception = {
5221                         .inode = hdr->inode,
5222                         .state = hdr->args.context->state,
5223                         .stateid = &hdr->args.stateid,
5224                 };
5225                 task->tk_status = nfs4_async_handle_exception(task,
5226                                 NFS_SERVER(inode), task->tk_status,
5227                                 &exception);
5228                 if (exception.retry) {
5229                         rpc_restart_call_prepare(task);
5230                         return -EAGAIN;
5231                 }
5232         }
5233         if (task->tk_status >= 0) {
5234                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5235                 nfs_writeback_update_inode(hdr);
5236         }
5237         return 0;
5238 }
5239
5240 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5241                 struct nfs_pgio_args *args)
5242 {
5243
5244         if (!nfs4_error_stateid_expired(task->tk_status) ||
5245                 nfs4_stateid_is_current(&args->stateid,
5246                                 args->context,
5247                                 args->lock_context,
5248                                 FMODE_WRITE))
5249                 return false;
5250         rpc_restart_call_prepare(task);
5251         return true;
5252 }
5253
5254 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5255 {
5256         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5257                 return -EAGAIN;
5258         if (nfs4_write_stateid_changed(task, &hdr->args))
5259                 return -EAGAIN;
5260         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5261                 nfs4_write_done_cb(task, hdr);
5262 }
5263
5264 static
5265 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5266 {
5267         /* Don't request attributes for pNFS or O_DIRECT writes */
5268         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5269                 return false;
5270         /* Otherwise, request attributes if and only if we don't hold
5271          * a delegation
5272          */
5273         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5274 }
5275
5276 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5277                                   struct rpc_message *msg,
5278                                   struct rpc_clnt **clnt)
5279 {
5280         struct nfs_server *server = NFS_SERVER(hdr->inode);
5281
5282         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5283                 hdr->args.bitmask = NULL;
5284                 hdr->res.fattr = NULL;
5285         } else
5286                 hdr->args.bitmask = server->cache_consistency_bitmask;
5287
5288         if (!hdr->pgio_done_cb)
5289                 hdr->pgio_done_cb = nfs4_write_done_cb;
5290         hdr->res.server = server;
5291         hdr->timestamp   = jiffies;
5292
5293         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5294         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1, 0);
5295         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5296 }
5297
5298 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5299 {
5300         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5301                         &data->args.seq_args,
5302                         &data->res.seq_res,
5303                         task);
5304 }
5305
5306 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5307 {
5308         struct inode *inode = data->inode;
5309
5310         trace_nfs4_commit(data, task->tk_status);
5311         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5312                                     NULL, NULL) == -EAGAIN) {
5313                 rpc_restart_call_prepare(task);
5314                 return -EAGAIN;
5315         }
5316         return 0;
5317 }
5318
5319 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5320 {
5321         if (!nfs4_sequence_done(task, &data->res.seq_res))
5322                 return -EAGAIN;
5323         return data->commit_done_cb(task, data);
5324 }
5325
5326 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5327                                    struct rpc_clnt **clnt)
5328 {
5329         struct nfs_server *server = NFS_SERVER(data->inode);
5330
5331         if (data->commit_done_cb == NULL)
5332                 data->commit_done_cb = nfs4_commit_done_cb;
5333         data->res.server = server;
5334         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5335         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5336         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5337 }
5338
5339 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5340                                 struct nfs_commitres *res)
5341 {
5342         struct inode *dst_inode = file_inode(dst);
5343         struct nfs_server *server = NFS_SERVER(dst_inode);
5344         struct rpc_message msg = {
5345                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5346                 .rpc_argp = args,
5347                 .rpc_resp = res,
5348         };
5349
5350         args->fh = NFS_FH(dst_inode);
5351         return nfs4_call_sync(server->client, server, &msg,
5352                         &args->seq_args, &res->seq_res, 1);
5353 }
5354
5355 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5356 {
5357         struct nfs_commitargs args = {
5358                 .offset = offset,
5359                 .count = count,
5360         };
5361         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5362         struct nfs4_exception exception = { };
5363         int status;
5364
5365         do {
5366                 status = _nfs4_proc_commit(dst, &args, res);
5367                 status = nfs4_handle_exception(dst_server, status, &exception);
5368         } while (exception.retry);
5369
5370         return status;
5371 }
5372
5373 struct nfs4_renewdata {
5374         struct nfs_client       *client;
5375         unsigned long           timestamp;
5376 };
5377
5378 /*
5379  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5380  * standalone procedure for queueing an asynchronous RENEW.
5381  */
5382 static void nfs4_renew_release(void *calldata)
5383 {
5384         struct nfs4_renewdata *data = calldata;
5385         struct nfs_client *clp = data->client;
5386
5387         if (refcount_read(&clp->cl_count) > 1)
5388                 nfs4_schedule_state_renewal(clp);
5389         nfs_put_client(clp);
5390         kfree(data);
5391 }
5392
5393 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5394 {
5395         struct nfs4_renewdata *data = calldata;
5396         struct nfs_client *clp = data->client;
5397         unsigned long timestamp = data->timestamp;
5398
5399         trace_nfs4_renew_async(clp, task->tk_status);
5400         switch (task->tk_status) {
5401         case 0:
5402                 break;
5403         case -NFS4ERR_LEASE_MOVED:
5404                 nfs4_schedule_lease_moved_recovery(clp);
5405                 break;
5406         default:
5407                 /* Unless we're shutting down, schedule state recovery! */
5408                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5409                         return;
5410                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5411                         nfs4_schedule_lease_recovery(clp);
5412                         return;
5413                 }
5414                 nfs4_schedule_path_down_recovery(clp);
5415         }
5416         do_renew_lease(clp, timestamp);
5417 }
5418
5419 static const struct rpc_call_ops nfs4_renew_ops = {
5420         .rpc_call_done = nfs4_renew_done,
5421         .rpc_release = nfs4_renew_release,
5422 };
5423
5424 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5425 {
5426         struct rpc_message msg = {
5427                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5428                 .rpc_argp       = clp,
5429                 .rpc_cred       = cred,
5430         };
5431         struct nfs4_renewdata *data;
5432
5433         if (renew_flags == 0)
5434                 return 0;
5435         if (!refcount_inc_not_zero(&clp->cl_count))
5436                 return -EIO;
5437         data = kmalloc(sizeof(*data), GFP_NOFS);
5438         if (data == NULL) {
5439                 nfs_put_client(clp);
5440                 return -ENOMEM;
5441         }
5442         data->client = clp;
5443         data->timestamp = jiffies;
5444         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5445                         &nfs4_renew_ops, data);
5446 }
5447
5448 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5449 {
5450         struct rpc_message msg = {
5451                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5452                 .rpc_argp       = clp,
5453                 .rpc_cred       = cred,
5454         };
5455         unsigned long now = jiffies;
5456         int status;
5457
5458         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5459         if (status < 0)
5460                 return status;
5461         do_renew_lease(clp, now);
5462         return 0;
5463 }
5464
5465 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5466 {
5467         return server->caps & NFS_CAP_ACLS;
5468 }
5469
5470 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5471  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5472  * the stack.
5473  */
5474 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5475
5476 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5477                 struct page **pages)
5478 {
5479         struct page *newpage, **spages;
5480         int rc = 0;
5481         size_t len;
5482         spages = pages;
5483
5484         do {
5485                 len = min_t(size_t, PAGE_SIZE, buflen);
5486                 newpage = alloc_page(GFP_KERNEL);
5487
5488                 if (newpage == NULL)
5489                         goto unwind;
5490                 memcpy(page_address(newpage), buf, len);
5491                 buf += len;
5492                 buflen -= len;
5493                 *pages++ = newpage;
5494                 rc++;
5495         } while (buflen != 0);
5496
5497         return rc;
5498
5499 unwind:
5500         for(; rc > 0; rc--)
5501                 __free_page(spages[rc-1]);
5502         return -ENOMEM;
5503 }
5504
5505 struct nfs4_cached_acl {
5506         int cached;
5507         size_t len;
5508         char data[0];
5509 };
5510
5511 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5512 {
5513         struct nfs_inode *nfsi = NFS_I(inode);
5514
5515         spin_lock(&inode->i_lock);
5516         kfree(nfsi->nfs4_acl);
5517         nfsi->nfs4_acl = acl;
5518         spin_unlock(&inode->i_lock);
5519 }
5520
5521 static void nfs4_zap_acl_attr(struct inode *inode)
5522 {
5523         nfs4_set_cached_acl(inode, NULL);
5524 }
5525
5526 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5527 {
5528         struct nfs_inode *nfsi = NFS_I(inode);
5529         struct nfs4_cached_acl *acl;
5530         int ret = -ENOENT;
5531
5532         spin_lock(&inode->i_lock);
5533         acl = nfsi->nfs4_acl;
5534         if (acl == NULL)
5535                 goto out;
5536         if (buf == NULL) /* user is just asking for length */
5537                 goto out_len;
5538         if (acl->cached == 0)
5539                 goto out;
5540         ret = -ERANGE; /* see getxattr(2) man page */
5541         if (acl->len > buflen)
5542                 goto out;
5543         memcpy(buf, acl->data, acl->len);
5544 out_len:
5545         ret = acl->len;
5546 out:
5547         spin_unlock(&inode->i_lock);
5548         return ret;
5549 }
5550
5551 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5552 {
5553         struct nfs4_cached_acl *acl;
5554         size_t buflen = sizeof(*acl) + acl_len;
5555
5556         if (buflen <= PAGE_SIZE) {
5557                 acl = kmalloc(buflen, GFP_KERNEL);
5558                 if (acl == NULL)
5559                         goto out;
5560                 acl->cached = 1;
5561                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5562         } else {
5563                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5564                 if (acl == NULL)
5565                         goto out;
5566                 acl->cached = 0;
5567         }
5568         acl->len = acl_len;
5569 out:
5570         nfs4_set_cached_acl(inode, acl);
5571 }
5572
5573 /*
5574  * The getxattr API returns the required buffer length when called with a
5575  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5576  * the required buf.  On a NULL buf, we send a page of data to the server
5577  * guessing that the ACL request can be serviced by a page. If so, we cache
5578  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5579  * the cache. If not so, we throw away the page, and cache the required
5580  * length. The next getxattr call will then produce another round trip to
5581  * the server, this time with the input buf of the required size.
5582  */
5583 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5584 {
5585         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5586         struct nfs_getaclargs args = {
5587                 .fh = NFS_FH(inode),
5588                 .acl_pages = pages,
5589                 .acl_len = buflen,
5590         };
5591         struct nfs_getaclres res = {
5592                 .acl_len = buflen,
5593         };
5594         struct rpc_message msg = {
5595                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5596                 .rpc_argp = &args,
5597                 .rpc_resp = &res,
5598         };
5599         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5600         int ret = -ENOMEM, i;
5601
5602         if (npages > ARRAY_SIZE(pages))
5603                 return -ERANGE;
5604
5605         for (i = 0; i < npages; i++) {
5606                 pages[i] = alloc_page(GFP_KERNEL);
5607                 if (!pages[i])
5608                         goto out_free;
5609         }
5610
5611         /* for decoding across pages */
5612         res.acl_scratch = alloc_page(GFP_KERNEL);
5613         if (!res.acl_scratch)
5614                 goto out_free;
5615
5616         args.acl_len = npages * PAGE_SIZE;
5617
5618         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5619                 __func__, buf, buflen, npages, args.acl_len);
5620         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5621                              &msg, &args.seq_args, &res.seq_res, 0);
5622         if (ret)
5623                 goto out_free;
5624
5625         /* Handle the case where the passed-in buffer is too short */
5626         if (res.acl_flags & NFS4_ACL_TRUNC) {
5627                 /* Did the user only issue a request for the acl length? */
5628                 if (buf == NULL)
5629                         goto out_ok;
5630                 ret = -ERANGE;
5631                 goto out_free;
5632         }
5633         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5634         if (buf) {
5635                 if (res.acl_len > buflen) {
5636                         ret = -ERANGE;
5637                         goto out_free;
5638                 }
5639                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5640         }
5641 out_ok:
5642         ret = res.acl_len;
5643 out_free:
5644         for (i = 0; i < npages; i++)
5645                 if (pages[i])
5646                         __free_page(pages[i]);
5647         if (res.acl_scratch)
5648                 __free_page(res.acl_scratch);
5649         return ret;
5650 }
5651
5652 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5653 {
5654         struct nfs4_exception exception = {
5655                 .interruptible = true,
5656         };
5657         ssize_t ret;
5658         do {
5659                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5660                 trace_nfs4_get_acl(inode, ret);
5661                 if (ret >= 0)
5662                         break;
5663                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5664         } while (exception.retry);
5665         return ret;
5666 }
5667
5668 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5669 {
5670         struct nfs_server *server = NFS_SERVER(inode);
5671         int ret;
5672
5673         if (!nfs4_server_supports_acls(server))
5674                 return -EOPNOTSUPP;
5675         ret = nfs_revalidate_inode(server, inode);
5676         if (ret < 0)
5677                 return ret;
5678         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5679                 nfs_zap_acl_cache(inode);
5680         ret = nfs4_read_cached_acl(inode, buf, buflen);
5681         if (ret != -ENOENT)
5682                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5683                  * but no cached acl data, just the acl length */
5684                 return ret;
5685         return nfs4_get_acl_uncached(inode, buf, buflen);
5686 }
5687
5688 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5689 {
5690         struct nfs_server *server = NFS_SERVER(inode);
5691         struct page *pages[NFS4ACL_MAXPAGES];
5692         struct nfs_setaclargs arg = {
5693                 .fh             = NFS_FH(inode),
5694                 .acl_pages      = pages,
5695                 .acl_len        = buflen,
5696         };
5697         struct nfs_setaclres res;
5698         struct rpc_message msg = {
5699                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5700                 .rpc_argp       = &arg,
5701                 .rpc_resp       = &res,
5702         };
5703         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5704         int ret, i;
5705
5706         if (!nfs4_server_supports_acls(server))
5707                 return -EOPNOTSUPP;
5708         if (npages > ARRAY_SIZE(pages))
5709                 return -ERANGE;
5710         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5711         if (i < 0)
5712                 return i;
5713         nfs4_inode_make_writeable(inode);
5714         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5715
5716         /*
5717          * Free each page after tx, so the only ref left is
5718          * held by the network stack
5719          */
5720         for (; i > 0; i--)
5721                 put_page(pages[i-1]);
5722
5723         /*
5724          * Acl update can result in inode attribute update.
5725          * so mark the attribute cache invalid.
5726          */
5727         spin_lock(&inode->i_lock);
5728         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5729                 | NFS_INO_INVALID_CTIME
5730                 | NFS_INO_REVAL_FORCED;
5731         spin_unlock(&inode->i_lock);
5732         nfs_access_zap_cache(inode);
5733         nfs_zap_acl_cache(inode);
5734         return ret;
5735 }
5736
5737 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5738 {
5739         struct nfs4_exception exception = { };
5740         int err;
5741         do {
5742                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5743                 trace_nfs4_set_acl(inode, err);
5744                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5745                                 &exception);
5746         } while (exception.retry);
5747         return err;
5748 }
5749
5750 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5751 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5752                                         size_t buflen)
5753 {
5754         struct nfs_server *server = NFS_SERVER(inode);
5755         struct nfs_fattr fattr;
5756         struct nfs4_label label = {0, 0, buflen, buf};
5757
5758         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5759         struct nfs4_getattr_arg arg = {
5760                 .fh             = NFS_FH(inode),
5761                 .bitmask        = bitmask,
5762         };
5763         struct nfs4_getattr_res res = {
5764                 .fattr          = &fattr,
5765                 .label          = &label,
5766                 .server         = server,
5767         };
5768         struct rpc_message msg = {
5769                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5770                 .rpc_argp       = &arg,
5771                 .rpc_resp       = &res,
5772         };
5773         int ret;
5774
5775         nfs_fattr_init(&fattr);
5776
5777         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5778         if (ret)
5779                 return ret;
5780         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5781                 return -ENOENT;
5782         if (buflen < label.len)
5783                 return -ERANGE;
5784         return 0;
5785 }
5786
5787 static int nfs4_get_security_label(struct inode *inode, void *buf,
5788                                         size_t buflen)
5789 {
5790         struct nfs4_exception exception = {
5791                 .interruptible = true,
5792         };
5793         int err;
5794
5795         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5796                 return -EOPNOTSUPP;
5797
5798         do {
5799                 err = _nfs4_get_security_label(inode, buf, buflen);
5800                 trace_nfs4_get_security_label(inode, err);
5801                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5802                                 &exception);
5803         } while (exception.retry);
5804         return err;
5805 }
5806
5807 static int _nfs4_do_set_security_label(struct inode *inode,
5808                 struct nfs4_label *ilabel,
5809                 struct nfs_fattr *fattr,
5810                 struct nfs4_label *olabel)
5811 {
5812
5813         struct iattr sattr = {0};
5814         struct nfs_server *server = NFS_SERVER(inode);
5815         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5816         struct nfs_setattrargs arg = {
5817                 .fh             = NFS_FH(inode),
5818                 .iap            = &sattr,
5819                 .server         = server,
5820                 .bitmask        = bitmask,
5821                 .label          = ilabel,
5822         };
5823         struct nfs_setattrres res = {
5824                 .fattr          = fattr,
5825                 .label          = olabel,
5826                 .server         = server,
5827         };
5828         struct rpc_message msg = {
5829                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5830                 .rpc_argp       = &arg,
5831                 .rpc_resp       = &res,
5832         };
5833         int status;
5834
5835         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5836
5837         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5838         if (status)
5839                 dprintk("%s failed: %d\n", __func__, status);
5840
5841         return status;
5842 }
5843
5844 static int nfs4_do_set_security_label(struct inode *inode,
5845                 struct nfs4_label *ilabel,
5846                 struct nfs_fattr *fattr,
5847                 struct nfs4_label *olabel)
5848 {
5849         struct nfs4_exception exception = { };
5850         int err;
5851
5852         do {
5853                 err = _nfs4_do_set_security_label(inode, ilabel,
5854                                 fattr, olabel);
5855                 trace_nfs4_set_security_label(inode, err);
5856                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5857                                 &exception);
5858         } while (exception.retry);
5859         return err;
5860 }
5861
5862 static int
5863 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5864 {
5865         struct nfs4_label ilabel, *olabel = NULL;
5866         struct nfs_fattr fattr;
5867         int status;
5868
5869         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5870                 return -EOPNOTSUPP;
5871
5872         nfs_fattr_init(&fattr);
5873
5874         ilabel.pi = 0;
5875         ilabel.lfs = 0;
5876         ilabel.label = (char *)buf;
5877         ilabel.len = buflen;
5878
5879         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5880         if (IS_ERR(olabel)) {
5881                 status = -PTR_ERR(olabel);
5882                 goto out;
5883         }
5884
5885         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5886         if (status == 0)
5887                 nfs_setsecurity(inode, &fattr, olabel);
5888
5889         nfs4_label_free(olabel);
5890 out:
5891         return status;
5892 }
5893 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5894
5895
5896 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5897                                     nfs4_verifier *bootverf)
5898 {
5899         __be32 verf[2];
5900
5901         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5902                 /* An impossible timestamp guarantees this value
5903                  * will never match a generated boot time. */
5904                 verf[0] = cpu_to_be32(U32_MAX);
5905                 verf[1] = cpu_to_be32(U32_MAX);
5906         } else {
5907                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5908                 u64 ns = ktime_to_ns(nn->boot_time);
5909
5910                 verf[0] = cpu_to_be32(ns >> 32);
5911                 verf[1] = cpu_to_be32(ns);
5912         }
5913         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5914 }
5915
5916 static int
5917 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5918 {
5919         size_t len;
5920         char *str;
5921
5922         if (clp->cl_owner_id != NULL)
5923                 return 0;
5924
5925         rcu_read_lock();
5926         len = 14 +
5927                 strlen(clp->cl_rpcclient->cl_nodename) +
5928                 1 +
5929                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5930                 1;
5931         rcu_read_unlock();
5932         if (nfs4_client_id_uniquifier[0] != '\0')
5933                 len += strlen(nfs4_client_id_uniquifier) + 1;
5934         if (len > NFS4_OPAQUE_LIMIT + 1)
5935                 return -EINVAL;
5936
5937         /*
5938          * Since this string is allocated at mount time, and held until the
5939          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5940          * about a memory-reclaim deadlock.
5941          */
5942         str = kmalloc(len, GFP_KERNEL);
5943         if (!str)
5944                 return -ENOMEM;
5945
5946         rcu_read_lock();
5947         if (nfs4_client_id_uniquifier[0] != '\0')
5948                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5949                           clp->cl_rpcclient->cl_nodename,
5950                           nfs4_client_id_uniquifier,
5951                           rpc_peeraddr2str(clp->cl_rpcclient,
5952                                            RPC_DISPLAY_ADDR));
5953         else
5954                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
5955                           clp->cl_rpcclient->cl_nodename,
5956                           rpc_peeraddr2str(clp->cl_rpcclient,
5957                                            RPC_DISPLAY_ADDR));
5958         rcu_read_unlock();
5959
5960         clp->cl_owner_id = str;
5961         return 0;
5962 }
5963
5964 static int
5965 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5966 {
5967         size_t len;
5968         char *str;
5969
5970         len = 10 + 10 + 1 + 10 + 1 +
5971                 strlen(nfs4_client_id_uniquifier) + 1 +
5972                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5973
5974         if (len > NFS4_OPAQUE_LIMIT + 1)
5975                 return -EINVAL;
5976
5977         /*
5978          * Since this string is allocated at mount time, and held until the
5979          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5980          * about a memory-reclaim deadlock.
5981          */
5982         str = kmalloc(len, GFP_KERNEL);
5983         if (!str)
5984                 return -ENOMEM;
5985
5986         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5987                         clp->rpc_ops->version, clp->cl_minorversion,
5988                         nfs4_client_id_uniquifier,
5989                         clp->cl_rpcclient->cl_nodename);
5990         clp->cl_owner_id = str;
5991         return 0;
5992 }
5993
5994 static int
5995 nfs4_init_uniform_client_string(struct nfs_client *clp)
5996 {
5997         size_t len;
5998         char *str;
5999
6000         if (clp->cl_owner_id != NULL)
6001                 return 0;
6002
6003         if (nfs4_client_id_uniquifier[0] != '\0')
6004                 return nfs4_init_uniquifier_client_string(clp);
6005
6006         len = 10 + 10 + 1 + 10 + 1 +
6007                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6008
6009         if (len > NFS4_OPAQUE_LIMIT + 1)
6010                 return -EINVAL;
6011
6012         /*
6013          * Since this string is allocated at mount time, and held until the
6014          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6015          * about a memory-reclaim deadlock.
6016          */
6017         str = kmalloc(len, GFP_KERNEL);
6018         if (!str)
6019                 return -ENOMEM;
6020
6021         scnprintf(str, len, "Linux NFSv%u.%u %s",
6022                         clp->rpc_ops->version, clp->cl_minorversion,
6023                         clp->cl_rpcclient->cl_nodename);
6024         clp->cl_owner_id = str;
6025         return 0;
6026 }
6027
6028 /*
6029  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6030  * services.  Advertise one based on the address family of the
6031  * clientaddr.
6032  */
6033 static unsigned int
6034 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6035 {
6036         if (strchr(clp->cl_ipaddr, ':') != NULL)
6037                 return scnprintf(buf, len, "tcp6");
6038         else
6039                 return scnprintf(buf, len, "tcp");
6040 }
6041
6042 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6043 {
6044         struct nfs4_setclientid *sc = calldata;
6045
6046         if (task->tk_status == 0)
6047                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6048 }
6049
6050 static const struct rpc_call_ops nfs4_setclientid_ops = {
6051         .rpc_call_done = nfs4_setclientid_done,
6052 };
6053
6054 /**
6055  * nfs4_proc_setclientid - Negotiate client ID
6056  * @clp: state data structure
6057  * @program: RPC program for NFSv4 callback service
6058  * @port: IP port number for NFS4 callback service
6059  * @cred: credential to use for this call
6060  * @res: where to place the result
6061  *
6062  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6063  */
6064 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6065                 unsigned short port, const struct cred *cred,
6066                 struct nfs4_setclientid_res *res)
6067 {
6068         nfs4_verifier sc_verifier;
6069         struct nfs4_setclientid setclientid = {
6070                 .sc_verifier = &sc_verifier,
6071                 .sc_prog = program,
6072                 .sc_clnt = clp,
6073         };
6074         struct rpc_message msg = {
6075                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6076                 .rpc_argp = &setclientid,
6077                 .rpc_resp = res,
6078                 .rpc_cred = cred,
6079         };
6080         struct rpc_task_setup task_setup_data = {
6081                 .rpc_client = clp->cl_rpcclient,
6082                 .rpc_message = &msg,
6083                 .callback_ops = &nfs4_setclientid_ops,
6084                 .callback_data = &setclientid,
6085                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6086         };
6087         int status;
6088
6089         /* nfs_client_id4 */
6090         nfs4_init_boot_verifier(clp, &sc_verifier);
6091
6092         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6093                 status = nfs4_init_uniform_client_string(clp);
6094         else
6095                 status = nfs4_init_nonuniform_client_string(clp);
6096
6097         if (status)
6098                 goto out;
6099
6100         /* cb_client4 */
6101         setclientid.sc_netid_len =
6102                                 nfs4_init_callback_netid(clp,
6103                                                 setclientid.sc_netid,
6104                                                 sizeof(setclientid.sc_netid));
6105         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6106                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6107                                 clp->cl_ipaddr, port >> 8, port & 255);
6108
6109         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6110                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6111                 clp->cl_owner_id);
6112
6113         status = nfs4_call_sync_custom(&task_setup_data);
6114         if (setclientid.sc_cred) {
6115                 kfree(clp->cl_acceptor);
6116                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6117                 put_rpccred(setclientid.sc_cred);
6118         }
6119 out:
6120         trace_nfs4_setclientid(clp, status);
6121         dprintk("NFS reply setclientid: %d\n", status);
6122         return status;
6123 }
6124
6125 /**
6126  * nfs4_proc_setclientid_confirm - Confirm client ID
6127  * @clp: state data structure
6128  * @arg: result of a previous SETCLIENTID
6129  * @cred: credential to use for this call
6130  *
6131  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6132  */
6133 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6134                 struct nfs4_setclientid_res *arg,
6135                 const struct cred *cred)
6136 {
6137         struct rpc_message msg = {
6138                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6139                 .rpc_argp = arg,
6140                 .rpc_cred = cred,
6141         };
6142         int status;
6143
6144         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6145                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6146                 clp->cl_clientid);
6147         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6148                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6149         trace_nfs4_setclientid_confirm(clp, status);
6150         dprintk("NFS reply setclientid_confirm: %d\n", status);
6151         return status;
6152 }
6153
6154 struct nfs4_delegreturndata {
6155         struct nfs4_delegreturnargs args;
6156         struct nfs4_delegreturnres res;
6157         struct nfs_fh fh;
6158         nfs4_stateid stateid;
6159         unsigned long timestamp;
6160         struct {
6161                 struct nfs4_layoutreturn_args arg;
6162                 struct nfs4_layoutreturn_res res;
6163                 struct nfs4_xdr_opaque_data ld_private;
6164                 u32 roc_barrier;
6165                 bool roc;
6166         } lr;
6167         struct nfs_fattr fattr;
6168         int rpc_status;
6169         struct inode *inode;
6170 };
6171
6172 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6173 {
6174         struct nfs4_delegreturndata *data = calldata;
6175         struct nfs4_exception exception = {
6176                 .inode = data->inode,
6177                 .stateid = &data->stateid,
6178         };
6179
6180         if (!nfs4_sequence_done(task, &data->res.seq_res))
6181                 return;
6182
6183         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6184
6185         /* Handle Layoutreturn errors */
6186         if (pnfs_roc_done(task, data->inode,
6187                                 &data->args.lr_args,
6188                                 &data->res.lr_res,
6189                                 &data->res.lr_ret) == -EAGAIN)
6190                 goto out_restart;
6191
6192         switch (task->tk_status) {
6193         case 0:
6194                 renew_lease(data->res.server, data->timestamp);
6195                 break;
6196         case -NFS4ERR_ADMIN_REVOKED:
6197         case -NFS4ERR_DELEG_REVOKED:
6198         case -NFS4ERR_EXPIRED:
6199                 nfs4_free_revoked_stateid(data->res.server,
6200                                 data->args.stateid,
6201                                 task->tk_msg.rpc_cred);
6202                 /* Fallthrough */
6203         case -NFS4ERR_BAD_STATEID:
6204         case -NFS4ERR_STALE_STATEID:
6205                 task->tk_status = 0;
6206                 break;
6207         case -NFS4ERR_OLD_STATEID:
6208                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6209                         nfs4_stateid_seqid_inc(&data->stateid);
6210                 if (data->args.bitmask) {
6211                         data->args.bitmask = NULL;
6212                         data->res.fattr = NULL;
6213                 }
6214                 goto out_restart;
6215         case -NFS4ERR_ACCESS:
6216                 if (data->args.bitmask) {
6217                         data->args.bitmask = NULL;
6218                         data->res.fattr = NULL;
6219                         goto out_restart;
6220                 }
6221                 /* Fallthrough */
6222         default:
6223                 task->tk_status = nfs4_async_handle_exception(task,
6224                                 data->res.server, task->tk_status,
6225                                 &exception);
6226                 if (exception.retry)
6227                         goto out_restart;
6228         }
6229         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6230         data->rpc_status = task->tk_status;
6231         return;
6232 out_restart:
6233         task->tk_status = 0;
6234         rpc_restart_call_prepare(task);
6235 }
6236
6237 static void nfs4_delegreturn_release(void *calldata)
6238 {
6239         struct nfs4_delegreturndata *data = calldata;
6240         struct inode *inode = data->inode;
6241
6242         if (inode) {
6243                 if (data->lr.roc)
6244                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6245                                         data->res.lr_ret);
6246                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6247                 nfs_iput_and_deactive(inode);
6248         }
6249         kfree(calldata);
6250 }
6251
6252 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6253 {
6254         struct nfs4_delegreturndata *d_data;
6255         struct pnfs_layout_hdr *lo;
6256
6257         d_data = (struct nfs4_delegreturndata *)data;
6258
6259         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6260                 nfs4_sequence_done(task, &d_data->res.seq_res);
6261                 return;
6262         }
6263
6264         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6265         if (lo && !pnfs_layout_is_valid(lo)) {
6266                 d_data->args.lr_args = NULL;
6267                 d_data->res.lr_res = NULL;
6268         }
6269
6270         nfs4_setup_sequence(d_data->res.server->nfs_client,
6271                         &d_data->args.seq_args,
6272                         &d_data->res.seq_res,
6273                         task);
6274 }
6275
6276 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6277         .rpc_call_prepare = nfs4_delegreturn_prepare,
6278         .rpc_call_done = nfs4_delegreturn_done,
6279         .rpc_release = nfs4_delegreturn_release,
6280 };
6281
6282 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6283 {
6284         struct nfs4_delegreturndata *data;
6285         struct nfs_server *server = NFS_SERVER(inode);
6286         struct rpc_task *task;
6287         struct rpc_message msg = {
6288                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6289                 .rpc_cred = cred,
6290         };
6291         struct rpc_task_setup task_setup_data = {
6292                 .rpc_client = server->client,
6293                 .rpc_message = &msg,
6294                 .callback_ops = &nfs4_delegreturn_ops,
6295                 .flags = RPC_TASK_ASYNC,
6296         };
6297         int status = 0;
6298
6299         data = kzalloc(sizeof(*data), GFP_NOFS);
6300         if (data == NULL)
6301                 return -ENOMEM;
6302         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6303
6304         nfs4_state_protect(server->nfs_client,
6305                         NFS_SP4_MACH_CRED_CLEANUP,
6306                         &task_setup_data.rpc_client, &msg);
6307
6308         data->args.fhandle = &data->fh;
6309         data->args.stateid = &data->stateid;
6310         data->args.bitmask = server->cache_consistency_bitmask;
6311         nfs_copy_fh(&data->fh, NFS_FH(inode));
6312         nfs4_stateid_copy(&data->stateid, stateid);
6313         data->res.fattr = &data->fattr;
6314         data->res.server = server;
6315         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6316         data->lr.arg.ld_private = &data->lr.ld_private;
6317         nfs_fattr_init(data->res.fattr);
6318         data->timestamp = jiffies;
6319         data->rpc_status = 0;
6320         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6321         data->inode = nfs_igrab_and_active(inode);
6322         if (data->inode) {
6323                 if (data->lr.roc) {
6324                         data->args.lr_args = &data->lr.arg;
6325                         data->res.lr_res = &data->lr.res;
6326                 }
6327         } else if (data->lr.roc) {
6328                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6329                 data->lr.roc = false;
6330         }
6331
6332         task_setup_data.callback_data = data;
6333         msg.rpc_argp = &data->args;
6334         msg.rpc_resp = &data->res;
6335         task = rpc_run_task(&task_setup_data);
6336         if (IS_ERR(task))
6337                 return PTR_ERR(task);
6338         if (!issync)
6339                 goto out;
6340         status = rpc_wait_for_completion_task(task);
6341         if (status != 0)
6342                 goto out;
6343         status = data->rpc_status;
6344 out:
6345         rpc_put_task(task);
6346         return status;
6347 }
6348
6349 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6350 {
6351         struct nfs_server *server = NFS_SERVER(inode);
6352         struct nfs4_exception exception = { };
6353         int err;
6354         do {
6355                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6356                 trace_nfs4_delegreturn(inode, stateid, err);
6357                 switch (err) {
6358                         case -NFS4ERR_STALE_STATEID:
6359                         case -NFS4ERR_EXPIRED:
6360                         case 0:
6361                                 return 0;
6362                 }
6363                 err = nfs4_handle_exception(server, err, &exception);
6364         } while (exception.retry);
6365         return err;
6366 }
6367
6368 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6369 {
6370         struct inode *inode = state->inode;
6371         struct nfs_server *server = NFS_SERVER(inode);
6372         struct nfs_client *clp = server->nfs_client;
6373         struct nfs_lockt_args arg = {
6374                 .fh = NFS_FH(inode),
6375                 .fl = request,
6376         };
6377         struct nfs_lockt_res res = {
6378                 .denied = request,
6379         };
6380         struct rpc_message msg = {
6381                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6382                 .rpc_argp       = &arg,
6383                 .rpc_resp       = &res,
6384                 .rpc_cred       = state->owner->so_cred,
6385         };
6386         struct nfs4_lock_state *lsp;
6387         int status;
6388
6389         arg.lock_owner.clientid = clp->cl_clientid;
6390         status = nfs4_set_lock_state(state, request);
6391         if (status != 0)
6392                 goto out;
6393         lsp = request->fl_u.nfs4_fl.owner;
6394         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6395         arg.lock_owner.s_dev = server->s_dev;
6396         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6397         switch (status) {
6398                 case 0:
6399                         request->fl_type = F_UNLCK;
6400                         break;
6401                 case -NFS4ERR_DENIED:
6402                         status = 0;
6403         }
6404         request->fl_ops->fl_release_private(request);
6405         request->fl_ops = NULL;
6406 out:
6407         return status;
6408 }
6409
6410 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6411 {
6412         struct nfs4_exception exception = {
6413                 .interruptible = true,
6414         };
6415         int err;
6416
6417         do {
6418                 err = _nfs4_proc_getlk(state, cmd, request);
6419                 trace_nfs4_get_lock(request, state, cmd, err);
6420                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6421                                 &exception);
6422         } while (exception.retry);
6423         return err;
6424 }
6425
6426 /*
6427  * Update the seqid of a lock stateid after receiving
6428  * NFS4ERR_OLD_STATEID
6429  */
6430 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6431                 struct nfs4_lock_state *lsp)
6432 {
6433         struct nfs4_state *state = lsp->ls_state;
6434         bool ret = false;
6435
6436         spin_lock(&state->state_lock);
6437         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6438                 goto out;
6439         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6440                 nfs4_stateid_seqid_inc(dst);
6441         else
6442                 dst->seqid = lsp->ls_stateid.seqid;
6443         ret = true;
6444 out:
6445         spin_unlock(&state->state_lock);
6446         return ret;
6447 }
6448
6449 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6450                 struct nfs4_lock_state *lsp)
6451 {
6452         struct nfs4_state *state = lsp->ls_state;
6453         bool ret;
6454
6455         spin_lock(&state->state_lock);
6456         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6457         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6458         spin_unlock(&state->state_lock);
6459         return ret;
6460 }
6461
6462 struct nfs4_unlockdata {
6463         struct nfs_locku_args arg;
6464         struct nfs_locku_res res;
6465         struct nfs4_lock_state *lsp;
6466         struct nfs_open_context *ctx;
6467         struct nfs_lock_context *l_ctx;
6468         struct file_lock fl;
6469         struct nfs_server *server;
6470         unsigned long timestamp;
6471 };
6472
6473 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6474                 struct nfs_open_context *ctx,
6475                 struct nfs4_lock_state *lsp,
6476                 struct nfs_seqid *seqid)
6477 {
6478         struct nfs4_unlockdata *p;
6479         struct nfs4_state *state = lsp->ls_state;
6480         struct inode *inode = state->inode;
6481
6482         p = kzalloc(sizeof(*p), GFP_NOFS);
6483         if (p == NULL)
6484                 return NULL;
6485         p->arg.fh = NFS_FH(inode);
6486         p->arg.fl = &p->fl;
6487         p->arg.seqid = seqid;
6488         p->res.seqid = seqid;
6489         p->lsp = lsp;
6490         /* Ensure we don't close file until we're done freeing locks! */
6491         p->ctx = get_nfs_open_context(ctx);
6492         p->l_ctx = nfs_get_lock_context(ctx);
6493         locks_init_lock(&p->fl);
6494         locks_copy_lock(&p->fl, fl);
6495         p->server = NFS_SERVER(inode);
6496         spin_lock(&state->state_lock);
6497         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6498         spin_unlock(&state->state_lock);
6499         return p;
6500 }
6501
6502 static void nfs4_locku_release_calldata(void *data)
6503 {
6504         struct nfs4_unlockdata *calldata = data;
6505         nfs_free_seqid(calldata->arg.seqid);
6506         nfs4_put_lock_state(calldata->lsp);
6507         nfs_put_lock_context(calldata->l_ctx);
6508         put_nfs_open_context(calldata->ctx);
6509         kfree(calldata);
6510 }
6511
6512 static void nfs4_locku_done(struct rpc_task *task, void *data)
6513 {
6514         struct nfs4_unlockdata *calldata = data;
6515         struct nfs4_exception exception = {
6516                 .inode = calldata->lsp->ls_state->inode,
6517                 .stateid = &calldata->arg.stateid,
6518         };
6519
6520         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6521                 return;
6522         switch (task->tk_status) {
6523                 case 0:
6524                         renew_lease(calldata->server, calldata->timestamp);
6525                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6526                         if (nfs4_update_lock_stateid(calldata->lsp,
6527                                         &calldata->res.stateid))
6528                                 break;
6529                         /* Fall through */
6530                 case -NFS4ERR_ADMIN_REVOKED:
6531                 case -NFS4ERR_EXPIRED:
6532                         nfs4_free_revoked_stateid(calldata->server,
6533                                         &calldata->arg.stateid,
6534                                         task->tk_msg.rpc_cred);
6535                         /* Fall through */
6536                 case -NFS4ERR_BAD_STATEID:
6537                 case -NFS4ERR_STALE_STATEID:
6538                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6539                                                 calldata->lsp))
6540                                 rpc_restart_call_prepare(task);
6541                         break;
6542                 case -NFS4ERR_OLD_STATEID:
6543                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6544                                                 calldata->lsp))
6545                                 rpc_restart_call_prepare(task);
6546                         break;
6547                 default:
6548                         task->tk_status = nfs4_async_handle_exception(task,
6549                                         calldata->server, task->tk_status,
6550                                         &exception);
6551                         if (exception.retry)
6552                                 rpc_restart_call_prepare(task);
6553         }
6554         nfs_release_seqid(calldata->arg.seqid);
6555 }
6556
6557 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6558 {
6559         struct nfs4_unlockdata *calldata = data;
6560
6561         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6562                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6563                 return;
6564
6565         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6566                 goto out_wait;
6567         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6568                 /* Note: exit _without_ running nfs4_locku_done */
6569                 goto out_no_action;
6570         }
6571         calldata->timestamp = jiffies;
6572         if (nfs4_setup_sequence(calldata->server->nfs_client,
6573                                 &calldata->arg.seq_args,
6574                                 &calldata->res.seq_res,
6575                                 task) != 0)
6576                 nfs_release_seqid(calldata->arg.seqid);
6577         return;
6578 out_no_action:
6579         task->tk_action = NULL;
6580 out_wait:
6581         nfs4_sequence_done(task, &calldata->res.seq_res);
6582 }
6583
6584 static const struct rpc_call_ops nfs4_locku_ops = {
6585         .rpc_call_prepare = nfs4_locku_prepare,
6586         .rpc_call_done = nfs4_locku_done,
6587         .rpc_release = nfs4_locku_release_calldata,
6588 };
6589
6590 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6591                 struct nfs_open_context *ctx,
6592                 struct nfs4_lock_state *lsp,
6593                 struct nfs_seqid *seqid)
6594 {
6595         struct nfs4_unlockdata *data;
6596         struct rpc_message msg = {
6597                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6598                 .rpc_cred = ctx->cred,
6599         };
6600         struct rpc_task_setup task_setup_data = {
6601                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6602                 .rpc_message = &msg,
6603                 .callback_ops = &nfs4_locku_ops,
6604                 .workqueue = nfsiod_workqueue,
6605                 .flags = RPC_TASK_ASYNC,
6606         };
6607
6608         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6609                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6610
6611         /* Ensure this is an unlock - when canceling a lock, the
6612          * canceled lock is passed in, and it won't be an unlock.
6613          */
6614         fl->fl_type = F_UNLCK;
6615         if (fl->fl_flags & FL_CLOSE)
6616                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6617
6618         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6619         if (data == NULL) {
6620                 nfs_free_seqid(seqid);
6621                 return ERR_PTR(-ENOMEM);
6622         }
6623
6624         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6625         msg.rpc_argp = &data->arg;
6626         msg.rpc_resp = &data->res;
6627         task_setup_data.callback_data = data;
6628         return rpc_run_task(&task_setup_data);
6629 }
6630
6631 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6632 {
6633         struct inode *inode = state->inode;
6634         struct nfs4_state_owner *sp = state->owner;
6635         struct nfs_inode *nfsi = NFS_I(inode);
6636         struct nfs_seqid *seqid;
6637         struct nfs4_lock_state *lsp;
6638         struct rpc_task *task;
6639         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6640         int status = 0;
6641         unsigned char fl_flags = request->fl_flags;
6642
6643         status = nfs4_set_lock_state(state, request);
6644         /* Unlock _before_ we do the RPC call */
6645         request->fl_flags |= FL_EXISTS;
6646         /* Exclude nfs_delegation_claim_locks() */
6647         mutex_lock(&sp->so_delegreturn_mutex);
6648         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6649         down_read(&nfsi->rwsem);
6650         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6651                 up_read(&nfsi->rwsem);
6652                 mutex_unlock(&sp->so_delegreturn_mutex);
6653                 goto out;
6654         }
6655         up_read(&nfsi->rwsem);
6656         mutex_unlock(&sp->so_delegreturn_mutex);
6657         if (status != 0)
6658                 goto out;
6659         /* Is this a delegated lock? */
6660         lsp = request->fl_u.nfs4_fl.owner;
6661         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6662                 goto out;
6663         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6664         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6665         status = -ENOMEM;
6666         if (IS_ERR(seqid))
6667                 goto out;
6668         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6669         status = PTR_ERR(task);
6670         if (IS_ERR(task))
6671                 goto out;
6672         status = rpc_wait_for_completion_task(task);
6673         rpc_put_task(task);
6674 out:
6675         request->fl_flags = fl_flags;
6676         trace_nfs4_unlock(request, state, F_SETLK, status);
6677         return status;
6678 }
6679
6680 struct nfs4_lockdata {
6681         struct nfs_lock_args arg;
6682         struct nfs_lock_res res;
6683         struct nfs4_lock_state *lsp;
6684         struct nfs_open_context *ctx;
6685         struct file_lock fl;
6686         unsigned long timestamp;
6687         int rpc_status;
6688         int cancelled;
6689         struct nfs_server *server;
6690 };
6691
6692 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6693                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6694                 gfp_t gfp_mask)
6695 {
6696         struct nfs4_lockdata *p;
6697         struct inode *inode = lsp->ls_state->inode;
6698         struct nfs_server *server = NFS_SERVER(inode);
6699         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6700
6701         p = kzalloc(sizeof(*p), gfp_mask);
6702         if (p == NULL)
6703                 return NULL;
6704
6705         p->arg.fh = NFS_FH(inode);
6706         p->arg.fl = &p->fl;
6707         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6708         if (IS_ERR(p->arg.open_seqid))
6709                 goto out_free;
6710         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6711         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6712         if (IS_ERR(p->arg.lock_seqid))
6713                 goto out_free_seqid;
6714         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6715         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6716         p->arg.lock_owner.s_dev = server->s_dev;
6717         p->res.lock_seqid = p->arg.lock_seqid;
6718         p->lsp = lsp;
6719         p->server = server;
6720         p->ctx = get_nfs_open_context(ctx);
6721         locks_init_lock(&p->fl);
6722         locks_copy_lock(&p->fl, fl);
6723         return p;
6724 out_free_seqid:
6725         nfs_free_seqid(p->arg.open_seqid);
6726 out_free:
6727         kfree(p);
6728         return NULL;
6729 }
6730
6731 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6732 {
6733         struct nfs4_lockdata *data = calldata;
6734         struct nfs4_state *state = data->lsp->ls_state;
6735
6736         dprintk("%s: begin!\n", __func__);
6737         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6738                 goto out_wait;
6739         /* Do we need to do an open_to_lock_owner? */
6740         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6741                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6742                         goto out_release_lock_seqid;
6743                 }
6744                 nfs4_stateid_copy(&data->arg.open_stateid,
6745                                 &state->open_stateid);
6746                 data->arg.new_lock_owner = 1;
6747                 data->res.open_seqid = data->arg.open_seqid;
6748         } else {
6749                 data->arg.new_lock_owner = 0;
6750                 nfs4_stateid_copy(&data->arg.lock_stateid,
6751                                 &data->lsp->ls_stateid);
6752         }
6753         if (!nfs4_valid_open_stateid(state)) {
6754                 data->rpc_status = -EBADF;
6755                 task->tk_action = NULL;
6756                 goto out_release_open_seqid;
6757         }
6758         data->timestamp = jiffies;
6759         if (nfs4_setup_sequence(data->server->nfs_client,
6760                                 &data->arg.seq_args,
6761                                 &data->res.seq_res,
6762                                 task) == 0)
6763                 return;
6764 out_release_open_seqid:
6765         nfs_release_seqid(data->arg.open_seqid);
6766 out_release_lock_seqid:
6767         nfs_release_seqid(data->arg.lock_seqid);
6768 out_wait:
6769         nfs4_sequence_done(task, &data->res.seq_res);
6770         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6771 }
6772
6773 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6774 {
6775         struct nfs4_lockdata *data = calldata;
6776         struct nfs4_lock_state *lsp = data->lsp;
6777
6778         dprintk("%s: begin!\n", __func__);
6779
6780         if (!nfs4_sequence_done(task, &data->res.seq_res))
6781                 return;
6782
6783         data->rpc_status = task->tk_status;
6784         switch (task->tk_status) {
6785         case 0:
6786                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6787                                 data->timestamp);
6788                 if (data->arg.new_lock && !data->cancelled) {
6789                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6790                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6791                                 goto out_restart;
6792                 }
6793                 if (data->arg.new_lock_owner != 0) {
6794                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6795                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6796                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6797                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6798                         goto out_restart;
6799                 break;
6800         case -NFS4ERR_BAD_STATEID:
6801         case -NFS4ERR_OLD_STATEID:
6802         case -NFS4ERR_STALE_STATEID:
6803         case -NFS4ERR_EXPIRED:
6804                 if (data->arg.new_lock_owner != 0) {
6805                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6806                                                 &lsp->ls_state->open_stateid))
6807                                 goto out_restart;
6808                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6809                                                 &lsp->ls_stateid))
6810                                 goto out_restart;
6811         }
6812 out_done:
6813         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6814         return;
6815 out_restart:
6816         if (!data->cancelled)
6817                 rpc_restart_call_prepare(task);
6818         goto out_done;
6819 }
6820
6821 static void nfs4_lock_release(void *calldata)
6822 {
6823         struct nfs4_lockdata *data = calldata;
6824
6825         dprintk("%s: begin!\n", __func__);
6826         nfs_free_seqid(data->arg.open_seqid);
6827         if (data->cancelled && data->rpc_status == 0) {
6828                 struct rpc_task *task;
6829                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6830                                 data->arg.lock_seqid);
6831                 if (!IS_ERR(task))
6832                         rpc_put_task_async(task);
6833                 dprintk("%s: cancelling lock!\n", __func__);
6834         } else
6835                 nfs_free_seqid(data->arg.lock_seqid);
6836         nfs4_put_lock_state(data->lsp);
6837         put_nfs_open_context(data->ctx);
6838         kfree(data);
6839         dprintk("%s: done!\n", __func__);
6840 }
6841
6842 static const struct rpc_call_ops nfs4_lock_ops = {
6843         .rpc_call_prepare = nfs4_lock_prepare,
6844         .rpc_call_done = nfs4_lock_done,
6845         .rpc_release = nfs4_lock_release,
6846 };
6847
6848 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6849 {
6850         switch (error) {
6851         case -NFS4ERR_ADMIN_REVOKED:
6852         case -NFS4ERR_EXPIRED:
6853         case -NFS4ERR_BAD_STATEID:
6854                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6855                 if (new_lock_owner != 0 ||
6856                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6857                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6858                 break;
6859         case -NFS4ERR_STALE_STATEID:
6860                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6861                 nfs4_schedule_lease_recovery(server->nfs_client);
6862         };
6863 }
6864
6865 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6866 {
6867         struct nfs4_lockdata *data;
6868         struct rpc_task *task;
6869         struct rpc_message msg = {
6870                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6871                 .rpc_cred = state->owner->so_cred,
6872         };
6873         struct rpc_task_setup task_setup_data = {
6874                 .rpc_client = NFS_CLIENT(state->inode),
6875                 .rpc_message = &msg,
6876                 .callback_ops = &nfs4_lock_ops,
6877                 .workqueue = nfsiod_workqueue,
6878                 .flags = RPC_TASK_ASYNC,
6879         };
6880         int ret;
6881
6882         dprintk("%s: begin!\n", __func__);
6883         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6884                         fl->fl_u.nfs4_fl.owner,
6885                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6886         if (data == NULL)
6887                 return -ENOMEM;
6888         if (IS_SETLKW(cmd))
6889                 data->arg.block = 1;
6890         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6891                                 recovery_type > NFS_LOCK_NEW);
6892         msg.rpc_argp = &data->arg;
6893         msg.rpc_resp = &data->res;
6894         task_setup_data.callback_data = data;
6895         if (recovery_type > NFS_LOCK_NEW) {
6896                 if (recovery_type == NFS_LOCK_RECLAIM)
6897                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6898         } else
6899                 data->arg.new_lock = 1;
6900         task = rpc_run_task(&task_setup_data);
6901         if (IS_ERR(task))
6902                 return PTR_ERR(task);
6903         ret = rpc_wait_for_completion_task(task);
6904         if (ret == 0) {
6905                 ret = data->rpc_status;
6906                 if (ret)
6907                         nfs4_handle_setlk_error(data->server, data->lsp,
6908                                         data->arg.new_lock_owner, ret);
6909         } else
6910                 data->cancelled = true;
6911         rpc_put_task(task);
6912         dprintk("%s: done, ret = %d!\n", __func__, ret);
6913         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6914         return ret;
6915 }
6916
6917 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6918 {
6919         struct nfs_server *server = NFS_SERVER(state->inode);
6920         struct nfs4_exception exception = {
6921                 .inode = state->inode,
6922         };
6923         int err;
6924
6925         do {
6926                 /* Cache the lock if possible... */
6927                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6928                         return 0;
6929                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6930                 if (err != -NFS4ERR_DELAY)
6931                         break;
6932                 nfs4_handle_exception(server, err, &exception);
6933         } while (exception.retry);
6934         return err;
6935 }
6936
6937 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6938 {
6939         struct nfs_server *server = NFS_SERVER(state->inode);
6940         struct nfs4_exception exception = {
6941                 .inode = state->inode,
6942         };
6943         int err;
6944
6945         err = nfs4_set_lock_state(state, request);
6946         if (err != 0)
6947                 return err;
6948         if (!recover_lost_locks) {
6949                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6950                 return 0;
6951         }
6952         do {
6953                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6954                         return 0;
6955                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6956                 switch (err) {
6957                 default:
6958                         goto out;
6959                 case -NFS4ERR_GRACE:
6960                 case -NFS4ERR_DELAY:
6961                         nfs4_handle_exception(server, err, &exception);
6962                         err = 0;
6963                 }
6964         } while (exception.retry);
6965 out:
6966         return err;
6967 }
6968
6969 #if defined(CONFIG_NFS_V4_1)
6970 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6971 {
6972         struct nfs4_lock_state *lsp;
6973         int status;
6974
6975         status = nfs4_set_lock_state(state, request);
6976         if (status != 0)
6977                 return status;
6978         lsp = request->fl_u.nfs4_fl.owner;
6979         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6980             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6981                 return 0;
6982         return nfs4_lock_expired(state, request);
6983 }
6984 #endif
6985
6986 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6987 {
6988         struct nfs_inode *nfsi = NFS_I(state->inode);
6989         struct nfs4_state_owner *sp = state->owner;
6990         unsigned char fl_flags = request->fl_flags;
6991         int status;
6992
6993         request->fl_flags |= FL_ACCESS;
6994         status = locks_lock_inode_wait(state->inode, request);
6995         if (status < 0)
6996                 goto out;
6997         mutex_lock(&sp->so_delegreturn_mutex);
6998         down_read(&nfsi->rwsem);
6999         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7000                 /* Yes: cache locks! */
7001                 /* ...but avoid races with delegation recall... */
7002                 request->fl_flags = fl_flags & ~FL_SLEEP;
7003                 status = locks_lock_inode_wait(state->inode, request);
7004                 up_read(&nfsi->rwsem);
7005                 mutex_unlock(&sp->so_delegreturn_mutex);
7006                 goto out;
7007         }
7008         up_read(&nfsi->rwsem);
7009         mutex_unlock(&sp->so_delegreturn_mutex);
7010         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7011 out:
7012         request->fl_flags = fl_flags;
7013         return status;
7014 }
7015
7016 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7017 {
7018         struct nfs4_exception exception = {
7019                 .state = state,
7020                 .inode = state->inode,
7021                 .interruptible = true,
7022         };
7023         int err;
7024
7025         do {
7026                 err = _nfs4_proc_setlk(state, cmd, request);
7027                 if (err == -NFS4ERR_DENIED)
7028                         err = -EAGAIN;
7029                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7030                                 err, &exception);
7031         } while (exception.retry);
7032         return err;
7033 }
7034
7035 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7036 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7037
7038 static int
7039 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7040                         struct file_lock *request)
7041 {
7042         int             status = -ERESTARTSYS;
7043         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7044
7045         while(!signalled()) {
7046                 status = nfs4_proc_setlk(state, cmd, request);
7047                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7048                         break;
7049                 freezable_schedule_timeout_interruptible(timeout);
7050                 timeout *= 2;
7051                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7052                 status = -ERESTARTSYS;
7053         }
7054         return status;
7055 }
7056
7057 #ifdef CONFIG_NFS_V4_1
7058 struct nfs4_lock_waiter {
7059         struct task_struct      *task;
7060         struct inode            *inode;
7061         struct nfs_lowner       *owner;
7062 };
7063
7064 static int
7065 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7066 {
7067         int ret;
7068         struct nfs4_lock_waiter *waiter = wait->private;
7069
7070         /* NULL key means to wake up everyone */
7071         if (key) {
7072                 struct cb_notify_lock_args      *cbnl = key;
7073                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7074                                                 *wowner = waiter->owner;
7075
7076                 /* Only wake if the callback was for the same owner. */
7077                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7078                         return 0;
7079
7080                 /* Make sure it's for the right inode */
7081                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7082                         return 0;
7083         }
7084
7085         /* override "private" so we can use default_wake_function */
7086         wait->private = waiter->task;
7087         ret = woken_wake_function(wait, mode, flags, key);
7088         if (ret)
7089                 list_del_init(&wait->entry);
7090         wait->private = waiter;
7091         return ret;
7092 }
7093
7094 static int
7095 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7096 {
7097         int status = -ERESTARTSYS;
7098         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7099         struct nfs_server *server = NFS_SERVER(state->inode);
7100         struct nfs_client *clp = server->nfs_client;
7101         wait_queue_head_t *q = &clp->cl_lock_waitq;
7102         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7103                                     .id = lsp->ls_seqid.owner_id,
7104                                     .s_dev = server->s_dev };
7105         struct nfs4_lock_waiter waiter = { .task  = current,
7106                                            .inode = state->inode,
7107                                            .owner = &owner};
7108         wait_queue_entry_t wait;
7109
7110         /* Don't bother with waitqueue if we don't expect a callback */
7111         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7112                 return nfs4_retry_setlk_simple(state, cmd, request);
7113
7114         init_wait(&wait);
7115         wait.private = &waiter;
7116         wait.func = nfs4_wake_lock_waiter;
7117
7118         while(!signalled()) {
7119                 add_wait_queue(q, &wait);
7120                 status = nfs4_proc_setlk(state, cmd, request);
7121                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7122                         finish_wait(q, &wait);
7123                         break;
7124                 }
7125
7126                 status = -ERESTARTSYS;
7127                 freezer_do_not_count();
7128                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7129                 freezer_count();
7130                 finish_wait(q, &wait);
7131         }
7132
7133         return status;
7134 }
7135 #else /* !CONFIG_NFS_V4_1 */
7136 static inline int
7137 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7138 {
7139         return nfs4_retry_setlk_simple(state, cmd, request);
7140 }
7141 #endif
7142
7143 static int
7144 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7145 {
7146         struct nfs_open_context *ctx;
7147         struct nfs4_state *state;
7148         int status;
7149
7150         /* verify open state */
7151         ctx = nfs_file_open_context(filp);
7152         state = ctx->state;
7153
7154         if (IS_GETLK(cmd)) {
7155                 if (state != NULL)
7156                         return nfs4_proc_getlk(state, F_GETLK, request);
7157                 return 0;
7158         }
7159
7160         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7161                 return -EINVAL;
7162
7163         if (request->fl_type == F_UNLCK) {
7164                 if (state != NULL)
7165                         return nfs4_proc_unlck(state, cmd, request);
7166                 return 0;
7167         }
7168
7169         if (state == NULL)
7170                 return -ENOLCK;
7171
7172         if ((request->fl_flags & FL_POSIX) &&
7173             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7174                 return -ENOLCK;
7175
7176         /*
7177          * Don't rely on the VFS having checked the file open mode,
7178          * since it won't do this for flock() locks.
7179          */
7180         switch (request->fl_type) {
7181         case F_RDLCK:
7182                 if (!(filp->f_mode & FMODE_READ))
7183                         return -EBADF;
7184                 break;
7185         case F_WRLCK:
7186                 if (!(filp->f_mode & FMODE_WRITE))
7187                         return -EBADF;
7188         }
7189
7190         status = nfs4_set_lock_state(state, request);
7191         if (status != 0)
7192                 return status;
7193
7194         return nfs4_retry_setlk(state, cmd, request);
7195 }
7196
7197 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7198 {
7199         struct nfs_server *server = NFS_SERVER(state->inode);
7200         int err;
7201
7202         err = nfs4_set_lock_state(state, fl);
7203         if (err != 0)
7204                 return err;
7205         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7206         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7207 }
7208
7209 struct nfs_release_lockowner_data {
7210         struct nfs4_lock_state *lsp;
7211         struct nfs_server *server;
7212         struct nfs_release_lockowner_args args;
7213         struct nfs_release_lockowner_res res;
7214         unsigned long timestamp;
7215 };
7216
7217 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7218 {
7219         struct nfs_release_lockowner_data *data = calldata;
7220         struct nfs_server *server = data->server;
7221         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7222                            &data->res.seq_res, task);
7223         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7224         data->timestamp = jiffies;
7225 }
7226
7227 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7228 {
7229         struct nfs_release_lockowner_data *data = calldata;
7230         struct nfs_server *server = data->server;
7231
7232         nfs40_sequence_done(task, &data->res.seq_res);
7233
7234         switch (task->tk_status) {
7235         case 0:
7236                 renew_lease(server, data->timestamp);
7237                 break;
7238         case -NFS4ERR_STALE_CLIENTID:
7239         case -NFS4ERR_EXPIRED:
7240                 nfs4_schedule_lease_recovery(server->nfs_client);
7241                 break;
7242         case -NFS4ERR_LEASE_MOVED:
7243         case -NFS4ERR_DELAY:
7244                 if (nfs4_async_handle_error(task, server,
7245                                             NULL, NULL) == -EAGAIN)
7246                         rpc_restart_call_prepare(task);
7247         }
7248 }
7249
7250 static void nfs4_release_lockowner_release(void *calldata)
7251 {
7252         struct nfs_release_lockowner_data *data = calldata;
7253         nfs4_free_lock_state(data->server, data->lsp);
7254         kfree(calldata);
7255 }
7256
7257 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7258         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7259         .rpc_call_done = nfs4_release_lockowner_done,
7260         .rpc_release = nfs4_release_lockowner_release,
7261 };
7262
7263 static void
7264 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7265 {
7266         struct nfs_release_lockowner_data *data;
7267         struct rpc_message msg = {
7268                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7269         };
7270
7271         if (server->nfs_client->cl_mvops->minor_version != 0)
7272                 return;
7273
7274         data = kmalloc(sizeof(*data), GFP_NOFS);
7275         if (!data)
7276                 return;
7277         data->lsp = lsp;
7278         data->server = server;
7279         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7280         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7281         data->args.lock_owner.s_dev = server->s_dev;
7282
7283         msg.rpc_argp = &data->args;
7284         msg.rpc_resp = &data->res;
7285         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7286         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7287 }
7288
7289 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7290
7291 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7292                                    struct dentry *unused, struct inode *inode,
7293                                    const char *key, const void *buf,
7294                                    size_t buflen, int flags)
7295 {
7296         return nfs4_proc_set_acl(inode, buf, buflen);
7297 }
7298
7299 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7300                                    struct dentry *unused, struct inode *inode,
7301                                    const char *key, void *buf, size_t buflen)
7302 {
7303         return nfs4_proc_get_acl(inode, buf, buflen);
7304 }
7305
7306 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7307 {
7308         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7309 }
7310
7311 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7312
7313 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7314                                      struct dentry *unused, struct inode *inode,
7315                                      const char *key, const void *buf,
7316                                      size_t buflen, int flags)
7317 {
7318         if (security_ismaclabel(key))
7319                 return nfs4_set_security_label(inode, buf, buflen);
7320
7321         return -EOPNOTSUPP;
7322 }
7323
7324 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7325                                      struct dentry *unused, struct inode *inode,
7326                                      const char *key, void *buf, size_t buflen)
7327 {
7328         if (security_ismaclabel(key))
7329                 return nfs4_get_security_label(inode, buf, buflen);
7330         return -EOPNOTSUPP;
7331 }
7332
7333 static ssize_t
7334 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7335 {
7336         int len = 0;
7337
7338         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7339                 len = security_inode_listsecurity(inode, list, list_len);
7340                 if (list_len && len > list_len)
7341                         return -ERANGE;
7342         }
7343         return len;
7344 }
7345
7346 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7347         .prefix = XATTR_SECURITY_PREFIX,
7348         .get    = nfs4_xattr_get_nfs4_label,
7349         .set    = nfs4_xattr_set_nfs4_label,
7350 };
7351
7352 #else
7353
7354 static ssize_t
7355 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7356 {
7357         return 0;
7358 }
7359
7360 #endif
7361
7362 /*
7363  * nfs_fhget will use either the mounted_on_fileid or the fileid
7364  */
7365 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7366 {
7367         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7368                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7369               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7370               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7371                 return;
7372
7373         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7374                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7375         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7376         fattr->nlink = 2;
7377 }
7378
7379 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7380                                    const struct qstr *name,
7381                                    struct nfs4_fs_locations *fs_locations,
7382                                    struct page *page)
7383 {
7384         struct nfs_server *server = NFS_SERVER(dir);
7385         u32 bitmask[3];
7386         struct nfs4_fs_locations_arg args = {
7387                 .dir_fh = NFS_FH(dir),
7388                 .name = name,
7389                 .page = page,
7390                 .bitmask = bitmask,
7391         };
7392         struct nfs4_fs_locations_res res = {
7393                 .fs_locations = fs_locations,
7394         };
7395         struct rpc_message msg = {
7396                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7397                 .rpc_argp = &args,
7398                 .rpc_resp = &res,
7399         };
7400         int status;
7401
7402         dprintk("%s: start\n", __func__);
7403
7404         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7405         bitmask[1] = nfs4_fattr_bitmap[1];
7406
7407         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7408          * is not supported */
7409         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7410                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7411         else
7412                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7413
7414         nfs_fattr_init(&fs_locations->fattr);
7415         fs_locations->server = server;
7416         fs_locations->nlocations = 0;
7417         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7418         dprintk("%s: returned status = %d\n", __func__, status);
7419         return status;
7420 }
7421
7422 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7423                            const struct qstr *name,
7424                            struct nfs4_fs_locations *fs_locations,
7425                            struct page *page)
7426 {
7427         struct nfs4_exception exception = {
7428                 .interruptible = true,
7429         };
7430         int err;
7431         do {
7432                 err = _nfs4_proc_fs_locations(client, dir, name,
7433                                 fs_locations, page);
7434                 trace_nfs4_get_fs_locations(dir, name, err);
7435                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7436                                 &exception);
7437         } while (exception.retry);
7438         return err;
7439 }
7440
7441 /*
7442  * This operation also signals the server that this client is
7443  * performing migration recovery.  The server can stop returning
7444  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7445  * appended to this compound to identify the client ID which is
7446  * performing recovery.
7447  */
7448 static int _nfs40_proc_get_locations(struct inode *inode,
7449                                      struct nfs4_fs_locations *locations,
7450                                      struct page *page, const struct cred *cred)
7451 {
7452         struct nfs_server *server = NFS_SERVER(inode);
7453         struct rpc_clnt *clnt = server->client;
7454         u32 bitmask[2] = {
7455                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7456         };
7457         struct nfs4_fs_locations_arg args = {
7458                 .clientid       = server->nfs_client->cl_clientid,
7459                 .fh             = NFS_FH(inode),
7460                 .page           = page,
7461                 .bitmask        = bitmask,
7462                 .migration      = 1,            /* skip LOOKUP */
7463                 .renew          = 1,            /* append RENEW */
7464         };
7465         struct nfs4_fs_locations_res res = {
7466                 .fs_locations   = locations,
7467                 .migration      = 1,
7468                 .renew          = 1,
7469         };
7470         struct rpc_message msg = {
7471                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7472                 .rpc_argp       = &args,
7473                 .rpc_resp       = &res,
7474                 .rpc_cred       = cred,
7475         };
7476         unsigned long now = jiffies;
7477         int status;
7478
7479         nfs_fattr_init(&locations->fattr);
7480         locations->server = server;
7481         locations->nlocations = 0;
7482
7483         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7484         status = nfs4_call_sync_sequence(clnt, server, &msg,
7485                                         &args.seq_args, &res.seq_res);
7486         if (status)
7487                 return status;
7488
7489         renew_lease(server, now);
7490         return 0;
7491 }
7492
7493 #ifdef CONFIG_NFS_V4_1
7494
7495 /*
7496  * This operation also signals the server that this client is
7497  * performing migration recovery.  The server can stop asserting
7498  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7499  * performing this operation is identified in the SEQUENCE
7500  * operation in this compound.
7501  *
7502  * When the client supports GETATTR(fs_locations_info), it can
7503  * be plumbed in here.
7504  */
7505 static int _nfs41_proc_get_locations(struct inode *inode,
7506                                      struct nfs4_fs_locations *locations,
7507                                      struct page *page, const struct cred *cred)
7508 {
7509         struct nfs_server *server = NFS_SERVER(inode);
7510         struct rpc_clnt *clnt = server->client;
7511         u32 bitmask[2] = {
7512                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7513         };
7514         struct nfs4_fs_locations_arg args = {
7515                 .fh             = NFS_FH(inode),
7516                 .page           = page,
7517                 .bitmask        = bitmask,
7518                 .migration      = 1,            /* skip LOOKUP */
7519         };
7520         struct nfs4_fs_locations_res res = {
7521                 .fs_locations   = locations,
7522                 .migration      = 1,
7523         };
7524         struct rpc_message msg = {
7525                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7526                 .rpc_argp       = &args,
7527                 .rpc_resp       = &res,
7528                 .rpc_cred       = cred,
7529         };
7530         int status;
7531
7532         nfs_fattr_init(&locations->fattr);
7533         locations->server = server;
7534         locations->nlocations = 0;
7535
7536         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7537         status = nfs4_call_sync_sequence(clnt, server, &msg,
7538                                         &args.seq_args, &res.seq_res);
7539         if (status == NFS4_OK &&
7540             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7541                 status = -NFS4ERR_LEASE_MOVED;
7542         return status;
7543 }
7544
7545 #endif  /* CONFIG_NFS_V4_1 */
7546
7547 /**
7548  * nfs4_proc_get_locations - discover locations for a migrated FSID
7549  * @inode: inode on FSID that is migrating
7550  * @locations: result of query
7551  * @page: buffer
7552  * @cred: credential to use for this operation
7553  *
7554  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7555  * operation failed, or a negative errno if a local error occurred.
7556  *
7557  * On success, "locations" is filled in, but if the server has
7558  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7559  * asserted.
7560  *
7561  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7562  * from this client that require migration recovery.
7563  */
7564 int nfs4_proc_get_locations(struct inode *inode,
7565                             struct nfs4_fs_locations *locations,
7566                             struct page *page, const struct cred *cred)
7567 {
7568         struct nfs_server *server = NFS_SERVER(inode);
7569         struct nfs_client *clp = server->nfs_client;
7570         const struct nfs4_mig_recovery_ops *ops =
7571                                         clp->cl_mvops->mig_recovery_ops;
7572         struct nfs4_exception exception = {
7573                 .interruptible = true,
7574         };
7575         int status;
7576
7577         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7578                 (unsigned long long)server->fsid.major,
7579                 (unsigned long long)server->fsid.minor,
7580                 clp->cl_hostname);
7581         nfs_display_fhandle(NFS_FH(inode), __func__);
7582
7583         do {
7584                 status = ops->get_locations(inode, locations, page, cred);
7585                 if (status != -NFS4ERR_DELAY)
7586                         break;
7587                 nfs4_handle_exception(server, status, &exception);
7588         } while (exception.retry);
7589         return status;
7590 }
7591
7592 /*
7593  * This operation also signals the server that this client is
7594  * performing "lease moved" recovery.  The server can stop
7595  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7596  * is appended to this compound to identify the client ID which is
7597  * performing recovery.
7598  */
7599 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7600 {
7601         struct nfs_server *server = NFS_SERVER(inode);
7602         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7603         struct rpc_clnt *clnt = server->client;
7604         struct nfs4_fsid_present_arg args = {
7605                 .fh             = NFS_FH(inode),
7606                 .clientid       = clp->cl_clientid,
7607                 .renew          = 1,            /* append RENEW */
7608         };
7609         struct nfs4_fsid_present_res res = {
7610                 .renew          = 1,
7611         };
7612         struct rpc_message msg = {
7613                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7614                 .rpc_argp       = &args,
7615                 .rpc_resp       = &res,
7616                 .rpc_cred       = cred,
7617         };
7618         unsigned long now = jiffies;
7619         int status;
7620
7621         res.fh = nfs_alloc_fhandle();
7622         if (res.fh == NULL)
7623                 return -ENOMEM;
7624
7625         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7626         status = nfs4_call_sync_sequence(clnt, server, &msg,
7627                                                 &args.seq_args, &res.seq_res);
7628         nfs_free_fhandle(res.fh);
7629         if (status)
7630                 return status;
7631
7632         do_renew_lease(clp, now);
7633         return 0;
7634 }
7635
7636 #ifdef CONFIG_NFS_V4_1
7637
7638 /*
7639  * This operation also signals the server that this client is
7640  * performing "lease moved" recovery.  The server can stop asserting
7641  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7642  * this operation is identified in the SEQUENCE operation in this
7643  * compound.
7644  */
7645 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7646 {
7647         struct nfs_server *server = NFS_SERVER(inode);
7648         struct rpc_clnt *clnt = server->client;
7649         struct nfs4_fsid_present_arg args = {
7650                 .fh             = NFS_FH(inode),
7651         };
7652         struct nfs4_fsid_present_res res = {
7653         };
7654         struct rpc_message msg = {
7655                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7656                 .rpc_argp       = &args,
7657                 .rpc_resp       = &res,
7658                 .rpc_cred       = cred,
7659         };
7660         int status;
7661
7662         res.fh = nfs_alloc_fhandle();
7663         if (res.fh == NULL)
7664                 return -ENOMEM;
7665
7666         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7667         status = nfs4_call_sync_sequence(clnt, server, &msg,
7668                                                 &args.seq_args, &res.seq_res);
7669         nfs_free_fhandle(res.fh);
7670         if (status == NFS4_OK &&
7671             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7672                 status = -NFS4ERR_LEASE_MOVED;
7673         return status;
7674 }
7675
7676 #endif  /* CONFIG_NFS_V4_1 */
7677
7678 /**
7679  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7680  * @inode: inode on FSID to check
7681  * @cred: credential to use for this operation
7682  *
7683  * Server indicates whether the FSID is present, moved, or not
7684  * recognized.  This operation is necessary to clear a LEASE_MOVED
7685  * condition for this client ID.
7686  *
7687  * Returns NFS4_OK if the FSID is present on this server,
7688  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7689  *  NFS4ERR code if some error occurred on the server, or a
7690  *  negative errno if a local failure occurred.
7691  */
7692 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
7693 {
7694         struct nfs_server *server = NFS_SERVER(inode);
7695         struct nfs_client *clp = server->nfs_client;
7696         const struct nfs4_mig_recovery_ops *ops =
7697                                         clp->cl_mvops->mig_recovery_ops;
7698         struct nfs4_exception exception = {
7699                 .interruptible = true,
7700         };
7701         int status;
7702
7703         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7704                 (unsigned long long)server->fsid.major,
7705                 (unsigned long long)server->fsid.minor,
7706                 clp->cl_hostname);
7707         nfs_display_fhandle(NFS_FH(inode), __func__);
7708
7709         do {
7710                 status = ops->fsid_present(inode, cred);
7711                 if (status != -NFS4ERR_DELAY)
7712                         break;
7713                 nfs4_handle_exception(server, status, &exception);
7714         } while (exception.retry);
7715         return status;
7716 }
7717
7718 /*
7719  * If 'use_integrity' is true and the state managment nfs_client
7720  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7721  * and the machine credential as per RFC3530bis and RFC5661 Security
7722  * Considerations sections. Otherwise, just use the user cred with the
7723  * filesystem's rpc_client.
7724  */
7725 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7726 {
7727         int status;
7728         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7729         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
7730         struct nfs4_secinfo_arg args = {
7731                 .dir_fh = NFS_FH(dir),
7732                 .name   = name,
7733         };
7734         struct nfs4_secinfo_res res = {
7735                 .flavors     = flavors,
7736         };
7737         struct rpc_message msg = {
7738                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7739                 .rpc_argp = &args,
7740                 .rpc_resp = &res,
7741         };
7742         struct nfs4_call_sync_data data = {
7743                 .seq_server = NFS_SERVER(dir),
7744                 .seq_args = &args.seq_args,
7745                 .seq_res = &res.seq_res,
7746         };
7747         struct rpc_task_setup task_setup = {
7748                 .rpc_client = clnt,
7749                 .rpc_message = &msg,
7750                 .callback_ops = clp->cl_mvops->call_sync_ops,
7751                 .callback_data = &data,
7752                 .flags = RPC_TASK_NO_ROUND_ROBIN,
7753         };
7754         const struct cred *cred = NULL;
7755
7756         if (use_integrity) {
7757                 clnt = clp->cl_rpcclient;
7758                 task_setup.rpc_client = clnt;
7759
7760                 cred = nfs4_get_clid_cred(clp);
7761                 msg.rpc_cred = cred;
7762         }
7763
7764         dprintk("NFS call  secinfo %s\n", name->name);
7765
7766         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7767         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
7768         status = nfs4_call_sync_custom(&task_setup);
7769
7770         dprintk("NFS reply  secinfo: %d\n", status);
7771
7772         put_cred(cred);
7773         return status;
7774 }
7775
7776 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7777                       struct nfs4_secinfo_flavors *flavors)
7778 {
7779         struct nfs4_exception exception = {
7780                 .interruptible = true,
7781         };
7782         int err;
7783         do {
7784                 err = -NFS4ERR_WRONGSEC;
7785
7786                 /* try to use integrity protection with machine cred */
7787                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7788                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7789
7790                 /*
7791                  * if unable to use integrity protection, or SECINFO with
7792                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7793                  * disallowed by spec, but exists in deployed servers) use
7794                  * the current filesystem's rpc_client and the user cred.
7795                  */
7796                 if (err == -NFS4ERR_WRONGSEC)
7797                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7798
7799                 trace_nfs4_secinfo(dir, name, err);
7800                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7801                                 &exception);
7802         } while (exception.retry);
7803         return err;
7804 }
7805
7806 #ifdef CONFIG_NFS_V4_1
7807 /*
7808  * Check the exchange flags returned by the server for invalid flags, having
7809  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7810  * DS flags set.
7811  */
7812 static int nfs4_check_cl_exchange_flags(u32 flags)
7813 {
7814         if (flags & ~EXCHGID4_FLAG_MASK_R)
7815                 goto out_inval;
7816         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7817             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7818                 goto out_inval;
7819         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7820                 goto out_inval;
7821         return NFS_OK;
7822 out_inval:
7823         return -NFS4ERR_INVAL;
7824 }
7825
7826 static bool
7827 nfs41_same_server_scope(struct nfs41_server_scope *a,
7828                         struct nfs41_server_scope *b)
7829 {
7830         if (a->server_scope_sz != b->server_scope_sz)
7831                 return false;
7832         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7833 }
7834
7835 static void
7836 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7837 {
7838         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
7839         struct nfs_client *clp = args->client;
7840
7841         switch (task->tk_status) {
7842         case -NFS4ERR_BADSESSION:
7843         case -NFS4ERR_DEADSESSION:
7844                 nfs4_schedule_session_recovery(clp->cl_session,
7845                                 task->tk_status);
7846         }
7847 }
7848
7849 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7850         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7851 };
7852
7853 /*
7854  * nfs4_proc_bind_one_conn_to_session()
7855  *
7856  * The 4.1 client currently uses the same TCP connection for the
7857  * fore and backchannel.
7858  */
7859 static
7860 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7861                 struct rpc_xprt *xprt,
7862                 struct nfs_client *clp,
7863                 const struct cred *cred)
7864 {
7865         int status;
7866         struct nfs41_bind_conn_to_session_args args = {
7867                 .client = clp,
7868                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7869         };
7870         struct nfs41_bind_conn_to_session_res res;
7871         struct rpc_message msg = {
7872                 .rpc_proc =
7873                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7874                 .rpc_argp = &args,
7875                 .rpc_resp = &res,
7876                 .rpc_cred = cred,
7877         };
7878         struct rpc_task_setup task_setup_data = {
7879                 .rpc_client = clnt,
7880                 .rpc_xprt = xprt,
7881                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7882                 .rpc_message = &msg,
7883                 .flags = RPC_TASK_TIMEOUT,
7884         };
7885         struct rpc_task *task;
7886
7887         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7888         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7889                 args.dir = NFS4_CDFC4_FORE;
7890
7891         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7892         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7893                 args.dir = NFS4_CDFC4_FORE;
7894
7895         task = rpc_run_task(&task_setup_data);
7896         if (!IS_ERR(task)) {
7897                 status = task->tk_status;
7898                 rpc_put_task(task);
7899         } else
7900                 status = PTR_ERR(task);
7901         trace_nfs4_bind_conn_to_session(clp, status);
7902         if (status == 0) {
7903                 if (memcmp(res.sessionid.data,
7904                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7905                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7906                         return -EIO;
7907                 }
7908                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7909                         dprintk("NFS: %s: Unexpected direction from server\n",
7910                                 __func__);
7911                         return -EIO;
7912                 }
7913                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7914                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7915                                 __func__);
7916                         return -EIO;
7917                 }
7918         }
7919
7920         return status;
7921 }
7922
7923 struct rpc_bind_conn_calldata {
7924         struct nfs_client *clp;
7925         const struct cred *cred;
7926 };
7927
7928 static int
7929 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7930                 struct rpc_xprt *xprt,
7931                 void *calldata)
7932 {
7933         struct rpc_bind_conn_calldata *p = calldata;
7934
7935         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7936 }
7937
7938 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
7939 {
7940         struct rpc_bind_conn_calldata data = {
7941                 .clp = clp,
7942                 .cred = cred,
7943         };
7944         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7945                         nfs4_proc_bind_conn_to_session_callback, &data);
7946 }
7947
7948 /*
7949  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7950  * and operations we'd like to see to enable certain features in the allow map
7951  */
7952 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7953         .how = SP4_MACH_CRED,
7954         .enforce.u.words = {
7955                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7956                       1 << (OP_EXCHANGE_ID - 32) |
7957                       1 << (OP_CREATE_SESSION - 32) |
7958                       1 << (OP_DESTROY_SESSION - 32) |
7959                       1 << (OP_DESTROY_CLIENTID - 32)
7960         },
7961         .allow.u.words = {
7962                 [0] = 1 << (OP_CLOSE) |
7963                       1 << (OP_OPEN_DOWNGRADE) |
7964                       1 << (OP_LOCKU) |
7965                       1 << (OP_DELEGRETURN) |
7966                       1 << (OP_COMMIT),
7967                 [1] = 1 << (OP_SECINFO - 32) |
7968                       1 << (OP_SECINFO_NO_NAME - 32) |
7969                       1 << (OP_LAYOUTRETURN - 32) |
7970                       1 << (OP_TEST_STATEID - 32) |
7971                       1 << (OP_FREE_STATEID - 32) |
7972                       1 << (OP_WRITE - 32)
7973         }
7974 };
7975
7976 /*
7977  * Select the state protection mode for client `clp' given the server results
7978  * from exchange_id in `sp'.
7979  *
7980  * Returns 0 on success, negative errno otherwise.
7981  */
7982 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7983                                  struct nfs41_state_protection *sp)
7984 {
7985         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7986                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7987                       1 << (OP_EXCHANGE_ID - 32) |
7988                       1 << (OP_CREATE_SESSION - 32) |
7989                       1 << (OP_DESTROY_SESSION - 32) |
7990                       1 << (OP_DESTROY_CLIENTID - 32)
7991         };
7992         unsigned long flags = 0;
7993         unsigned int i;
7994         int ret = 0;
7995
7996         if (sp->how == SP4_MACH_CRED) {
7997                 /* Print state protect result */
7998                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7999                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8000                         if (test_bit(i, sp->enforce.u.longs))
8001                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8002                         if (test_bit(i, sp->allow.u.longs))
8003                                 dfprintk(MOUNT, "  allow op %d\n", i);
8004                 }
8005
8006                 /* make sure nothing is on enforce list that isn't supported */
8007                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8008                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8009                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8010                                 ret = -EINVAL;
8011                                 goto out;
8012                         }
8013                 }
8014
8015                 /*
8016                  * Minimal mode - state operations are allowed to use machine
8017                  * credential.  Note this already happens by default, so the
8018                  * client doesn't have to do anything more than the negotiation.
8019                  *
8020                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8021                  *       we're already using the machine cred for exchange_id
8022                  *       and will never use a different cred.
8023                  */
8024                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8025                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8026                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8027                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8028                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8029                         dfprintk(MOUNT, "  minimal mode enabled\n");
8030                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8031                 } else {
8032                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8033                         ret = -EINVAL;
8034                         goto out;
8035                 }
8036
8037                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8038                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8039                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8040                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8041                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8042                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8043                 }
8044
8045                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8046                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8047                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8048                 }
8049
8050                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8051                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8052                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8053                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8054                 }
8055
8056                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8057                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8058                         dfprintk(MOUNT, "  stateid mode enabled\n");
8059                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8060                 }
8061
8062                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8063                         dfprintk(MOUNT, "  write mode enabled\n");
8064                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8065                 }
8066
8067                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8068                         dfprintk(MOUNT, "  commit mode enabled\n");
8069                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8070                 }
8071         }
8072 out:
8073         clp->cl_sp4_flags = flags;
8074         return ret;
8075 }
8076
8077 struct nfs41_exchange_id_data {
8078         struct nfs41_exchange_id_res res;
8079         struct nfs41_exchange_id_args args;
8080 };
8081
8082 static void nfs4_exchange_id_release(void *data)
8083 {
8084         struct nfs41_exchange_id_data *cdata =
8085                                         (struct nfs41_exchange_id_data *)data;
8086
8087         nfs_put_client(cdata->args.client);
8088         kfree(cdata->res.impl_id);
8089         kfree(cdata->res.server_scope);
8090         kfree(cdata->res.server_owner);
8091         kfree(cdata);
8092 }
8093
8094 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8095         .rpc_release = nfs4_exchange_id_release,
8096 };
8097
8098 /*
8099  * _nfs4_proc_exchange_id()
8100  *
8101  * Wrapper for EXCHANGE_ID operation.
8102  */
8103 static struct rpc_task *
8104 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8105                         u32 sp4_how, struct rpc_xprt *xprt)
8106 {
8107         struct rpc_message msg = {
8108                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8109                 .rpc_cred = cred,
8110         };
8111         struct rpc_task_setup task_setup_data = {
8112                 .rpc_client = clp->cl_rpcclient,
8113                 .callback_ops = &nfs4_exchange_id_call_ops,
8114                 .rpc_message = &msg,
8115                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8116         };
8117         struct nfs41_exchange_id_data *calldata;
8118         int status;
8119
8120         if (!refcount_inc_not_zero(&clp->cl_count))
8121                 return ERR_PTR(-EIO);
8122
8123         status = -ENOMEM;
8124         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8125         if (!calldata)
8126                 goto out;
8127
8128         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8129
8130         status = nfs4_init_uniform_client_string(clp);
8131         if (status)
8132                 goto out_calldata;
8133
8134         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8135                                                 GFP_NOFS);
8136         status = -ENOMEM;
8137         if (unlikely(calldata->res.server_owner == NULL))
8138                 goto out_calldata;
8139
8140         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8141                                         GFP_NOFS);
8142         if (unlikely(calldata->res.server_scope == NULL))
8143                 goto out_server_owner;
8144
8145         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8146         if (unlikely(calldata->res.impl_id == NULL))
8147                 goto out_server_scope;
8148
8149         switch (sp4_how) {
8150         case SP4_NONE:
8151                 calldata->args.state_protect.how = SP4_NONE;
8152                 break;
8153
8154         case SP4_MACH_CRED:
8155                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8156                 break;
8157
8158         default:
8159                 /* unsupported! */
8160                 WARN_ON_ONCE(1);
8161                 status = -EINVAL;
8162                 goto out_impl_id;
8163         }
8164         if (xprt) {
8165                 task_setup_data.rpc_xprt = xprt;
8166                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8167                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8168                                 sizeof(calldata->args.verifier.data));
8169         }
8170         calldata->args.client = clp;
8171         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8172         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8173 #ifdef CONFIG_NFS_V4_1_MIGRATION
8174         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8175 #endif
8176         msg.rpc_argp = &calldata->args;
8177         msg.rpc_resp = &calldata->res;
8178         task_setup_data.callback_data = calldata;
8179
8180         return rpc_run_task(&task_setup_data);
8181
8182 out_impl_id:
8183         kfree(calldata->res.impl_id);
8184 out_server_scope:
8185         kfree(calldata->res.server_scope);
8186 out_server_owner:
8187         kfree(calldata->res.server_owner);
8188 out_calldata:
8189         kfree(calldata);
8190 out:
8191         nfs_put_client(clp);
8192         return ERR_PTR(status);
8193 }
8194
8195 /*
8196  * _nfs4_proc_exchange_id()
8197  *
8198  * Wrapper for EXCHANGE_ID operation.
8199  */
8200 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8201                         u32 sp4_how)
8202 {
8203         struct rpc_task *task;
8204         struct nfs41_exchange_id_args *argp;
8205         struct nfs41_exchange_id_res *resp;
8206         int status;
8207
8208         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8209         if (IS_ERR(task))
8210                 return PTR_ERR(task);
8211
8212         argp = task->tk_msg.rpc_argp;
8213         resp = task->tk_msg.rpc_resp;
8214         status = task->tk_status;
8215         if (status  != 0)
8216                 goto out;
8217
8218         status = nfs4_check_cl_exchange_flags(resp->flags);
8219         if (status  != 0)
8220                 goto out;
8221
8222         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8223         if (status != 0)
8224                 goto out;
8225
8226         clp->cl_clientid = resp->clientid;
8227         clp->cl_exchange_flags = resp->flags;
8228         clp->cl_seqid = resp->seqid;
8229         /* Client ID is not confirmed */
8230         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8231                 clear_bit(NFS4_SESSION_ESTABLISHED,
8232                           &clp->cl_session->session_state);
8233
8234         if (clp->cl_serverscope != NULL &&
8235             !nfs41_same_server_scope(clp->cl_serverscope,
8236                                 resp->server_scope)) {
8237                 dprintk("%s: server_scope mismatch detected\n",
8238                         __func__);
8239                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8240         }
8241
8242         swap(clp->cl_serverowner, resp->server_owner);
8243         swap(clp->cl_serverscope, resp->server_scope);
8244         swap(clp->cl_implid, resp->impl_id);
8245
8246         /* Save the EXCHANGE_ID verifier session trunk tests */
8247         memcpy(clp->cl_confirm.data, argp->verifier.data,
8248                sizeof(clp->cl_confirm.data));
8249 out:
8250         trace_nfs4_exchange_id(clp, status);
8251         rpc_put_task(task);
8252         return status;
8253 }
8254
8255 /*
8256  * nfs4_proc_exchange_id()
8257  *
8258  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8259  *
8260  * Since the clientid has expired, all compounds using sessions
8261  * associated with the stale clientid will be returning
8262  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8263  * be in some phase of session reset.
8264  *
8265  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8266  */
8267 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8268 {
8269         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8270         int status;
8271
8272         /* try SP4_MACH_CRED if krb5i/p */
8273         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8274             authflavor == RPC_AUTH_GSS_KRB5P) {
8275                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8276                 if (!status)
8277                         return 0;
8278         }
8279
8280         /* try SP4_NONE */
8281         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8282 }
8283
8284 /**
8285  * nfs4_test_session_trunk
8286  *
8287  * This is an add_xprt_test() test function called from
8288  * rpc_clnt_setup_test_and_add_xprt.
8289  *
8290  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8291  * and is dereferrenced in nfs4_exchange_id_release
8292  *
8293  * Upon success, add the new transport to the rpc_clnt
8294  *
8295  * @clnt: struct rpc_clnt to get new transport
8296  * @xprt: the rpc_xprt to test
8297  * @data: call data for _nfs4_proc_exchange_id.
8298  */
8299 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8300                             void *data)
8301 {
8302         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8303         struct rpc_task *task;
8304         int status;
8305
8306         u32 sp4_how;
8307
8308         dprintk("--> %s try %s\n", __func__,
8309                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8310
8311         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8312
8313         /* Test connection for session trunking. Async exchange_id call */
8314         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8315         if (IS_ERR(task))
8316                 return;
8317
8318         status = task->tk_status;
8319         if (status == 0)
8320                 status = nfs4_detect_session_trunking(adata->clp,
8321                                 task->tk_msg.rpc_resp, xprt);
8322
8323         if (status == 0)
8324                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8325
8326         rpc_put_task(task);
8327 }
8328 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8329
8330 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8331                 const struct cred *cred)
8332 {
8333         struct rpc_message msg = {
8334                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8335                 .rpc_argp = clp,
8336                 .rpc_cred = cred,
8337         };
8338         int status;
8339
8340         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8341                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8342         trace_nfs4_destroy_clientid(clp, status);
8343         if (status)
8344                 dprintk("NFS: Got error %d from the server %s on "
8345                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8346         return status;
8347 }
8348
8349 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8350                 const struct cred *cred)
8351 {
8352         unsigned int loop;
8353         int ret;
8354
8355         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8356                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8357                 switch (ret) {
8358                 case -NFS4ERR_DELAY:
8359                 case -NFS4ERR_CLIENTID_BUSY:
8360                         ssleep(1);
8361                         break;
8362                 default:
8363                         return ret;
8364                 }
8365         }
8366         return 0;
8367 }
8368
8369 int nfs4_destroy_clientid(struct nfs_client *clp)
8370 {
8371         const struct cred *cred;
8372         int ret = 0;
8373
8374         if (clp->cl_mvops->minor_version < 1)
8375                 goto out;
8376         if (clp->cl_exchange_flags == 0)
8377                 goto out;
8378         if (clp->cl_preserve_clid)
8379                 goto out;
8380         cred = nfs4_get_clid_cred(clp);
8381         ret = nfs4_proc_destroy_clientid(clp, cred);
8382         put_cred(cred);
8383         switch (ret) {
8384         case 0:
8385         case -NFS4ERR_STALE_CLIENTID:
8386                 clp->cl_exchange_flags = 0;
8387         }
8388 out:
8389         return ret;
8390 }
8391
8392 #endif /* CONFIG_NFS_V4_1 */
8393
8394 struct nfs4_get_lease_time_data {
8395         struct nfs4_get_lease_time_args *args;
8396         struct nfs4_get_lease_time_res *res;
8397         struct nfs_client *clp;
8398 };
8399
8400 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8401                                         void *calldata)
8402 {
8403         struct nfs4_get_lease_time_data *data =
8404                         (struct nfs4_get_lease_time_data *)calldata;
8405
8406         dprintk("--> %s\n", __func__);
8407         /* just setup sequence, do not trigger session recovery
8408            since we're invoked within one */
8409         nfs4_setup_sequence(data->clp,
8410                         &data->args->la_seq_args,
8411                         &data->res->lr_seq_res,
8412                         task);
8413         dprintk("<-- %s\n", __func__);
8414 }
8415
8416 /*
8417  * Called from nfs4_state_manager thread for session setup, so don't recover
8418  * from sequence operation or clientid errors.
8419  */
8420 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8421 {
8422         struct nfs4_get_lease_time_data *data =
8423                         (struct nfs4_get_lease_time_data *)calldata;
8424
8425         dprintk("--> %s\n", __func__);
8426         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8427                 return;
8428         switch (task->tk_status) {
8429         case -NFS4ERR_DELAY:
8430         case -NFS4ERR_GRACE:
8431                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8432                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8433                 task->tk_status = 0;
8434                 /* fall through */
8435         case -NFS4ERR_RETRY_UNCACHED_REP:
8436                 rpc_restart_call_prepare(task);
8437                 return;
8438         }
8439         dprintk("<-- %s\n", __func__);
8440 }
8441
8442 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8443         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8444         .rpc_call_done = nfs4_get_lease_time_done,
8445 };
8446
8447 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8448 {
8449         struct nfs4_get_lease_time_args args;
8450         struct nfs4_get_lease_time_res res = {
8451                 .lr_fsinfo = fsinfo,
8452         };
8453         struct nfs4_get_lease_time_data data = {
8454                 .args = &args,
8455                 .res = &res,
8456                 .clp = clp,
8457         };
8458         struct rpc_message msg = {
8459                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8460                 .rpc_argp = &args,
8461                 .rpc_resp = &res,
8462         };
8463         struct rpc_task_setup task_setup = {
8464                 .rpc_client = clp->cl_rpcclient,
8465                 .rpc_message = &msg,
8466                 .callback_ops = &nfs4_get_lease_time_ops,
8467                 .callback_data = &data,
8468                 .flags = RPC_TASK_TIMEOUT,
8469         };
8470
8471         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8472         return nfs4_call_sync_custom(&task_setup);
8473 }
8474
8475 #ifdef CONFIG_NFS_V4_1
8476
8477 /*
8478  * Initialize the values to be used by the client in CREATE_SESSION
8479  * If nfs4_init_session set the fore channel request and response sizes,
8480  * use them.
8481  *
8482  * Set the back channel max_resp_sz_cached to zero to force the client to
8483  * always set csa_cachethis to FALSE because the current implementation
8484  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8485  */
8486 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8487                                     struct rpc_clnt *clnt)
8488 {
8489         unsigned int max_rqst_sz, max_resp_sz;
8490         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8491         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8492
8493         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8494         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8495
8496         /* Fore channel attributes */
8497         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8498         args->fc_attrs.max_resp_sz = max_resp_sz;
8499         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8500         args->fc_attrs.max_reqs = max_session_slots;
8501
8502         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8503                 "max_ops=%u max_reqs=%u\n",
8504                 __func__,
8505                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8506                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8507
8508         /* Back channel attributes */
8509         args->bc_attrs.max_rqst_sz = max_bc_payload;
8510         args->bc_attrs.max_resp_sz = max_bc_payload;
8511         args->bc_attrs.max_resp_sz_cached = 0;
8512         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8513         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8514         if (args->bc_attrs.max_reqs > max_bc_slots)
8515                 args->bc_attrs.max_reqs = max_bc_slots;
8516
8517         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8518                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8519                 __func__,
8520                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8521                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8522                 args->bc_attrs.max_reqs);
8523 }
8524
8525 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8526                 struct nfs41_create_session_res *res)
8527 {
8528         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8529         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8530
8531         if (rcvd->max_resp_sz > sent->max_resp_sz)
8532                 return -EINVAL;
8533         /*
8534          * Our requested max_ops is the minimum we need; we're not
8535          * prepared to break up compounds into smaller pieces than that.
8536          * So, no point even trying to continue if the server won't
8537          * cooperate:
8538          */
8539         if (rcvd->max_ops < sent->max_ops)
8540                 return -EINVAL;
8541         if (rcvd->max_reqs == 0)
8542                 return -EINVAL;
8543         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8544                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8545         return 0;
8546 }
8547
8548 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8549                 struct nfs41_create_session_res *res)
8550 {
8551         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8552         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8553
8554         if (!(res->flags & SESSION4_BACK_CHAN))
8555                 goto out;
8556         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8557                 return -EINVAL;
8558         if (rcvd->max_resp_sz < sent->max_resp_sz)
8559                 return -EINVAL;
8560         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8561                 return -EINVAL;
8562         if (rcvd->max_ops > sent->max_ops)
8563                 return -EINVAL;
8564         if (rcvd->max_reqs > sent->max_reqs)
8565                 return -EINVAL;
8566 out:
8567         return 0;
8568 }
8569
8570 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8571                                      struct nfs41_create_session_res *res)
8572 {
8573         int ret;
8574
8575         ret = nfs4_verify_fore_channel_attrs(args, res);
8576         if (ret)
8577                 return ret;
8578         return nfs4_verify_back_channel_attrs(args, res);
8579 }
8580
8581 static void nfs4_update_session(struct nfs4_session *session,
8582                 struct nfs41_create_session_res *res)
8583 {
8584         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8585         /* Mark client id and session as being confirmed */
8586         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8587         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8588         session->flags = res->flags;
8589         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8590         if (res->flags & SESSION4_BACK_CHAN)
8591                 memcpy(&session->bc_attrs, &res->bc_attrs,
8592                                 sizeof(session->bc_attrs));
8593 }
8594
8595 static int _nfs4_proc_create_session(struct nfs_client *clp,
8596                 const struct cred *cred)
8597 {
8598         struct nfs4_session *session = clp->cl_session;
8599         struct nfs41_create_session_args args = {
8600                 .client = clp,
8601                 .clientid = clp->cl_clientid,
8602                 .seqid = clp->cl_seqid,
8603                 .cb_program = NFS4_CALLBACK,
8604         };
8605         struct nfs41_create_session_res res;
8606
8607         struct rpc_message msg = {
8608                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8609                 .rpc_argp = &args,
8610                 .rpc_resp = &res,
8611                 .rpc_cred = cred,
8612         };
8613         int status;
8614
8615         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8616         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8617
8618         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8619                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8620         trace_nfs4_create_session(clp, status);
8621
8622         switch (status) {
8623         case -NFS4ERR_STALE_CLIENTID:
8624         case -NFS4ERR_DELAY:
8625         case -ETIMEDOUT:
8626         case -EACCES:
8627         case -EAGAIN:
8628                 goto out;
8629         };
8630
8631         clp->cl_seqid++;
8632         if (!status) {
8633                 /* Verify the session's negotiated channel_attrs values */
8634                 status = nfs4_verify_channel_attrs(&args, &res);
8635                 /* Increment the clientid slot sequence id */
8636                 if (status)
8637                         goto out;
8638                 nfs4_update_session(session, &res);
8639         }
8640 out:
8641         return status;
8642 }
8643
8644 /*
8645  * Issues a CREATE_SESSION operation to the server.
8646  * It is the responsibility of the caller to verify the session is
8647  * expired before calling this routine.
8648  */
8649 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8650 {
8651         int status;
8652         unsigned *ptr;
8653         struct nfs4_session *session = clp->cl_session;
8654
8655         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8656
8657         status = _nfs4_proc_create_session(clp, cred);
8658         if (status)
8659                 goto out;
8660
8661         /* Init or reset the session slot tables */
8662         status = nfs4_setup_session_slot_tables(session);
8663         dprintk("slot table setup returned %d\n", status);
8664         if (status)
8665                 goto out;
8666
8667         ptr = (unsigned *)&session->sess_id.data[0];
8668         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8669                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8670 out:
8671         dprintk("<-- %s\n", __func__);
8672         return status;
8673 }
8674
8675 /*
8676  * Issue the over-the-wire RPC DESTROY_SESSION.
8677  * The caller must serialize access to this routine.
8678  */
8679 int nfs4_proc_destroy_session(struct nfs4_session *session,
8680                 const struct cred *cred)
8681 {
8682         struct rpc_message msg = {
8683                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8684                 .rpc_argp = session,
8685                 .rpc_cred = cred,
8686         };
8687         int status = 0;
8688
8689         dprintk("--> nfs4_proc_destroy_session\n");
8690
8691         /* session is still being setup */
8692         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8693                 return 0;
8694
8695         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8696                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8697         trace_nfs4_destroy_session(session->clp, status);
8698
8699         if (status)
8700                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8701                         "Session has been destroyed regardless...\n", status);
8702
8703         dprintk("<-- nfs4_proc_destroy_session\n");
8704         return status;
8705 }
8706
8707 /*
8708  * Renew the cl_session lease.
8709  */
8710 struct nfs4_sequence_data {
8711         struct nfs_client *clp;
8712         struct nfs4_sequence_args args;
8713         struct nfs4_sequence_res res;
8714 };
8715
8716 static void nfs41_sequence_release(void *data)
8717 {
8718         struct nfs4_sequence_data *calldata = data;
8719         struct nfs_client *clp = calldata->clp;
8720
8721         if (refcount_read(&clp->cl_count) > 1)
8722                 nfs4_schedule_state_renewal(clp);
8723         nfs_put_client(clp);
8724         kfree(calldata);
8725 }
8726
8727 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8728 {
8729         switch(task->tk_status) {
8730         case -NFS4ERR_DELAY:
8731                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8732                 return -EAGAIN;
8733         default:
8734                 nfs4_schedule_lease_recovery(clp);
8735         }
8736         return 0;
8737 }
8738
8739 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8740 {
8741         struct nfs4_sequence_data *calldata = data;
8742         struct nfs_client *clp = calldata->clp;
8743
8744         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8745                 return;
8746
8747         trace_nfs4_sequence(clp, task->tk_status);
8748         if (task->tk_status < 0) {
8749                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8750                 if (refcount_read(&clp->cl_count) == 1)
8751                         goto out;
8752
8753                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8754                         rpc_restart_call_prepare(task);
8755                         return;
8756                 }
8757         }
8758         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8759 out:
8760         dprintk("<-- %s\n", __func__);
8761 }
8762
8763 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8764 {
8765         struct nfs4_sequence_data *calldata = data;
8766         struct nfs_client *clp = calldata->clp;
8767         struct nfs4_sequence_args *args;
8768         struct nfs4_sequence_res *res;
8769
8770         args = task->tk_msg.rpc_argp;
8771         res = task->tk_msg.rpc_resp;
8772
8773         nfs4_setup_sequence(clp, args, res, task);
8774 }
8775
8776 static const struct rpc_call_ops nfs41_sequence_ops = {
8777         .rpc_call_done = nfs41_sequence_call_done,
8778         .rpc_call_prepare = nfs41_sequence_prepare,
8779         .rpc_release = nfs41_sequence_release,
8780 };
8781
8782 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8783                 const struct cred *cred,
8784                 struct nfs4_slot *slot,
8785                 bool is_privileged)
8786 {
8787         struct nfs4_sequence_data *calldata;
8788         struct rpc_message msg = {
8789                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8790                 .rpc_cred = cred,
8791         };
8792         struct rpc_task_setup task_setup_data = {
8793                 .rpc_client = clp->cl_rpcclient,
8794                 .rpc_message = &msg,
8795                 .callback_ops = &nfs41_sequence_ops,
8796                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8797         };
8798         struct rpc_task *ret;
8799
8800         ret = ERR_PTR(-EIO);
8801         if (!refcount_inc_not_zero(&clp->cl_count))
8802                 goto out_err;
8803
8804         ret = ERR_PTR(-ENOMEM);
8805         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8806         if (calldata == NULL)
8807                 goto out_put_clp;
8808         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8809         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8810         msg.rpc_argp = &calldata->args;
8811         msg.rpc_resp = &calldata->res;
8812         calldata->clp = clp;
8813         task_setup_data.callback_data = calldata;
8814
8815         ret = rpc_run_task(&task_setup_data);
8816         if (IS_ERR(ret))
8817                 goto out_err;
8818         return ret;
8819 out_put_clp:
8820         nfs_put_client(clp);
8821 out_err:
8822         nfs41_release_slot(slot);
8823         return ret;
8824 }
8825
8826 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
8827 {
8828         struct rpc_task *task;
8829         int ret = 0;
8830
8831         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8832                 return -EAGAIN;
8833         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8834         if (IS_ERR(task))
8835                 ret = PTR_ERR(task);
8836         else
8837                 rpc_put_task_async(task);
8838         dprintk("<-- %s status=%d\n", __func__, ret);
8839         return ret;
8840 }
8841
8842 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
8843 {
8844         struct rpc_task *task;
8845         int ret;
8846
8847         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8848         if (IS_ERR(task)) {
8849                 ret = PTR_ERR(task);
8850                 goto out;
8851         }
8852         ret = rpc_wait_for_completion_task(task);
8853         if (!ret)
8854                 ret = task->tk_status;
8855         rpc_put_task(task);
8856 out:
8857         dprintk("<-- %s status=%d\n", __func__, ret);
8858         return ret;
8859 }
8860
8861 struct nfs4_reclaim_complete_data {
8862         struct nfs_client *clp;
8863         struct nfs41_reclaim_complete_args arg;
8864         struct nfs41_reclaim_complete_res res;
8865 };
8866
8867 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8868 {
8869         struct nfs4_reclaim_complete_data *calldata = data;
8870
8871         nfs4_setup_sequence(calldata->clp,
8872                         &calldata->arg.seq_args,
8873                         &calldata->res.seq_res,
8874                         task);
8875 }
8876
8877 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8878 {
8879         switch(task->tk_status) {
8880         case 0:
8881                 wake_up_all(&clp->cl_lock_waitq);
8882                 /* Fallthrough */
8883         case -NFS4ERR_COMPLETE_ALREADY:
8884         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8885                 break;
8886         case -NFS4ERR_DELAY:
8887                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8888                 /* fall through */
8889         case -NFS4ERR_RETRY_UNCACHED_REP:
8890                 return -EAGAIN;
8891         case -NFS4ERR_BADSESSION:
8892         case -NFS4ERR_DEADSESSION:
8893         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8894                 break;
8895         default:
8896                 nfs4_schedule_lease_recovery(clp);
8897         }
8898         return 0;
8899 }
8900
8901 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8902 {
8903         struct nfs4_reclaim_complete_data *calldata = data;
8904         struct nfs_client *clp = calldata->clp;
8905         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8906
8907         dprintk("--> %s\n", __func__);
8908         if (!nfs41_sequence_done(task, res))
8909                 return;
8910
8911         trace_nfs4_reclaim_complete(clp, task->tk_status);
8912         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8913                 rpc_restart_call_prepare(task);
8914                 return;
8915         }
8916         dprintk("<-- %s\n", __func__);
8917 }
8918
8919 static void nfs4_free_reclaim_complete_data(void *data)
8920 {
8921         struct nfs4_reclaim_complete_data *calldata = data;
8922
8923         kfree(calldata);
8924 }
8925
8926 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8927         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8928         .rpc_call_done = nfs4_reclaim_complete_done,
8929         .rpc_release = nfs4_free_reclaim_complete_data,
8930 };
8931
8932 /*
8933  * Issue a global reclaim complete.
8934  */
8935 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8936                 const struct cred *cred)
8937 {
8938         struct nfs4_reclaim_complete_data *calldata;
8939         struct rpc_message msg = {
8940                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8941                 .rpc_cred = cred,
8942         };
8943         struct rpc_task_setup task_setup_data = {
8944                 .rpc_client = clp->cl_rpcclient,
8945                 .rpc_message = &msg,
8946                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8947                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8948         };
8949         int status = -ENOMEM;
8950
8951         dprintk("--> %s\n", __func__);
8952         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8953         if (calldata == NULL)
8954                 goto out;
8955         calldata->clp = clp;
8956         calldata->arg.one_fs = 0;
8957
8958         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
8959         msg.rpc_argp = &calldata->arg;
8960         msg.rpc_resp = &calldata->res;
8961         task_setup_data.callback_data = calldata;
8962         status = nfs4_call_sync_custom(&task_setup_data);
8963 out:
8964         dprintk("<-- %s status=%d\n", __func__, status);
8965         return status;
8966 }
8967
8968 static void
8969 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8970 {
8971         struct nfs4_layoutget *lgp = calldata;
8972         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8973
8974         dprintk("--> %s\n", __func__);
8975         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8976                                 &lgp->res.seq_res, task);
8977         dprintk("<-- %s\n", __func__);
8978 }
8979
8980 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8981 {
8982         struct nfs4_layoutget *lgp = calldata;
8983
8984         dprintk("--> %s\n", __func__);
8985         nfs41_sequence_process(task, &lgp->res.seq_res);
8986         dprintk("<-- %s\n", __func__);
8987 }
8988
8989 static int
8990 nfs4_layoutget_handle_exception(struct rpc_task *task,
8991                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8992 {
8993         struct inode *inode = lgp->args.inode;
8994         struct nfs_server *server = NFS_SERVER(inode);
8995         struct pnfs_layout_hdr *lo;
8996         int nfs4err = task->tk_status;
8997         int err, status = 0;
8998         LIST_HEAD(head);
8999
9000         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9001
9002         nfs4_sequence_free_slot(&lgp->res.seq_res);
9003
9004         switch (nfs4err) {
9005         case 0:
9006                 goto out;
9007
9008         /*
9009          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9010          * on the file. set tk_status to -ENODATA to tell upper layer to
9011          * retry go inband.
9012          */
9013         case -NFS4ERR_LAYOUTUNAVAILABLE:
9014                 status = -ENODATA;
9015                 goto out;
9016         /*
9017          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9018          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9019          */
9020         case -NFS4ERR_BADLAYOUT:
9021                 status = -EOVERFLOW;
9022                 goto out;
9023         /*
9024          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9025          * (or clients) writing to the same RAID stripe except when
9026          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9027          *
9028          * Treat it like we would RECALLCONFLICT -- we retry for a little
9029          * while, and then eventually give up.
9030          */
9031         case -NFS4ERR_LAYOUTTRYLATER:
9032                 if (lgp->args.minlength == 0) {
9033                         status = -EOVERFLOW;
9034                         goto out;
9035                 }
9036                 status = -EBUSY;
9037                 break;
9038         case -NFS4ERR_RECALLCONFLICT:
9039                 status = -ERECALLCONFLICT;
9040                 break;
9041         case -NFS4ERR_DELEG_REVOKED:
9042         case -NFS4ERR_ADMIN_REVOKED:
9043         case -NFS4ERR_EXPIRED:
9044         case -NFS4ERR_BAD_STATEID:
9045                 exception->timeout = 0;
9046                 spin_lock(&inode->i_lock);
9047                 lo = NFS_I(inode)->layout;
9048                 /* If the open stateid was bad, then recover it. */
9049                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9050                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9051                         spin_unlock(&inode->i_lock);
9052                         exception->state = lgp->args.ctx->state;
9053                         exception->stateid = &lgp->args.stateid;
9054                         break;
9055                 }
9056
9057                 /*
9058                  * Mark the bad layout state as invalid, then retry
9059                  */
9060                 pnfs_mark_layout_stateid_invalid(lo, &head);
9061                 spin_unlock(&inode->i_lock);
9062                 nfs_commit_inode(inode, 0);
9063                 pnfs_free_lseg_list(&head);
9064                 status = -EAGAIN;
9065                 goto out;
9066         }
9067
9068         err = nfs4_handle_exception(server, nfs4err, exception);
9069         if (!status) {
9070                 if (exception->retry)
9071                         status = -EAGAIN;
9072                 else
9073                         status = err;
9074         }
9075 out:
9076         dprintk("<-- %s\n", __func__);
9077         return status;
9078 }
9079
9080 size_t max_response_pages(struct nfs_server *server)
9081 {
9082         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9083         return nfs_page_array_len(0, max_resp_sz);
9084 }
9085
9086 static void nfs4_layoutget_release(void *calldata)
9087 {
9088         struct nfs4_layoutget *lgp = calldata;
9089
9090         dprintk("--> %s\n", __func__);
9091         nfs4_sequence_free_slot(&lgp->res.seq_res);
9092         pnfs_layoutget_free(lgp);
9093         dprintk("<-- %s\n", __func__);
9094 }
9095
9096 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9097         .rpc_call_prepare = nfs4_layoutget_prepare,
9098         .rpc_call_done = nfs4_layoutget_done,
9099         .rpc_release = nfs4_layoutget_release,
9100 };
9101
9102 struct pnfs_layout_segment *
9103 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9104 {
9105         struct inode *inode = lgp->args.inode;
9106         struct nfs_server *server = NFS_SERVER(inode);
9107         struct rpc_task *task;
9108         struct rpc_message msg = {
9109                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9110                 .rpc_argp = &lgp->args,
9111                 .rpc_resp = &lgp->res,
9112                 .rpc_cred = lgp->cred,
9113         };
9114         struct rpc_task_setup task_setup_data = {
9115                 .rpc_client = server->client,
9116                 .rpc_message = &msg,
9117                 .callback_ops = &nfs4_layoutget_call_ops,
9118                 .callback_data = lgp,
9119                 .flags = RPC_TASK_ASYNC,
9120         };
9121         struct pnfs_layout_segment *lseg = NULL;
9122         struct nfs4_exception exception = {
9123                 .inode = inode,
9124                 .timeout = *timeout,
9125         };
9126         int status = 0;
9127
9128         dprintk("--> %s\n", __func__);
9129
9130         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9131         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9132
9133         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9134
9135         task = rpc_run_task(&task_setup_data);
9136         if (IS_ERR(task))
9137                 return ERR_CAST(task);
9138         status = rpc_wait_for_completion_task(task);
9139         if (status != 0)
9140                 goto out;
9141
9142         if (task->tk_status < 0) {
9143                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9144                 *timeout = exception.timeout;
9145         } else if (lgp->res.layoutp->len == 0) {
9146                 status = -EAGAIN;
9147                 *timeout = nfs4_update_delay(&exception.timeout);
9148         } else
9149                 lseg = pnfs_layout_process(lgp);
9150 out:
9151         trace_nfs4_layoutget(lgp->args.ctx,
9152                         &lgp->args.range,
9153                         &lgp->res.range,
9154                         &lgp->res.stateid,
9155                         status);
9156
9157         rpc_put_task(task);
9158         dprintk("<-- %s status=%d\n", __func__, status);
9159         if (status)
9160                 return ERR_PTR(status);
9161         return lseg;
9162 }
9163
9164 static void
9165 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9166 {
9167         struct nfs4_layoutreturn *lrp = calldata;
9168
9169         dprintk("--> %s\n", __func__);
9170         nfs4_setup_sequence(lrp->clp,
9171                         &lrp->args.seq_args,
9172                         &lrp->res.seq_res,
9173                         task);
9174         if (!pnfs_layout_is_valid(lrp->args.layout))
9175                 rpc_exit(task, 0);
9176 }
9177
9178 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9179 {
9180         struct nfs4_layoutreturn *lrp = calldata;
9181         struct nfs_server *server;
9182
9183         dprintk("--> %s\n", __func__);
9184
9185         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9186                 return;
9187
9188         /*
9189          * Was there an RPC level error? Assume the call succeeded,
9190          * and that we need to release the layout
9191          */
9192         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9193                 lrp->res.lrs_present = 0;
9194                 return;
9195         }
9196
9197         server = NFS_SERVER(lrp->args.inode);
9198         switch (task->tk_status) {
9199         case -NFS4ERR_OLD_STATEID:
9200                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9201                                         &lrp->args.range,
9202                                         lrp->args.inode))
9203                         goto out_restart;
9204                 /* Fallthrough */
9205         default:
9206                 task->tk_status = 0;
9207                 /* Fallthrough */
9208         case 0:
9209                 break;
9210         case -NFS4ERR_DELAY:
9211                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9212                         break;
9213                 goto out_restart;
9214         }
9215         dprintk("<-- %s\n", __func__);
9216         return;
9217 out_restart:
9218         task->tk_status = 0;
9219         nfs4_sequence_free_slot(&lrp->res.seq_res);
9220         rpc_restart_call_prepare(task);
9221 }
9222
9223 static void nfs4_layoutreturn_release(void *calldata)
9224 {
9225         struct nfs4_layoutreturn *lrp = calldata;
9226         struct pnfs_layout_hdr *lo = lrp->args.layout;
9227
9228         dprintk("--> %s\n", __func__);
9229         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9230                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9231         nfs4_sequence_free_slot(&lrp->res.seq_res);
9232         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9233                 lrp->ld_private.ops->free(&lrp->ld_private);
9234         pnfs_put_layout_hdr(lrp->args.layout);
9235         nfs_iput_and_deactive(lrp->inode);
9236         kfree(calldata);
9237         dprintk("<-- %s\n", __func__);
9238 }
9239
9240 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9241         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9242         .rpc_call_done = nfs4_layoutreturn_done,
9243         .rpc_release = nfs4_layoutreturn_release,
9244 };
9245
9246 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9247 {
9248         struct rpc_task *task;
9249         struct rpc_message msg = {
9250                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9251                 .rpc_argp = &lrp->args,
9252                 .rpc_resp = &lrp->res,
9253                 .rpc_cred = lrp->cred,
9254         };
9255         struct rpc_task_setup task_setup_data = {
9256                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9257                 .rpc_message = &msg,
9258                 .callback_ops = &nfs4_layoutreturn_call_ops,
9259                 .callback_data = lrp,
9260         };
9261         int status = 0;
9262
9263         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9264                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9265                         &task_setup_data.rpc_client, &msg);
9266
9267         dprintk("--> %s\n", __func__);
9268         if (!sync) {
9269                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9270                 if (!lrp->inode) {
9271                         nfs4_layoutreturn_release(lrp);
9272                         return -EAGAIN;
9273                 }
9274                 task_setup_data.flags |= RPC_TASK_ASYNC;
9275         }
9276         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9277         task = rpc_run_task(&task_setup_data);
9278         if (IS_ERR(task))
9279                 return PTR_ERR(task);
9280         if (sync)
9281                 status = task->tk_status;
9282         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9283         dprintk("<-- %s status=%d\n", __func__, status);
9284         rpc_put_task(task);
9285         return status;
9286 }
9287
9288 static int
9289 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9290                 struct pnfs_device *pdev,
9291                 const struct cred *cred)
9292 {
9293         struct nfs4_getdeviceinfo_args args = {
9294                 .pdev = pdev,
9295                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9296                         NOTIFY_DEVICEID4_DELETE,
9297         };
9298         struct nfs4_getdeviceinfo_res res = {
9299                 .pdev = pdev,
9300         };
9301         struct rpc_message msg = {
9302                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9303                 .rpc_argp = &args,
9304                 .rpc_resp = &res,
9305                 .rpc_cred = cred,
9306         };
9307         int status;
9308
9309         dprintk("--> %s\n", __func__);
9310         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9311         if (res.notification & ~args.notify_types)
9312                 dprintk("%s: unsupported notification\n", __func__);
9313         if (res.notification != args.notify_types)
9314                 pdev->nocache = 1;
9315
9316         dprintk("<-- %s status=%d\n", __func__, status);
9317
9318         return status;
9319 }
9320
9321 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9322                 struct pnfs_device *pdev,
9323                 const struct cred *cred)
9324 {
9325         struct nfs4_exception exception = { };
9326         int err;
9327
9328         do {
9329                 err = nfs4_handle_exception(server,
9330                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9331                                         &exception);
9332         } while (exception.retry);
9333         return err;
9334 }
9335 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9336
9337 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9338 {
9339         struct nfs4_layoutcommit_data *data = calldata;
9340         struct nfs_server *server = NFS_SERVER(data->args.inode);
9341
9342         nfs4_setup_sequence(server->nfs_client,
9343                         &data->args.seq_args,
9344                         &data->res.seq_res,
9345                         task);
9346 }
9347
9348 static void
9349 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9350 {
9351         struct nfs4_layoutcommit_data *data = calldata;
9352         struct nfs_server *server = NFS_SERVER(data->args.inode);
9353
9354         if (!nfs41_sequence_done(task, &data->res.seq_res))
9355                 return;
9356
9357         switch (task->tk_status) { /* Just ignore these failures */
9358         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9359         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9360         case -NFS4ERR_BADLAYOUT:     /* no layout */
9361         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9362                 task->tk_status = 0;
9363         case 0:
9364                 break;
9365         default:
9366                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9367                         rpc_restart_call_prepare(task);
9368                         return;
9369                 }
9370         }
9371 }
9372
9373 static void nfs4_layoutcommit_release(void *calldata)
9374 {
9375         struct nfs4_layoutcommit_data *data = calldata;
9376
9377         pnfs_cleanup_layoutcommit(data);
9378         nfs_post_op_update_inode_force_wcc(data->args.inode,
9379                                            data->res.fattr);
9380         put_cred(data->cred);
9381         nfs_iput_and_deactive(data->inode);
9382         kfree(data);
9383 }
9384
9385 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9386         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9387         .rpc_call_done = nfs4_layoutcommit_done,
9388         .rpc_release = nfs4_layoutcommit_release,
9389 };
9390
9391 int
9392 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9393 {
9394         struct rpc_message msg = {
9395                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9396                 .rpc_argp = &data->args,
9397                 .rpc_resp = &data->res,
9398                 .rpc_cred = data->cred,
9399         };
9400         struct rpc_task_setup task_setup_data = {
9401                 .task = &data->task,
9402                 .rpc_client = NFS_CLIENT(data->args.inode),
9403                 .rpc_message = &msg,
9404                 .callback_ops = &nfs4_layoutcommit_ops,
9405                 .callback_data = data,
9406         };
9407         struct rpc_task *task;
9408         int status = 0;
9409
9410         dprintk("NFS: initiating layoutcommit call. sync %d "
9411                 "lbw: %llu inode %lu\n", sync,
9412                 data->args.lastbytewritten,
9413                 data->args.inode->i_ino);
9414
9415         if (!sync) {
9416                 data->inode = nfs_igrab_and_active(data->args.inode);
9417                 if (data->inode == NULL) {
9418                         nfs4_layoutcommit_release(data);
9419                         return -EAGAIN;
9420                 }
9421                 task_setup_data.flags = RPC_TASK_ASYNC;
9422         }
9423         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9424         task = rpc_run_task(&task_setup_data);
9425         if (IS_ERR(task))
9426                 return PTR_ERR(task);
9427         if (sync)
9428                 status = task->tk_status;
9429         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9430         dprintk("%s: status %d\n", __func__, status);
9431         rpc_put_task(task);
9432         return status;
9433 }
9434
9435 /*
9436  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9437  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9438  */
9439 static int
9440 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9441                     struct nfs_fsinfo *info,
9442                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9443 {
9444         struct nfs41_secinfo_no_name_args args = {
9445                 .style = SECINFO_STYLE_CURRENT_FH,
9446         };
9447         struct nfs4_secinfo_res res = {
9448                 .flavors = flavors,
9449         };
9450         struct rpc_message msg = {
9451                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9452                 .rpc_argp = &args,
9453                 .rpc_resp = &res,
9454         };
9455         struct rpc_clnt *clnt = server->client;
9456         struct nfs4_call_sync_data data = {
9457                 .seq_server = server,
9458                 .seq_args = &args.seq_args,
9459                 .seq_res = &res.seq_res,
9460         };
9461         struct rpc_task_setup task_setup = {
9462                 .rpc_client = server->client,
9463                 .rpc_message = &msg,
9464                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9465                 .callback_data = &data,
9466                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9467         };
9468         const struct cred *cred = NULL;
9469         int status;
9470
9471         if (use_integrity) {
9472                 clnt = server->nfs_client->cl_rpcclient;
9473                 task_setup.rpc_client = clnt;
9474
9475                 cred = nfs4_get_clid_cred(server->nfs_client);
9476                 msg.rpc_cred = cred;
9477         }
9478
9479         dprintk("--> %s\n", __func__);
9480         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9481         status = nfs4_call_sync_custom(&task_setup);
9482         dprintk("<-- %s status=%d\n", __func__, status);
9483
9484         put_cred(cred);
9485
9486         return status;
9487 }
9488
9489 static int
9490 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9491                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9492 {
9493         struct nfs4_exception exception = {
9494                 .interruptible = true,
9495         };
9496         int err;
9497         do {
9498                 /* first try using integrity protection */
9499                 err = -NFS4ERR_WRONGSEC;
9500
9501                 /* try to use integrity protection with machine cred */
9502                 if (_nfs4_is_integrity_protected(server->nfs_client))
9503                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9504                                                           flavors, true);
9505
9506                 /*
9507                  * if unable to use integrity protection, or SECINFO with
9508                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9509                  * disallowed by spec, but exists in deployed servers) use
9510                  * the current filesystem's rpc_client and the user cred.
9511                  */
9512                 if (err == -NFS4ERR_WRONGSEC)
9513                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9514                                                           flavors, false);
9515
9516                 switch (err) {
9517                 case 0:
9518                 case -NFS4ERR_WRONGSEC:
9519                 case -ENOTSUPP:
9520                         goto out;
9521                 default:
9522                         err = nfs4_handle_exception(server, err, &exception);
9523                 }
9524         } while (exception.retry);
9525 out:
9526         return err;
9527 }
9528
9529 static int
9530 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9531                     struct nfs_fsinfo *info)
9532 {
9533         int err;
9534         struct page *page;
9535         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9536         struct nfs4_secinfo_flavors *flavors;
9537         struct nfs4_secinfo4 *secinfo;
9538         int i;
9539
9540         page = alloc_page(GFP_KERNEL);
9541         if (!page) {
9542                 err = -ENOMEM;
9543                 goto out;
9544         }
9545
9546         flavors = page_address(page);
9547         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9548
9549         /*
9550          * Fall back on "guess and check" method if
9551          * the server doesn't support SECINFO_NO_NAME
9552          */
9553         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9554                 err = nfs4_find_root_sec(server, fhandle, info);
9555                 goto out_freepage;
9556         }
9557         if (err)
9558                 goto out_freepage;
9559
9560         for (i = 0; i < flavors->num_flavors; i++) {
9561                 secinfo = &flavors->flavors[i];
9562
9563                 switch (secinfo->flavor) {
9564                 case RPC_AUTH_NULL:
9565                 case RPC_AUTH_UNIX:
9566                 case RPC_AUTH_GSS:
9567                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9568                                         &secinfo->flavor_info);
9569                         break;
9570                 default:
9571                         flavor = RPC_AUTH_MAXFLAVOR;
9572                         break;
9573                 }
9574
9575                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9576                         flavor = RPC_AUTH_MAXFLAVOR;
9577
9578                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9579                         err = nfs4_lookup_root_sec(server, fhandle,
9580                                                    info, flavor);
9581                         if (!err)
9582                                 break;
9583                 }
9584         }
9585
9586         if (flavor == RPC_AUTH_MAXFLAVOR)
9587                 err = -EPERM;
9588
9589 out_freepage:
9590         put_page(page);
9591         if (err == -EACCES)
9592                 return -EPERM;
9593 out:
9594         return err;
9595 }
9596
9597 static int _nfs41_test_stateid(struct nfs_server *server,
9598                 nfs4_stateid *stateid,
9599                 const struct cred *cred)
9600 {
9601         int status;
9602         struct nfs41_test_stateid_args args = {
9603                 .stateid = stateid,
9604         };
9605         struct nfs41_test_stateid_res res;
9606         struct rpc_message msg = {
9607                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9608                 .rpc_argp = &args,
9609                 .rpc_resp = &res,
9610                 .rpc_cred = cred,
9611         };
9612         struct rpc_clnt *rpc_client = server->client;
9613
9614         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9615                 &rpc_client, &msg);
9616
9617         dprintk("NFS call  test_stateid %p\n", stateid);
9618         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9619         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9620                         &args.seq_args, &res.seq_res);
9621         if (status != NFS_OK) {
9622                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9623                 return status;
9624         }
9625         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9626         return -res.status;
9627 }
9628
9629 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9630                 int err, struct nfs4_exception *exception)
9631 {
9632         exception->retry = 0;
9633         switch(err) {
9634         case -NFS4ERR_DELAY:
9635         case -NFS4ERR_RETRY_UNCACHED_REP:
9636                 nfs4_handle_exception(server, err, exception);
9637                 break;
9638         case -NFS4ERR_BADSESSION:
9639         case -NFS4ERR_BADSLOT:
9640         case -NFS4ERR_BAD_HIGH_SLOT:
9641         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9642         case -NFS4ERR_DEADSESSION:
9643                 nfs4_do_handle_exception(server, err, exception);
9644         }
9645 }
9646
9647 /**
9648  * nfs41_test_stateid - perform a TEST_STATEID operation
9649  *
9650  * @server: server / transport on which to perform the operation
9651  * @stateid: state ID to test
9652  * @cred: credential
9653  *
9654  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9655  * Otherwise a negative NFS4ERR value is returned if the operation
9656  * failed or the state ID is not currently valid.
9657  */
9658 static int nfs41_test_stateid(struct nfs_server *server,
9659                 nfs4_stateid *stateid,
9660                 const struct cred *cred)
9661 {
9662         struct nfs4_exception exception = {
9663                 .interruptible = true,
9664         };
9665         int err;
9666         do {
9667                 err = _nfs41_test_stateid(server, stateid, cred);
9668                 nfs4_handle_delay_or_session_error(server, err, &exception);
9669         } while (exception.retry);
9670         return err;
9671 }
9672
9673 struct nfs_free_stateid_data {
9674         struct nfs_server *server;
9675         struct nfs41_free_stateid_args args;
9676         struct nfs41_free_stateid_res res;
9677 };
9678
9679 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9680 {
9681         struct nfs_free_stateid_data *data = calldata;
9682         nfs4_setup_sequence(data->server->nfs_client,
9683                         &data->args.seq_args,
9684                         &data->res.seq_res,
9685                         task);
9686 }
9687
9688 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9689 {
9690         struct nfs_free_stateid_data *data = calldata;
9691
9692         nfs41_sequence_done(task, &data->res.seq_res);
9693
9694         switch (task->tk_status) {
9695         case -NFS4ERR_DELAY:
9696                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9697                         rpc_restart_call_prepare(task);
9698         }
9699 }
9700
9701 static void nfs41_free_stateid_release(void *calldata)
9702 {
9703         kfree(calldata);
9704 }
9705
9706 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9707         .rpc_call_prepare = nfs41_free_stateid_prepare,
9708         .rpc_call_done = nfs41_free_stateid_done,
9709         .rpc_release = nfs41_free_stateid_release,
9710 };
9711
9712 /**
9713  * nfs41_free_stateid - perform a FREE_STATEID operation
9714  *
9715  * @server: server / transport on which to perform the operation
9716  * @stateid: state ID to release
9717  * @cred: credential
9718  * @privileged: set to true if this call needs to be privileged
9719  *
9720  * Note: this function is always asynchronous.
9721  */
9722 static int nfs41_free_stateid(struct nfs_server *server,
9723                 const nfs4_stateid *stateid,
9724                 const struct cred *cred,
9725                 bool privileged)
9726 {
9727         struct rpc_message msg = {
9728                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9729                 .rpc_cred = cred,
9730         };
9731         struct rpc_task_setup task_setup = {
9732                 .rpc_client = server->client,
9733                 .rpc_message = &msg,
9734                 .callback_ops = &nfs41_free_stateid_ops,
9735                 .flags = RPC_TASK_ASYNC,
9736         };
9737         struct nfs_free_stateid_data *data;
9738         struct rpc_task *task;
9739
9740         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9741                 &task_setup.rpc_client, &msg);
9742
9743         dprintk("NFS call  free_stateid %p\n", stateid);
9744         data = kmalloc(sizeof(*data), GFP_NOFS);
9745         if (!data)
9746                 return -ENOMEM;
9747         data->server = server;
9748         nfs4_stateid_copy(&data->args.stateid, stateid);
9749
9750         task_setup.callback_data = data;
9751
9752         msg.rpc_argp = &data->args;
9753         msg.rpc_resp = &data->res;
9754         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9755         task = rpc_run_task(&task_setup);
9756         if (IS_ERR(task))
9757                 return PTR_ERR(task);
9758         rpc_put_task(task);
9759         return 0;
9760 }
9761
9762 static void
9763 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9764 {
9765         const struct cred *cred = lsp->ls_state->owner->so_cred;
9766
9767         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9768         nfs4_free_lock_state(server, lsp);
9769 }
9770
9771 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9772                 const nfs4_stateid *s2)
9773 {
9774         if (s1->type != s2->type)
9775                 return false;
9776
9777         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9778                 return false;
9779
9780         if (s1->seqid == s2->seqid)
9781                 return true;
9782
9783         return s1->seqid == 0 || s2->seqid == 0;
9784 }
9785
9786 #endif /* CONFIG_NFS_V4_1 */
9787
9788 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9789                 const nfs4_stateid *s2)
9790 {
9791         return nfs4_stateid_match(s1, s2);
9792 }
9793
9794
9795 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9796         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9797         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9798         .recover_open   = nfs4_open_reclaim,
9799         .recover_lock   = nfs4_lock_reclaim,
9800         .establish_clid = nfs4_init_clientid,
9801         .detect_trunking = nfs40_discover_server_trunking,
9802 };
9803
9804 #if defined(CONFIG_NFS_V4_1)
9805 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9806         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9807         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9808         .recover_open   = nfs4_open_reclaim,
9809         .recover_lock   = nfs4_lock_reclaim,
9810         .establish_clid = nfs41_init_clientid,
9811         .reclaim_complete = nfs41_proc_reclaim_complete,
9812         .detect_trunking = nfs41_discover_server_trunking,
9813 };
9814 #endif /* CONFIG_NFS_V4_1 */
9815
9816 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9817         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9818         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9819         .recover_open   = nfs40_open_expired,
9820         .recover_lock   = nfs4_lock_expired,
9821         .establish_clid = nfs4_init_clientid,
9822 };
9823
9824 #if defined(CONFIG_NFS_V4_1)
9825 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9826         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9827         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9828         .recover_open   = nfs41_open_expired,
9829         .recover_lock   = nfs41_lock_expired,
9830         .establish_clid = nfs41_init_clientid,
9831 };
9832 #endif /* CONFIG_NFS_V4_1 */
9833
9834 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9835         .sched_state_renewal = nfs4_proc_async_renew,
9836         .get_state_renewal_cred = nfs4_get_renew_cred,
9837         .renew_lease = nfs4_proc_renew,
9838 };
9839
9840 #if defined(CONFIG_NFS_V4_1)
9841 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9842         .sched_state_renewal = nfs41_proc_async_sequence,
9843         .get_state_renewal_cred = nfs4_get_machine_cred,
9844         .renew_lease = nfs4_proc_sequence,
9845 };
9846 #endif
9847
9848 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9849         .get_locations = _nfs40_proc_get_locations,
9850         .fsid_present = _nfs40_proc_fsid_present,
9851 };
9852
9853 #if defined(CONFIG_NFS_V4_1)
9854 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9855         .get_locations = _nfs41_proc_get_locations,
9856         .fsid_present = _nfs41_proc_fsid_present,
9857 };
9858 #endif  /* CONFIG_NFS_V4_1 */
9859
9860 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9861         .minor_version = 0,
9862         .init_caps = NFS_CAP_READDIRPLUS
9863                 | NFS_CAP_ATOMIC_OPEN
9864                 | NFS_CAP_POSIX_LOCK,
9865         .init_client = nfs40_init_client,
9866         .shutdown_client = nfs40_shutdown_client,
9867         .match_stateid = nfs4_match_stateid,
9868         .find_root_sec = nfs4_find_root_sec,
9869         .free_lock_state = nfs4_release_lockowner,
9870         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9871         .alloc_seqid = nfs_alloc_seqid,
9872         .call_sync_ops = &nfs40_call_sync_ops,
9873         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9874         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9875         .state_renewal_ops = &nfs40_state_renewal_ops,
9876         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9877 };
9878
9879 #if defined(CONFIG_NFS_V4_1)
9880 static struct nfs_seqid *
9881 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9882 {
9883         return NULL;
9884 }
9885
9886 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9887         .minor_version = 1,
9888         .init_caps = NFS_CAP_READDIRPLUS
9889                 | NFS_CAP_ATOMIC_OPEN
9890                 | NFS_CAP_POSIX_LOCK
9891                 | NFS_CAP_STATEID_NFSV41
9892                 | NFS_CAP_ATOMIC_OPEN_V1
9893                 | NFS_CAP_LGOPEN,
9894         .init_client = nfs41_init_client,
9895         .shutdown_client = nfs41_shutdown_client,
9896         .match_stateid = nfs41_match_stateid,
9897         .find_root_sec = nfs41_find_root_sec,
9898         .free_lock_state = nfs41_free_lock_state,
9899         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9900         .alloc_seqid = nfs_alloc_no_seqid,
9901         .session_trunk = nfs4_test_session_trunk,
9902         .call_sync_ops = &nfs41_call_sync_ops,
9903         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9904         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9905         .state_renewal_ops = &nfs41_state_renewal_ops,
9906         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9907 };
9908 #endif
9909
9910 #if defined(CONFIG_NFS_V4_2)
9911 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9912         .minor_version = 2,
9913         .init_caps = NFS_CAP_READDIRPLUS
9914                 | NFS_CAP_ATOMIC_OPEN
9915                 | NFS_CAP_POSIX_LOCK
9916                 | NFS_CAP_STATEID_NFSV41
9917                 | NFS_CAP_ATOMIC_OPEN_V1
9918                 | NFS_CAP_LGOPEN
9919                 | NFS_CAP_ALLOCATE
9920                 | NFS_CAP_COPY
9921                 | NFS_CAP_OFFLOAD_CANCEL
9922                 | NFS_CAP_COPY_NOTIFY
9923                 | NFS_CAP_DEALLOCATE
9924                 | NFS_CAP_SEEK
9925                 | NFS_CAP_LAYOUTSTATS
9926                 | NFS_CAP_CLONE
9927                 | NFS_CAP_LAYOUTERROR,
9928         .init_client = nfs41_init_client,
9929         .shutdown_client = nfs41_shutdown_client,
9930         .match_stateid = nfs41_match_stateid,
9931         .find_root_sec = nfs41_find_root_sec,
9932         .free_lock_state = nfs41_free_lock_state,
9933         .call_sync_ops = &nfs41_call_sync_ops,
9934         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9935         .alloc_seqid = nfs_alloc_no_seqid,
9936         .session_trunk = nfs4_test_session_trunk,
9937         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9938         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9939         .state_renewal_ops = &nfs41_state_renewal_ops,
9940         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9941 };
9942 #endif
9943
9944 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9945         [0] = &nfs_v4_0_minor_ops,
9946 #if defined(CONFIG_NFS_V4_1)
9947         [1] = &nfs_v4_1_minor_ops,
9948 #endif
9949 #if defined(CONFIG_NFS_V4_2)
9950         [2] = &nfs_v4_2_minor_ops,
9951 #endif
9952 };
9953
9954 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9955 {
9956         ssize_t error, error2;
9957
9958         error = generic_listxattr(dentry, list, size);
9959         if (error < 0)
9960                 return error;
9961         if (list) {
9962                 list += error;
9963                 size -= error;
9964         }
9965
9966         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9967         if (error2 < 0)
9968                 return error2;
9969         return error + error2;
9970 }
9971
9972 static const struct inode_operations nfs4_dir_inode_operations = {
9973         .create         = nfs_create,
9974         .lookup         = nfs_lookup,
9975         .atomic_open    = nfs_atomic_open,
9976         .link           = nfs_link,
9977         .unlink         = nfs_unlink,
9978         .symlink        = nfs_symlink,
9979         .mkdir          = nfs_mkdir,
9980         .rmdir          = nfs_rmdir,
9981         .mknod          = nfs_mknod,
9982         .rename         = nfs_rename,
9983         .permission     = nfs_permission,
9984         .getattr        = nfs_getattr,
9985         .setattr        = nfs_setattr,
9986         .listxattr      = nfs4_listxattr,
9987 };
9988
9989 static const struct inode_operations nfs4_file_inode_operations = {
9990         .permission     = nfs_permission,
9991         .getattr        = nfs_getattr,
9992         .setattr        = nfs_setattr,
9993         .listxattr      = nfs4_listxattr,
9994 };
9995
9996 const struct nfs_rpc_ops nfs_v4_clientops = {
9997         .version        = 4,                    /* protocol version */
9998         .dentry_ops     = &nfs4_dentry_operations,
9999         .dir_inode_ops  = &nfs4_dir_inode_operations,
10000         .file_inode_ops = &nfs4_file_inode_operations,
10001         .file_ops       = &nfs4_file_operations,
10002         .getroot        = nfs4_proc_get_root,
10003         .submount       = nfs4_submount,
10004         .try_mount      = nfs4_try_mount,
10005         .getattr        = nfs4_proc_getattr,
10006         .setattr        = nfs4_proc_setattr,
10007         .lookup         = nfs4_proc_lookup,
10008         .lookupp        = nfs4_proc_lookupp,
10009         .access         = nfs4_proc_access,
10010         .readlink       = nfs4_proc_readlink,
10011         .create         = nfs4_proc_create,
10012         .remove         = nfs4_proc_remove,
10013         .unlink_setup   = nfs4_proc_unlink_setup,
10014         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10015         .unlink_done    = nfs4_proc_unlink_done,
10016         .rename_setup   = nfs4_proc_rename_setup,
10017         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10018         .rename_done    = nfs4_proc_rename_done,
10019         .link           = nfs4_proc_link,
10020         .symlink        = nfs4_proc_symlink,
10021         .mkdir          = nfs4_proc_mkdir,
10022         .rmdir          = nfs4_proc_rmdir,
10023         .readdir        = nfs4_proc_readdir,
10024         .mknod          = nfs4_proc_mknod,
10025         .statfs         = nfs4_proc_statfs,
10026         .fsinfo         = nfs4_proc_fsinfo,
10027         .pathconf       = nfs4_proc_pathconf,
10028         .set_capabilities = nfs4_server_capabilities,
10029         .decode_dirent  = nfs4_decode_dirent,
10030         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10031         .read_setup     = nfs4_proc_read_setup,
10032         .read_done      = nfs4_read_done,
10033         .write_setup    = nfs4_proc_write_setup,
10034         .write_done     = nfs4_write_done,
10035         .commit_setup   = nfs4_proc_commit_setup,
10036         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10037         .commit_done    = nfs4_commit_done,
10038         .lock           = nfs4_proc_lock,
10039         .clear_acl_cache = nfs4_zap_acl_attr,
10040         .close_context  = nfs4_close_context,
10041         .open_context   = nfs4_atomic_open,
10042         .have_delegation = nfs4_have_delegation,
10043         .alloc_client   = nfs4_alloc_client,
10044         .init_client    = nfs4_init_client,
10045         .free_client    = nfs4_free_client,
10046         .create_server  = nfs4_create_server,
10047         .clone_server   = nfs_clone_server,
10048 };
10049
10050 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10051         .name   = XATTR_NAME_NFSV4_ACL,
10052         .list   = nfs4_xattr_list_nfs4_acl,
10053         .get    = nfs4_xattr_get_nfs4_acl,
10054         .set    = nfs4_xattr_set_nfs4_acl,
10055 };
10056
10057 const struct xattr_handler *nfs4_xattr_handlers[] = {
10058         &nfs4_xattr_nfs4_acl_handler,
10059 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10060         &nfs4_xattr_nfs4_label_handler,
10061 #endif
10062         NULL
10063 };
10064
10065 /*
10066  * Local variables:
10067  *  c-basic-offset: 8
10068  * End:
10069  */