fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <[email protected]>
  10  *  Andy Adamson   <[email protected]>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57 #include <linux/iversion.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_BITMASK_SZ         3
  75
  76 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  77 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  78
  79 /* file attributes which can be mapped to nfs attributes */
  80 #define NFS4_VALID_ATTRS (ATTR_MODE \
  81         | ATTR_UID \
  82         | ATTR_GID \
  83         | ATTR_SIZE \
  84         | ATTR_ATIME \
  85         | ATTR_MTIME \
  86         | ATTR_CTIME \
  87         | ATTR_ATIME_SET \
  88         | ATTR_MTIME_SET)
  89
  90 struct nfs4_opendata;
  91 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  92 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  93 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  94 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label, struct inode *inode);
  95 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  96 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  97                             struct nfs_fattr *fattr, struct iattr *sattr,
  98                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  99                             struct nfs4_label *olabel);
 100 #ifdef CONFIG_NFS_V4_1
 101 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 102                 const struct cred *cred,
 103                 struct nfs4_slot *slot,
 104                 bool is_privileged);
 105 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 106                 const struct cred *);
 107 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 108                 const struct cred *, bool);
 109 #endif
 110
 111 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 112 static inline struct nfs4_label *
 113 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 114         struct iattr *sattr, struct nfs4_label *label)
 115 {
 116         int err;
 117
 118         if (label == NULL)
 119                 return NULL;
 120
 121         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 122                 return NULL;
 123
 124         err = security_dentry_init_security(dentry, sattr->ia_mode,
 125                                 &dentry->d_name, (void **)&label->label, &label->len);
 126         if (err == 0)
 127                 return label;
 128
 129         return NULL;
 130 }
 131 static inline void
 132 nfs4_label_release_security(struct nfs4_label *label)
 133 {
 134         if (label)
 135                 security_release_secctx(label->label, label->len);
 136 }
 137 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 138 {
 139         if (label)
 140                 return server->attr_bitmask;
 141
 142         return server->attr_bitmask_nl;
 143 }
 144 #else
 145 static inline struct nfs4_label *
 146 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 147         struct iattr *sattr, struct nfs4_label *l)
 148 { return NULL; }
 149 static inline void
 150 nfs4_label_release_security(struct nfs4_label *label)
 151 { return; }
 152 static inline u32 *
 153 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 154 { return server->attr_bitmask; }
 155 #endif
 156
 157 /* Prevent leaks of NFSv4 errors into userland */
 158 static int nfs4_map_errors(int err)
 159 {
 160         if (err >= -1000)
 161                 return err;
 162         switch (err) {
 163         case -NFS4ERR_RESOURCE:
 164         case -NFS4ERR_LAYOUTTRYLATER:
 165         case -NFS4ERR_RECALLCONFLICT:
 166                 return -EREMOTEIO;
 167         case -NFS4ERR_WRONGSEC:
 168         case -NFS4ERR_WRONG_CRED:
 169                 return -EPERM;
 170         case -NFS4ERR_BADOWNER:
 171         case -NFS4ERR_BADNAME:
 172                 return -EINVAL;
 173         case -NFS4ERR_SHARE_DENIED:
 174                 return -EACCES;
 175         case -NFS4ERR_MINOR_VERS_MISMATCH:
 176                 return -EPROTONOSUPPORT;
 177         case -NFS4ERR_FILE_OPEN:
 178                 return -EBUSY;
 179         default:
 180                 dprintk("%s could not handle NFSv4 error %d\n",
 181                                 __func__, -err);
 182                 break;
 183         }
 184         return -EIO;
 185 }
 186
 187 /*
 188  * This is our standard bitmap for GETATTR requests.
 189  */
 190 const u32 nfs4_fattr_bitmap[3] = {
 191         FATTR4_WORD0_TYPE
 192         | FATTR4_WORD0_CHANGE
 193         | FATTR4_WORD0_SIZE
 194         | FATTR4_WORD0_FSID
 195         | FATTR4_WORD0_FILEID,
 196         FATTR4_WORD1_MODE
 197         | FATTR4_WORD1_NUMLINKS
 198         | FATTR4_WORD1_OWNER
 199         | FATTR4_WORD1_OWNER_GROUP
 200         | FATTR4_WORD1_RAWDEV
 201         | FATTR4_WORD1_SPACE_USED
 202         | FATTR4_WORD1_TIME_ACCESS
 203         | FATTR4_WORD1_TIME_METADATA
 204         | FATTR4_WORD1_TIME_MODIFY
 205         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 206 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 207         FATTR4_WORD2_SECURITY_LABEL
 208 #endif
 209 };
 210
 211 static const u32 nfs4_pnfs_open_bitmap[3] = {
 212         FATTR4_WORD0_TYPE
 213         | FATTR4_WORD0_CHANGE
 214         | FATTR4_WORD0_SIZE
 215         | FATTR4_WORD0_FSID
 216         | FATTR4_WORD0_FILEID,
 217         FATTR4_WORD1_MODE
 218         | FATTR4_WORD1_NUMLINKS
 219         | FATTR4_WORD1_OWNER
 220         | FATTR4_WORD1_OWNER_GROUP
 221         | FATTR4_WORD1_RAWDEV
 222         | FATTR4_WORD1_SPACE_USED
 223         | FATTR4_WORD1_TIME_ACCESS
 224         | FATTR4_WORD1_TIME_METADATA
 225         | FATTR4_WORD1_TIME_MODIFY,
 226         FATTR4_WORD2_MDSTHRESHOLD
 227 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 228         | FATTR4_WORD2_SECURITY_LABEL
 229 #endif
 230 };
 231
 232 static const u32 nfs4_open_noattr_bitmap[3] = {
 233         FATTR4_WORD0_TYPE
 234         | FATTR4_WORD0_FILEID,
 235 };
 236
 237 const u32 nfs4_statfs_bitmap[3] = {
 238         FATTR4_WORD0_FILES_AVAIL
 239         | FATTR4_WORD0_FILES_FREE
 240         | FATTR4_WORD0_FILES_TOTAL,
 241         FATTR4_WORD1_SPACE_AVAIL
 242         | FATTR4_WORD1_SPACE_FREE
 243         | FATTR4_WORD1_SPACE_TOTAL
 244 };
 245
 246 const u32 nfs4_pathconf_bitmap[3] = {
 247         FATTR4_WORD0_MAXLINK
 248         | FATTR4_WORD0_MAXNAME,
 249         0
 250 };
 251
 252 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 253                         | FATTR4_WORD0_MAXREAD
 254                         | FATTR4_WORD0_MAXWRITE
 255                         | FATTR4_WORD0_LEASE_TIME,
 256                         FATTR4_WORD1_TIME_DELTA
 257                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 258                         FATTR4_WORD2_LAYOUT_BLKSIZE
 259                         | FATTR4_WORD2_CLONE_BLKSIZE
 260 };
 261
 262 const u32 nfs4_fs_locations_bitmap[3] = {
 263         FATTR4_WORD0_CHANGE
 264         | FATTR4_WORD0_SIZE
 265         | FATTR4_WORD0_FSID
 266         | FATTR4_WORD0_FILEID
 267         | FATTR4_WORD0_FS_LOCATIONS,
 268         FATTR4_WORD1_OWNER
 269         | FATTR4_WORD1_OWNER_GROUP
 270         | FATTR4_WORD1_RAWDEV
 271         | FATTR4_WORD1_SPACE_USED
 272         | FATTR4_WORD1_TIME_ACCESS
 273         | FATTR4_WORD1_TIME_METADATA
 274         | FATTR4_WORD1_TIME_MODIFY
 275         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 276 };
 277
 278 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 279                 struct inode *inode)
 280 {
 281         unsigned long cache_validity;
 282
 283         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 284         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 285                 return;
 286
 287         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 288         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 289                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 290                                 | NFS_INO_INVALID_SIZE);
 291
 292         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 293                 dst[0] &= ~FATTR4_WORD0_SIZE;
 294
 295         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 296                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 297 }
 298
 299 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 300                 const __u32 *src, struct inode *inode)
 301 {
 302         nfs4_bitmap_copy_adjust(dst, src, inode);
 303 }
 304
 305 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 306                 struct nfs4_readdir_arg *readdir)
 307 {
 308         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 309         __be32 *start, *p;
 310
 311         if (cookie > 2) {
 312                 readdir->cookie = cookie;
 313                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 314                 return;
 315         }
 316
 317         readdir->cookie = 0;
 318         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 319         if (cookie == 2)
 320                 return;
 321
 322         /*
 323          * NFSv4 servers do not return entries for '.' and '..'
 324          * Therefore, we fake these entries here.  We let '.'
 325          * have cookie 0 and '..' have cookie 1.  Note that
 326          * when talking to the server, we always send cookie 0
 327          * instead of 1 or 2.
 328          */
 329         start = p = kmap_atomic(*readdir->pages);
 330
 331         if (cookie == 0) {
 332                 *p++ = xdr_one;                                  /* next */
 333                 *p++ = xdr_zero;                   /* cookie, first word */
 334                 *p++ = xdr_one;                   /* cookie, second word */
 335                 *p++ = xdr_one;                             /* entry len */
 336                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 337                 p++;
 338                 *p++ = xdr_one;                         /* bitmap length */
 339                 *p++ = htonl(attrs);                           /* bitmap */
 340                 *p++ = htonl(12);             /* attribute buffer length */
 341                 *p++ = htonl(NF4DIR);
 342                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 343         }
 344
 345         *p++ = xdr_one;                                  /* next */
 346         *p++ = xdr_zero;                   /* cookie, first word */
 347         *p++ = xdr_two;                   /* cookie, second word */
 348         *p++ = xdr_two;                             /* entry len */
 349         memcpy(p, "..\0\0", 4);                         /* entry */
 350         p++;
 351         *p++ = xdr_one;                         /* bitmap length */
 352         *p++ = htonl(attrs);                           /* bitmap */
 353         *p++ = htonl(12);             /* attribute buffer length */
 354         *p++ = htonl(NF4DIR);
 355         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 356
 357         readdir->pgbase = (char *)p - (char *)start;
 358         readdir->count -= readdir->pgbase;
 359         kunmap_atomic(start);
 360 }
 361
 362 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 363                 nfs4_stateid *stateid,
 364                 const struct cred *cred)
 365 {
 366         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 367
 368         ops->test_and_free_expired(server, stateid, cred);
 369 }
 370
 371 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 372                 nfs4_stateid *stateid,
 373                 const struct cred *cred)
 374 {
 375         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 376         nfs4_test_and_free_stateid(server, stateid, cred);
 377 }
 378
 379 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 380                 const nfs4_stateid *stateid,
 381                 const struct cred *cred)
 382 {
 383         nfs4_stateid tmp;
 384
 385         nfs4_stateid_copy(&tmp, stateid);
 386         __nfs4_free_revoked_stateid(server, &tmp, cred);
 387 }
 388
 389 static long nfs4_update_delay(long *timeout)
 390 {
 391         long ret;
 392         if (!timeout)
 393                 return NFS4_POLL_RETRY_MAX;
 394         if (*timeout <= 0)
 395                 *timeout = NFS4_POLL_RETRY_MIN;
 396         if (*timeout > NFS4_POLL_RETRY_MAX)
 397                 *timeout = NFS4_POLL_RETRY_MAX;
 398         ret = *timeout;
 399         *timeout <<= 1;
 400         return ret;
 401 }
 402
 403 static int nfs4_delay_killable(long *timeout)
 404 {
 405         might_sleep();
 406
 407         freezable_schedule_timeout_killable_unsafe(
 408                 nfs4_update_delay(timeout));
 409         if (!__fatal_signal_pending(current))
 410                 return 0;
 411         return -EINTR;
 412 }
 413
 414 static int nfs4_delay_interruptible(long *timeout)
 415 {
 416         might_sleep();
 417
 418         freezable_schedule_timeout_interruptible(nfs4_update_delay(timeout));
 419         if (!signal_pending(current))
 420                 return 0;
 421         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 422 }
 423
 424 static int nfs4_delay(long *timeout, bool interruptible)
 425 {
 426         if (interruptible)
 427                 return nfs4_delay_interruptible(timeout);
 428         return nfs4_delay_killable(timeout);
 429 }
 430
 431 static const nfs4_stateid *
 432 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 433 {
 434         if (!stateid)
 435                 return NULL;
 436         switch (stateid->type) {
 437         case NFS4_OPEN_STATEID_TYPE:
 438         case NFS4_LOCK_STATEID_TYPE:
 439         case NFS4_DELEGATION_STATEID_TYPE:
 440                 return stateid;
 441         default:
 442                 break;
 443         }
 444         return NULL;
 445 }
 446
 447 /* This is the error handling routine for processes that are allowed
 448  * to sleep.
 449  */
 450 static int nfs4_do_handle_exception(struct nfs_server *server,
 451                 int errorcode, struct nfs4_exception *exception)
 452 {
 453         struct nfs_client *clp = server->nfs_client;
 454         struct nfs4_state *state = exception->state;
 455         const nfs4_stateid *stateid;
 456         struct inode *inode = exception->inode;
 457         int ret = errorcode;
 458
 459         exception->delay = 0;
 460         exception->recovering = 0;
 461         exception->retry = 0;
 462
 463         stateid = nfs4_recoverable_stateid(exception->stateid);
 464         if (stateid == NULL && state != NULL)
 465                 stateid = nfs4_recoverable_stateid(&state->stateid);
 466
 467         switch(errorcode) {
 468                 case 0:
 469                         return 0;
 470                 case -NFS4ERR_BADHANDLE:
 471                 case -ESTALE:
 472                         if (inode != NULL && S_ISREG(inode->i_mode))
 473                                 pnfs_destroy_layout(NFS_I(inode));
 474                         break;
 475                 case -NFS4ERR_DELEG_REVOKED:
 476                 case -NFS4ERR_ADMIN_REVOKED:
 477                 case -NFS4ERR_EXPIRED:
 478                 case -NFS4ERR_BAD_STATEID:
 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                         dprintk("%s ERROR: %d Reset session\n", __func__,
 525                                 errorcode);
 526                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 527                         goto wait_on_recovery;
 528 #endif /* defined(CONFIG_NFS_V4_1) */
 529                 case -NFS4ERR_FILE_OPEN:
 530                         if (exception->timeout > HZ) {
 531                                 /* We have retried a decent amount, time to
 532                                  * fail
 533                                  */
 534                                 ret = -EBUSY;
 535                                 break;
 536                         }
 537                         /* Fall through */
 538                 case -NFS4ERR_DELAY:
 539                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 540                         /* Fall through */
 541                 case -NFS4ERR_GRACE:
 542                 case -NFS4ERR_LAYOUTTRYLATER:
 543                 case -NFS4ERR_RECALLCONFLICT:
 544                         exception->delay = 1;
 545                         return 0;
 546
 547                 case -NFS4ERR_RETRY_UNCACHED_REP:
 548                 case -NFS4ERR_OLD_STATEID:
 549                         exception->retry = 1;
 550                         break;
 551                 case -NFS4ERR_BADOWNER:
 552                         /* The following works around a Linux server bug! */
 553                 case -NFS4ERR_BADNAME:
 554                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 555                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 556                                 exception->retry = 1;
 557                                 printk(KERN_WARNING "NFS: v4 server %s "
 558                                                 "does not accept raw "
 559                                                 "uid/gids. "
 560                                                 "Reenabling the idmapper.\n",
 561                                                 server->nfs_client->cl_hostname);
 562                         }
 563         }
 564         /* We failed to handle the error */
 565         return nfs4_map_errors(ret);
 566 wait_on_recovery:
 567         exception->recovering = 1;
 568         return 0;
 569 }
 570
 571 /* This is the error handling routine for processes that are allowed
 572  * to sleep.
 573  */
 574 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 575 {
 576         struct nfs_client *clp = server->nfs_client;
 577         int ret;
 578
 579         ret = nfs4_do_handle_exception(server, errorcode, exception);
 580         if (exception->delay) {
 581                 ret = nfs4_delay(&exception->timeout,
 582                                 exception->interruptible);
 583                 goto out_retry;
 584         }
 585         if (exception->recovering) {
 586                 ret = nfs4_wait_clnt_recover(clp);
 587                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 588                         return -EIO;
 589                 goto out_retry;
 590         }
 591         return ret;
 592 out_retry:
 593         if (ret == 0)
 594                 exception->retry = 1;
 595         return ret;
 596 }
 597
 598 static int
 599 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 600                 int errorcode, struct nfs4_exception *exception)
 601 {
 602         struct nfs_client *clp = server->nfs_client;
 603         int ret;
 604
 605         ret = nfs4_do_handle_exception(server, errorcode, exception);
 606         if (exception->delay) {
 607                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 608                 goto out_retry;
 609         }
 610         if (exception->recovering) {
 611                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 612                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 613                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 614                 goto out_retry;
 615         }
 616         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 617                 ret = -EIO;
 618         return ret;
 619 out_retry:
 620         if (ret == 0) {
 621                 exception->retry = 1;
 622                 /*
 623                  * For NFS4ERR_MOVED, the client transport will need to
 624                  * be recomputed after migration recovery has completed.
 625                  */
 626                 if (errorcode == -NFS4ERR_MOVED)
 627                         rpc_task_release_transport(task);
 628         }
 629         return ret;
 630 }
 631
 632 int
 633 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 634                         struct nfs4_state *state, long *timeout)
 635 {
 636         struct nfs4_exception exception = {
 637                 .state = state,
 638         };
 639
 640         if (task->tk_status >= 0)
 641                 return 0;
 642         if (timeout)
 643                 exception.timeout = *timeout;
 644         task->tk_status = nfs4_async_handle_exception(task, server,
 645                         task->tk_status,
 646                         &exception);
 647         if (exception.delay && timeout)
 648                 *timeout = exception.timeout;
 649         if (exception.retry)
 650                 return -EAGAIN;
 651         return 0;
 652 }
 653
 654 /*
 655  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 656  * or 'false' otherwise.
 657  */
 658 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 659 {
 660         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 661         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 662 }
 663
 664 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 665 {
 666         spin_lock(&clp->cl_lock);
 667         if (time_before(clp->cl_last_renewal,timestamp))
 668                 clp->cl_last_renewal = timestamp;
 669         spin_unlock(&clp->cl_lock);
 670 }
 671
 672 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 673 {
 674         struct nfs_client *clp = server->nfs_client;
 675
 676         if (!nfs4_has_session(clp))
 677                 do_renew_lease(clp, timestamp);
 678 }
 679
 680 struct nfs4_call_sync_data {
 681         const struct nfs_server *seq_server;
 682         struct nfs4_sequence_args *seq_args;
 683         struct nfs4_sequence_res *seq_res;
 684 };
 685
 686 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 687                         struct nfs4_sequence_res *res, int cache_reply,
 688                         int privileged)
 689 {
 690         args->sa_slot = NULL;
 691         args->sa_cache_this = cache_reply;
 692         args->sa_privileged = privileged;
 693
 694         res->sr_slot = NULL;
 695 }
 696
 697 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 698 {
 699         struct nfs4_slot *slot = res->sr_slot;
 700         struct nfs4_slot_table *tbl;
 701
 702         tbl = slot->table;
 703         spin_lock(&tbl->slot_tbl_lock);
 704         if (!nfs41_wake_and_assign_slot(tbl, slot))
 705                 nfs4_free_slot(tbl, slot);
 706         spin_unlock(&tbl->slot_tbl_lock);
 707
 708         res->sr_slot = NULL;
 709 }
 710
 711 static int nfs40_sequence_done(struct rpc_task *task,
 712                                struct nfs4_sequence_res *res)
 713 {
 714         if (res->sr_slot != NULL)
 715                 nfs40_sequence_free_slot(res);
 716         return 1;
 717 }
 718
 719 #if defined(CONFIG_NFS_V4_1)
 720
 721 static void nfs41_release_slot(struct nfs4_slot *slot)
 722 {
 723         struct nfs4_session *session;
 724         struct nfs4_slot_table *tbl;
 725         bool send_new_highest_used_slotid = false;
 726
 727         if (!slot)
 728                 return;
 729         tbl = slot->table;
 730         session = tbl->session;
 731
 732         /* Bump the slot sequence number */
 733         if (slot->seq_done)
 734                 slot->seq_nr++;
 735         slot->seq_done = 0;
 736
 737         spin_lock(&tbl->slot_tbl_lock);
 738         /* Be nice to the server: try to ensure that the last transmitted
 739          * value for highest_user_slotid <= target_highest_slotid
 740          */
 741         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 742                 send_new_highest_used_slotid = true;
 743
 744         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 745                 send_new_highest_used_slotid = false;
 746                 goto out_unlock;
 747         }
 748         nfs4_free_slot(tbl, slot);
 749
 750         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 751                 send_new_highest_used_slotid = false;
 752 out_unlock:
 753         spin_unlock(&tbl->slot_tbl_lock);
 754         if (send_new_highest_used_slotid)
 755                 nfs41_notify_server(session->clp);
 756         if (waitqueue_active(&tbl->slot_waitq))
 757                 wake_up_all(&tbl->slot_waitq);
 758 }
 759
 760 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 761 {
 762         nfs41_release_slot(res->sr_slot);
 763         res->sr_slot = NULL;
 764 }
 765
 766 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 767                 u32 seqnr)
 768 {
 769         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 770                 slot->seq_nr_highest_sent = seqnr;
 771 }
 772 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 773                 u32 seqnr)
 774 {
 775         slot->seq_nr_highest_sent = seqnr;
 776         slot->seq_nr_last_acked = seqnr;
 777 }
 778
 779 static int nfs41_sequence_process(struct rpc_task *task,
 780                 struct nfs4_sequence_res *res)
 781 {
 782         struct nfs4_session *session;
 783         struct nfs4_slot *slot = res->sr_slot;
 784         struct nfs_client *clp;
 785         int ret = 1;
 786
 787         if (slot == NULL)
 788                 goto out_noaction;
 789         /* don't increment the sequence number if the task wasn't sent */
 790         if (!RPC_WAS_SENT(task) || slot->seq_done)
 791                 goto out;
 792
 793         session = slot->table->session;
 794
 795         trace_nfs4_sequence_done(session, res);
 796         /* Check the SEQUENCE operation status */
 797         switch (res->sr_status) {
 798         case 0:
 799                 /* Mark this sequence number as having been acked */
 800                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 801                 /* Update the slot's sequence and clientid lease timer */
 802                 slot->seq_done = 1;
 803                 clp = session->clp;
 804                 do_renew_lease(clp, res->sr_timestamp);
 805                 /* Check sequence flags */
 806                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 807                                 !!slot->privileged);
 808                 nfs41_update_target_slotid(slot->table, slot, res);
 809                 break;
 810         case 1:
 811                 /*
 812                  * sr_status remains 1 if an RPC level error occurred.
 813                  * The server may or may not have processed the sequence
 814                  * operation..
 815                  */
 816                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 817                 slot->seq_done = 1;
 818                 goto out;
 819         case -NFS4ERR_DELAY:
 820                 /* The server detected a resend of the RPC call and
 821                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 822                  * of RFC5661.
 823                  */
 824                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 825                         __func__,
 826                         slot->slot_nr,
 827                         slot->seq_nr);
 828                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 829                 goto out_retry;
 830         case -NFS4ERR_RETRY_UNCACHED_REP:
 831         case -NFS4ERR_SEQ_FALSE_RETRY:
 832                 /*
 833                  * The server thinks we tried to replay a request.
 834                  * Retry the call after bumping the sequence ID.
 835                  */
 836                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 837                 goto retry_new_seq;
 838         case -NFS4ERR_BADSLOT:
 839                 /*
 840                  * The slot id we used was probably retired. Try again
 841                  * using a different slot id.
 842                  */
 843                 if (slot->slot_nr < slot->table->target_highest_slotid)
 844                         goto session_recover;
 845                 goto retry_nowait;
 846         case -NFS4ERR_SEQ_MISORDERED:
 847                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 848                 /*
 849                  * Were one or more calls using this slot interrupted?
 850                  * If the server never received the request, then our
 851                  * transmitted slot sequence number may be too high.
 852                  */
 853                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 854                         slot->seq_nr--;
 855                         goto retry_nowait;
 856                 }
 857                 /*
 858                  * RFC5661:
 859                  * A retry might be sent while the original request is
 860                  * still in progress on the replier. The replier SHOULD
 861                  * deal with the issue by returning NFS4ERR_DELAY as the
 862                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 863                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 864                  *
 865                  * Restart the search after a delay.
 866                  */
 867                 slot->seq_nr = slot->seq_nr_highest_sent;
 868                 goto out_retry;
 869         default:
 870                 /* Just update the slot sequence no. */
 871                 slot->seq_done = 1;
 872         }
 873 out:
 874         /* The session may be reset by one of the error handlers. */
 875         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 876 out_noaction:
 877         return ret;
 878 session_recover:
 879         nfs4_schedule_session_recovery(session, res->sr_status);
 880         goto retry_nowait;
 881 retry_new_seq:
 882         ++slot->seq_nr;
 883 retry_nowait:
 884         if (rpc_restart_call_prepare(task)) {
 885                 nfs41_sequence_free_slot(res);
 886                 task->tk_status = 0;
 887                 ret = 0;
 888         }
 889         goto out;
 890 out_retry:
 891         if (!rpc_restart_call(task))
 892                 goto out;
 893         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 894         return 0;
 895 }
 896
 897 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 898 {
 899         if (!nfs41_sequence_process(task, res))
 900                 return 0;
 901         if (res->sr_slot != NULL)
 902                 nfs41_sequence_free_slot(res);
 903         return 1;
 904
 905 }
 906 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 907
 908 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 909 {
 910         if (res->sr_slot == NULL)
 911                 return 1;
 912         if (res->sr_slot->table->session != NULL)
 913                 return nfs41_sequence_process(task, res);
 914         return nfs40_sequence_done(task, res);
 915 }
 916
 917 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 918 {
 919         if (res->sr_slot != NULL) {
 920                 if (res->sr_slot->table->session != NULL)
 921                         nfs41_sequence_free_slot(res);
 922                 else
 923                         nfs40_sequence_free_slot(res);
 924         }
 925 }
 926
 927 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 928 {
 929         if (res->sr_slot == NULL)
 930                 return 1;
 931         if (!res->sr_slot->table->session)
 932                 return nfs40_sequence_done(task, res);
 933         return nfs41_sequence_done(task, res);
 934 }
 935 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 936
 937 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 938 {
 939         struct nfs4_call_sync_data *data = calldata;
 940
 941         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 942
 943         nfs4_setup_sequence(data->seq_server->nfs_client,
 944                             data->seq_args, data->seq_res, task);
 945 }
 946
 947 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 948 {
 949         struct nfs4_call_sync_data *data = calldata;
 950
 951         nfs41_sequence_done(task, data->seq_res);
 952 }
 953
 954 static const struct rpc_call_ops nfs41_call_sync_ops = {
 955         .rpc_call_prepare = nfs41_call_sync_prepare,
 956         .rpc_call_done = nfs41_call_sync_done,
 957 };
 958
 959 #else   /* !CONFIG_NFS_V4_1 */
 960
 961 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 962 {
 963         return nfs40_sequence_done(task, res);
 964 }
 965
 966 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 967 {
 968         if (res->sr_slot != NULL)
 969                 nfs40_sequence_free_slot(res);
 970 }
 971
 972 int nfs4_sequence_done(struct rpc_task *task,
 973                        struct nfs4_sequence_res *res)
 974 {
 975         return nfs40_sequence_done(task, res);
 976 }
 977 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 978
 979 #endif  /* !CONFIG_NFS_V4_1 */
 980
 981 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
 982 {
 983         res->sr_timestamp = jiffies;
 984         res->sr_status_flags = 0;
 985         res->sr_status = 1;
 986 }
 987
 988 static
 989 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
 990                 struct nfs4_sequence_res *res,
 991                 struct nfs4_slot *slot)
 992 {
 993         if (!slot)
 994                 return;
 995         slot->privileged = args->sa_privileged ? 1 : 0;
 996         args->sa_slot = slot;
 997
 998         res->sr_slot = slot;
 999 }
1000
1001 int nfs4_setup_sequence(struct nfs_client *client,
1002                         struct nfs4_sequence_args *args,
1003                         struct nfs4_sequence_res *res,
1004                         struct rpc_task *task)
1005 {
1006         struct nfs4_session *session = nfs4_get_session(client);
1007         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1008         struct nfs4_slot *slot;
1009
1010         /* slot already allocated? */
1011         if (res->sr_slot != NULL)
1012                 goto out_start;
1013
1014         if (session)
1015                 tbl = &session->fc_slot_table;
1016
1017         spin_lock(&tbl->slot_tbl_lock);
1018         /* The state manager will wait until the slot table is empty */
1019         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1020                 goto out_sleep;
1021
1022         slot = nfs4_alloc_slot(tbl);
1023         if (IS_ERR(slot)) {
1024                 if (slot == ERR_PTR(-ENOMEM))
1025                         goto out_sleep_timeout;
1026                 goto out_sleep;
1027         }
1028         spin_unlock(&tbl->slot_tbl_lock);
1029
1030         nfs4_sequence_attach_slot(args, res, slot);
1031
1032         trace_nfs4_setup_sequence(session, args);
1033 out_start:
1034         nfs41_sequence_res_init(res);
1035         rpc_call_start(task);
1036         return 0;
1037 out_sleep_timeout:
1038         /* Try again in 1/4 second */
1039         if (args->sa_privileged)
1040                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1041                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1042         else
1043                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1044                                 NULL, jiffies + (HZ >> 2));
1045         spin_unlock(&tbl->slot_tbl_lock);
1046         return -EAGAIN;
1047 out_sleep:
1048         if (args->sa_privileged)
1049                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1050                                 RPC_PRIORITY_PRIVILEGED);
1051         else
1052                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1053         spin_unlock(&tbl->slot_tbl_lock);
1054         return -EAGAIN;
1055 }
1056 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1057
1058 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1059 {
1060         struct nfs4_call_sync_data *data = calldata;
1061         nfs4_setup_sequence(data->seq_server->nfs_client,
1062                                 data->seq_args, data->seq_res, task);
1063 }
1064
1065 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1066 {
1067         struct nfs4_call_sync_data *data = calldata;
1068         nfs4_sequence_done(task, data->seq_res);
1069 }
1070
1071 static const struct rpc_call_ops nfs40_call_sync_ops = {
1072         .rpc_call_prepare = nfs40_call_sync_prepare,
1073         .rpc_call_done = nfs40_call_sync_done,
1074 };
1075
1076 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1077 {
1078         int ret;
1079         struct rpc_task *task;
1080
1081         task = rpc_run_task(task_setup);
1082         if (IS_ERR(task))
1083                 return PTR_ERR(task);
1084
1085         ret = task->tk_status;
1086         rpc_put_task(task);
1087         return ret;
1088 }
1089
1090 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1091                                    struct nfs_server *server,
1092                                    struct rpc_message *msg,
1093                                    struct nfs4_sequence_args *args,
1094                                    struct nfs4_sequence_res *res)
1095 {
1096         struct nfs_client *clp = server->nfs_client;
1097         struct nfs4_call_sync_data data = {
1098                 .seq_server = server,
1099                 .seq_args = args,
1100                 .seq_res = res,
1101         };
1102         struct rpc_task_setup task_setup = {
1103                 .rpc_client = clnt,
1104                 .rpc_message = msg,
1105                 .callback_ops = clp->cl_mvops->call_sync_ops,
1106                 .callback_data = &data
1107         };
1108
1109         return nfs4_call_sync_custom(&task_setup);
1110 }
1111
1112 int nfs4_call_sync(struct rpc_clnt *clnt,
1113                    struct nfs_server *server,
1114                    struct rpc_message *msg,
1115                    struct nfs4_sequence_args *args,
1116                    struct nfs4_sequence_res *res,
1117                    int cache_reply)
1118 {
1119         nfs4_init_sequence(args, res, cache_reply, 0);
1120         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1121 }
1122
1123 static void
1124 nfs4_inc_nlink_locked(struct inode *inode)
1125 {
1126         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1127         inc_nlink(inode);
1128 }
1129
1130 static void
1131 nfs4_dec_nlink_locked(struct inode *inode)
1132 {
1133         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1134         drop_nlink(inode);
1135 }
1136
1137 static void
1138 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1139                 unsigned long timestamp, unsigned long cache_validity)
1140 {
1141         struct nfs_inode *nfsi = NFS_I(dir);
1142
1143         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1144                 | NFS_INO_INVALID_MTIME
1145                 | NFS_INO_INVALID_DATA
1146                 | cache_validity;
1147         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1148                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1149                 nfsi->attrtimeo_timestamp = jiffies;
1150         } else {
1151                 nfs_force_lookup_revalidate(dir);
1152                 if (cinfo->before != inode_peek_iversion_raw(dir))
1153                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1154                                 NFS_INO_INVALID_ACL;
1155         }
1156         inode_set_iversion_raw(dir, cinfo->after);
1157         nfsi->read_cache_jiffies = timestamp;
1158         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1159         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1160         nfs_fscache_invalidate(dir);
1161 }
1162
1163 static void
1164 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1165                 unsigned long timestamp, unsigned long cache_validity)
1166 {
1167         spin_lock(&dir->i_lock);
1168         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1169         spin_unlock(&dir->i_lock);
1170 }
1171
1172 struct nfs4_open_createattrs {
1173         struct nfs4_label *label;
1174         struct iattr *sattr;
1175         const __u32 verf[2];
1176 };
1177
1178 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1179                 int err, struct nfs4_exception *exception)
1180 {
1181         if (err != -EINVAL)
1182                 return false;
1183         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1184                 return false;
1185         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1186         exception->retry = 1;
1187         return true;
1188 }
1189
1190 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1191 {
1192          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1193 }
1194
1195 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1196 {
1197         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1198
1199         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1200 }
1201
1202 static u32
1203 nfs4_map_atomic_open_share(struct nfs_server *server,
1204                 fmode_t fmode, int openflags)
1205 {
1206         u32 res = 0;
1207
1208         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1209         case FMODE_READ:
1210                 res = NFS4_SHARE_ACCESS_READ;
1211                 break;
1212         case FMODE_WRITE:
1213                 res = NFS4_SHARE_ACCESS_WRITE;
1214                 break;
1215         case FMODE_READ|FMODE_WRITE:
1216                 res = NFS4_SHARE_ACCESS_BOTH;
1217         }
1218         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1219                 goto out;
1220         /* Want no delegation if we're using O_DIRECT */
1221         if (openflags & O_DIRECT)
1222                 res |= NFS4_SHARE_WANT_NO_DELEG;
1223 out:
1224         return res;
1225 }
1226
1227 static enum open_claim_type4
1228 nfs4_map_atomic_open_claim(struct nfs_server *server,
1229                 enum open_claim_type4 claim)
1230 {
1231         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1232                 return claim;
1233         switch (claim) {
1234         default:
1235                 return claim;
1236         case NFS4_OPEN_CLAIM_FH:
1237                 return NFS4_OPEN_CLAIM_NULL;
1238         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1239                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1240         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1241                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1242         }
1243 }
1244
1245 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1246 {
1247         p->o_res.f_attr = &p->f_attr;
1248         p->o_res.f_label = p->f_label;
1249         p->o_res.seqid = p->o_arg.seqid;
1250         p->c_res.seqid = p->c_arg.seqid;
1251         p->o_res.server = p->o_arg.server;
1252         p->o_res.access_request = p->o_arg.access;
1253         nfs_fattr_init(&p->f_attr);
1254         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1255 }
1256
1257 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1258                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1259                 const struct nfs4_open_createattrs *c,
1260                 enum open_claim_type4 claim,
1261                 gfp_t gfp_mask)
1262 {
1263         struct dentry *parent = dget_parent(dentry);
1264         struct inode *dir = d_inode(parent);
1265         struct nfs_server *server = NFS_SERVER(dir);
1266         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1267         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1268         struct nfs4_opendata *p;
1269
1270         p = kzalloc(sizeof(*p), gfp_mask);
1271         if (p == NULL)
1272                 goto err;
1273
1274         p->f_label = nfs4_label_alloc(server, gfp_mask);
1275         if (IS_ERR(p->f_label))
1276                 goto err_free_p;
1277
1278         p->a_label = nfs4_label_alloc(server, gfp_mask);
1279         if (IS_ERR(p->a_label))
1280                 goto err_free_f;
1281
1282         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1283         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1284         if (IS_ERR(p->o_arg.seqid))
1285                 goto err_free_label;
1286         nfs_sb_active(dentry->d_sb);
1287         p->dentry = dget(dentry);
1288         p->dir = parent;
1289         p->owner = sp;
1290         atomic_inc(&sp->so_count);
1291         p->o_arg.open_flags = flags;
1292         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1293         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1294         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1295                         fmode, flags);
1296         if (flags & O_CREAT) {
1297                 p->o_arg.umask = current_umask();
1298                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1299                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1300                         p->o_arg.u.attrs = &p->attrs;
1301                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1302
1303                         memcpy(p->o_arg.u.verifier.data, c->verf,
1304                                         sizeof(p->o_arg.u.verifier.data));
1305                 }
1306         }
1307         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1308          * will return permission denied for all bits until close */
1309         if (!(flags & O_EXCL)) {
1310                 /* ask server to check for all possible rights as results
1311                  * are cached */
1312                 switch (p->o_arg.claim) {
1313                 default:
1314                         break;
1315                 case NFS4_OPEN_CLAIM_NULL:
1316                 case NFS4_OPEN_CLAIM_FH:
1317                         p->o_arg.access = NFS4_ACCESS_READ |
1318                                 NFS4_ACCESS_MODIFY |
1319                                 NFS4_ACCESS_EXTEND |
1320                                 NFS4_ACCESS_EXECUTE;
1321                 }
1322         }
1323         p->o_arg.clientid = server->nfs_client->cl_clientid;
1324         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1325         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1326         p->o_arg.name = &dentry->d_name;
1327         p->o_arg.server = server;
1328         p->o_arg.bitmask = nfs4_bitmask(server, label);
1329         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1330         switch (p->o_arg.claim) {
1331         case NFS4_OPEN_CLAIM_NULL:
1332         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1333         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1334                 p->o_arg.fh = NFS_FH(dir);
1335                 break;
1336         case NFS4_OPEN_CLAIM_PREVIOUS:
1337         case NFS4_OPEN_CLAIM_FH:
1338         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1339         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1340                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1341         }
1342         p->c_arg.fh = &p->o_res.fh;
1343         p->c_arg.stateid = &p->o_res.stateid;
1344         p->c_arg.seqid = p->o_arg.seqid;
1345         nfs4_init_opendata_res(p);
1346         kref_init(&p->kref);
1347         return p;
1348
1349 err_free_label:
1350         nfs4_label_free(p->a_label);
1351 err_free_f:
1352         nfs4_label_free(p->f_label);
1353 err_free_p:
1354         kfree(p);
1355 err:
1356         dput(parent);
1357         return NULL;
1358 }
1359
1360 static void nfs4_opendata_free(struct kref *kref)
1361 {
1362         struct nfs4_opendata *p = container_of(kref,
1363                         struct nfs4_opendata, kref);
1364         struct super_block *sb = p->dentry->d_sb;
1365
1366         nfs4_lgopen_release(p->lgp);
1367         nfs_free_seqid(p->o_arg.seqid);
1368         nfs4_sequence_free_slot(&p->o_res.seq_res);
1369         if (p->state != NULL)
1370                 nfs4_put_open_state(p->state);
1371         nfs4_put_state_owner(p->owner);
1372
1373         nfs4_label_free(p->a_label);
1374         nfs4_label_free(p->f_label);
1375
1376         dput(p->dir);
1377         dput(p->dentry);
1378         nfs_sb_deactive(sb);
1379         nfs_fattr_free_names(&p->f_attr);
1380         kfree(p->f_attr.mdsthreshold);
1381         kfree(p);
1382 }
1383
1384 static void nfs4_opendata_put(struct nfs4_opendata *p)
1385 {
1386         if (p != NULL)
1387                 kref_put(&p->kref, nfs4_opendata_free);
1388 }
1389
1390 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1391                 fmode_t fmode)
1392 {
1393         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1394         case FMODE_READ|FMODE_WRITE:
1395                 return state->n_rdwr != 0;
1396         case FMODE_WRITE:
1397                 return state->n_wronly != 0;
1398         case FMODE_READ:
1399                 return state->n_rdonly != 0;
1400         }
1401         WARN_ON_ONCE(1);
1402         return false;
1403 }
1404
1405 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1406                 int open_mode, enum open_claim_type4 claim)
1407 {
1408         int ret = 0;
1409
1410         if (open_mode & (O_EXCL|O_TRUNC))
1411                 goto out;
1412         switch (claim) {
1413         case NFS4_OPEN_CLAIM_NULL:
1414         case NFS4_OPEN_CLAIM_FH:
1415                 goto out;
1416         default:
1417                 break;
1418         }
1419         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1420                 case FMODE_READ:
1421                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1422                                 && state->n_rdonly != 0;
1423                         break;
1424                 case FMODE_WRITE:
1425                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1426                                 && state->n_wronly != 0;
1427                         break;
1428                 case FMODE_READ|FMODE_WRITE:
1429                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1430                                 && state->n_rdwr != 0;
1431         }
1432 out:
1433         return ret;
1434 }
1435
1436 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1437                 enum open_claim_type4 claim)
1438 {
1439         if (delegation == NULL)
1440                 return 0;
1441         if ((delegation->type & fmode) != fmode)
1442                 return 0;
1443         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1444                 return 0;
1445         switch (claim) {
1446         case NFS4_OPEN_CLAIM_NULL:
1447         case NFS4_OPEN_CLAIM_FH:
1448                 break;
1449         case NFS4_OPEN_CLAIM_PREVIOUS:
1450                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1451                         break;
1452                 /* Fall through */
1453         default:
1454                 return 0;
1455         }
1456         nfs_mark_delegation_referenced(delegation);
1457         return 1;
1458 }
1459
1460 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1461 {
1462         switch (fmode) {
1463                 case FMODE_WRITE:
1464                         state->n_wronly++;
1465                         break;
1466                 case FMODE_READ:
1467                         state->n_rdonly++;
1468                         break;
1469                 case FMODE_READ|FMODE_WRITE:
1470                         state->n_rdwr++;
1471         }
1472         nfs4_state_set_mode_locked(state, state->state | fmode);
1473 }
1474
1475 #ifdef CONFIG_NFS_V4_1
1476 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1477 {
1478         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1479                 return true;
1480         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1481                 return true;
1482         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1483                 return true;
1484         return false;
1485 }
1486 #endif /* CONFIG_NFS_V4_1 */
1487
1488 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1489 {
1490         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1491                 wake_up_all(&state->waitq);
1492 }
1493
1494 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1495                 const nfs4_stateid *stateid)
1496 {
1497         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1498         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1499
1500         if (stateid_seqid == state_seqid + 1U ||
1501             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1502                 nfs_state_log_update_open_stateid(state);
1503         else
1504                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1505 }
1506
1507 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1508 {
1509         struct nfs_client *clp = state->owner->so_server->nfs_client;
1510         bool need_recover = false;
1511
1512         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1513                 need_recover = true;
1514         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1515                 need_recover = true;
1516         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1517                 need_recover = true;
1518         if (need_recover)
1519                 nfs4_state_mark_reclaim_nograce(clp, state);
1520 }
1521
1522 /*
1523  * Check for whether or not the caller may update the open stateid
1524  * to the value passed in by stateid.
1525  *
1526  * Note: This function relies heavily on the server implementing
1527  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1528  * correctly.
1529  * i.e. The stateid seqids have to be initialised to 1, and
1530  * are then incremented on every state transition.
1531  */
1532 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1533                 const nfs4_stateid *stateid)
1534 {
1535         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1536             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1537                 if (stateid->seqid == cpu_to_be32(1))
1538                         nfs_state_log_update_open_stateid(state);
1539                 else
1540                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1541                 return true;
1542         }
1543
1544         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1545                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1546                 return true;
1547         }
1548         return false;
1549 }
1550
1551 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1552 {
1553         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1554                 return;
1555         if (state->n_wronly)
1556                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1557         if (state->n_rdonly)
1558                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1559         if (state->n_rdwr)
1560                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1561         set_bit(NFS_OPEN_STATE, &state->flags);
1562 }
1563
1564 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1565                 nfs4_stateid *stateid, fmode_t fmode)
1566 {
1567         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1568         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1569         case FMODE_WRITE:
1570                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1571                 break;
1572         case FMODE_READ:
1573                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1574                 break;
1575         case 0:
1576                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1577                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1578                 clear_bit(NFS_OPEN_STATE, &state->flags);
1579         }
1580         if (stateid == NULL)
1581                 return;
1582         /* Handle OPEN+OPEN_DOWNGRADE races */
1583         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1584             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1585                 nfs_resync_open_stateid_locked(state);
1586                 goto out;
1587         }
1588         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1589                 nfs4_stateid_copy(&state->stateid, stateid);
1590         nfs4_stateid_copy(&state->open_stateid, stateid);
1591         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1592 out:
1593         nfs_state_log_update_open_stateid(state);
1594 }
1595
1596 static void nfs_clear_open_stateid(struct nfs4_state *state,
1597         nfs4_stateid *arg_stateid,
1598         nfs4_stateid *stateid, fmode_t fmode)
1599 {
1600         write_seqlock(&state->seqlock);
1601         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1602         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1603                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1604         write_sequnlock(&state->seqlock);
1605         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1606                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1607 }
1608
1609 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1610                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1611         __must_hold(&state->owner->so_lock)
1612         __must_hold(&state->seqlock)
1613         __must_hold(RCU)
1614
1615 {
1616         DEFINE_WAIT(wait);
1617         int status = 0;
1618         for (;;) {
1619
1620                 if (!nfs_need_update_open_stateid(state, stateid))
1621                         return;
1622                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1623                         break;
1624                 if (status)
1625                         break;
1626                 /* Rely on seqids for serialisation with NFSv4.0 */
1627                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1628                         break;
1629
1630                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1631                 /*
1632                  * Ensure we process the state changes in the same order
1633                  * in which the server processed them by delaying the
1634                  * update of the stateid until we are in sequence.
1635                  */
1636                 write_sequnlock(&state->seqlock);
1637                 spin_unlock(&state->owner->so_lock);
1638                 rcu_read_unlock();
1639                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1640                 if (!signal_pending(current)) {
1641                         if (schedule_timeout(5*HZ) == 0)
1642                                 status = -EAGAIN;
1643                         else
1644                                 status = 0;
1645                 } else
1646                         status = -EINTR;
1647                 finish_wait(&state->waitq, &wait);
1648                 rcu_read_lock();
1649                 spin_lock(&state->owner->so_lock);
1650                 write_seqlock(&state->seqlock);
1651         }
1652
1653         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1654             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1655                 nfs4_stateid_copy(freeme, &state->open_stateid);
1656                 nfs_test_and_clear_all_open_stateid(state);
1657         }
1658
1659         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1660                 nfs4_stateid_copy(&state->stateid, stateid);
1661         nfs4_stateid_copy(&state->open_stateid, stateid);
1662         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1663         nfs_state_log_update_open_stateid(state);
1664 }
1665
1666 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1667                 const nfs4_stateid *open_stateid,
1668                 fmode_t fmode,
1669                 nfs4_stateid *freeme)
1670 {
1671         /*
1672          * Protect the call to nfs4_state_set_mode_locked and
1673          * serialise the stateid update
1674          */
1675         write_seqlock(&state->seqlock);
1676         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1677         switch (fmode) {
1678         case FMODE_READ:
1679                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1680                 break;
1681         case FMODE_WRITE:
1682                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1683                 break;
1684         case FMODE_READ|FMODE_WRITE:
1685                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1686         }
1687         set_bit(NFS_OPEN_STATE, &state->flags);
1688         write_sequnlock(&state->seqlock);
1689 }
1690
1691 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1692 {
1693         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1694         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1695         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1696         clear_bit(NFS_OPEN_STATE, &state->flags);
1697 }
1698
1699 static void nfs_state_set_delegation(struct nfs4_state *state,
1700                 const nfs4_stateid *deleg_stateid,
1701                 fmode_t fmode)
1702 {
1703         /*
1704          * Protect the call to nfs4_state_set_mode_locked and
1705          * serialise the stateid update
1706          */
1707         write_seqlock(&state->seqlock);
1708         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1709         set_bit(NFS_DELEGATED_STATE, &state->flags);
1710         write_sequnlock(&state->seqlock);
1711 }
1712
1713 static void nfs_state_clear_delegation(struct nfs4_state *state)
1714 {
1715         write_seqlock(&state->seqlock);
1716         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1717         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1718         write_sequnlock(&state->seqlock);
1719 }
1720
1721 static int update_open_stateid(struct nfs4_state *state,
1722                 const nfs4_stateid *open_stateid,
1723                 const nfs4_stateid *delegation,
1724                 fmode_t fmode)
1725 {
1726         struct nfs_server *server = NFS_SERVER(state->inode);
1727         struct nfs_client *clp = server->nfs_client;
1728         struct nfs_inode *nfsi = NFS_I(state->inode);
1729         struct nfs_delegation *deleg_cur;
1730         nfs4_stateid freeme = { };
1731         int ret = 0;
1732
1733         fmode &= (FMODE_READ|FMODE_WRITE);
1734
1735         rcu_read_lock();
1736         spin_lock(&state->owner->so_lock);
1737         if (open_stateid != NULL) {
1738                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1739                 ret = 1;
1740         }
1741
1742         deleg_cur = rcu_dereference(nfsi->delegation);
1743         if (deleg_cur == NULL)
1744                 goto no_delegation;
1745
1746         spin_lock(&deleg_cur->lock);
1747         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1748            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1749             (deleg_cur->type & fmode) != fmode)
1750                 goto no_delegation_unlock;
1751
1752         if (delegation == NULL)
1753                 delegation = &deleg_cur->stateid;
1754         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1755                 goto no_delegation_unlock;
1756
1757         nfs_mark_delegation_referenced(deleg_cur);
1758         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1759         ret = 1;
1760 no_delegation_unlock:
1761         spin_unlock(&deleg_cur->lock);
1762 no_delegation:
1763         if (ret)
1764                 update_open_stateflags(state, fmode);
1765         spin_unlock(&state->owner->so_lock);
1766         rcu_read_unlock();
1767
1768         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1769                 nfs4_schedule_state_manager(clp);
1770         if (freeme.type != 0)
1771                 nfs4_test_and_free_stateid(server, &freeme,
1772                                 state->owner->so_cred);
1773
1774         return ret;
1775 }
1776
1777 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1778                 const nfs4_stateid *stateid)
1779 {
1780         struct nfs4_state *state = lsp->ls_state;
1781         bool ret = false;
1782
1783         spin_lock(&state->state_lock);
1784         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1785                 goto out_noupdate;
1786         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1787                 goto out_noupdate;
1788         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1789         ret = true;
1790 out_noupdate:
1791         spin_unlock(&state->state_lock);
1792         return ret;
1793 }
1794
1795 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1796 {
1797         struct nfs_delegation *delegation;
1798
1799         fmode &= FMODE_READ|FMODE_WRITE;
1800         rcu_read_lock();
1801         delegation = rcu_dereference(NFS_I(inode)->delegation);
1802         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1803                 rcu_read_unlock();
1804                 return;
1805         }
1806         rcu_read_unlock();
1807         nfs4_inode_return_delegation(inode);
1808 }
1809
1810 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1811 {
1812         struct nfs4_state *state = opendata->state;
1813         struct nfs_inode *nfsi = NFS_I(state->inode);
1814         struct nfs_delegation *delegation;
1815         int open_mode = opendata->o_arg.open_flags;
1816         fmode_t fmode = opendata->o_arg.fmode;
1817         enum open_claim_type4 claim = opendata->o_arg.claim;
1818         nfs4_stateid stateid;
1819         int ret = -EAGAIN;
1820
1821         for (;;) {
1822                 spin_lock(&state->owner->so_lock);
1823                 if (can_open_cached(state, fmode, open_mode, claim)) {
1824                         update_open_stateflags(state, fmode);
1825                         spin_unlock(&state->owner->so_lock);
1826                         goto out_return_state;
1827                 }
1828                 spin_unlock(&state->owner->so_lock);
1829                 rcu_read_lock();
1830                 delegation = rcu_dereference(nfsi->delegation);
1831                 if (!can_open_delegated(delegation, fmode, claim)) {
1832                         rcu_read_unlock();
1833                         break;
1834                 }
1835                 /* Save the delegation */
1836                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1837                 rcu_read_unlock();
1838                 nfs_release_seqid(opendata->o_arg.seqid);
1839                 if (!opendata->is_recover) {
1840                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1841                         if (ret != 0)
1842                                 goto out;
1843                 }
1844                 ret = -EAGAIN;
1845
1846                 /* Try to update the stateid using the delegation */
1847                 if (update_open_stateid(state, NULL, &stateid, fmode))
1848                         goto out_return_state;
1849         }
1850 out:
1851         return ERR_PTR(ret);
1852 out_return_state:
1853         refcount_inc(&state->count);
1854         return state;
1855 }
1856
1857 static void
1858 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1859 {
1860         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1861         struct nfs_delegation *delegation;
1862         int delegation_flags = 0;
1863
1864         rcu_read_lock();
1865         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1866         if (delegation)
1867                 delegation_flags = delegation->flags;
1868         rcu_read_unlock();
1869         switch (data->o_arg.claim) {
1870         default:
1871                 break;
1872         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1873         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1874                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1875                                    "returning a delegation for "
1876                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1877                                    clp->cl_hostname);
1878                 return;
1879         }
1880         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1881                 nfs_inode_set_delegation(state->inode,
1882                                 data->owner->so_cred,
1883                                 data->o_res.delegation_type,
1884                                 &data->o_res.delegation,
1885                                 data->o_res.pagemod_limit);
1886         else
1887                 nfs_inode_reclaim_delegation(state->inode,
1888                                 data->owner->so_cred,
1889                                 data->o_res.delegation_type,
1890                                 &data->o_res.delegation,
1891                                 data->o_res.pagemod_limit);
1892
1893         if (data->o_res.do_recall)
1894                 nfs_async_inode_return_delegation(state->inode,
1895                                                   &data->o_res.delegation);
1896 }
1897
1898 /*
1899  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1900  * and update the nfs4_state.
1901  */
1902 static struct nfs4_state *
1903 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1904 {
1905         struct inode *inode = data->state->inode;
1906         struct nfs4_state *state = data->state;
1907         int ret;
1908
1909         if (!data->rpc_done) {
1910                 if (data->rpc_status)
1911                         return ERR_PTR(data->rpc_status);
1912                 /* cached opens have already been processed */
1913                 goto update;
1914         }
1915
1916         ret = nfs_refresh_inode(inode, &data->f_attr);
1917         if (ret)
1918                 return ERR_PTR(ret);
1919
1920         if (data->o_res.delegation_type != 0)
1921                 nfs4_opendata_check_deleg(data, state);
1922 update:
1923         if (!update_open_stateid(state, &data->o_res.stateid,
1924                                 NULL, data->o_arg.fmode))
1925                 return ERR_PTR(-EAGAIN);
1926         refcount_inc(&state->count);
1927
1928         return state;
1929 }
1930
1931 static struct inode *
1932 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1933 {
1934         struct inode *inode;
1935
1936         switch (data->o_arg.claim) {
1937         case NFS4_OPEN_CLAIM_NULL:
1938         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1939         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1940                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1941                         return ERR_PTR(-EAGAIN);
1942                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1943                                 &data->f_attr, data->f_label);
1944                 break;
1945         default:
1946                 inode = d_inode(data->dentry);
1947                 ihold(inode);
1948                 nfs_refresh_inode(inode, &data->f_attr);
1949         }
1950         return inode;
1951 }
1952
1953 static struct nfs4_state *
1954 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1955 {
1956         struct nfs4_state *state;
1957         struct inode *inode;
1958
1959         inode = nfs4_opendata_get_inode(data);
1960         if (IS_ERR(inode))
1961                 return ERR_CAST(inode);
1962         if (data->state != NULL && data->state->inode == inode) {
1963                 state = data->state;
1964                 refcount_inc(&state->count);
1965         } else
1966                 state = nfs4_get_open_state(inode, data->owner);
1967         iput(inode);
1968         if (state == NULL)
1969                 state = ERR_PTR(-ENOMEM);
1970         return state;
1971 }
1972
1973 static struct nfs4_state *
1974 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1975 {
1976         struct nfs4_state *state;
1977
1978         if (!data->rpc_done) {
1979                 state = nfs4_try_open_cached(data);
1980                 trace_nfs4_cached_open(data->state);
1981                 goto out;
1982         }
1983
1984         state = nfs4_opendata_find_nfs4_state(data);
1985         if (IS_ERR(state))
1986                 goto out;
1987
1988         if (data->o_res.delegation_type != 0)
1989                 nfs4_opendata_check_deleg(data, state);
1990         if (!update_open_stateid(state, &data->o_res.stateid,
1991                                 NULL, data->o_arg.fmode)) {
1992                 nfs4_put_open_state(state);
1993                 state = ERR_PTR(-EAGAIN);
1994         }
1995 out:
1996         nfs_release_seqid(data->o_arg.seqid);
1997         return state;
1998 }
1999
2000 static struct nfs4_state *
2001 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2002 {
2003         struct nfs4_state *ret;
2004
2005         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2006                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2007         else
2008                 ret = _nfs4_opendata_to_nfs4_state(data);
2009         nfs4_sequence_free_slot(&data->o_res.seq_res);
2010         return ret;
2011 }
2012
2013 static struct nfs_open_context *
2014 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2015 {
2016         struct nfs_inode *nfsi = NFS_I(state->inode);
2017         struct nfs_open_context *ctx;
2018
2019         rcu_read_lock();
2020         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2021                 if (ctx->state != state)
2022                         continue;
2023                 if ((ctx->mode & mode) != mode)
2024                         continue;
2025                 if (!get_nfs_open_context(ctx))
2026                         continue;
2027                 rcu_read_unlock();
2028                 return ctx;
2029         }
2030         rcu_read_unlock();
2031         return ERR_PTR(-ENOENT);
2032 }
2033
2034 static struct nfs_open_context *
2035 nfs4_state_find_open_context(struct nfs4_state *state)
2036 {
2037         struct nfs_open_context *ctx;
2038
2039         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2040         if (!IS_ERR(ctx))
2041                 return ctx;
2042         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2043         if (!IS_ERR(ctx))
2044                 return ctx;
2045         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2046 }
2047
2048 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2049                 struct nfs4_state *state, enum open_claim_type4 claim)
2050 {
2051         struct nfs4_opendata *opendata;
2052
2053         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2054                         NULL, claim, GFP_NOFS);
2055         if (opendata == NULL)
2056                 return ERR_PTR(-ENOMEM);
2057         opendata->state = state;
2058         refcount_inc(&state->count);
2059         return opendata;
2060 }
2061
2062 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2063                 fmode_t fmode)
2064 {
2065         struct nfs4_state *newstate;
2066         int ret;
2067
2068         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2069                 return 0;
2070         opendata->o_arg.open_flags = 0;
2071         opendata->o_arg.fmode = fmode;
2072         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2073                         NFS_SB(opendata->dentry->d_sb),
2074                         fmode, 0);
2075         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2076         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2077         nfs4_init_opendata_res(opendata);
2078         ret = _nfs4_recover_proc_open(opendata);
2079         if (ret != 0)
2080                 return ret;
2081         newstate = nfs4_opendata_to_nfs4_state(opendata);
2082         if (IS_ERR(newstate))
2083                 return PTR_ERR(newstate);
2084         if (newstate != opendata->state)
2085                 ret = -ESTALE;
2086         nfs4_close_state(newstate, fmode);
2087         return ret;
2088 }
2089
2090 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2091 {
2092         int ret;
2093
2094         /* memory barrier prior to reading state->n_* */
2095         smp_rmb();
2096         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2097         if (ret != 0)
2098                 return ret;
2099         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2100         if (ret != 0)
2101                 return ret;
2102         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2103         if (ret != 0)
2104                 return ret;
2105         /*
2106          * We may have performed cached opens for all three recoveries.
2107          * Check if we need to update the current stateid.
2108          */
2109         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2110             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2111                 write_seqlock(&state->seqlock);
2112                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2113                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2114                 write_sequnlock(&state->seqlock);
2115         }
2116         return 0;
2117 }
2118
2119 /*
2120  * OPEN_RECLAIM:
2121  *      reclaim state on the server after a reboot.
2122  */
2123 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2124 {
2125         struct nfs_delegation *delegation;
2126         struct nfs4_opendata *opendata;
2127         fmode_t delegation_type = 0;
2128         int status;
2129
2130         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2131                         NFS4_OPEN_CLAIM_PREVIOUS);
2132         if (IS_ERR(opendata))
2133                 return PTR_ERR(opendata);
2134         rcu_read_lock();
2135         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2136         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2137                 delegation_type = delegation->type;
2138         rcu_read_unlock();
2139         opendata->o_arg.u.delegation_type = delegation_type;
2140         status = nfs4_open_recover(opendata, state);
2141         nfs4_opendata_put(opendata);
2142         return status;
2143 }
2144
2145 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2146 {
2147         struct nfs_server *server = NFS_SERVER(state->inode);
2148         struct nfs4_exception exception = { };
2149         int err;
2150         do {
2151                 err = _nfs4_do_open_reclaim(ctx, state);
2152                 trace_nfs4_open_reclaim(ctx, 0, err);
2153                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2154                         continue;
2155                 if (err != -NFS4ERR_DELAY)
2156                         break;
2157                 nfs4_handle_exception(server, err, &exception);
2158         } while (exception.retry);
2159         return err;
2160 }
2161
2162 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2163 {
2164         struct nfs_open_context *ctx;
2165         int ret;
2166
2167         ctx = nfs4_state_find_open_context(state);
2168         if (IS_ERR(ctx))
2169                 return -EAGAIN;
2170         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2171         nfs_state_clear_open_state_flags(state);
2172         ret = nfs4_do_open_reclaim(ctx, state);
2173         put_nfs_open_context(ctx);
2174         return ret;
2175 }
2176
2177 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)
2178 {
2179         switch (err) {
2180                 default:
2181                         printk(KERN_ERR "NFS: %s: unhandled error "
2182                                         "%d.\n", __func__, err);
2183                 case 0:
2184                 case -ENOENT:
2185                 case -EAGAIN:
2186                 case -ESTALE:
2187                 case -ETIMEDOUT:
2188                         break;
2189                 case -NFS4ERR_BADSESSION:
2190                 case -NFS4ERR_BADSLOT:
2191                 case -NFS4ERR_BAD_HIGH_SLOT:
2192                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2193                 case -NFS4ERR_DEADSESSION:
2194                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
2195                         return -EAGAIN;
2196                 case -NFS4ERR_STALE_CLIENTID:
2197                 case -NFS4ERR_STALE_STATEID:
2198                         /* Don't recall a delegation if it was lost */
2199                         nfs4_schedule_lease_recovery(server->nfs_client);
2200                         return -EAGAIN;
2201                 case -NFS4ERR_MOVED:
2202                         nfs4_schedule_migration_recovery(server);
2203                         return -EAGAIN;
2204                 case -NFS4ERR_LEASE_MOVED:
2205                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2206                         return -EAGAIN;
2207                 case -NFS4ERR_DELEG_REVOKED:
2208                 case -NFS4ERR_ADMIN_REVOKED:
2209                 case -NFS4ERR_EXPIRED:
2210                 case -NFS4ERR_BAD_STATEID:
2211                 case -NFS4ERR_OPENMODE:
2212                         nfs_inode_find_state_and_recover(state->inode,
2213                                         stateid);
2214                         nfs4_schedule_stateid_recovery(server, state);
2215                         return -EAGAIN;
2216                 case -NFS4ERR_DELAY:
2217                 case -NFS4ERR_GRACE:
2218                         ssleep(1);
2219                         return -EAGAIN;
2220                 case -ENOMEM:
2221                 case -NFS4ERR_DENIED:
2222                         if (fl) {
2223                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2224                                 if (lsp)
2225                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2226                         }
2227                         return 0;
2228         }
2229         return err;
2230 }
2231
2232 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2233                 struct nfs4_state *state, const nfs4_stateid *stateid)
2234 {
2235         struct nfs_server *server = NFS_SERVER(state->inode);
2236         struct nfs4_opendata *opendata;
2237         int err = 0;
2238
2239         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2240                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2241         if (IS_ERR(opendata))
2242                 return PTR_ERR(opendata);
2243         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2244         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2245                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2246                 if (err)
2247                         goto out;
2248         }
2249         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2250                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2251                 if (err)
2252                         goto out;
2253         }
2254         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2255                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2256                 if (err)
2257                         goto out;
2258         }
2259         nfs_state_clear_delegation(state);
2260 out:
2261         nfs4_opendata_put(opendata);
2262         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2263 }
2264
2265 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2266 {
2267         struct nfs4_opendata *data = calldata;
2268
2269         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2270                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2271 }
2272
2273 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2274 {
2275         struct nfs4_opendata *data = calldata;
2276
2277         nfs40_sequence_done(task, &data->c_res.seq_res);
2278
2279         data->rpc_status = task->tk_status;
2280         if (data->rpc_status == 0) {
2281                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2282                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2283                 renew_lease(data->o_res.server, data->timestamp);
2284                 data->rpc_done = true;
2285         }
2286 }
2287
2288 static void nfs4_open_confirm_release(void *calldata)
2289 {
2290         struct nfs4_opendata *data = calldata;
2291         struct nfs4_state *state = NULL;
2292
2293         /* If this request hasn't been cancelled, do nothing */
2294         if (!data->cancelled)
2295                 goto out_free;
2296         /* In case of error, no cleanup! */
2297         if (!data->rpc_done)
2298                 goto out_free;
2299         state = nfs4_opendata_to_nfs4_state(data);
2300         if (!IS_ERR(state))
2301                 nfs4_close_state(state, data->o_arg.fmode);
2302 out_free:
2303         nfs4_opendata_put(data);
2304 }
2305
2306 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2307         .rpc_call_prepare = nfs4_open_confirm_prepare,
2308         .rpc_call_done = nfs4_open_confirm_done,
2309         .rpc_release = nfs4_open_confirm_release,
2310 };
2311
2312 /*
2313  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2314  */
2315 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2316 {
2317         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2318         struct rpc_task *task;
2319         struct  rpc_message msg = {
2320                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2321                 .rpc_argp = &data->c_arg,
2322                 .rpc_resp = &data->c_res,
2323                 .rpc_cred = data->owner->so_cred,
2324         };
2325         struct rpc_task_setup task_setup_data = {
2326                 .rpc_client = server->client,
2327                 .rpc_message = &msg,
2328                 .callback_ops = &nfs4_open_confirm_ops,
2329                 .callback_data = data,
2330                 .workqueue = nfsiod_workqueue,
2331                 .flags = RPC_TASK_ASYNC,
2332         };
2333         int status;
2334
2335         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2336                                 data->is_recover);
2337         kref_get(&data->kref);
2338         data->rpc_done = false;
2339         data->rpc_status = 0;
2340         data->timestamp = jiffies;
2341         task = rpc_run_task(&task_setup_data);
2342         if (IS_ERR(task))
2343                 return PTR_ERR(task);
2344         status = rpc_wait_for_completion_task(task);
2345         if (status != 0) {
2346                 data->cancelled = true;
2347                 smp_wmb();
2348         } else
2349                 status = data->rpc_status;
2350         rpc_put_task(task);
2351         return status;
2352 }
2353
2354 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2355 {
2356         struct nfs4_opendata *data = calldata;
2357         struct nfs4_state_owner *sp = data->owner;
2358         struct nfs_client *clp = sp->so_server->nfs_client;
2359         enum open_claim_type4 claim = data->o_arg.claim;
2360
2361         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2362                 goto out_wait;
2363         /*
2364          * Check if we still need to send an OPEN call, or if we can use
2365          * a delegation instead.
2366          */
2367         if (data->state != NULL) {
2368                 struct nfs_delegation *delegation;
2369
2370                 if (can_open_cached(data->state, data->o_arg.fmode,
2371                                         data->o_arg.open_flags, claim))
2372                         goto out_no_action;
2373                 rcu_read_lock();
2374                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2375                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2376                         goto unlock_no_action;
2377                 rcu_read_unlock();
2378         }
2379         /* Update client id. */
2380         data->o_arg.clientid = clp->cl_clientid;
2381         switch (claim) {
2382         default:
2383                 break;
2384         case NFS4_OPEN_CLAIM_PREVIOUS:
2385         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2386         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2387                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2388                 /* Fall through */
2389         case NFS4_OPEN_CLAIM_FH:
2390                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2391         }
2392         data->timestamp = jiffies;
2393         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2394                                 &data->o_arg.seq_args,
2395                                 &data->o_res.seq_res,
2396                                 task) != 0)
2397                 nfs_release_seqid(data->o_arg.seqid);
2398
2399         /* Set the create mode (note dependency on the session type) */
2400         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2401         if (data->o_arg.open_flags & O_EXCL) {
2402                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2403                 if (nfs4_has_persistent_session(clp))
2404                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2405                 else if (clp->cl_mvops->minor_version > 0)
2406                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2407         }
2408         return;
2409 unlock_no_action:
2410         trace_nfs4_cached_open(data->state);
2411         rcu_read_unlock();
2412 out_no_action:
2413         task->tk_action = NULL;
2414 out_wait:
2415         nfs4_sequence_done(task, &data->o_res.seq_res);
2416 }
2417
2418 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2419 {
2420         struct nfs4_opendata *data = calldata;
2421
2422         data->rpc_status = task->tk_status;
2423
2424         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2425                 return;
2426
2427         if (task->tk_status == 0) {
2428                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2429                         switch (data->o_res.f_attr->mode & S_IFMT) {
2430                         case S_IFREG:
2431                                 break;
2432                         case S_IFLNK:
2433                                 data->rpc_status = -ELOOP;
2434                                 break;
2435                         case S_IFDIR:
2436                                 data->rpc_status = -EISDIR;
2437                                 break;
2438                         default:
2439                                 data->rpc_status = -ENOTDIR;
2440                         }
2441                 }
2442                 renew_lease(data->o_res.server, data->timestamp);
2443                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2444                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2445         }
2446         data->rpc_done = true;
2447 }
2448
2449 static void nfs4_open_release(void *calldata)
2450 {
2451         struct nfs4_opendata *data = calldata;
2452         struct nfs4_state *state = NULL;
2453
2454         /* If this request hasn't been cancelled, do nothing */
2455         if (!data->cancelled)
2456                 goto out_free;
2457         /* In case of error, no cleanup! */
2458         if (data->rpc_status != 0 || !data->rpc_done)
2459                 goto out_free;
2460         /* In case we need an open_confirm, no cleanup! */
2461         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2462                 goto out_free;
2463         state = nfs4_opendata_to_nfs4_state(data);
2464         if (!IS_ERR(state))
2465                 nfs4_close_state(state, data->o_arg.fmode);
2466 out_free:
2467         nfs4_opendata_put(data);
2468 }
2469
2470 static const struct rpc_call_ops nfs4_open_ops = {
2471         .rpc_call_prepare = nfs4_open_prepare,
2472         .rpc_call_done = nfs4_open_done,
2473         .rpc_release = nfs4_open_release,
2474 };
2475
2476 static int nfs4_run_open_task(struct nfs4_opendata *data,
2477                               struct nfs_open_context *ctx)
2478 {
2479         struct inode *dir = d_inode(data->dir);
2480         struct nfs_server *server = NFS_SERVER(dir);
2481         struct nfs_openargs *o_arg = &data->o_arg;
2482         struct nfs_openres *o_res = &data->o_res;
2483         struct rpc_task *task;
2484         struct rpc_message msg = {
2485                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2486                 .rpc_argp = o_arg,
2487                 .rpc_resp = o_res,
2488                 .rpc_cred = data->owner->so_cred,
2489         };
2490         struct rpc_task_setup task_setup_data = {
2491                 .rpc_client = server->client,
2492                 .rpc_message = &msg,
2493                 .callback_ops = &nfs4_open_ops,
2494                 .callback_data = data,
2495                 .workqueue = nfsiod_workqueue,
2496                 .flags = RPC_TASK_ASYNC,
2497         };
2498         int status;
2499
2500         kref_get(&data->kref);
2501         data->rpc_done = false;
2502         data->rpc_status = 0;
2503         data->cancelled = false;
2504         data->is_recover = false;
2505         if (!ctx) {
2506                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2507                 data->is_recover = true;
2508                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2509         } else {
2510                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2511                 pnfs_lgopen_prepare(data, ctx);
2512         }
2513         task = rpc_run_task(&task_setup_data);
2514         if (IS_ERR(task))
2515                 return PTR_ERR(task);
2516         status = rpc_wait_for_completion_task(task);
2517         if (status != 0) {
2518                 data->cancelled = true;
2519                 smp_wmb();
2520         } else
2521                 status = data->rpc_status;
2522         rpc_put_task(task);
2523
2524         return status;
2525 }
2526
2527 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2528 {
2529         struct inode *dir = d_inode(data->dir);
2530         struct nfs_openres *o_res = &data->o_res;
2531         int status;
2532
2533         status = nfs4_run_open_task(data, NULL);
2534         if (status != 0 || !data->rpc_done)
2535                 return status;
2536
2537         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2538
2539         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2540                 status = _nfs4_proc_open_confirm(data);
2541
2542         return status;
2543 }
2544
2545 /*
2546  * Additional permission checks in order to distinguish between an
2547  * open for read, and an open for execute. This works around the
2548  * fact that NFSv4 OPEN treats read and execute permissions as being
2549  * the same.
2550  * Note that in the non-execute case, we want to turn off permission
2551  * checking if we just created a new file (POSIX open() semantics).
2552  */
2553 static int nfs4_opendata_access(const struct cred *cred,
2554                                 struct nfs4_opendata *opendata,
2555                                 struct nfs4_state *state, fmode_t fmode,
2556                                 int openflags)
2557 {
2558         struct nfs_access_entry cache;
2559         u32 mask, flags;
2560
2561         /* access call failed or for some reason the server doesn't
2562          * support any access modes -- defer access call until later */
2563         if (opendata->o_res.access_supported == 0)
2564                 return 0;
2565
2566         mask = 0;
2567         /*
2568          * Use openflags to check for exec, because fmode won't
2569          * always have FMODE_EXEC set when file open for exec.
2570          */
2571         if (openflags & __FMODE_EXEC) {
2572                 /* ONLY check for exec rights */
2573                 if (S_ISDIR(state->inode->i_mode))
2574                         mask = NFS4_ACCESS_LOOKUP;
2575                 else
2576                         mask = NFS4_ACCESS_EXECUTE;
2577         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2578                 mask = NFS4_ACCESS_READ;
2579
2580         cache.cred = cred;
2581         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2582         nfs_access_add_cache(state->inode, &cache);
2583
2584         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2585         if ((mask & ~cache.mask & flags) == 0)
2586                 return 0;
2587
2588         return -EACCES;
2589 }
2590
2591 /*
2592  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2593  */
2594 static int _nfs4_proc_open(struct nfs4_opendata *data,
2595                            struct nfs_open_context *ctx)
2596 {
2597         struct inode *dir = d_inode(data->dir);
2598         struct nfs_server *server = NFS_SERVER(dir);
2599         struct nfs_openargs *o_arg = &data->o_arg;
2600         struct nfs_openres *o_res = &data->o_res;
2601         int status;
2602
2603         status = nfs4_run_open_task(data, ctx);
2604         if (!data->rpc_done)
2605                 return status;
2606         if (status != 0) {
2607                 if (status == -NFS4ERR_BADNAME &&
2608                                 !(o_arg->open_flags & O_CREAT))
2609                         return -ENOENT;
2610                 return status;
2611         }
2612
2613         nfs_fattr_map_and_free_names(server, &data->f_attr);
2614
2615         if (o_arg->open_flags & O_CREAT) {
2616                 if (o_arg->open_flags & O_EXCL)
2617                         data->file_created = true;
2618                 else if (o_res->cinfo.before != o_res->cinfo.after)
2619                         data->file_created = true;
2620                 if (data->file_created ||
2621                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2622                         update_changeattr(dir, &o_res->cinfo,
2623                                         o_res->f_attr->time_start, 0);
2624         }
2625         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2626                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2627         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2628                 status = _nfs4_proc_open_confirm(data);
2629                 if (status != 0)
2630                         return status;
2631         }
2632         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2633                 nfs4_sequence_free_slot(&o_res->seq_res);
2634                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2635                                 o_res->f_label, NULL);
2636         }
2637         return 0;
2638 }
2639
2640 /*
2641  * OPEN_EXPIRED:
2642  *      reclaim state on the server after a network partition.
2643  *      Assumes caller holds the appropriate lock
2644  */
2645 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2646 {
2647         struct nfs4_opendata *opendata;
2648         int ret;
2649
2650         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2651                         NFS4_OPEN_CLAIM_FH);
2652         if (IS_ERR(opendata))
2653                 return PTR_ERR(opendata);
2654         ret = nfs4_open_recover(opendata, state);
2655         if (ret == -ESTALE)
2656                 d_drop(ctx->dentry);
2657         nfs4_opendata_put(opendata);
2658         return ret;
2659 }
2660
2661 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2662 {
2663         struct nfs_server *server = NFS_SERVER(state->inode);
2664         struct nfs4_exception exception = { };
2665         int err;
2666
2667         do {
2668                 err = _nfs4_open_expired(ctx, state);
2669                 trace_nfs4_open_expired(ctx, 0, err);
2670                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2671                         continue;
2672                 switch (err) {
2673                 default:
2674                         goto out;
2675                 case -NFS4ERR_GRACE:
2676                 case -NFS4ERR_DELAY:
2677                         nfs4_handle_exception(server, err, &exception);
2678                         err = 0;
2679                 }
2680         } while (exception.retry);
2681 out:
2682         return err;
2683 }
2684
2685 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2686 {
2687         struct nfs_open_context *ctx;
2688         int ret;
2689
2690         ctx = nfs4_state_find_open_context(state);
2691         if (IS_ERR(ctx))
2692                 return -EAGAIN;
2693         ret = nfs4_do_open_expired(ctx, state);
2694         put_nfs_open_context(ctx);
2695         return ret;
2696 }
2697
2698 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2699                 const nfs4_stateid *stateid)
2700 {
2701         nfs_remove_bad_delegation(state->inode, stateid);
2702         nfs_state_clear_delegation(state);
2703 }
2704
2705 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2706 {
2707         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2708                 nfs_finish_clear_delegation_stateid(state, NULL);
2709 }
2710
2711 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2712 {
2713         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2714         nfs40_clear_delegation_stateid(state);
2715         nfs_state_clear_open_state_flags(state);
2716         return nfs4_open_expired(sp, state);
2717 }
2718
2719 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2720                 nfs4_stateid *stateid,
2721                 const struct cred *cred)
2722 {
2723         return -NFS4ERR_BAD_STATEID;
2724 }
2725
2726 #if defined(CONFIG_NFS_V4_1)
2727 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2728                 nfs4_stateid *stateid,
2729                 const struct cred *cred)
2730 {
2731         int status;
2732
2733         switch (stateid->type) {
2734         default:
2735                 break;
2736         case NFS4_INVALID_STATEID_TYPE:
2737         case NFS4_SPECIAL_STATEID_TYPE:
2738                 return -NFS4ERR_BAD_STATEID;
2739         case NFS4_REVOKED_STATEID_TYPE:
2740                 goto out_free;
2741         }
2742
2743         status = nfs41_test_stateid(server, stateid, cred);
2744         switch (status) {
2745         case -NFS4ERR_EXPIRED:
2746         case -NFS4ERR_ADMIN_REVOKED:
2747         case -NFS4ERR_DELEG_REVOKED:
2748                 break;
2749         default:
2750                 return status;
2751         }
2752 out_free:
2753         /* Ack the revoked state to the server */
2754         nfs41_free_stateid(server, stateid, cred, true);
2755         return -NFS4ERR_EXPIRED;
2756 }
2757
2758 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2759 {
2760         struct nfs_server *server = NFS_SERVER(state->inode);
2761         nfs4_stateid stateid;
2762         struct nfs_delegation *delegation;
2763         const struct cred *cred = NULL;
2764         int status, ret = NFS_OK;
2765
2766         /* Get the delegation credential for use by test/free_stateid */
2767         rcu_read_lock();
2768         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2769         if (delegation == NULL) {
2770                 rcu_read_unlock();
2771                 nfs_state_clear_delegation(state);
2772                 return NFS_OK;
2773         }
2774
2775         nfs4_stateid_copy(&stateid, &delegation->stateid);
2776
2777         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2778                                 &delegation->flags)) {
2779                 rcu_read_unlock();
2780                 return NFS_OK;
2781         }
2782
2783         if (delegation->cred)
2784                 cred = get_cred(delegation->cred);
2785         rcu_read_unlock();
2786         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2787         trace_nfs4_test_delegation_stateid(state, NULL, status);
2788         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2789                 nfs_finish_clear_delegation_stateid(state, &stateid);
2790         else
2791                 ret = status;
2792
2793         put_cred(cred);
2794         return ret;
2795 }
2796
2797 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2798 {
2799         nfs4_stateid tmp;
2800
2801         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2802             nfs4_copy_delegation_stateid(state->inode, state->state,
2803                                 &tmp, NULL) &&
2804             nfs4_stateid_match_other(&state->stateid, &tmp))
2805                 nfs_state_set_delegation(state, &tmp, state->state);
2806         else
2807                 nfs_state_clear_delegation(state);
2808 }
2809
2810 /**
2811  * nfs41_check_expired_locks - possibly free a lock stateid
2812  *
2813  * @state: NFSv4 state for an inode
2814  *
2815  * Returns NFS_OK if recovery for this stateid is now finished.
2816  * Otherwise a negative NFS4ERR value is returned.
2817  */
2818 static int nfs41_check_expired_locks(struct nfs4_state *state)
2819 {
2820         int status, ret = NFS_OK;
2821         struct nfs4_lock_state *lsp, *prev = NULL;
2822         struct nfs_server *server = NFS_SERVER(state->inode);
2823
2824         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2825                 goto out;
2826
2827         spin_lock(&state->state_lock);
2828         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2829                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2830                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2831
2832                         refcount_inc(&lsp->ls_count);
2833                         spin_unlock(&state->state_lock);
2834
2835                         nfs4_put_lock_state(prev);
2836                         prev = lsp;
2837
2838                         status = nfs41_test_and_free_expired_stateid(server,
2839                                         &lsp->ls_stateid,
2840                                         cred);
2841                         trace_nfs4_test_lock_stateid(state, lsp, status);
2842                         if (status == -NFS4ERR_EXPIRED ||
2843                             status == -NFS4ERR_BAD_STATEID) {
2844                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2845                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2846                                 if (!recover_lost_locks)
2847                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2848                         } else if (status != NFS_OK) {
2849                                 ret = status;
2850                                 nfs4_put_lock_state(prev);
2851                                 goto out;
2852                         }
2853                         spin_lock(&state->state_lock);
2854                 }
2855         }
2856         spin_unlock(&state->state_lock);
2857         nfs4_put_lock_state(prev);
2858 out:
2859         return ret;
2860 }
2861
2862 /**
2863  * nfs41_check_open_stateid - possibly free an open stateid
2864  *
2865  * @state: NFSv4 state for an inode
2866  *
2867  * Returns NFS_OK if recovery for this stateid is now finished.
2868  * Otherwise a negative NFS4ERR value is returned.
2869  */
2870 static int nfs41_check_open_stateid(struct nfs4_state *state)
2871 {
2872         struct nfs_server *server = NFS_SERVER(state->inode);
2873         nfs4_stateid *stateid = &state->open_stateid;
2874         const struct cred *cred = state->owner->so_cred;
2875         int status;
2876
2877         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2878                 return -NFS4ERR_BAD_STATEID;
2879         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2880         trace_nfs4_test_open_stateid(state, NULL, status);
2881         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2882                 nfs_state_clear_open_state_flags(state);
2883                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2884                 return status;
2885         }
2886         if (nfs_open_stateid_recover_openmode(state))
2887                 return -NFS4ERR_OPENMODE;
2888         return NFS_OK;
2889 }
2890
2891 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2892 {
2893         int status;
2894
2895         status = nfs41_check_delegation_stateid(state);
2896         if (status != NFS_OK)
2897                 return status;
2898         nfs41_delegation_recover_stateid(state);
2899
2900         status = nfs41_check_expired_locks(state);
2901         if (status != NFS_OK)
2902                 return status;
2903         status = nfs41_check_open_stateid(state);
2904         if (status != NFS_OK)
2905                 status = nfs4_open_expired(sp, state);
2906         return status;
2907 }
2908 #endif
2909
2910 /*
2911  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2912  * fields corresponding to attributes that were used to store the verifier.
2913  * Make sure we clobber those fields in the later setattr call
2914  */
2915 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2916                                 struct iattr *sattr, struct nfs4_label **label)
2917 {
2918         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2919         __u32 attrset[3];
2920         unsigned ret;
2921         unsigned i;
2922
2923         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2924                 attrset[i] = opendata->o_res.attrset[i];
2925                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2926                         attrset[i] &= ~bitmask[i];
2927         }
2928
2929         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2930                 sattr->ia_valid : 0;
2931
2932         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2933                 if (sattr->ia_valid & ATTR_ATIME_SET)
2934                         ret |= ATTR_ATIME_SET;
2935                 else
2936                         ret |= ATTR_ATIME;
2937         }
2938
2939         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2940                 if (sattr->ia_valid & ATTR_MTIME_SET)
2941                         ret |= ATTR_MTIME_SET;
2942                 else
2943                         ret |= ATTR_MTIME;
2944         }
2945
2946         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2947                 *label = NULL;
2948         return ret;
2949 }
2950
2951 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2952                 int flags, struct nfs_open_context *ctx)
2953 {
2954         struct nfs4_state_owner *sp = opendata->owner;
2955         struct nfs_server *server = sp->so_server;
2956         struct dentry *dentry;
2957         struct nfs4_state *state;
2958         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
2959         unsigned int seq;
2960         int ret;
2961
2962         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2963
2964         ret = _nfs4_proc_open(opendata, ctx);
2965         if (ret != 0)
2966                 goto out;
2967
2968         state = _nfs4_opendata_to_nfs4_state(opendata);
2969         ret = PTR_ERR(state);
2970         if (IS_ERR(state))
2971                 goto out;
2972         ctx->state = state;
2973         if (server->caps & NFS_CAP_POSIX_LOCK)
2974                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2975         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2976                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2977
2978         dentry = opendata->dentry;
2979         if (d_really_is_negative(dentry)) {
2980                 struct dentry *alias;
2981                 d_drop(dentry);
2982                 alias = d_exact_alias(dentry, state->inode);
2983                 if (!alias)
2984                         alias = d_splice_alias(igrab(state->inode), dentry);
2985                 /* d_splice_alias() can't fail here - it's a non-directory */
2986                 if (alias) {
2987                         dput(ctx->dentry);
2988                         ctx->dentry = dentry = alias;
2989                 }
2990                 nfs_set_verifier(dentry,
2991                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2992         }
2993
2994         /* Parse layoutget results before we check for access */
2995         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
2996
2997         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
2998                         acc_mode, flags);
2999         if (ret != 0)
3000                 goto out;
3001
3002         if (d_inode(dentry) == state->inode) {
3003                 nfs_inode_attach_open_context(ctx);
3004                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3005                         nfs4_schedule_stateid_recovery(server, state);
3006         }
3007
3008 out:
3009         if (!opendata->cancelled)
3010                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3011         return ret;
3012 }
3013
3014 /*
3015  * Returns a referenced nfs4_state
3016  */
3017 static int _nfs4_do_open(struct inode *dir,
3018                         struct nfs_open_context *ctx,
3019                         int flags,
3020                         const struct nfs4_open_createattrs *c,
3021                         int *opened)
3022 {
3023         struct nfs4_state_owner  *sp;
3024         struct nfs4_state     *state = NULL;
3025         struct nfs_server       *server = NFS_SERVER(dir);
3026         struct nfs4_opendata *opendata;
3027         struct dentry *dentry = ctx->dentry;
3028         const struct cred *cred = ctx->cred;
3029         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3030         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3031         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3032         struct iattr *sattr = c->sattr;
3033         struct nfs4_label *label = c->label;
3034         struct nfs4_label *olabel = NULL;
3035         int status;
3036
3037         /* Protect against reboot recovery conflicts */
3038         status = -ENOMEM;
3039         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3040         if (sp == NULL) {
3041                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3042                 goto out_err;
3043         }
3044         status = nfs4_client_recover_expired_lease(server->nfs_client);
3045         if (status != 0)
3046                 goto err_put_state_owner;
3047         if (d_really_is_positive(dentry))
3048                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3049         status = -ENOMEM;
3050         if (d_really_is_positive(dentry))
3051                 claim = NFS4_OPEN_CLAIM_FH;
3052         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3053                         c, claim, GFP_KERNEL);
3054         if (opendata == NULL)
3055                 goto err_put_state_owner;
3056
3057         if (label) {
3058                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3059                 if (IS_ERR(olabel)) {
3060                         status = PTR_ERR(olabel);
3061                         goto err_opendata_put;
3062                 }
3063         }
3064
3065         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3066                 if (!opendata->f_attr.mdsthreshold) {
3067                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3068                         if (!opendata->f_attr.mdsthreshold)
3069                                 goto err_free_label;
3070                 }
3071                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3072         }
3073         if (d_really_is_positive(dentry))
3074                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3075
3076         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3077         if (status != 0)
3078                 goto err_free_label;
3079         state = ctx->state;
3080
3081         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3082             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3083                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3084                 /*
3085                  * send create attributes which was not set by open
3086                  * with an extra setattr.
3087                  */
3088                 if (attrs || label) {
3089                         unsigned ia_old = sattr->ia_valid;
3090
3091                         sattr->ia_valid = attrs;
3092                         nfs_fattr_init(opendata->o_res.f_attr);
3093                         status = nfs4_do_setattr(state->inode, cred,
3094                                         opendata->o_res.f_attr, sattr,
3095                                         ctx, label, olabel);
3096                         if (status == 0) {
3097                                 nfs_setattr_update_inode(state->inode, sattr,
3098                                                 opendata->o_res.f_attr);
3099                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3100                         }
3101                         sattr->ia_valid = ia_old;
3102                 }
3103         }
3104         if (opened && opendata->file_created)
3105                 *opened = 1;
3106
3107         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3108                 *ctx_th = opendata->f_attr.mdsthreshold;
3109                 opendata->f_attr.mdsthreshold = NULL;
3110         }
3111
3112         nfs4_label_free(olabel);
3113
3114         nfs4_opendata_put(opendata);
3115         nfs4_put_state_owner(sp);
3116         return 0;
3117 err_free_label:
3118         nfs4_label_free(olabel);
3119 err_opendata_put:
3120         nfs4_opendata_put(opendata);
3121 err_put_state_owner:
3122         nfs4_put_state_owner(sp);
3123 out_err:
3124         return status;
3125 }
3126
3127
3128 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3129                                         struct nfs_open_context *ctx,
3130                                         int flags,
3131                                         struct iattr *sattr,
3132                                         struct nfs4_label *label,
3133                                         int *opened)
3134 {
3135         struct nfs_server *server = NFS_SERVER(dir);
3136         struct nfs4_exception exception = {
3137                 .interruptible = true,
3138         };
3139         struct nfs4_state *res;
3140         struct nfs4_open_createattrs c = {
3141                 .label = label,
3142                 .sattr = sattr,
3143                 .verf = {
3144                         [0] = (__u32)jiffies,
3145                         [1] = (__u32)current->pid,
3146                 },
3147         };
3148         int status;
3149
3150         do {
3151                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3152                 res = ctx->state;
3153                 trace_nfs4_open_file(ctx, flags, status);
3154                 if (status == 0)
3155                         break;
3156                 /* NOTE: BAD_SEQID means the server and client disagree about the
3157                  * book-keeping w.r.t. state-changing operations
3158                  * (OPEN/CLOSE/LOCK/LOCKU...)
3159                  * It is actually a sign of a bug on the client or on the server.
3160                  *
3161                  * If we receive a BAD_SEQID error in the particular case of
3162                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3163                  * have unhashed the old state_owner for us, and that we can
3164                  * therefore safely retry using a new one. We should still warn
3165                  * the user though...
3166                  */
3167                 if (status == -NFS4ERR_BAD_SEQID) {
3168                         pr_warn_ratelimited("NFS: v4 server %s "
3169                                         " returned a bad sequence-id error!\n",
3170                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3171                         exception.retry = 1;
3172                         continue;
3173                 }
3174                 /*
3175                  * BAD_STATEID on OPEN means that the server cancelled our
3176                  * state before it received the OPEN_CONFIRM.
3177                  * Recover by retrying the request as per the discussion
3178                  * on Page 181 of RFC3530.
3179                  */
3180                 if (status == -NFS4ERR_BAD_STATEID) {
3181                         exception.retry = 1;
3182                         continue;
3183                 }
3184                 if (status == -EAGAIN) {
3185                         /* We must have found a delegation */
3186                         exception.retry = 1;
3187                         continue;
3188                 }
3189                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3190                         continue;
3191                 res = ERR_PTR(nfs4_handle_exception(server,
3192                                         status, &exception));
3193         } while (exception.retry);
3194         return res;
3195 }
3196
3197 static int _nfs4_do_setattr(struct inode *inode,
3198                             struct nfs_setattrargs *arg,
3199                             struct nfs_setattrres *res,
3200                             const struct cred *cred,
3201                             struct nfs_open_context *ctx)
3202 {
3203         struct nfs_server *server = NFS_SERVER(inode);
3204         struct rpc_message msg = {
3205                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3206                 .rpc_argp       = arg,
3207                 .rpc_resp       = res,
3208                 .rpc_cred       = cred,
3209         };
3210         const struct cred *delegation_cred = NULL;
3211         unsigned long timestamp = jiffies;
3212         bool truncate;
3213         int status;
3214
3215         nfs_fattr_init(res->fattr);
3216
3217         /* Servers should only apply open mode checks for file size changes */
3218         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3219         if (!truncate)
3220                 goto zero_stateid;
3221
3222         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3223                 /* Use that stateid */
3224         } else if (ctx != NULL && ctx->state) {
3225                 struct nfs_lock_context *l_ctx;
3226                 if (!nfs4_valid_open_stateid(ctx->state))
3227                         return -EBADF;
3228                 l_ctx = nfs_get_lock_context(ctx);
3229                 if (IS_ERR(l_ctx))
3230                         return PTR_ERR(l_ctx);
3231                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3232                                                 &arg->stateid, &delegation_cred);
3233                 nfs_put_lock_context(l_ctx);
3234                 if (status == -EIO)
3235                         return -EBADF;
3236         } else {
3237 zero_stateid:
3238                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3239         }
3240         if (delegation_cred)
3241                 msg.rpc_cred = delegation_cred;
3242
3243         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3244
3245         put_cred(delegation_cred);
3246         if (status == 0 && ctx != NULL)
3247                 renew_lease(server, timestamp);
3248         trace_nfs4_setattr(inode, &arg->stateid, status);
3249         return status;
3250 }
3251
3252 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3253                            struct nfs_fattr *fattr, struct iattr *sattr,
3254                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3255                            struct nfs4_label *olabel)
3256 {
3257         struct nfs_server *server = NFS_SERVER(inode);
3258         __u32 bitmask[NFS4_BITMASK_SZ];
3259         struct nfs4_state *state = ctx ? ctx->state : NULL;
3260         struct nfs_setattrargs  arg = {
3261                 .fh             = NFS_FH(inode),
3262                 .iap            = sattr,
3263                 .server         = server,
3264                 .bitmask = bitmask,
3265                 .label          = ilabel,
3266         };
3267         struct nfs_setattrres  res = {
3268                 .fattr          = fattr,
3269                 .label          = olabel,
3270                 .server         = server,
3271         };
3272         struct nfs4_exception exception = {
3273                 .state = state,
3274                 .inode = inode,
3275                 .stateid = &arg.stateid,
3276         };
3277         int err;
3278
3279         do {
3280                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3281                                 nfs4_bitmask(server, olabel),
3282                                 inode);
3283
3284                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3285                 switch (err) {
3286                 case -NFS4ERR_OPENMODE:
3287                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3288                                 pr_warn_once("NFSv4: server %s is incorrectly "
3289                                                 "applying open mode checks to "
3290                                                 "a SETATTR that is not "
3291                                                 "changing file size.\n",
3292                                                 server->nfs_client->cl_hostname);
3293                         }
3294                         if (state && !(state->state & FMODE_WRITE)) {
3295                                 err = -EBADF;
3296                                 if (sattr->ia_valid & ATTR_OPEN)
3297                                         err = -EACCES;
3298                                 goto out;
3299                         }
3300                 }
3301                 err = nfs4_handle_exception(server, err, &exception);
3302         } while (exception.retry);
3303 out:
3304         return err;
3305 }
3306
3307 static bool
3308 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3309 {
3310         if (inode == NULL || !nfs_have_layout(inode))
3311                 return false;
3312
3313         return pnfs_wait_on_layoutreturn(inode, task);
3314 }
3315
3316 /*
3317  * Update the seqid of an open stateid
3318  */
3319 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3320                 struct nfs4_state *state)
3321 {
3322         __be32 seqid_open;
3323         u32 dst_seqid;
3324         int seq;
3325
3326         for (;;) {
3327                 if (!nfs4_valid_open_stateid(state))
3328                         break;
3329                 seq = read_seqbegin(&state->seqlock);
3330                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3331                         nfs4_stateid_copy(dst, &state->open_stateid);
3332                         if (read_seqretry(&state->seqlock, seq))
3333                                 continue;
3334                         break;
3335                 }
3336                 seqid_open = state->open_stateid.seqid;
3337                 if (read_seqretry(&state->seqlock, seq))
3338                         continue;
3339
3340                 dst_seqid = be32_to_cpu(dst->seqid);
3341                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3342                         dst->seqid = seqid_open;
3343                 break;
3344         }
3345 }
3346
3347 /*
3348  * Update the seqid of an open stateid after receiving
3349  * NFS4ERR_OLD_STATEID
3350  */
3351 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3352                 struct nfs4_state *state)
3353 {
3354         __be32 seqid_open;
3355         u32 dst_seqid;
3356         bool ret;
3357         int seq;
3358
3359         for (;;) {
3360                 ret = false;
3361                 if (!nfs4_valid_open_stateid(state))
3362                         break;
3363                 seq = read_seqbegin(&state->seqlock);
3364                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3365                         if (read_seqretry(&state->seqlock, seq))
3366                                 continue;
3367                         break;
3368                 }
3369                 seqid_open = state->open_stateid.seqid;
3370                 if (read_seqretry(&state->seqlock, seq))
3371                         continue;
3372
3373                 dst_seqid = be32_to_cpu(dst->seqid);
3374                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) >= 0)
3375                         dst->seqid = cpu_to_be32(dst_seqid + 1);
3376                 else
3377                         dst->seqid = seqid_open;
3378                 ret = true;
3379                 break;
3380         }
3381
3382         return ret;
3383 }
3384
3385 struct nfs4_closedata {
3386         struct inode *inode;
3387         struct nfs4_state *state;
3388         struct nfs_closeargs arg;
3389         struct nfs_closeres res;
3390         struct {
3391                 struct nfs4_layoutreturn_args arg;
3392                 struct nfs4_layoutreturn_res res;
3393                 struct nfs4_xdr_opaque_data ld_private;
3394                 u32 roc_barrier;
3395                 bool roc;
3396         } lr;
3397         struct nfs_fattr fattr;
3398         unsigned long timestamp;
3399 };
3400
3401 static void nfs4_free_closedata(void *data)
3402 {
3403         struct nfs4_closedata *calldata = data;
3404         struct nfs4_state_owner *sp = calldata->state->owner;
3405         struct super_block *sb = calldata->state->inode->i_sb;
3406
3407         if (calldata->lr.roc)
3408                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3409                                 calldata->res.lr_ret);
3410         nfs4_put_open_state(calldata->state);
3411         nfs_free_seqid(calldata->arg.seqid);
3412         nfs4_put_state_owner(sp);
3413         nfs_sb_deactive(sb);
3414         kfree(calldata);
3415 }
3416
3417 static void nfs4_close_done(struct rpc_task *task, void *data)
3418 {
3419         struct nfs4_closedata *calldata = data;
3420         struct nfs4_state *state = calldata->state;
3421         struct nfs_server *server = NFS_SERVER(calldata->inode);
3422         nfs4_stateid *res_stateid = NULL;
3423         struct nfs4_exception exception = {
3424                 .state = state,
3425                 .inode = calldata->inode,
3426                 .stateid = &calldata->arg.stateid,
3427         };
3428
3429         dprintk("%s: begin!\n", __func__);
3430         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3431                 return;
3432         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3433
3434         /* Handle Layoutreturn errors */
3435         if (pnfs_roc_done(task, calldata->inode,
3436                                 &calldata->arg.lr_args,
3437                                 &calldata->res.lr_res,
3438                                 &calldata->res.lr_ret) == -EAGAIN)
3439                 goto out_restart;
3440
3441         /* hmm. we are done with the inode, and in the process of freeing
3442          * the state_owner. we keep this around to process errors
3443          */
3444         switch (task->tk_status) {
3445                 case 0:
3446                         res_stateid = &calldata->res.stateid;
3447                         renew_lease(server, calldata->timestamp);
3448                         break;
3449                 case -NFS4ERR_ACCESS:
3450                         if (calldata->arg.bitmask != NULL) {
3451                                 calldata->arg.bitmask = NULL;
3452                                 calldata->res.fattr = NULL;
3453                                 goto out_restart;
3454
3455                         }
3456                         break;
3457                 case -NFS4ERR_OLD_STATEID:
3458                         /* Did we race with OPEN? */
3459                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3460                                                 state))
3461                                 goto out_restart;
3462                         goto out_release;
3463                 case -NFS4ERR_ADMIN_REVOKED:
3464                 case -NFS4ERR_STALE_STATEID:
3465                 case -NFS4ERR_EXPIRED:
3466                         nfs4_free_revoked_stateid(server,
3467                                         &calldata->arg.stateid,
3468                                         task->tk_msg.rpc_cred);
3469                         /* Fallthrough */
3470                 case -NFS4ERR_BAD_STATEID:
3471                         if (calldata->arg.fmode == 0)
3472                                 break;
3473                         /* Fallthrough */
3474                 default:
3475                         task->tk_status = nfs4_async_handle_exception(task,
3476                                         server, task->tk_status, &exception);
3477                         if (exception.retry)
3478                                 goto out_restart;
3479         }
3480         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3481                         res_stateid, calldata->arg.fmode);
3482 out_release:
3483         task->tk_status = 0;
3484         nfs_release_seqid(calldata->arg.seqid);
3485         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3486         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3487         return;
3488 out_restart:
3489         task->tk_status = 0;
3490         rpc_restart_call_prepare(task);
3491         goto out_release;
3492 }
3493
3494 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3495 {
3496         struct nfs4_closedata *calldata = data;
3497         struct nfs4_state *state = calldata->state;
3498         struct inode *inode = calldata->inode;
3499         struct pnfs_layout_hdr *lo;
3500         bool is_rdonly, is_wronly, is_rdwr;
3501         int call_close = 0;
3502
3503         dprintk("%s: begin!\n", __func__);
3504         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3505                 goto out_wait;
3506
3507         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3508         spin_lock(&state->owner->so_lock);
3509         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3510         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3511         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3512         /* Calculate the change in open mode */
3513         calldata->arg.fmode = 0;
3514         if (state->n_rdwr == 0) {
3515                 if (state->n_rdonly == 0)
3516                         call_close |= is_rdonly;
3517                 else if (is_rdonly)
3518                         calldata->arg.fmode |= FMODE_READ;
3519                 if (state->n_wronly == 0)
3520                         call_close |= is_wronly;
3521                 else if (is_wronly)
3522                         calldata->arg.fmode |= FMODE_WRITE;
3523                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3524                         call_close |= is_rdwr;
3525         } else if (is_rdwr)
3526                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3527
3528         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3529         if (!nfs4_valid_open_stateid(state))
3530                 call_close = 0;
3531         spin_unlock(&state->owner->so_lock);
3532
3533         if (!call_close) {
3534                 /* Note: exit _without_ calling nfs4_close_done */
3535                 goto out_no_action;
3536         }
3537
3538         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3539                 nfs_release_seqid(calldata->arg.seqid);
3540                 goto out_wait;
3541         }
3542
3543         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3544         if (lo && !pnfs_layout_is_valid(lo)) {
3545                 calldata->arg.lr_args = NULL;
3546                 calldata->res.lr_res = NULL;
3547         }
3548
3549         if (calldata->arg.fmode == 0)
3550                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3551
3552         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3553                 /* Close-to-open cache consistency revalidation */
3554                 if (!nfs4_have_delegation(inode, FMODE_READ))
3555                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3556                 else
3557                         calldata->arg.bitmask = NULL;
3558         }
3559
3560         calldata->arg.share_access =
3561                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3562                                 calldata->arg.fmode, 0);
3563
3564         if (calldata->res.fattr == NULL)
3565                 calldata->arg.bitmask = NULL;
3566         else if (calldata->arg.bitmask == NULL)
3567                 calldata->res.fattr = NULL;
3568         calldata->timestamp = jiffies;
3569         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3570                                 &calldata->arg.seq_args,
3571                                 &calldata->res.seq_res,
3572                                 task) != 0)
3573                 nfs_release_seqid(calldata->arg.seqid);
3574         dprintk("%s: done!\n", __func__);
3575         return;
3576 out_no_action:
3577         task->tk_action = NULL;
3578 out_wait:
3579         nfs4_sequence_done(task, &calldata->res.seq_res);
3580 }
3581
3582 static const struct rpc_call_ops nfs4_close_ops = {
3583         .rpc_call_prepare = nfs4_close_prepare,
3584         .rpc_call_done = nfs4_close_done,
3585         .rpc_release = nfs4_free_closedata,
3586 };
3587
3588 /*
3589  * It is possible for data to be read/written from a mem-mapped file
3590  * after the sys_close call (which hits the vfs layer as a flush).
3591  * This means that we can't safely call nfsv4 close on a file until
3592  * the inode is cleared. This in turn means that we are not good
3593  * NFSv4 citizens - we do not indicate to the server to update the file's
3594  * share state even when we are done with one of the three share
3595  * stateid's in the inode.
3596  *
3597  * NOTE: Caller must be holding the sp->so_owner semaphore!
3598  */
3599 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3600 {
3601         struct nfs_server *server = NFS_SERVER(state->inode);
3602         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3603         struct nfs4_closedata *calldata;
3604         struct nfs4_state_owner *sp = state->owner;
3605         struct rpc_task *task;
3606         struct rpc_message msg = {
3607                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3608                 .rpc_cred = state->owner->so_cred,
3609         };
3610         struct rpc_task_setup task_setup_data = {
3611                 .rpc_client = server->client,
3612                 .rpc_message = &msg,
3613                 .callback_ops = &nfs4_close_ops,
3614                 .workqueue = nfsiod_workqueue,
3615                 .flags = RPC_TASK_ASYNC,
3616         };
3617         int status = -ENOMEM;
3618
3619         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3620                 &task_setup_data.rpc_client, &msg);
3621
3622         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3623         if (calldata == NULL)
3624                 goto out;
3625         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3626         calldata->inode = state->inode;
3627         calldata->state = state;
3628         calldata->arg.fh = NFS_FH(state->inode);
3629         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3630                 goto out_free_calldata;
3631         /* Serialization for the sequence id */
3632         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3633         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3634         if (IS_ERR(calldata->arg.seqid))
3635                 goto out_free_calldata;
3636         nfs_fattr_init(&calldata->fattr);
3637         calldata->arg.fmode = 0;
3638         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3639         calldata->res.fattr = &calldata->fattr;
3640         calldata->res.seqid = calldata->arg.seqid;
3641         calldata->res.server = server;
3642         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3643         calldata->lr.roc = pnfs_roc(state->inode,
3644                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3645         if (calldata->lr.roc) {
3646                 calldata->arg.lr_args = &calldata->lr.arg;
3647                 calldata->res.lr_res = &calldata->lr.res;
3648         }
3649         nfs_sb_active(calldata->inode->i_sb);
3650
3651         msg.rpc_argp = &calldata->arg;
3652         msg.rpc_resp = &calldata->res;
3653         task_setup_data.callback_data = calldata;
3654         task = rpc_run_task(&task_setup_data);
3655         if (IS_ERR(task))
3656                 return PTR_ERR(task);
3657         status = 0;
3658         if (wait)
3659                 status = rpc_wait_for_completion_task(task);
3660         rpc_put_task(task);
3661         return status;
3662 out_free_calldata:
3663         kfree(calldata);
3664 out:
3665         nfs4_put_open_state(state);
3666         nfs4_put_state_owner(sp);
3667         return status;
3668 }
3669
3670 static struct inode *
3671 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3672                 int open_flags, struct iattr *attr, int *opened)
3673 {
3674         struct nfs4_state *state;
3675         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3676
3677         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3678
3679         /* Protect against concurrent sillydeletes */
3680         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3681
3682         nfs4_label_release_security(label);
3683
3684         if (IS_ERR(state))
3685                 return ERR_CAST(state);
3686         return state->inode;
3687 }
3688
3689 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3690 {
3691         if (ctx->state == NULL)
3692                 return;
3693         if (is_sync)
3694                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3695         else
3696                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3697 }
3698
3699 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3700 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3701 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3702
3703 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3704 {
3705         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3706         struct nfs4_server_caps_arg args = {
3707                 .fhandle = fhandle,
3708                 .bitmask = bitmask,
3709         };
3710         struct nfs4_server_caps_res res = {};
3711         struct rpc_message msg = {
3712                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3713                 .rpc_argp = &args,
3714                 .rpc_resp = &res,
3715         };
3716         int status;
3717         int i;
3718
3719         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3720                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3721                      FATTR4_WORD0_LINK_SUPPORT |
3722                      FATTR4_WORD0_SYMLINK_SUPPORT |
3723                      FATTR4_WORD0_ACLSUPPORT;
3724         if (minorversion)
3725                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3726
3727         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3728         if (status == 0) {
3729                 /* Sanity check the server answers */
3730                 switch (minorversion) {
3731                 case 0:
3732                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3733                         res.attr_bitmask[2] = 0;
3734                         break;
3735                 case 1:
3736                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3737                         break;
3738                 case 2:
3739                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3740                 }
3741                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3742                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3743                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3744                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3745                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3746                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3747                                 NFS_CAP_SECURITY_LABEL);
3748                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3749                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3750                         server->caps |= NFS_CAP_ACLS;
3751                 if (res.has_links != 0)
3752                         server->caps |= NFS_CAP_HARDLINKS;
3753                 if (res.has_symlinks != 0)
3754                         server->caps |= NFS_CAP_SYMLINKS;
3755                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3756                         server->caps |= NFS_CAP_FILEID;
3757                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3758                         server->caps |= NFS_CAP_MODE;
3759                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3760                         server->caps |= NFS_CAP_NLINK;
3761                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3762                         server->caps |= NFS_CAP_OWNER;
3763                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3764                         server->caps |= NFS_CAP_OWNER_GROUP;
3765                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3766                         server->caps |= NFS_CAP_ATIME;
3767                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3768                         server->caps |= NFS_CAP_CTIME;
3769                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3770                         server->caps |= NFS_CAP_MTIME;
3771 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3772                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3773                         server->caps |= NFS_CAP_SECURITY_LABEL;
3774 #endif
3775                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3776                                 sizeof(server->attr_bitmask));
3777                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3778
3779                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3780                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3781                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3782                 server->cache_consistency_bitmask[2] = 0;
3783
3784                 /* Avoid a regression due to buggy server */
3785                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3786                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3787                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3788                         sizeof(server->exclcreat_bitmask));
3789
3790                 server->acl_bitmask = res.acl_bitmask;
3791                 server->fh_expire_type = res.fh_expire_type;
3792         }
3793
3794         return status;
3795 }
3796
3797 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3798 {
3799         struct nfs4_exception exception = {
3800                 .interruptible = true,
3801         };
3802         int err;
3803         do {
3804                 err = nfs4_handle_exception(server,
3805                                 _nfs4_server_capabilities(server, fhandle),
3806                                 &exception);
3807         } while (exception.retry);
3808         return err;
3809 }
3810
3811 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3812                 struct nfs_fsinfo *info)
3813 {
3814         u32 bitmask[3];
3815         struct nfs4_lookup_root_arg args = {
3816                 .bitmask = bitmask,
3817         };
3818         struct nfs4_lookup_res res = {
3819                 .server = server,
3820                 .fattr = info->fattr,
3821                 .fh = fhandle,
3822         };
3823         struct rpc_message msg = {
3824                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3825                 .rpc_argp = &args,
3826                 .rpc_resp = &res,
3827         };
3828
3829         bitmask[0] = nfs4_fattr_bitmap[0];
3830         bitmask[1] = nfs4_fattr_bitmap[1];
3831         /*
3832          * Process the label in the upcoming getfattr
3833          */
3834         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3835
3836         nfs_fattr_init(info->fattr);
3837         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3838 }
3839
3840 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3841                 struct nfs_fsinfo *info)
3842 {
3843         struct nfs4_exception exception = {
3844                 .interruptible = true,
3845         };
3846         int err;
3847         do {
3848                 err = _nfs4_lookup_root(server, fhandle, info);
3849                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3850                 switch (err) {
3851                 case 0:
3852                 case -NFS4ERR_WRONGSEC:
3853                         goto out;
3854                 default:
3855                         err = nfs4_handle_exception(server, err, &exception);
3856                 }
3857         } while (exception.retry);
3858 out:
3859         return err;
3860 }
3861
3862 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3863                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3864 {
3865         struct rpc_auth_create_args auth_args = {
3866                 .pseudoflavor = flavor,
3867         };
3868         struct rpc_auth *auth;
3869
3870         auth = rpcauth_create(&auth_args, server->client);
3871         if (IS_ERR(auth))
3872                 return -EACCES;
3873         return nfs4_lookup_root(server, fhandle, info);
3874 }
3875
3876 /*
3877  * Retry pseudoroot lookup with various security flavors.  We do this when:
3878  *
3879  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3880  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3881  *
3882  * Returns zero on success, or a negative NFS4ERR value, or a
3883  * negative errno value.
3884  */
3885 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3886                               struct nfs_fsinfo *info)
3887 {
3888         /* Per 3530bis 15.33.5 */
3889         static const rpc_authflavor_t flav_array[] = {
3890                 RPC_AUTH_GSS_KRB5P,
3891                 RPC_AUTH_GSS_KRB5I,
3892                 RPC_AUTH_GSS_KRB5,
3893                 RPC_AUTH_UNIX,                  /* courtesy */
3894                 RPC_AUTH_NULL,
3895         };
3896         int status = -EPERM;
3897         size_t i;
3898
3899         if (server->auth_info.flavor_len > 0) {
3900                 /* try each flavor specified by user */
3901                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3902                         status = nfs4_lookup_root_sec(server, fhandle, info,
3903                                                 server->auth_info.flavors[i]);
3904                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3905                                 continue;
3906                         break;
3907                 }
3908         } else {
3909                 /* no flavors specified by user, try default list */
3910                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3911                         status = nfs4_lookup_root_sec(server, fhandle, info,
3912                                                       flav_array[i]);
3913                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3914                                 continue;
3915                         break;
3916                 }
3917         }
3918
3919         /*
3920          * -EACCES could mean that the user doesn't have correct permissions
3921          * to access the mount.  It could also mean that we tried to mount
3922          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3923          * existing mount programs don't handle -EACCES very well so it should
3924          * be mapped to -EPERM instead.
3925          */
3926         if (status == -EACCES)
3927                 status = -EPERM;
3928         return status;
3929 }
3930
3931 /**
3932  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3933  * @server: initialized nfs_server handle
3934  * @fhandle: we fill in the pseudo-fs root file handle
3935  * @info: we fill in an FSINFO struct
3936  * @auth_probe: probe the auth flavours
3937  *
3938  * Returns zero on success, or a negative errno.
3939  */
3940 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3941                          struct nfs_fsinfo *info,
3942                          bool auth_probe)
3943 {
3944         int status = 0;
3945
3946         if (!auth_probe)
3947                 status = nfs4_lookup_root(server, fhandle, info);
3948
3949         if (auth_probe || status == NFS4ERR_WRONGSEC)
3950                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3951                                 fhandle, info);
3952
3953         if (status == 0)
3954                 status = nfs4_server_capabilities(server, fhandle);
3955         if (status == 0)
3956                 status = nfs4_do_fsinfo(server, fhandle, info);
3957
3958         return nfs4_map_errors(status);
3959 }
3960
3961 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3962                               struct nfs_fsinfo *info)
3963 {
3964         int error;
3965         struct nfs_fattr *fattr = info->fattr;
3966         struct nfs4_label *label = NULL;
3967
3968         error = nfs4_server_capabilities(server, mntfh);
3969         if (error < 0) {
3970                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3971                 return error;
3972         }
3973
3974         label = nfs4_label_alloc(server, GFP_KERNEL);
3975         if (IS_ERR(label))
3976                 return PTR_ERR(label);
3977
3978         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
3979         if (error < 0) {
3980                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3981                 goto err_free_label;
3982         }
3983
3984         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3985             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3986                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3987
3988 err_free_label:
3989         nfs4_label_free(label);
3990
3991         return error;
3992 }
3993
3994 /*
3995  * Get locations and (maybe) other attributes of a referral.
3996  * Note that we'll actually follow the referral later when
3997  * we detect fsid mismatch in inode revalidation
3998  */
3999 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4000                              const struct qstr *name, struct nfs_fattr *fattr,
4001                              struct nfs_fh *fhandle)
4002 {
4003         int status = -ENOMEM;
4004         struct page *page = NULL;
4005         struct nfs4_fs_locations *locations = NULL;
4006
4007         page = alloc_page(GFP_KERNEL);
4008         if (page == NULL)
4009                 goto out;
4010         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4011         if (locations == NULL)
4012                 goto out;
4013
4014         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4015         if (status != 0)
4016                 goto out;
4017
4018         /*
4019          * If the fsid didn't change, this is a migration event, not a
4020          * referral.  Cause us to drop into the exception handler, which
4021          * will kick off migration recovery.
4022          */
4023         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4024                 dprintk("%s: server did not return a different fsid for"
4025                         " a referral at %s\n", __func__, name->name);
4026                 status = -NFS4ERR_MOVED;
4027                 goto out;
4028         }
4029         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4030         nfs_fixup_referral_attributes(&locations->fattr);
4031
4032         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4033         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4034         memset(fhandle, 0, sizeof(struct nfs_fh));
4035 out:
4036         if (page)
4037                 __free_page(page);
4038         kfree(locations);
4039         return status;
4040 }
4041
4042 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4043                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4044                                 struct inode *inode)
4045 {
4046         __u32 bitmask[NFS4_BITMASK_SZ];
4047         struct nfs4_getattr_arg args = {
4048                 .fh = fhandle,
4049                 .bitmask = bitmask,
4050         };
4051         struct nfs4_getattr_res res = {
4052                 .fattr = fattr,
4053                 .label = label,
4054                 .server = server,
4055         };
4056         struct rpc_message msg = {
4057                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4058                 .rpc_argp = &args,
4059                 .rpc_resp = &res,
4060         };
4061
4062         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4063
4064         nfs_fattr_init(fattr);
4065         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4066 }
4067
4068 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4069                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4070                                 struct inode *inode)
4071 {
4072         struct nfs4_exception exception = {
4073                 .interruptible = true,
4074         };
4075         int err;
4076         do {
4077                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4078                 trace_nfs4_getattr(server, fhandle, fattr, err);
4079                 err = nfs4_handle_exception(server, err,
4080                                 &exception);
4081         } while (exception.retry);
4082         return err;
4083 }
4084
4085 /*
4086  * The file is not closed if it is opened due to the a request to change
4087  * the size of the file. The open call will not be needed once the
4088  * VFS layer lookup-intents are implemented.
4089  *
4090  * Close is called when the inode is destroyed.
4091  * If we haven't opened the file for O_WRONLY, we
4092  * need to in the size_change case to obtain a stateid.
4093  *
4094  * Got race?
4095  * Because OPEN is always done by name in nfsv4, it is
4096  * possible that we opened a different file by the same
4097  * name.  We can recognize this race condition, but we
4098  * can't do anything about it besides returning an error.
4099  *
4100  * This will be fixed with VFS changes (lookup-intent).
4101  */
4102 static int
4103 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4104                   struct iattr *sattr)
4105 {
4106         struct inode *inode = d_inode(dentry);
4107         const struct cred *cred = NULL;
4108         struct nfs_open_context *ctx = NULL;
4109         struct nfs4_label *label = NULL;
4110         int status;
4111
4112         if (pnfs_ld_layoutret_on_setattr(inode) &&
4113             sattr->ia_valid & ATTR_SIZE &&
4114             sattr->ia_size < i_size_read(inode))
4115                 pnfs_commit_and_return_layout(inode);
4116
4117         nfs_fattr_init(fattr);
4118
4119         /* Deal with open(O_TRUNC) */
4120         if (sattr->ia_valid & ATTR_OPEN)
4121                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4122
4123         /* Optimization: if the end result is no change, don't RPC */
4124         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4125                 return 0;
4126
4127         /* Search for an existing open(O_WRITE) file */
4128         if (sattr->ia_valid & ATTR_FILE) {
4129
4130                 ctx = nfs_file_open_context(sattr->ia_file);
4131                 if (ctx)
4132                         cred = ctx->cred;
4133         }
4134
4135         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4136         if (IS_ERR(label))
4137                 return PTR_ERR(label);
4138
4139         /* Return any delegations if we're going to change ACLs */
4140         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4141                 nfs4_inode_make_writeable(inode);
4142
4143         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4144         if (status == 0) {
4145                 nfs_setattr_update_inode(inode, sattr, fattr);
4146                 nfs_setsecurity(inode, fattr, label);
4147         }
4148         nfs4_label_free(label);
4149         return status;
4150 }
4151
4152 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4153                 const struct qstr *name, struct nfs_fh *fhandle,
4154                 struct nfs_fattr *fattr, struct nfs4_label *label)
4155 {
4156         struct nfs_server *server = NFS_SERVER(dir);
4157         int                    status;
4158         struct nfs4_lookup_arg args = {
4159                 .bitmask = server->attr_bitmask,
4160                 .dir_fh = NFS_FH(dir),
4161                 .name = name,
4162         };
4163         struct nfs4_lookup_res res = {
4164                 .server = server,
4165                 .fattr = fattr,
4166                 .label = label,
4167                 .fh = fhandle,
4168         };
4169         struct rpc_message msg = {
4170                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4171                 .rpc_argp = &args,
4172                 .rpc_resp = &res,
4173         };
4174
4175         args.bitmask = nfs4_bitmask(server, label);
4176
4177         nfs_fattr_init(fattr);
4178
4179         dprintk("NFS call  lookup %s\n", name->name);
4180         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
4181         dprintk("NFS reply lookup: %d\n", status);
4182         return status;
4183 }
4184
4185 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4186 {
4187         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4188                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4189         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4190         fattr->nlink = 2;
4191 }
4192
4193 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4194                                    const struct qstr *name, struct nfs_fh *fhandle,
4195                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4196 {
4197         struct nfs4_exception exception = {
4198                 .interruptible = true,
4199         };
4200         struct rpc_clnt *client = *clnt;
4201         int err;
4202         do {
4203                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
4204                 trace_nfs4_lookup(dir, name, err);
4205                 switch (err) {
4206                 case -NFS4ERR_BADNAME:
4207                         err = -ENOENT;
4208                         goto out;
4209                 case -NFS4ERR_MOVED:
4210                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4211                         if (err == -NFS4ERR_MOVED)
4212                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4213                         goto out;
4214                 case -NFS4ERR_WRONGSEC:
4215                         err = -EPERM;
4216                         if (client != *clnt)
4217                                 goto out;
4218                         client = nfs4_negotiate_security(client, dir, name);
4219                         if (IS_ERR(client))
4220                                 return PTR_ERR(client);
4221
4222                         exception.retry = 1;
4223                         break;
4224                 default:
4225                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4226                 }
4227         } while (exception.retry);
4228
4229 out:
4230         if (err == 0)
4231                 *clnt = client;
4232         else if (client != *clnt)
4233                 rpc_shutdown_client(client);
4234
4235         return err;
4236 }
4237
4238 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
4239                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4240                             struct nfs4_label *label)
4241 {
4242         int status;
4243         struct rpc_clnt *client = NFS_CLIENT(dir);
4244
4245         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
4246         if (client != NFS_CLIENT(dir)) {
4247                 rpc_shutdown_client(client);
4248                 nfs_fixup_secinfo_attributes(fattr);
4249         }
4250         return status;
4251 }
4252
4253 struct rpc_clnt *
4254 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
4255                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4256 {
4257         struct rpc_clnt *client = NFS_CLIENT(dir);
4258         int status;
4259
4260         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
4261         if (status < 0)
4262                 return ERR_PTR(status);
4263         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4264 }
4265
4266 static int _nfs4_proc_lookupp(struct inode *inode,
4267                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4268                 struct nfs4_label *label)
4269 {
4270         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4271         struct nfs_server *server = NFS_SERVER(inode);
4272         int                    status;
4273         struct nfs4_lookupp_arg args = {
4274                 .bitmask = server->attr_bitmask,
4275                 .fh = NFS_FH(inode),
4276         };
4277         struct nfs4_lookupp_res res = {
4278                 .server = server,
4279                 .fattr = fattr,
4280                 .label = label,
4281                 .fh = fhandle,
4282         };
4283         struct rpc_message msg = {
4284                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4285                 .rpc_argp = &args,
4286                 .rpc_resp = &res,
4287         };
4288
4289         args.bitmask = nfs4_bitmask(server, label);
4290
4291         nfs_fattr_init(fattr);
4292
4293         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4294         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4295                                 &res.seq_res, 0);
4296         dprintk("NFS reply lookupp: %d\n", status);
4297         return status;
4298 }
4299
4300 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4301                              struct nfs_fattr *fattr, struct nfs4_label *label)
4302 {
4303         struct nfs4_exception exception = {
4304                 .interruptible = true,
4305         };
4306         int err;
4307         do {
4308                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4309                 trace_nfs4_lookupp(inode, err);
4310                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4311                                 &exception);
4312         } while (exception.retry);
4313         return err;
4314 }
4315
4316 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4317 {
4318         struct nfs_server *server = NFS_SERVER(inode);
4319         struct nfs4_accessargs args = {
4320                 .fh = NFS_FH(inode),
4321                 .access = entry->mask,
4322         };
4323         struct nfs4_accessres res = {
4324                 .server = server,
4325         };
4326         struct rpc_message msg = {
4327                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4328                 .rpc_argp = &args,
4329                 .rpc_resp = &res,
4330                 .rpc_cred = entry->cred,
4331         };
4332         int status = 0;
4333
4334         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4335                 res.fattr = nfs_alloc_fattr();
4336                 if (res.fattr == NULL)
4337                         return -ENOMEM;
4338                 args.bitmask = server->cache_consistency_bitmask;
4339         }
4340         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4341         if (!status) {
4342                 nfs_access_set_mask(entry, res.access);
4343                 if (res.fattr)
4344                         nfs_refresh_inode(inode, res.fattr);
4345         }
4346         nfs_free_fattr(res.fattr);
4347         return status;
4348 }
4349
4350 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4351 {
4352         struct nfs4_exception exception = {
4353                 .interruptible = true,
4354         };
4355         int err;
4356         do {
4357                 err = _nfs4_proc_access(inode, entry);
4358                 trace_nfs4_access(inode, err);
4359                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4360                                 &exception);
4361         } while (exception.retry);
4362         return err;
4363 }
4364
4365 /*
4366  * TODO: For the time being, we don't try to get any attributes
4367  * along with any of the zero-copy operations READ, READDIR,
4368  * READLINK, WRITE.
4369  *
4370  * In the case of the first three, we want to put the GETATTR
4371  * after the read-type operation -- this is because it is hard
4372  * to predict the length of a GETATTR response in v4, and thus
4373  * align the READ data correctly.  This means that the GETATTR
4374  * may end up partially falling into the page cache, and we should
4375  * shift it into the 'tail' of the xdr_buf before processing.
4376  * To do this efficiently, we need to know the total length
4377  * of data received, which doesn't seem to be available outside
4378  * of the RPC layer.
4379  *
4380  * In the case of WRITE, we also want to put the GETATTR after
4381  * the operation -- in this case because we want to make sure
4382  * we get the post-operation mtime and size.
4383  *
4384  * Both of these changes to the XDR layer would in fact be quite
4385  * minor, but I decided to leave them for a subsequent patch.
4386  */
4387 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4388                 unsigned int pgbase, unsigned int pglen)
4389 {
4390         struct nfs4_readlink args = {
4391                 .fh       = NFS_FH(inode),
4392                 .pgbase   = pgbase,
4393                 .pglen    = pglen,
4394                 .pages    = &page,
4395         };
4396         struct nfs4_readlink_res res;
4397         struct rpc_message msg = {
4398                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4399                 .rpc_argp = &args,
4400                 .rpc_resp = &res,
4401         };
4402
4403         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4404 }
4405
4406 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4407                 unsigned int pgbase, unsigned int pglen)
4408 {
4409         struct nfs4_exception exception = {
4410                 .interruptible = true,
4411         };
4412         int err;
4413         do {
4414                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4415                 trace_nfs4_readlink(inode, err);
4416                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4417                                 &exception);
4418         } while (exception.retry);
4419         return err;
4420 }
4421
4422 /*
4423  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4424  */
4425 static int
4426 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4427                  int flags)
4428 {
4429         struct nfs_server *server = NFS_SERVER(dir);
4430         struct nfs4_label l, *ilabel = NULL;
4431         struct nfs_open_context *ctx;
4432         struct nfs4_state *state;
4433         int status = 0;
4434
4435         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4436         if (IS_ERR(ctx))
4437                 return PTR_ERR(ctx);
4438
4439         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4440
4441         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4442                 sattr->ia_mode &= ~current_umask();
4443         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4444         if (IS_ERR(state)) {
4445                 status = PTR_ERR(state);
4446                 goto out;
4447         }
4448 out:
4449         nfs4_label_release_security(ilabel);
4450         put_nfs_open_context(ctx);
4451         return status;
4452 }
4453
4454 static int
4455 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4456 {
4457         struct nfs_server *server = NFS_SERVER(dir);
4458         struct nfs_removeargs args = {
4459                 .fh = NFS_FH(dir),
4460                 .name = *name,
4461         };
4462         struct nfs_removeres res = {
4463                 .server = server,
4464         };
4465         struct rpc_message msg = {
4466                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4467                 .rpc_argp = &args,
4468                 .rpc_resp = &res,
4469         };
4470         unsigned long timestamp = jiffies;
4471         int status;
4472
4473         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4474         if (status == 0) {
4475                 spin_lock(&dir->i_lock);
4476                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4477                 /* Removing a directory decrements nlink in the parent */
4478                 if (ftype == NF4DIR && dir->i_nlink > 2)
4479                         nfs4_dec_nlink_locked(dir);
4480                 spin_unlock(&dir->i_lock);
4481         }
4482         return status;
4483 }
4484
4485 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4486 {
4487         struct nfs4_exception exception = {
4488                 .interruptible = true,
4489         };
4490         struct inode *inode = d_inode(dentry);
4491         int err;
4492
4493         if (inode) {
4494                 if (inode->i_nlink == 1)
4495                         nfs4_inode_return_delegation(inode);
4496                 else
4497                         nfs4_inode_make_writeable(inode);
4498         }
4499         do {
4500                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4501                 trace_nfs4_remove(dir, &dentry->d_name, err);
4502                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4503                                 &exception);
4504         } while (exception.retry);
4505         return err;
4506 }
4507
4508 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4509 {
4510         struct nfs4_exception exception = {
4511                 .interruptible = true,
4512         };
4513         int err;
4514
4515         do {
4516                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4517                 trace_nfs4_remove(dir, name, err);
4518                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4519                                 &exception);
4520         } while (exception.retry);
4521         return err;
4522 }
4523
4524 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4525                 struct dentry *dentry,
4526                 struct inode *inode)
4527 {
4528         struct nfs_removeargs *args = msg->rpc_argp;
4529         struct nfs_removeres *res = msg->rpc_resp;
4530
4531         res->server = NFS_SB(dentry->d_sb);
4532         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4533         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4534
4535         nfs_fattr_init(res->dir_attr);
4536
4537         if (inode)
4538                 nfs4_inode_return_delegation(inode);
4539 }
4540
4541 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4542 {
4543         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4544                         &data->args.seq_args,
4545                         &data->res.seq_res,
4546                         task);
4547 }
4548
4549 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4550 {
4551         struct nfs_unlinkdata *data = task->tk_calldata;
4552         struct nfs_removeres *res = &data->res;
4553
4554         if (!nfs4_sequence_done(task, &res->seq_res))
4555                 return 0;
4556         if (nfs4_async_handle_error(task, res->server, NULL,
4557                                     &data->timeout) == -EAGAIN)
4558                 return 0;
4559         if (task->tk_status == 0)
4560                 update_changeattr(dir, &res->cinfo,
4561                                 res->dir_attr->time_start, 0);
4562         return 1;
4563 }
4564
4565 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4566                 struct dentry *old_dentry,
4567                 struct dentry *new_dentry)
4568 {
4569         struct nfs_renameargs *arg = msg->rpc_argp;
4570         struct nfs_renameres *res = msg->rpc_resp;
4571         struct inode *old_inode = d_inode(old_dentry);
4572         struct inode *new_inode = d_inode(new_dentry);
4573
4574         if (old_inode)
4575                 nfs4_inode_make_writeable(old_inode);
4576         if (new_inode)
4577                 nfs4_inode_return_delegation(new_inode);
4578         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4579         res->server = NFS_SB(old_dentry->d_sb);
4580         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4581 }
4582
4583 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4584 {
4585         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4586                         &data->args.seq_args,
4587                         &data->res.seq_res,
4588                         task);
4589 }
4590
4591 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4592                                  struct inode *new_dir)
4593 {
4594         struct nfs_renamedata *data = task->tk_calldata;
4595         struct nfs_renameres *res = &data->res;
4596
4597         if (!nfs4_sequence_done(task, &res->seq_res))
4598                 return 0;
4599         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4600                 return 0;
4601
4602         if (task->tk_status == 0) {
4603                 if (new_dir != old_dir) {
4604                         /* Note: If we moved a directory, nlink will change */
4605                         update_changeattr(old_dir, &res->old_cinfo,
4606                                         res->old_fattr->time_start,
4607                                         NFS_INO_INVALID_OTHER);
4608                         update_changeattr(new_dir, &res->new_cinfo,
4609                                         res->new_fattr->time_start,
4610                                         NFS_INO_INVALID_OTHER);
4611                 } else
4612                         update_changeattr(old_dir, &res->old_cinfo,
4613                                         res->old_fattr->time_start,
4614                                         0);
4615         }
4616         return 1;
4617 }
4618
4619 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4620 {
4621         struct nfs_server *server = NFS_SERVER(inode);
4622         __u32 bitmask[NFS4_BITMASK_SZ];
4623         struct nfs4_link_arg arg = {
4624                 .fh     = NFS_FH(inode),
4625                 .dir_fh = NFS_FH(dir),
4626                 .name   = name,
4627                 .bitmask = bitmask,
4628         };
4629         struct nfs4_link_res res = {
4630                 .server = server,
4631                 .label = NULL,
4632         };
4633         struct rpc_message msg = {
4634                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4635                 .rpc_argp = &arg,
4636                 .rpc_resp = &res,
4637         };
4638         int status = -ENOMEM;
4639
4640         res.fattr = nfs_alloc_fattr();
4641         if (res.fattr == NULL)
4642                 goto out;
4643
4644         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4645         if (IS_ERR(res.label)) {
4646                 status = PTR_ERR(res.label);
4647                 goto out;
4648         }
4649
4650         nfs4_inode_make_writeable(inode);
4651         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4652
4653         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4654         if (!status) {
4655                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4656                 status = nfs_post_op_update_inode(inode, res.fattr);
4657                 if (!status)
4658                         nfs_setsecurity(inode, res.fattr, res.label);
4659         }
4660
4661
4662         nfs4_label_free(res.label);
4663
4664 out:
4665         nfs_free_fattr(res.fattr);
4666         return status;
4667 }
4668
4669 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4670 {
4671         struct nfs4_exception exception = {
4672                 .interruptible = true,
4673         };
4674         int err;
4675         do {
4676                 err = nfs4_handle_exception(NFS_SERVER(inode),
4677                                 _nfs4_proc_link(inode, dir, name),
4678                                 &exception);
4679         } while (exception.retry);
4680         return err;
4681 }
4682
4683 struct nfs4_createdata {
4684         struct rpc_message msg;
4685         struct nfs4_create_arg arg;
4686         struct nfs4_create_res res;
4687         struct nfs_fh fh;
4688         struct nfs_fattr fattr;
4689         struct nfs4_label *label;
4690 };
4691
4692 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4693                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4694 {
4695         struct nfs4_createdata *data;
4696
4697         data = kzalloc(sizeof(*data), GFP_KERNEL);
4698         if (data != NULL) {
4699                 struct nfs_server *server = NFS_SERVER(dir);
4700
4701                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4702                 if (IS_ERR(data->label))
4703                         goto out_free;
4704
4705                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4706                 data->msg.rpc_argp = &data->arg;
4707                 data->msg.rpc_resp = &data->res;
4708                 data->arg.dir_fh = NFS_FH(dir);
4709                 data->arg.server = server;
4710                 data->arg.name = name;
4711                 data->arg.attrs = sattr;
4712                 data->arg.ftype = ftype;
4713                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4714                 data->arg.umask = current_umask();
4715                 data->res.server = server;
4716                 data->res.fh = &data->fh;
4717                 data->res.fattr = &data->fattr;
4718                 data->res.label = data->label;
4719                 nfs_fattr_init(data->res.fattr);
4720         }
4721         return data;
4722 out_free:
4723         kfree(data);
4724         return NULL;
4725 }
4726
4727 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4728 {
4729         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4730                                     &data->arg.seq_args, &data->res.seq_res, 1);
4731         if (status == 0) {
4732                 spin_lock(&dir->i_lock);
4733                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4734                                 data->res.fattr->time_start, 0);
4735                 /* Creating a directory bumps nlink in the parent */
4736                 if (data->arg.ftype == NF4DIR)
4737                         nfs4_inc_nlink_locked(dir);
4738                 spin_unlock(&dir->i_lock);
4739                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4740         }
4741         return status;
4742 }
4743
4744 static void nfs4_free_createdata(struct nfs4_createdata *data)
4745 {
4746         nfs4_label_free(data->label);
4747         kfree(data);
4748 }
4749
4750 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4751                 struct page *page, unsigned int len, struct iattr *sattr,
4752                 struct nfs4_label *label)
4753 {
4754         struct nfs4_createdata *data;
4755         int status = -ENAMETOOLONG;
4756
4757         if (len > NFS4_MAXPATHLEN)
4758                 goto out;
4759
4760         status = -ENOMEM;
4761         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4762         if (data == NULL)
4763                 goto out;
4764
4765         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4766         data->arg.u.symlink.pages = &page;
4767         data->arg.u.symlink.len = len;
4768         data->arg.label = label;
4769
4770         status = nfs4_do_create(dir, dentry, data);
4771
4772         nfs4_free_createdata(data);
4773 out:
4774         return status;
4775 }
4776
4777 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4778                 struct page *page, unsigned int len, struct iattr *sattr)
4779 {
4780         struct nfs4_exception exception = {
4781                 .interruptible = true,
4782         };
4783         struct nfs4_label l, *label = NULL;
4784         int err;
4785
4786         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4787
4788         do {
4789                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4790                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4791                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4792                                 &exception);
4793         } while (exception.retry);
4794
4795         nfs4_label_release_security(label);
4796         return err;
4797 }
4798
4799 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4800                 struct iattr *sattr, struct nfs4_label *label)
4801 {
4802         struct nfs4_createdata *data;
4803         int status = -ENOMEM;
4804
4805         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4806         if (data == NULL)
4807                 goto out;
4808
4809         data->arg.label = label;
4810         status = nfs4_do_create(dir, dentry, data);
4811
4812         nfs4_free_createdata(data);
4813 out:
4814         return status;
4815 }
4816
4817 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4818                 struct iattr *sattr)
4819 {
4820         struct nfs_server *server = NFS_SERVER(dir);
4821         struct nfs4_exception exception = {
4822                 .interruptible = true,
4823         };
4824         struct nfs4_label l, *label = NULL;
4825         int err;
4826
4827         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4828
4829         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4830                 sattr->ia_mode &= ~current_umask();
4831         do {
4832                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4833                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4834                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4835                                 &exception);
4836         } while (exception.retry);
4837         nfs4_label_release_security(label);
4838
4839         return err;
4840 }
4841
4842 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4843                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4844 {
4845         struct inode            *dir = d_inode(dentry);
4846         struct nfs4_readdir_arg args = {
4847                 .fh = NFS_FH(dir),
4848                 .pages = pages,
4849                 .pgbase = 0,
4850                 .count = count,
4851                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4852                 .plus = plus,
4853         };
4854         struct nfs4_readdir_res res;
4855         struct rpc_message msg = {
4856                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4857                 .rpc_argp = &args,
4858                 .rpc_resp = &res,
4859                 .rpc_cred = cred,
4860         };
4861         int                     status;
4862
4863         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4864                         dentry,
4865                         (unsigned long long)cookie);
4866         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4867         res.pgbase = args.pgbase;
4868         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4869         if (status >= 0) {
4870                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4871                 status += args.pgbase;
4872         }
4873
4874         nfs_invalidate_atime(dir);
4875
4876         dprintk("%s: returns %d\n", __func__, status);
4877         return status;
4878 }
4879
4880 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4881                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4882 {
4883         struct nfs4_exception exception = {
4884                 .interruptible = true,
4885         };
4886         int err;
4887         do {
4888                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4889                                 pages, count, plus);
4890                 trace_nfs4_readdir(d_inode(dentry), err);
4891                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4892                                 &exception);
4893         } while (exception.retry);
4894         return err;
4895 }
4896
4897 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4898                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4899 {
4900         struct nfs4_createdata *data;
4901         int mode = sattr->ia_mode;
4902         int status = -ENOMEM;
4903
4904         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4905         if (data == NULL)
4906                 goto out;
4907
4908         if (S_ISFIFO(mode))
4909                 data->arg.ftype = NF4FIFO;
4910         else if (S_ISBLK(mode)) {
4911                 data->arg.ftype = NF4BLK;
4912                 data->arg.u.device.specdata1 = MAJOR(rdev);
4913                 data->arg.u.device.specdata2 = MINOR(rdev);
4914         }
4915         else if (S_ISCHR(mode)) {
4916                 data->arg.ftype = NF4CHR;
4917                 data->arg.u.device.specdata1 = MAJOR(rdev);
4918                 data->arg.u.device.specdata2 = MINOR(rdev);
4919         } else if (!S_ISSOCK(mode)) {
4920                 status = -EINVAL;
4921                 goto out_free;
4922         }
4923
4924         data->arg.label = label;
4925         status = nfs4_do_create(dir, dentry, data);
4926 out_free:
4927         nfs4_free_createdata(data);
4928 out:
4929         return status;
4930 }
4931
4932 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4933                 struct iattr *sattr, dev_t rdev)
4934 {
4935         struct nfs_server *server = NFS_SERVER(dir);
4936         struct nfs4_exception exception = {
4937                 .interruptible = true,
4938         };
4939         struct nfs4_label l, *label = NULL;
4940         int err;
4941
4942         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4943
4944         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4945                 sattr->ia_mode &= ~current_umask();
4946         do {
4947                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4948                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4949                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4950                                 &exception);
4951         } while (exception.retry);
4952
4953         nfs4_label_release_security(label);
4954
4955         return err;
4956 }
4957
4958 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4959                  struct nfs_fsstat *fsstat)
4960 {
4961         struct nfs4_statfs_arg args = {
4962                 .fh = fhandle,
4963                 .bitmask = server->attr_bitmask,
4964         };
4965         struct nfs4_statfs_res res = {
4966                 .fsstat = fsstat,
4967         };
4968         struct rpc_message msg = {
4969                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4970                 .rpc_argp = &args,
4971                 .rpc_resp = &res,
4972         };
4973
4974         nfs_fattr_init(fsstat->fattr);
4975         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4976 }
4977
4978 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4979 {
4980         struct nfs4_exception exception = {
4981                 .interruptible = true,
4982         };
4983         int err;
4984         do {
4985                 err = nfs4_handle_exception(server,
4986                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4987                                 &exception);
4988         } while (exception.retry);
4989         return err;
4990 }
4991
4992 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4993                 struct nfs_fsinfo *fsinfo)
4994 {
4995         struct nfs4_fsinfo_arg args = {
4996                 .fh = fhandle,
4997                 .bitmask = server->attr_bitmask,
4998         };
4999         struct nfs4_fsinfo_res res = {
5000                 .fsinfo = fsinfo,
5001         };
5002         struct rpc_message msg = {
5003                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5004                 .rpc_argp = &args,
5005                 .rpc_resp = &res,
5006         };
5007
5008         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5009 }
5010
5011 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5012 {
5013         struct nfs4_exception exception = {
5014                 .interruptible = true,
5015         };
5016         unsigned long now = jiffies;
5017         int err;
5018
5019         do {
5020                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5021                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5022                 if (err == 0) {
5023                         nfs4_set_lease_period(server->nfs_client,
5024                                         fsinfo->lease_time * HZ,
5025                                         now);
5026                         break;
5027                 }
5028                 err = nfs4_handle_exception(server, err, &exception);
5029         } while (exception.retry);
5030         return err;
5031 }
5032
5033 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5034 {
5035         int error;
5036
5037         nfs_fattr_init(fsinfo->fattr);
5038         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5039         if (error == 0) {
5040                 /* block layout checks this! */
5041                 server->pnfs_blksize = fsinfo->blksize;
5042                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5043         }
5044
5045         return error;
5046 }
5047
5048 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5049                 struct nfs_pathconf *pathconf)
5050 {
5051         struct nfs4_pathconf_arg args = {
5052                 .fh = fhandle,
5053                 .bitmask = server->attr_bitmask,
5054         };
5055         struct nfs4_pathconf_res res = {
5056                 .pathconf = pathconf,
5057         };
5058         struct rpc_message msg = {
5059                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5060                 .rpc_argp = &args,
5061                 .rpc_resp = &res,
5062         };
5063
5064         /* None of the pathconf attributes are mandatory to implement */
5065         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5066                 memset(pathconf, 0, sizeof(*pathconf));
5067                 return 0;
5068         }
5069
5070         nfs_fattr_init(pathconf->fattr);
5071         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5072 }
5073
5074 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5075                 struct nfs_pathconf *pathconf)
5076 {
5077         struct nfs4_exception exception = {
5078                 .interruptible = true,
5079         };
5080         int err;
5081
5082         do {
5083                 err = nfs4_handle_exception(server,
5084                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5085                                 &exception);
5086         } while (exception.retry);
5087         return err;
5088 }
5089
5090 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5091                 const struct nfs_open_context *ctx,
5092                 const struct nfs_lock_context *l_ctx,
5093                 fmode_t fmode)
5094 {
5095         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5096 }
5097 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5098
5099 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5100                 const struct nfs_open_context *ctx,
5101                 const struct nfs_lock_context *l_ctx,
5102                 fmode_t fmode)
5103 {
5104         nfs4_stateid current_stateid;
5105
5106         /* If the current stateid represents a lost lock, then exit */
5107         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
5108                 return true;
5109         return nfs4_stateid_match(stateid, &current_stateid);
5110 }
5111
5112 static bool nfs4_error_stateid_expired(int err)
5113 {
5114         switch (err) {
5115         case -NFS4ERR_DELEG_REVOKED:
5116         case -NFS4ERR_ADMIN_REVOKED:
5117         case -NFS4ERR_BAD_STATEID:
5118         case -NFS4ERR_STALE_STATEID:
5119         case -NFS4ERR_OLD_STATEID:
5120         case -NFS4ERR_OPENMODE:
5121         case -NFS4ERR_EXPIRED:
5122                 return true;
5123         }
5124         return false;
5125 }
5126
5127 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5128 {
5129         struct nfs_server *server = NFS_SERVER(hdr->inode);
5130
5131         trace_nfs4_read(hdr, task->tk_status);
5132         if (task->tk_status < 0) {
5133                 struct nfs4_exception exception = {
5134                         .inode = hdr->inode,
5135                         .state = hdr->args.context->state,
5136                         .stateid = &hdr->args.stateid,
5137                 };
5138                 task->tk_status = nfs4_async_handle_exception(task,
5139                                 server, task->tk_status, &exception);
5140                 if (exception.retry) {
5141                         rpc_restart_call_prepare(task);
5142                         return -EAGAIN;
5143                 }
5144         }
5145
5146         if (task->tk_status > 0)
5147                 renew_lease(server, hdr->timestamp);
5148         return 0;
5149 }
5150
5151 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5152                 struct nfs_pgio_args *args)
5153 {
5154
5155         if (!nfs4_error_stateid_expired(task->tk_status) ||
5156                 nfs4_stateid_is_current(&args->stateid,
5157                                 args->context,
5158                                 args->lock_context,
5159                                 FMODE_READ))
5160                 return false;
5161         rpc_restart_call_prepare(task);
5162         return true;
5163 }
5164
5165 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5166 {
5167
5168         dprintk("--> %s\n", __func__);
5169
5170         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5171                 return -EAGAIN;
5172         if (nfs4_read_stateid_changed(task, &hdr->args))
5173                 return -EAGAIN;
5174         if (task->tk_status > 0)
5175                 nfs_invalidate_atime(hdr->inode);
5176         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5177                                     nfs4_read_done_cb(task, hdr);
5178 }
5179
5180 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5181                                  struct rpc_message *msg)
5182 {
5183         hdr->timestamp   = jiffies;
5184         if (!hdr->pgio_done_cb)
5185                 hdr->pgio_done_cb = nfs4_read_done_cb;
5186         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5187         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5188 }
5189
5190 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5191                                       struct nfs_pgio_header *hdr)
5192 {
5193         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5194                         &hdr->args.seq_args,
5195                         &hdr->res.seq_res,
5196                         task))
5197                 return 0;
5198         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5199                                 hdr->args.lock_context,
5200                                 hdr->rw_mode) == -EIO)
5201                 return -EIO;
5202         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5203                 return -EIO;
5204         return 0;
5205 }
5206
5207 static int nfs4_write_done_cb(struct rpc_task *task,
5208                               struct nfs_pgio_header *hdr)
5209 {
5210         struct inode *inode = hdr->inode;
5211
5212         trace_nfs4_write(hdr, task->tk_status);
5213         if (task->tk_status < 0) {
5214                 struct nfs4_exception exception = {
5215                         .inode = hdr->inode,
5216                         .state = hdr->args.context->state,
5217                         .stateid = &hdr->args.stateid,
5218                 };
5219                 task->tk_status = nfs4_async_handle_exception(task,
5220                                 NFS_SERVER(inode), task->tk_status,
5221                                 &exception);
5222                 if (exception.retry) {
5223                         rpc_restart_call_prepare(task);
5224                         return -EAGAIN;
5225                 }
5226         }
5227         if (task->tk_status >= 0) {
5228                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5229                 nfs_writeback_update_inode(hdr);
5230         }
5231         return 0;
5232 }
5233
5234 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5235                 struct nfs_pgio_args *args)
5236 {
5237
5238         if (!nfs4_error_stateid_expired(task->tk_status) ||
5239                 nfs4_stateid_is_current(&args->stateid,
5240                                 args->context,
5241                                 args->lock_context,
5242                                 FMODE_WRITE))
5243                 return false;
5244         rpc_restart_call_prepare(task);
5245         return true;
5246 }
5247
5248 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5249 {
5250         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5251                 return -EAGAIN;
5252         if (nfs4_write_stateid_changed(task, &hdr->args))
5253                 return -EAGAIN;
5254         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5255                 nfs4_write_done_cb(task, hdr);
5256 }
5257
5258 static
5259 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5260 {
5261         /* Don't request attributes for pNFS or O_DIRECT writes */
5262         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5263                 return false;
5264         /* Otherwise, request attributes if and only if we don't hold
5265          * a delegation
5266          */
5267         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5268 }
5269
5270 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5271                                   struct rpc_message *msg,
5272                                   struct rpc_clnt **clnt)
5273 {
5274         struct nfs_server *server = NFS_SERVER(hdr->inode);
5275
5276         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5277                 hdr->args.bitmask = NULL;
5278                 hdr->res.fattr = NULL;
5279         } else
5280                 hdr->args.bitmask = server->cache_consistency_bitmask;
5281
5282         if (!hdr->pgio_done_cb)
5283                 hdr->pgio_done_cb = nfs4_write_done_cb;
5284         hdr->res.server = server;
5285         hdr->timestamp   = jiffies;
5286
5287         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5288         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1, 0);
5289         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5290 }
5291
5292 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5293 {
5294         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5295                         &data->args.seq_args,
5296                         &data->res.seq_res,
5297                         task);
5298 }
5299
5300 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5301 {
5302         struct inode *inode = data->inode;
5303
5304         trace_nfs4_commit(data, task->tk_status);
5305         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5306                                     NULL, NULL) == -EAGAIN) {
5307                 rpc_restart_call_prepare(task);
5308                 return -EAGAIN;
5309         }
5310         return 0;
5311 }
5312
5313 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5314 {
5315         if (!nfs4_sequence_done(task, &data->res.seq_res))
5316                 return -EAGAIN;
5317         return data->commit_done_cb(task, data);
5318 }
5319
5320 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5321                                    struct rpc_clnt **clnt)
5322 {
5323         struct nfs_server *server = NFS_SERVER(data->inode);
5324
5325         if (data->commit_done_cb == NULL)
5326                 data->commit_done_cb = nfs4_commit_done_cb;
5327         data->res.server = server;
5328         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5329         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5330         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5331 }
5332
5333 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5334                                 struct nfs_commitres *res)
5335 {
5336         struct inode *dst_inode = file_inode(dst);
5337         struct nfs_server *server = NFS_SERVER(dst_inode);
5338         struct rpc_message msg = {
5339                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5340                 .rpc_argp = args,
5341                 .rpc_resp = res,
5342         };
5343
5344         args->fh = NFS_FH(dst_inode);
5345         return nfs4_call_sync(server->client, server, &msg,
5346                         &args->seq_args, &res->seq_res, 1);
5347 }
5348
5349 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5350 {
5351         struct nfs_commitargs args = {
5352                 .offset = offset,
5353                 .count = count,
5354         };
5355         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5356         struct nfs4_exception exception = { };
5357         int status;
5358
5359         do {
5360                 status = _nfs4_proc_commit(dst, &args, res);
5361                 status = nfs4_handle_exception(dst_server, status, &exception);
5362         } while (exception.retry);
5363
5364         return status;
5365 }
5366
5367 struct nfs4_renewdata {
5368         struct nfs_client       *client;
5369         unsigned long           timestamp;
5370 };
5371
5372 /*
5373  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5374  * standalone procedure for queueing an asynchronous RENEW.
5375  */
5376 static void nfs4_renew_release(void *calldata)
5377 {
5378         struct nfs4_renewdata *data = calldata;
5379         struct nfs_client *clp = data->client;
5380
5381         if (refcount_read(&clp->cl_count) > 1)
5382                 nfs4_schedule_state_renewal(clp);
5383         nfs_put_client(clp);
5384         kfree(data);
5385 }
5386
5387 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5388 {
5389         struct nfs4_renewdata *data = calldata;
5390         struct nfs_client *clp = data->client;
5391         unsigned long timestamp = data->timestamp;
5392
5393         trace_nfs4_renew_async(clp, task->tk_status);
5394         switch (task->tk_status) {
5395         case 0:
5396                 break;
5397         case -NFS4ERR_LEASE_MOVED:
5398                 nfs4_schedule_lease_moved_recovery(clp);
5399                 break;
5400         default:
5401                 /* Unless we're shutting down, schedule state recovery! */
5402                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5403                         return;
5404                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5405                         nfs4_schedule_lease_recovery(clp);
5406                         return;
5407                 }
5408                 nfs4_schedule_path_down_recovery(clp);
5409         }
5410         do_renew_lease(clp, timestamp);
5411 }
5412
5413 static const struct rpc_call_ops nfs4_renew_ops = {
5414         .rpc_call_done = nfs4_renew_done,
5415         .rpc_release = nfs4_renew_release,
5416 };
5417
5418 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5419 {
5420         struct rpc_message msg = {
5421                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5422                 .rpc_argp       = clp,
5423                 .rpc_cred       = cred,
5424         };
5425         struct nfs4_renewdata *data;
5426
5427         if (renew_flags == 0)
5428                 return 0;
5429         if (!refcount_inc_not_zero(&clp->cl_count))
5430                 return -EIO;
5431         data = kmalloc(sizeof(*data), GFP_NOFS);
5432         if (data == NULL) {
5433                 nfs_put_client(clp);
5434                 return -ENOMEM;
5435         }
5436         data->client = clp;
5437         data->timestamp = jiffies;
5438         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5439                         &nfs4_renew_ops, data);
5440 }
5441
5442 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5443 {
5444         struct rpc_message msg = {
5445                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5446                 .rpc_argp       = clp,
5447                 .rpc_cred       = cred,
5448         };
5449         unsigned long now = jiffies;
5450         int status;
5451
5452         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5453         if (status < 0)
5454                 return status;
5455         do_renew_lease(clp, now);
5456         return 0;
5457 }
5458
5459 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5460 {
5461         return server->caps & NFS_CAP_ACLS;
5462 }
5463
5464 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5465  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5466  * the stack.
5467  */
5468 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5469
5470 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5471                 struct page **pages)
5472 {
5473         struct page *newpage, **spages;
5474         int rc = 0;
5475         size_t len;
5476         spages = pages;
5477
5478         do {
5479                 len = min_t(size_t, PAGE_SIZE, buflen);
5480                 newpage = alloc_page(GFP_KERNEL);
5481
5482                 if (newpage == NULL)
5483                         goto unwind;
5484                 memcpy(page_address(newpage), buf, len);
5485                 buf += len;
5486                 buflen -= len;
5487                 *pages++ = newpage;
5488                 rc++;
5489         } while (buflen != 0);
5490
5491         return rc;
5492
5493 unwind:
5494         for(; rc > 0; rc--)
5495                 __free_page(spages[rc-1]);
5496         return -ENOMEM;
5497 }
5498
5499 struct nfs4_cached_acl {
5500         int cached;
5501         size_t len;
5502         char data[0];
5503 };
5504
5505 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5506 {
5507         struct nfs_inode *nfsi = NFS_I(inode);
5508
5509         spin_lock(&inode->i_lock);
5510         kfree(nfsi->nfs4_acl);
5511         nfsi->nfs4_acl = acl;
5512         spin_unlock(&inode->i_lock);
5513 }
5514
5515 static void nfs4_zap_acl_attr(struct inode *inode)
5516 {
5517         nfs4_set_cached_acl(inode, NULL);
5518 }
5519
5520 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5521 {
5522         struct nfs_inode *nfsi = NFS_I(inode);
5523         struct nfs4_cached_acl *acl;
5524         int ret = -ENOENT;
5525
5526         spin_lock(&inode->i_lock);
5527         acl = nfsi->nfs4_acl;
5528         if (acl == NULL)
5529                 goto out;
5530         if (buf == NULL) /* user is just asking for length */
5531                 goto out_len;
5532         if (acl->cached == 0)
5533                 goto out;
5534         ret = -ERANGE; /* see getxattr(2) man page */
5535         if (acl->len > buflen)
5536                 goto out;
5537         memcpy(buf, acl->data, acl->len);
5538 out_len:
5539         ret = acl->len;
5540 out:
5541         spin_unlock(&inode->i_lock);
5542         return ret;
5543 }
5544
5545 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5546 {
5547         struct nfs4_cached_acl *acl;
5548         size_t buflen = sizeof(*acl) + acl_len;
5549
5550         if (buflen <= PAGE_SIZE) {
5551                 acl = kmalloc(buflen, GFP_KERNEL);
5552                 if (acl == NULL)
5553                         goto out;
5554                 acl->cached = 1;
5555                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5556         } else {
5557                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5558                 if (acl == NULL)
5559                         goto out;
5560                 acl->cached = 0;
5561         }
5562         acl->len = acl_len;
5563 out:
5564         nfs4_set_cached_acl(inode, acl);
5565 }
5566
5567 /*
5568  * The getxattr API returns the required buffer length when called with a
5569  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5570  * the required buf.  On a NULL buf, we send a page of data to the server
5571  * guessing that the ACL request can be serviced by a page. If so, we cache
5572  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5573  * the cache. If not so, we throw away the page, and cache the required
5574  * length. The next getxattr call will then produce another round trip to
5575  * the server, this time with the input buf of the required size.
5576  */
5577 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5578 {
5579         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5580         struct nfs_getaclargs args = {
5581                 .fh = NFS_FH(inode),
5582                 .acl_pages = pages,
5583                 .acl_len = buflen,
5584         };
5585         struct nfs_getaclres res = {
5586                 .acl_len = buflen,
5587         };
5588         struct rpc_message msg = {
5589                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5590                 .rpc_argp = &args,
5591                 .rpc_resp = &res,
5592         };
5593         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5594         int ret = -ENOMEM, i;
5595
5596         if (npages > ARRAY_SIZE(pages))
5597                 return -ERANGE;
5598
5599         for (i = 0; i < npages; i++) {
5600                 pages[i] = alloc_page(GFP_KERNEL);
5601                 if (!pages[i])
5602                         goto out_free;
5603         }
5604
5605         /* for decoding across pages */
5606         res.acl_scratch = alloc_page(GFP_KERNEL);
5607         if (!res.acl_scratch)
5608                 goto out_free;
5609
5610         args.acl_len = npages * PAGE_SIZE;
5611
5612         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5613                 __func__, buf, buflen, npages, args.acl_len);
5614         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5615                              &msg, &args.seq_args, &res.seq_res, 0);
5616         if (ret)
5617                 goto out_free;
5618
5619         /* Handle the case where the passed-in buffer is too short */
5620         if (res.acl_flags & NFS4_ACL_TRUNC) {
5621                 /* Did the user only issue a request for the acl length? */
5622                 if (buf == NULL)
5623                         goto out_ok;
5624                 ret = -ERANGE;
5625                 goto out_free;
5626         }
5627         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5628         if (buf) {
5629                 if (res.acl_len > buflen) {
5630                         ret = -ERANGE;
5631                         goto out_free;
5632                 }
5633                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5634         }
5635 out_ok:
5636         ret = res.acl_len;
5637 out_free:
5638         for (i = 0; i < npages; i++)
5639                 if (pages[i])
5640                         __free_page(pages[i]);
5641         if (res.acl_scratch)
5642                 __free_page(res.acl_scratch);
5643         return ret;
5644 }
5645
5646 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5647 {
5648         struct nfs4_exception exception = {
5649                 .interruptible = true,
5650         };
5651         ssize_t ret;
5652         do {
5653                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5654                 trace_nfs4_get_acl(inode, ret);
5655                 if (ret >= 0)
5656                         break;
5657                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5658         } while (exception.retry);
5659         return ret;
5660 }
5661
5662 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5663 {
5664         struct nfs_server *server = NFS_SERVER(inode);
5665         int ret;
5666
5667         if (!nfs4_server_supports_acls(server))
5668                 return -EOPNOTSUPP;
5669         ret = nfs_revalidate_inode(server, inode);
5670         if (ret < 0)
5671                 return ret;
5672         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5673                 nfs_zap_acl_cache(inode);
5674         ret = nfs4_read_cached_acl(inode, buf, buflen);
5675         if (ret != -ENOENT)
5676                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5677                  * but no cached acl data, just the acl length */
5678                 return ret;
5679         return nfs4_get_acl_uncached(inode, buf, buflen);
5680 }
5681
5682 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5683 {
5684         struct nfs_server *server = NFS_SERVER(inode);
5685         struct page *pages[NFS4ACL_MAXPAGES];
5686         struct nfs_setaclargs arg = {
5687                 .fh             = NFS_FH(inode),
5688                 .acl_pages      = pages,
5689                 .acl_len        = buflen,
5690         };
5691         struct nfs_setaclres res;
5692         struct rpc_message msg = {
5693                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5694                 .rpc_argp       = &arg,
5695                 .rpc_resp       = &res,
5696         };
5697         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5698         int ret, i;
5699
5700         if (!nfs4_server_supports_acls(server))
5701                 return -EOPNOTSUPP;
5702         if (npages > ARRAY_SIZE(pages))
5703                 return -ERANGE;
5704         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5705         if (i < 0)
5706                 return i;
5707         nfs4_inode_make_writeable(inode);
5708         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5709
5710         /*
5711          * Free each page after tx, so the only ref left is
5712          * held by the network stack
5713          */
5714         for (; i > 0; i--)
5715                 put_page(pages[i-1]);
5716
5717         /*
5718          * Acl update can result in inode attribute update.
5719          * so mark the attribute cache invalid.
5720          */
5721         spin_lock(&inode->i_lock);
5722         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5723                 | NFS_INO_INVALID_CTIME
5724                 | NFS_INO_REVAL_FORCED;
5725         spin_unlock(&inode->i_lock);
5726         nfs_access_zap_cache(inode);
5727         nfs_zap_acl_cache(inode);
5728         return ret;
5729 }
5730
5731 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5732 {
5733         struct nfs4_exception exception = { };
5734         int err;
5735         do {
5736                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5737                 trace_nfs4_set_acl(inode, err);
5738                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5739                                 &exception);
5740         } while (exception.retry);
5741         return err;
5742 }
5743
5744 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5745 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5746                                         size_t buflen)
5747 {
5748         struct nfs_server *server = NFS_SERVER(inode);
5749         struct nfs_fattr fattr;
5750         struct nfs4_label label = {0, 0, buflen, buf};
5751
5752         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5753         struct nfs4_getattr_arg arg = {
5754                 .fh             = NFS_FH(inode),
5755                 .bitmask        = bitmask,
5756         };
5757         struct nfs4_getattr_res res = {
5758                 .fattr          = &fattr,
5759                 .label          = &label,
5760                 .server         = server,
5761         };
5762         struct rpc_message msg = {
5763                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5764                 .rpc_argp       = &arg,
5765                 .rpc_resp       = &res,
5766         };
5767         int ret;
5768
5769         nfs_fattr_init(&fattr);
5770
5771         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5772         if (ret)
5773                 return ret;
5774         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5775                 return -ENOENT;
5776         if (buflen < label.len)
5777                 return -ERANGE;
5778         return 0;
5779 }
5780
5781 static int nfs4_get_security_label(struct inode *inode, void *buf,
5782                                         size_t buflen)
5783 {
5784         struct nfs4_exception exception = {
5785                 .interruptible = true,
5786         };
5787         int err;
5788
5789         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5790                 return -EOPNOTSUPP;
5791
5792         do {
5793                 err = _nfs4_get_security_label(inode, buf, buflen);
5794                 trace_nfs4_get_security_label(inode, err);
5795                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5796                                 &exception);
5797         } while (exception.retry);
5798         return err;
5799 }
5800
5801 static int _nfs4_do_set_security_label(struct inode *inode,
5802                 struct nfs4_label *ilabel,
5803                 struct nfs_fattr *fattr,
5804                 struct nfs4_label *olabel)
5805 {
5806
5807         struct iattr sattr = {0};
5808         struct nfs_server *server = NFS_SERVER(inode);
5809         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5810         struct nfs_setattrargs arg = {
5811                 .fh             = NFS_FH(inode),
5812                 .iap            = &sattr,
5813                 .server         = server,
5814                 .bitmask        = bitmask,
5815                 .label          = ilabel,
5816         };
5817         struct nfs_setattrres res = {
5818                 .fattr          = fattr,
5819                 .label          = olabel,
5820                 .server         = server,
5821         };
5822         struct rpc_message msg = {
5823                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5824                 .rpc_argp       = &arg,
5825                 .rpc_resp       = &res,
5826         };
5827         int status;
5828
5829         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5830
5831         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5832         if (status)
5833                 dprintk("%s failed: %d\n", __func__, status);
5834
5835         return status;
5836 }
5837
5838 static int nfs4_do_set_security_label(struct inode *inode,
5839                 struct nfs4_label *ilabel,
5840                 struct nfs_fattr *fattr,
5841                 struct nfs4_label *olabel)
5842 {
5843         struct nfs4_exception exception = { };
5844         int err;
5845
5846         do {
5847                 err = _nfs4_do_set_security_label(inode, ilabel,
5848                                 fattr, olabel);
5849                 trace_nfs4_set_security_label(inode, err);
5850                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5851                                 &exception);
5852         } while (exception.retry);
5853         return err;
5854 }
5855
5856 static int
5857 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5858 {
5859         struct nfs4_label ilabel, *olabel = NULL;
5860         struct nfs_fattr fattr;
5861         int status;
5862
5863         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5864                 return -EOPNOTSUPP;
5865
5866         nfs_fattr_init(&fattr);
5867
5868         ilabel.pi = 0;
5869         ilabel.lfs = 0;
5870         ilabel.label = (char *)buf;
5871         ilabel.len = buflen;
5872
5873         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5874         if (IS_ERR(olabel)) {
5875                 status = -PTR_ERR(olabel);
5876                 goto out;
5877         }
5878
5879         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5880         if (status == 0)
5881                 nfs_setsecurity(inode, &fattr, olabel);
5882
5883         nfs4_label_free(olabel);
5884 out:
5885         return status;
5886 }
5887 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5888
5889
5890 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5891                                     nfs4_verifier *bootverf)
5892 {
5893         __be32 verf[2];
5894
5895         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5896                 /* An impossible timestamp guarantees this value
5897                  * will never match a generated boot time. */
5898                 verf[0] = cpu_to_be32(U32_MAX);
5899                 verf[1] = cpu_to_be32(U32_MAX);
5900         } else {
5901                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5902                 u64 ns = ktime_to_ns(nn->boot_time);
5903
5904                 verf[0] = cpu_to_be32(ns >> 32);
5905                 verf[1] = cpu_to_be32(ns);
5906         }
5907         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5908 }
5909
5910 static int
5911 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5912 {
5913         size_t len;
5914         char *str;
5915
5916         if (clp->cl_owner_id != NULL)
5917                 return 0;
5918
5919         rcu_read_lock();
5920         len = 14 +
5921                 strlen(clp->cl_rpcclient->cl_nodename) +
5922                 1 +
5923                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5924                 1;
5925         rcu_read_unlock();
5926         if (nfs4_client_id_uniquifier[0] != '\0')
5927                 len += strlen(nfs4_client_id_uniquifier) + 1;
5928         if (len > NFS4_OPAQUE_LIMIT + 1)
5929                 return -EINVAL;
5930
5931         /*
5932          * Since this string is allocated at mount time, and held until the
5933          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5934          * about a memory-reclaim deadlock.
5935          */
5936         str = kmalloc(len, GFP_KERNEL);
5937         if (!str)
5938                 return -ENOMEM;
5939
5940         rcu_read_lock();
5941         if (nfs4_client_id_uniquifier[0] != '\0')
5942                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5943                           clp->cl_rpcclient->cl_nodename,
5944                           nfs4_client_id_uniquifier,
5945                           rpc_peeraddr2str(clp->cl_rpcclient,
5946                                            RPC_DISPLAY_ADDR));
5947         else
5948                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
5949                           clp->cl_rpcclient->cl_nodename,
5950                           rpc_peeraddr2str(clp->cl_rpcclient,
5951                                            RPC_DISPLAY_ADDR));
5952         rcu_read_unlock();
5953
5954         clp->cl_owner_id = str;
5955         return 0;
5956 }
5957
5958 static int
5959 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5960 {
5961         size_t len;
5962         char *str;
5963
5964         len = 10 + 10 + 1 + 10 + 1 +
5965                 strlen(nfs4_client_id_uniquifier) + 1 +
5966                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5967
5968         if (len > NFS4_OPAQUE_LIMIT + 1)
5969                 return -EINVAL;
5970
5971         /*
5972          * Since this string is allocated at mount time, and held until the
5973          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5974          * about a memory-reclaim deadlock.
5975          */
5976         str = kmalloc(len, GFP_KERNEL);
5977         if (!str)
5978                 return -ENOMEM;
5979
5980         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5981                         clp->rpc_ops->version, clp->cl_minorversion,
5982                         nfs4_client_id_uniquifier,
5983                         clp->cl_rpcclient->cl_nodename);
5984         clp->cl_owner_id = str;
5985         return 0;
5986 }
5987
5988 static int
5989 nfs4_init_uniform_client_string(struct nfs_client *clp)
5990 {
5991         size_t len;
5992         char *str;
5993
5994         if (clp->cl_owner_id != NULL)
5995                 return 0;
5996
5997         if (nfs4_client_id_uniquifier[0] != '\0')
5998                 return nfs4_init_uniquifier_client_string(clp);
5999
6000         len = 10 + 10 + 1 + 10 + 1 +
6001                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6002
6003         if (len > NFS4_OPAQUE_LIMIT + 1)
6004                 return -EINVAL;
6005
6006         /*
6007          * Since this string is allocated at mount time, and held until the
6008          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6009          * about a memory-reclaim deadlock.
6010          */
6011         str = kmalloc(len, GFP_KERNEL);
6012         if (!str)
6013                 return -ENOMEM;
6014
6015         scnprintf(str, len, "Linux NFSv%u.%u %s",
6016                         clp->rpc_ops->version, clp->cl_minorversion,
6017                         clp->cl_rpcclient->cl_nodename);
6018         clp->cl_owner_id = str;
6019         return 0;
6020 }
6021
6022 /*
6023  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6024  * services.  Advertise one based on the address family of the
6025  * clientaddr.
6026  */
6027 static unsigned int
6028 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6029 {
6030         if (strchr(clp->cl_ipaddr, ':') != NULL)
6031                 return scnprintf(buf, len, "tcp6");
6032         else
6033                 return scnprintf(buf, len, "tcp");
6034 }
6035
6036 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6037 {
6038         struct nfs4_setclientid *sc = calldata;
6039
6040         if (task->tk_status == 0)
6041                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6042 }
6043
6044 static const struct rpc_call_ops nfs4_setclientid_ops = {
6045         .rpc_call_done = nfs4_setclientid_done,
6046 };
6047
6048 /**
6049  * nfs4_proc_setclientid - Negotiate client ID
6050  * @clp: state data structure
6051  * @program: RPC program for NFSv4 callback service
6052  * @port: IP port number for NFS4 callback service
6053  * @cred: credential to use for this call
6054  * @res: where to place the result
6055  *
6056  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6057  */
6058 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6059                 unsigned short port, const struct cred *cred,
6060                 struct nfs4_setclientid_res *res)
6061 {
6062         nfs4_verifier sc_verifier;
6063         struct nfs4_setclientid setclientid = {
6064                 .sc_verifier = &sc_verifier,
6065                 .sc_prog = program,
6066                 .sc_clnt = clp,
6067         };
6068         struct rpc_message msg = {
6069                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6070                 .rpc_argp = &setclientid,
6071                 .rpc_resp = res,
6072                 .rpc_cred = cred,
6073         };
6074         struct rpc_task_setup task_setup_data = {
6075                 .rpc_client = clp->cl_rpcclient,
6076                 .rpc_message = &msg,
6077                 .callback_ops = &nfs4_setclientid_ops,
6078                 .callback_data = &setclientid,
6079                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6080         };
6081         int status;
6082
6083         /* nfs_client_id4 */
6084         nfs4_init_boot_verifier(clp, &sc_verifier);
6085
6086         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6087                 status = nfs4_init_uniform_client_string(clp);
6088         else
6089                 status = nfs4_init_nonuniform_client_string(clp);
6090
6091         if (status)
6092                 goto out;
6093
6094         /* cb_client4 */
6095         setclientid.sc_netid_len =
6096                                 nfs4_init_callback_netid(clp,
6097                                                 setclientid.sc_netid,
6098                                                 sizeof(setclientid.sc_netid));
6099         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6100                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6101                                 clp->cl_ipaddr, port >> 8, port & 255);
6102
6103         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6104                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6105                 clp->cl_owner_id);
6106
6107         status = nfs4_call_sync_custom(&task_setup_data);
6108         if (setclientid.sc_cred) {
6109                 kfree(clp->cl_acceptor);
6110                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6111                 put_rpccred(setclientid.sc_cred);
6112         }
6113 out:
6114         trace_nfs4_setclientid(clp, status);
6115         dprintk("NFS reply setclientid: %d\n", status);
6116         return status;
6117 }
6118
6119 /**
6120  * nfs4_proc_setclientid_confirm - Confirm client ID
6121  * @clp: state data structure
6122  * @arg: result of a previous SETCLIENTID
6123  * @cred: credential to use for this call
6124  *
6125  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6126  */
6127 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6128                 struct nfs4_setclientid_res *arg,
6129                 const struct cred *cred)
6130 {
6131         struct rpc_message msg = {
6132                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6133                 .rpc_argp = arg,
6134                 .rpc_cred = cred,
6135         };
6136         int status;
6137
6138         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6139                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6140                 clp->cl_clientid);
6141         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6142                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6143         trace_nfs4_setclientid_confirm(clp, status);
6144         dprintk("NFS reply setclientid_confirm: %d\n", status);
6145         return status;
6146 }
6147
6148 struct nfs4_delegreturndata {
6149         struct nfs4_delegreturnargs args;
6150         struct nfs4_delegreturnres res;
6151         struct nfs_fh fh;
6152         nfs4_stateid stateid;
6153         unsigned long timestamp;
6154         struct {
6155                 struct nfs4_layoutreturn_args arg;
6156                 struct nfs4_layoutreturn_res res;
6157                 struct nfs4_xdr_opaque_data ld_private;
6158                 u32 roc_barrier;
6159                 bool roc;
6160         } lr;
6161         struct nfs_fattr fattr;
6162         int rpc_status;
6163         struct inode *inode;
6164 };
6165
6166 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6167 {
6168         struct nfs4_delegreturndata *data = calldata;
6169         struct nfs4_exception exception = {
6170                 .inode = data->inode,
6171                 .stateid = &data->stateid,
6172         };
6173
6174         if (!nfs4_sequence_done(task, &data->res.seq_res))
6175                 return;
6176
6177         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6178
6179         /* Handle Layoutreturn errors */
6180         if (pnfs_roc_done(task, data->inode,
6181                                 &data->args.lr_args,
6182                                 &data->res.lr_res,
6183                                 &data->res.lr_ret) == -EAGAIN)
6184                 goto out_restart;
6185
6186         switch (task->tk_status) {
6187         case 0:
6188                 renew_lease(data->res.server, data->timestamp);
6189                 break;
6190         case -NFS4ERR_ADMIN_REVOKED:
6191         case -NFS4ERR_DELEG_REVOKED:
6192         case -NFS4ERR_EXPIRED:
6193                 nfs4_free_revoked_stateid(data->res.server,
6194                                 data->args.stateid,
6195                                 task->tk_msg.rpc_cred);
6196                 /* Fallthrough */
6197         case -NFS4ERR_BAD_STATEID:
6198         case -NFS4ERR_STALE_STATEID:
6199                 task->tk_status = 0;
6200                 break;
6201         case -NFS4ERR_OLD_STATEID:
6202                 if (nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6203                         goto out_restart;
6204                 task->tk_status = 0;
6205                 break;
6206         case -NFS4ERR_ACCESS:
6207                 if (data->args.bitmask) {
6208                         data->args.bitmask = NULL;
6209                         data->res.fattr = NULL;
6210                         goto out_restart;
6211                 }
6212                 /* Fallthrough */
6213         default:
6214                 task->tk_status = nfs4_async_handle_exception(task,
6215                                 data->res.server, task->tk_status,
6216                                 &exception);
6217                 if (exception.retry)
6218                         goto out_restart;
6219         }
6220         data->rpc_status = task->tk_status;
6221         return;
6222 out_restart:
6223         task->tk_status = 0;
6224         rpc_restart_call_prepare(task);
6225 }
6226
6227 static void nfs4_delegreturn_release(void *calldata)
6228 {
6229         struct nfs4_delegreturndata *data = calldata;
6230         struct inode *inode = data->inode;
6231
6232         if (inode) {
6233                 if (data->lr.roc)
6234                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6235                                         data->res.lr_ret);
6236                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6237                 nfs_iput_and_deactive(inode);
6238         }
6239         kfree(calldata);
6240 }
6241
6242 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6243 {
6244         struct nfs4_delegreturndata *d_data;
6245         struct pnfs_layout_hdr *lo;
6246
6247         d_data = (struct nfs4_delegreturndata *)data;
6248
6249         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
6250                 return;
6251
6252         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6253         if (lo && !pnfs_layout_is_valid(lo)) {
6254                 d_data->args.lr_args = NULL;
6255                 d_data->res.lr_res = NULL;
6256         }
6257
6258         nfs4_setup_sequence(d_data->res.server->nfs_client,
6259                         &d_data->args.seq_args,
6260                         &d_data->res.seq_res,
6261                         task);
6262 }
6263
6264 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6265         .rpc_call_prepare = nfs4_delegreturn_prepare,
6266         .rpc_call_done = nfs4_delegreturn_done,
6267         .rpc_release = nfs4_delegreturn_release,
6268 };
6269
6270 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6271 {
6272         struct nfs4_delegreturndata *data;
6273         struct nfs_server *server = NFS_SERVER(inode);
6274         struct rpc_task *task;
6275         struct rpc_message msg = {
6276                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6277                 .rpc_cred = cred,
6278         };
6279         struct rpc_task_setup task_setup_data = {
6280                 .rpc_client = server->client,
6281                 .rpc_message = &msg,
6282                 .callback_ops = &nfs4_delegreturn_ops,
6283                 .flags = RPC_TASK_ASYNC,
6284         };
6285         int status = 0;
6286
6287         data = kzalloc(sizeof(*data), GFP_NOFS);
6288         if (data == NULL)
6289                 return -ENOMEM;
6290         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6291
6292         nfs4_state_protect(server->nfs_client,
6293                         NFS_SP4_MACH_CRED_CLEANUP,
6294                         &task_setup_data.rpc_client, &msg);
6295
6296         data->args.fhandle = &data->fh;
6297         data->args.stateid = &data->stateid;
6298         data->args.bitmask = server->cache_consistency_bitmask;
6299         nfs_copy_fh(&data->fh, NFS_FH(inode));
6300         nfs4_stateid_copy(&data->stateid, stateid);
6301         data->res.fattr = &data->fattr;
6302         data->res.server = server;
6303         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6304         data->lr.arg.ld_private = &data->lr.ld_private;
6305         nfs_fattr_init(data->res.fattr);
6306         data->timestamp = jiffies;
6307         data->rpc_status = 0;
6308         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6309         data->inode = nfs_igrab_and_active(inode);
6310         if (data->inode) {
6311                 if (data->lr.roc) {
6312                         data->args.lr_args = &data->lr.arg;
6313                         data->res.lr_res = &data->lr.res;
6314                 }
6315         } else if (data->lr.roc) {
6316                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6317                 data->lr.roc = false;
6318         }
6319
6320         task_setup_data.callback_data = data;
6321         msg.rpc_argp = &data->args;
6322         msg.rpc_resp = &data->res;
6323         task = rpc_run_task(&task_setup_data);
6324         if (IS_ERR(task))
6325                 return PTR_ERR(task);
6326         if (!issync)
6327                 goto out;
6328         status = rpc_wait_for_completion_task(task);
6329         if (status != 0)
6330                 goto out;
6331         status = data->rpc_status;
6332 out:
6333         rpc_put_task(task);
6334         return status;
6335 }
6336
6337 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6338 {
6339         struct nfs_server *server = NFS_SERVER(inode);
6340         struct nfs4_exception exception = { };
6341         int err;
6342         do {
6343                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6344                 trace_nfs4_delegreturn(inode, stateid, err);
6345                 switch (err) {
6346                         case -NFS4ERR_STALE_STATEID:
6347                         case -NFS4ERR_EXPIRED:
6348                         case 0:
6349                                 return 0;
6350                 }
6351                 err = nfs4_handle_exception(server, err, &exception);
6352         } while (exception.retry);
6353         return err;
6354 }
6355
6356 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6357 {
6358         struct inode *inode = state->inode;
6359         struct nfs_server *server = NFS_SERVER(inode);
6360         struct nfs_client *clp = server->nfs_client;
6361         struct nfs_lockt_args arg = {
6362                 .fh = NFS_FH(inode),
6363                 .fl = request,
6364         };
6365         struct nfs_lockt_res res = {
6366                 .denied = request,
6367         };
6368         struct rpc_message msg = {
6369                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6370                 .rpc_argp       = &arg,
6371                 .rpc_resp       = &res,
6372                 .rpc_cred       = state->owner->so_cred,
6373         };
6374         struct nfs4_lock_state *lsp;
6375         int status;
6376
6377         arg.lock_owner.clientid = clp->cl_clientid;
6378         status = nfs4_set_lock_state(state, request);
6379         if (status != 0)
6380                 goto out;
6381         lsp = request->fl_u.nfs4_fl.owner;
6382         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6383         arg.lock_owner.s_dev = server->s_dev;
6384         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6385         switch (status) {
6386                 case 0:
6387                         request->fl_type = F_UNLCK;
6388                         break;
6389                 case -NFS4ERR_DENIED:
6390                         status = 0;
6391         }
6392         request->fl_ops->fl_release_private(request);
6393         request->fl_ops = NULL;
6394 out:
6395         return status;
6396 }
6397
6398 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6399 {
6400         struct nfs4_exception exception = {
6401                 .interruptible = true,
6402         };
6403         int err;
6404
6405         do {
6406                 err = _nfs4_proc_getlk(state, cmd, request);
6407                 trace_nfs4_get_lock(request, state, cmd, err);
6408                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6409                                 &exception);
6410         } while (exception.retry);
6411         return err;
6412 }
6413
6414 /*
6415  * Update the seqid of a lock stateid after receiving
6416  * NFS4ERR_OLD_STATEID
6417  */
6418 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6419                 struct nfs4_lock_state *lsp)
6420 {
6421         struct nfs4_state *state = lsp->ls_state;
6422         bool ret = false;
6423
6424         spin_lock(&state->state_lock);
6425         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6426                 goto out;
6427         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6428                 nfs4_stateid_seqid_inc(dst);
6429         else
6430                 dst->seqid = lsp->ls_stateid.seqid;
6431         ret = true;
6432 out:
6433         spin_unlock(&state->state_lock);
6434         return ret;
6435 }
6436
6437 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6438                 struct nfs4_lock_state *lsp)
6439 {
6440         struct nfs4_state *state = lsp->ls_state;
6441         bool ret;
6442
6443         spin_lock(&state->state_lock);
6444         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6445         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6446         spin_unlock(&state->state_lock);
6447         return ret;
6448 }
6449
6450 struct nfs4_unlockdata {
6451         struct nfs_locku_args arg;
6452         struct nfs_locku_res res;
6453         struct nfs4_lock_state *lsp;
6454         struct nfs_open_context *ctx;
6455         struct nfs_lock_context *l_ctx;
6456         struct file_lock fl;
6457         struct nfs_server *server;
6458         unsigned long timestamp;
6459 };
6460
6461 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6462                 struct nfs_open_context *ctx,
6463                 struct nfs4_lock_state *lsp,
6464                 struct nfs_seqid *seqid)
6465 {
6466         struct nfs4_unlockdata *p;
6467         struct nfs4_state *state = lsp->ls_state;
6468         struct inode *inode = state->inode;
6469
6470         p = kzalloc(sizeof(*p), GFP_NOFS);
6471         if (p == NULL)
6472                 return NULL;
6473         p->arg.fh = NFS_FH(inode);
6474         p->arg.fl = &p->fl;
6475         p->arg.seqid = seqid;
6476         p->res.seqid = seqid;
6477         p->lsp = lsp;
6478         /* Ensure we don't close file until we're done freeing locks! */
6479         p->ctx = get_nfs_open_context(ctx);
6480         p->l_ctx = nfs_get_lock_context(ctx);
6481         locks_init_lock(&p->fl);
6482         locks_copy_lock(&p->fl, fl);
6483         p->server = NFS_SERVER(inode);
6484         spin_lock(&state->state_lock);
6485         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6486         spin_unlock(&state->state_lock);
6487         return p;
6488 }
6489
6490 static void nfs4_locku_release_calldata(void *data)
6491 {
6492         struct nfs4_unlockdata *calldata = data;
6493         nfs_free_seqid(calldata->arg.seqid);
6494         nfs4_put_lock_state(calldata->lsp);
6495         nfs_put_lock_context(calldata->l_ctx);
6496         put_nfs_open_context(calldata->ctx);
6497         kfree(calldata);
6498 }
6499
6500 static void nfs4_locku_done(struct rpc_task *task, void *data)
6501 {
6502         struct nfs4_unlockdata *calldata = data;
6503         struct nfs4_exception exception = {
6504                 .inode = calldata->lsp->ls_state->inode,
6505                 .stateid = &calldata->arg.stateid,
6506         };
6507
6508         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6509                 return;
6510         switch (task->tk_status) {
6511                 case 0:
6512                         renew_lease(calldata->server, calldata->timestamp);
6513                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6514                         if (nfs4_update_lock_stateid(calldata->lsp,
6515                                         &calldata->res.stateid))
6516                                 break;
6517                         /* Fall through */
6518                 case -NFS4ERR_ADMIN_REVOKED:
6519                 case -NFS4ERR_EXPIRED:
6520                         nfs4_free_revoked_stateid(calldata->server,
6521                                         &calldata->arg.stateid,
6522                                         task->tk_msg.rpc_cred);
6523                         /* Fall through */
6524                 case -NFS4ERR_BAD_STATEID:
6525                 case -NFS4ERR_STALE_STATEID:
6526                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6527                                                 calldata->lsp))
6528                                 rpc_restart_call_prepare(task);
6529                         break;
6530                 case -NFS4ERR_OLD_STATEID:
6531                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6532                                                 calldata->lsp))
6533                                 rpc_restart_call_prepare(task);
6534                         break;
6535                 default:
6536                         task->tk_status = nfs4_async_handle_exception(task,
6537                                         calldata->server, task->tk_status,
6538                                         &exception);
6539                         if (exception.retry)
6540                                 rpc_restart_call_prepare(task);
6541         }
6542         nfs_release_seqid(calldata->arg.seqid);
6543 }
6544
6545 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6546 {
6547         struct nfs4_unlockdata *calldata = data;
6548
6549         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6550                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6551                 return;
6552
6553         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6554                 goto out_wait;
6555         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6556                 /* Note: exit _without_ running nfs4_locku_done */
6557                 goto out_no_action;
6558         }
6559         calldata->timestamp = jiffies;
6560         if (nfs4_setup_sequence(calldata->server->nfs_client,
6561                                 &calldata->arg.seq_args,
6562                                 &calldata->res.seq_res,
6563                                 task) != 0)
6564                 nfs_release_seqid(calldata->arg.seqid);
6565         return;
6566 out_no_action:
6567         task->tk_action = NULL;
6568 out_wait:
6569         nfs4_sequence_done(task, &calldata->res.seq_res);
6570 }
6571
6572 static const struct rpc_call_ops nfs4_locku_ops = {
6573         .rpc_call_prepare = nfs4_locku_prepare,
6574         .rpc_call_done = nfs4_locku_done,
6575         .rpc_release = nfs4_locku_release_calldata,
6576 };
6577
6578 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6579                 struct nfs_open_context *ctx,
6580                 struct nfs4_lock_state *lsp,
6581                 struct nfs_seqid *seqid)
6582 {
6583         struct nfs4_unlockdata *data;
6584         struct rpc_message msg = {
6585                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6586                 .rpc_cred = ctx->cred,
6587         };
6588         struct rpc_task_setup task_setup_data = {
6589                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6590                 .rpc_message = &msg,
6591                 .callback_ops = &nfs4_locku_ops,
6592                 .workqueue = nfsiod_workqueue,
6593                 .flags = RPC_TASK_ASYNC,
6594         };
6595
6596         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6597                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6598
6599         /* Ensure this is an unlock - when canceling a lock, the
6600          * canceled lock is passed in, and it won't be an unlock.
6601          */
6602         fl->fl_type = F_UNLCK;
6603         if (fl->fl_flags & FL_CLOSE)
6604                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6605
6606         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6607         if (data == NULL) {
6608                 nfs_free_seqid(seqid);
6609                 return ERR_PTR(-ENOMEM);
6610         }
6611
6612         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6613         msg.rpc_argp = &data->arg;
6614         msg.rpc_resp = &data->res;
6615         task_setup_data.callback_data = data;
6616         return rpc_run_task(&task_setup_data);
6617 }
6618
6619 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6620 {
6621         struct inode *inode = state->inode;
6622         struct nfs4_state_owner *sp = state->owner;
6623         struct nfs_inode *nfsi = NFS_I(inode);
6624         struct nfs_seqid *seqid;
6625         struct nfs4_lock_state *lsp;
6626         struct rpc_task *task;
6627         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6628         int status = 0;
6629         unsigned char fl_flags = request->fl_flags;
6630
6631         status = nfs4_set_lock_state(state, request);
6632         /* Unlock _before_ we do the RPC call */
6633         request->fl_flags |= FL_EXISTS;
6634         /* Exclude nfs_delegation_claim_locks() */
6635         mutex_lock(&sp->so_delegreturn_mutex);
6636         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6637         down_read(&nfsi->rwsem);
6638         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6639                 up_read(&nfsi->rwsem);
6640                 mutex_unlock(&sp->so_delegreturn_mutex);
6641                 goto out;
6642         }
6643         up_read(&nfsi->rwsem);
6644         mutex_unlock(&sp->so_delegreturn_mutex);
6645         if (status != 0)
6646                 goto out;
6647         /* Is this a delegated lock? */
6648         lsp = request->fl_u.nfs4_fl.owner;
6649         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6650                 goto out;
6651         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6652         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6653         status = -ENOMEM;
6654         if (IS_ERR(seqid))
6655                 goto out;
6656         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6657         status = PTR_ERR(task);
6658         if (IS_ERR(task))
6659                 goto out;
6660         status = rpc_wait_for_completion_task(task);
6661         rpc_put_task(task);
6662 out:
6663         request->fl_flags = fl_flags;
6664         trace_nfs4_unlock(request, state, F_SETLK, status);
6665         return status;
6666 }
6667
6668 struct nfs4_lockdata {
6669         struct nfs_lock_args arg;
6670         struct nfs_lock_res res;
6671         struct nfs4_lock_state *lsp;
6672         struct nfs_open_context *ctx;
6673         struct file_lock fl;
6674         unsigned long timestamp;
6675         int rpc_status;
6676         int cancelled;
6677         struct nfs_server *server;
6678 };
6679
6680 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6681                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6682                 gfp_t gfp_mask)
6683 {
6684         struct nfs4_lockdata *p;
6685         struct inode *inode = lsp->ls_state->inode;
6686         struct nfs_server *server = NFS_SERVER(inode);
6687         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6688
6689         p = kzalloc(sizeof(*p), gfp_mask);
6690         if (p == NULL)
6691                 return NULL;
6692
6693         p->arg.fh = NFS_FH(inode);
6694         p->arg.fl = &p->fl;
6695         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6696         if (IS_ERR(p->arg.open_seqid))
6697                 goto out_free;
6698         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6699         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6700         if (IS_ERR(p->arg.lock_seqid))
6701                 goto out_free_seqid;
6702         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6703         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6704         p->arg.lock_owner.s_dev = server->s_dev;
6705         p->res.lock_seqid = p->arg.lock_seqid;
6706         p->lsp = lsp;
6707         p->server = server;
6708         p->ctx = get_nfs_open_context(ctx);
6709         locks_init_lock(&p->fl);
6710         locks_copy_lock(&p->fl, fl);
6711         return p;
6712 out_free_seqid:
6713         nfs_free_seqid(p->arg.open_seqid);
6714 out_free:
6715         kfree(p);
6716         return NULL;
6717 }
6718
6719 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6720 {
6721         struct nfs4_lockdata *data = calldata;
6722         struct nfs4_state *state = data->lsp->ls_state;
6723
6724         dprintk("%s: begin!\n", __func__);
6725         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6726                 goto out_wait;
6727         /* Do we need to do an open_to_lock_owner? */
6728         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6729                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6730                         goto out_release_lock_seqid;
6731                 }
6732                 nfs4_stateid_copy(&data->arg.open_stateid,
6733                                 &state->open_stateid);
6734                 data->arg.new_lock_owner = 1;
6735                 data->res.open_seqid = data->arg.open_seqid;
6736         } else {
6737                 data->arg.new_lock_owner = 0;
6738                 nfs4_stateid_copy(&data->arg.lock_stateid,
6739                                 &data->lsp->ls_stateid);
6740         }
6741         if (!nfs4_valid_open_stateid(state)) {
6742                 data->rpc_status = -EBADF;
6743                 task->tk_action = NULL;
6744                 goto out_release_open_seqid;
6745         }
6746         data->timestamp = jiffies;
6747         if (nfs4_setup_sequence(data->server->nfs_client,
6748                                 &data->arg.seq_args,
6749                                 &data->res.seq_res,
6750                                 task) == 0)
6751                 return;
6752 out_release_open_seqid:
6753         nfs_release_seqid(data->arg.open_seqid);
6754 out_release_lock_seqid:
6755         nfs_release_seqid(data->arg.lock_seqid);
6756 out_wait:
6757         nfs4_sequence_done(task, &data->res.seq_res);
6758         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6759 }
6760
6761 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6762 {
6763         struct nfs4_lockdata *data = calldata;
6764         struct nfs4_lock_state *lsp = data->lsp;
6765
6766         dprintk("%s: begin!\n", __func__);
6767
6768         if (!nfs4_sequence_done(task, &data->res.seq_res))
6769                 return;
6770
6771         data->rpc_status = task->tk_status;
6772         switch (task->tk_status) {
6773         case 0:
6774                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6775                                 data->timestamp);
6776                 if (data->arg.new_lock && !data->cancelled) {
6777                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6778                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6779                                 goto out_restart;
6780                 }
6781                 if (data->arg.new_lock_owner != 0) {
6782                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6783                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6784                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6785                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6786                         goto out_restart;
6787                 break;
6788         case -NFS4ERR_BAD_STATEID:
6789         case -NFS4ERR_OLD_STATEID:
6790         case -NFS4ERR_STALE_STATEID:
6791         case -NFS4ERR_EXPIRED:
6792                 if (data->arg.new_lock_owner != 0) {
6793                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6794                                                 &lsp->ls_state->open_stateid))
6795                                 goto out_restart;
6796                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6797                                                 &lsp->ls_stateid))
6798                                 goto out_restart;
6799         }
6800 out_done:
6801         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6802         return;
6803 out_restart:
6804         if (!data->cancelled)
6805                 rpc_restart_call_prepare(task);
6806         goto out_done;
6807 }
6808
6809 static void nfs4_lock_release(void *calldata)
6810 {
6811         struct nfs4_lockdata *data = calldata;
6812
6813         dprintk("%s: begin!\n", __func__);
6814         nfs_free_seqid(data->arg.open_seqid);
6815         if (data->cancelled && data->rpc_status == 0) {
6816                 struct rpc_task *task;
6817                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6818                                 data->arg.lock_seqid);
6819                 if (!IS_ERR(task))
6820                         rpc_put_task_async(task);
6821                 dprintk("%s: cancelling lock!\n", __func__);
6822         } else
6823                 nfs_free_seqid(data->arg.lock_seqid);
6824         nfs4_put_lock_state(data->lsp);
6825         put_nfs_open_context(data->ctx);
6826         kfree(data);
6827         dprintk("%s: done!\n", __func__);
6828 }
6829
6830 static const struct rpc_call_ops nfs4_lock_ops = {
6831         .rpc_call_prepare = nfs4_lock_prepare,
6832         .rpc_call_done = nfs4_lock_done,
6833         .rpc_release = nfs4_lock_release,
6834 };
6835
6836 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6837 {
6838         switch (error) {
6839         case -NFS4ERR_ADMIN_REVOKED:
6840         case -NFS4ERR_EXPIRED:
6841         case -NFS4ERR_BAD_STATEID:
6842                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6843                 if (new_lock_owner != 0 ||
6844                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6845                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6846                 break;
6847         case -NFS4ERR_STALE_STATEID:
6848                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6849                 nfs4_schedule_lease_recovery(server->nfs_client);
6850         };
6851 }
6852
6853 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6854 {
6855         struct nfs4_lockdata *data;
6856         struct rpc_task *task;
6857         struct rpc_message msg = {
6858                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6859                 .rpc_cred = state->owner->so_cred,
6860         };
6861         struct rpc_task_setup task_setup_data = {
6862                 .rpc_client = NFS_CLIENT(state->inode),
6863                 .rpc_message = &msg,
6864                 .callback_ops = &nfs4_lock_ops,
6865                 .workqueue = nfsiod_workqueue,
6866                 .flags = RPC_TASK_ASYNC,
6867         };
6868         int ret;
6869
6870         dprintk("%s: begin!\n", __func__);
6871         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6872                         fl->fl_u.nfs4_fl.owner,
6873                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6874         if (data == NULL)
6875                 return -ENOMEM;
6876         if (IS_SETLKW(cmd))
6877                 data->arg.block = 1;
6878         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6879                                 recovery_type > NFS_LOCK_NEW);
6880         msg.rpc_argp = &data->arg;
6881         msg.rpc_resp = &data->res;
6882         task_setup_data.callback_data = data;
6883         if (recovery_type > NFS_LOCK_NEW) {
6884                 if (recovery_type == NFS_LOCK_RECLAIM)
6885                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6886         } else
6887                 data->arg.new_lock = 1;
6888         task = rpc_run_task(&task_setup_data);
6889         if (IS_ERR(task))
6890                 return PTR_ERR(task);
6891         ret = rpc_wait_for_completion_task(task);
6892         if (ret == 0) {
6893                 ret = data->rpc_status;
6894                 if (ret)
6895                         nfs4_handle_setlk_error(data->server, data->lsp,
6896                                         data->arg.new_lock_owner, ret);
6897         } else
6898                 data->cancelled = true;
6899         rpc_put_task(task);
6900         dprintk("%s: done, ret = %d!\n", __func__, ret);
6901         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6902         return ret;
6903 }
6904
6905 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6906 {
6907         struct nfs_server *server = NFS_SERVER(state->inode);
6908         struct nfs4_exception exception = {
6909                 .inode = state->inode,
6910         };
6911         int err;
6912
6913         do {
6914                 /* Cache the lock if possible... */
6915                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6916                         return 0;
6917                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6918                 if (err != -NFS4ERR_DELAY)
6919                         break;
6920                 nfs4_handle_exception(server, err, &exception);
6921         } while (exception.retry);
6922         return err;
6923 }
6924
6925 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6926 {
6927         struct nfs_server *server = NFS_SERVER(state->inode);
6928         struct nfs4_exception exception = {
6929                 .inode = state->inode,
6930         };
6931         int err;
6932
6933         err = nfs4_set_lock_state(state, request);
6934         if (err != 0)
6935                 return err;
6936         if (!recover_lost_locks) {
6937                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6938                 return 0;
6939         }
6940         do {
6941                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6942                         return 0;
6943                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6944                 switch (err) {
6945                 default:
6946                         goto out;
6947                 case -NFS4ERR_GRACE:
6948                 case -NFS4ERR_DELAY:
6949                         nfs4_handle_exception(server, err, &exception);
6950                         err = 0;
6951                 }
6952         } while (exception.retry);
6953 out:
6954         return err;
6955 }
6956
6957 #if defined(CONFIG_NFS_V4_1)
6958 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6959 {
6960         struct nfs4_lock_state *lsp;
6961         int status;
6962
6963         status = nfs4_set_lock_state(state, request);
6964         if (status != 0)
6965                 return status;
6966         lsp = request->fl_u.nfs4_fl.owner;
6967         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6968             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6969                 return 0;
6970         return nfs4_lock_expired(state, request);
6971 }
6972 #endif
6973
6974 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6975 {
6976         struct nfs_inode *nfsi = NFS_I(state->inode);
6977         struct nfs4_state_owner *sp = state->owner;
6978         unsigned char fl_flags = request->fl_flags;
6979         int status;
6980
6981         request->fl_flags |= FL_ACCESS;
6982         status = locks_lock_inode_wait(state->inode, request);
6983         if (status < 0)
6984                 goto out;
6985         mutex_lock(&sp->so_delegreturn_mutex);
6986         down_read(&nfsi->rwsem);
6987         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6988                 /* Yes: cache locks! */
6989                 /* ...but avoid races with delegation recall... */
6990                 request->fl_flags = fl_flags & ~FL_SLEEP;
6991                 status = locks_lock_inode_wait(state->inode, request);
6992                 up_read(&nfsi->rwsem);
6993                 mutex_unlock(&sp->so_delegreturn_mutex);
6994                 goto out;
6995         }
6996         up_read(&nfsi->rwsem);
6997         mutex_unlock(&sp->so_delegreturn_mutex);
6998         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6999 out:
7000         request->fl_flags = fl_flags;
7001         return status;
7002 }
7003
7004 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7005 {
7006         struct nfs4_exception exception = {
7007                 .state = state,
7008                 .inode = state->inode,
7009                 .interruptible = true,
7010         };
7011         int err;
7012
7013         do {
7014                 err = _nfs4_proc_setlk(state, cmd, request);
7015                 if (err == -NFS4ERR_DENIED)
7016                         err = -EAGAIN;
7017                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7018                                 err, &exception);
7019         } while (exception.retry);
7020         return err;
7021 }
7022
7023 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7024 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7025
7026 static int
7027 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7028                         struct file_lock *request)
7029 {
7030         int             status = -ERESTARTSYS;
7031         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7032
7033         while(!signalled()) {
7034                 status = nfs4_proc_setlk(state, cmd, request);
7035                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7036                         break;
7037                 freezable_schedule_timeout_interruptible(timeout);
7038                 timeout *= 2;
7039                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7040                 status = -ERESTARTSYS;
7041         }
7042         return status;
7043 }
7044
7045 #ifdef CONFIG_NFS_V4_1
7046 struct nfs4_lock_waiter {
7047         struct task_struct      *task;
7048         struct inode            *inode;
7049         struct nfs_lowner       *owner;
7050 };
7051
7052 static int
7053 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7054 {
7055         int ret;
7056         struct nfs4_lock_waiter *waiter = wait->private;
7057
7058         /* NULL key means to wake up everyone */
7059         if (key) {
7060                 struct cb_notify_lock_args      *cbnl = key;
7061                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7062                                                 *wowner = waiter->owner;
7063
7064                 /* Only wake if the callback was for the same owner. */
7065                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7066                         return 0;
7067
7068                 /* Make sure it's for the right inode */
7069                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7070                         return 0;
7071         }
7072
7073         /* override "private" so we can use default_wake_function */
7074         wait->private = waiter->task;
7075         ret = woken_wake_function(wait, mode, flags, key);
7076         if (ret)
7077                 list_del_init(&wait->entry);
7078         wait->private = waiter;
7079         return ret;
7080 }
7081
7082 static int
7083 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7084 {
7085         int status = -ERESTARTSYS;
7086         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7087         struct nfs_server *server = NFS_SERVER(state->inode);
7088         struct nfs_client *clp = server->nfs_client;
7089         wait_queue_head_t *q = &clp->cl_lock_waitq;
7090         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7091                                     .id = lsp->ls_seqid.owner_id,
7092                                     .s_dev = server->s_dev };
7093         struct nfs4_lock_waiter waiter = { .task  = current,
7094                                            .inode = state->inode,
7095                                            .owner = &owner};
7096         wait_queue_entry_t wait;
7097
7098         /* Don't bother with waitqueue if we don't expect a callback */
7099         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7100                 return nfs4_retry_setlk_simple(state, cmd, request);
7101
7102         init_wait(&wait);
7103         wait.private = &waiter;
7104         wait.func = nfs4_wake_lock_waiter;
7105
7106         while(!signalled()) {
7107                 add_wait_queue(q, &wait);
7108                 status = nfs4_proc_setlk(state, cmd, request);
7109                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7110                         finish_wait(q, &wait);
7111                         break;
7112                 }
7113
7114                 status = -ERESTARTSYS;
7115                 freezer_do_not_count();
7116                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7117                 freezer_count();
7118                 finish_wait(q, &wait);
7119         }
7120
7121         return status;
7122 }
7123 #else /* !CONFIG_NFS_V4_1 */
7124 static inline int
7125 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7126 {
7127         return nfs4_retry_setlk_simple(state, cmd, request);
7128 }
7129 #endif
7130
7131 static int
7132 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7133 {
7134         struct nfs_open_context *ctx;
7135         struct nfs4_state *state;
7136         int status;
7137
7138         /* verify open state */
7139         ctx = nfs_file_open_context(filp);
7140         state = ctx->state;
7141
7142         if (IS_GETLK(cmd)) {
7143                 if (state != NULL)
7144                         return nfs4_proc_getlk(state, F_GETLK, request);
7145                 return 0;
7146         }
7147
7148         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7149                 return -EINVAL;
7150
7151         if (request->fl_type == F_UNLCK) {
7152                 if (state != NULL)
7153                         return nfs4_proc_unlck(state, cmd, request);
7154                 return 0;
7155         }
7156
7157         if (state == NULL)
7158                 return -ENOLCK;
7159
7160         if ((request->fl_flags & FL_POSIX) &&
7161             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7162                 return -ENOLCK;
7163
7164         /*
7165          * Don't rely on the VFS having checked the file open mode,
7166          * since it won't do this for flock() locks.
7167          */
7168         switch (request->fl_type) {
7169         case F_RDLCK:
7170                 if (!(filp->f_mode & FMODE_READ))
7171                         return -EBADF;
7172                 break;
7173         case F_WRLCK:
7174                 if (!(filp->f_mode & FMODE_WRITE))
7175                         return -EBADF;
7176         }
7177
7178         status = nfs4_set_lock_state(state, request);
7179         if (status != 0)
7180                 return status;
7181
7182         return nfs4_retry_setlk(state, cmd, request);
7183 }
7184
7185 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7186 {
7187         struct nfs_server *server = NFS_SERVER(state->inode);
7188         int err;
7189
7190         err = nfs4_set_lock_state(state, fl);
7191         if (err != 0)
7192                 return err;
7193         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7194         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7195 }
7196
7197 struct nfs_release_lockowner_data {
7198         struct nfs4_lock_state *lsp;
7199         struct nfs_server *server;
7200         struct nfs_release_lockowner_args args;
7201         struct nfs_release_lockowner_res res;
7202         unsigned long timestamp;
7203 };
7204
7205 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7206 {
7207         struct nfs_release_lockowner_data *data = calldata;
7208         struct nfs_server *server = data->server;
7209         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7210                            &data->res.seq_res, task);
7211         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7212         data->timestamp = jiffies;
7213 }
7214
7215 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7216 {
7217         struct nfs_release_lockowner_data *data = calldata;
7218         struct nfs_server *server = data->server;
7219
7220         nfs40_sequence_done(task, &data->res.seq_res);
7221
7222         switch (task->tk_status) {
7223         case 0:
7224                 renew_lease(server, data->timestamp);
7225                 break;
7226         case -NFS4ERR_STALE_CLIENTID:
7227         case -NFS4ERR_EXPIRED:
7228                 nfs4_schedule_lease_recovery(server->nfs_client);
7229                 break;
7230         case -NFS4ERR_LEASE_MOVED:
7231         case -NFS4ERR_DELAY:
7232                 if (nfs4_async_handle_error(task, server,
7233                                             NULL, NULL) == -EAGAIN)
7234                         rpc_restart_call_prepare(task);
7235         }
7236 }
7237
7238 static void nfs4_release_lockowner_release(void *calldata)
7239 {
7240         struct nfs_release_lockowner_data *data = calldata;
7241         nfs4_free_lock_state(data->server, data->lsp);
7242         kfree(calldata);
7243 }
7244
7245 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7246         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7247         .rpc_call_done = nfs4_release_lockowner_done,
7248         .rpc_release = nfs4_release_lockowner_release,
7249 };
7250
7251 static void
7252 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7253 {
7254         struct nfs_release_lockowner_data *data;
7255         struct rpc_message msg = {
7256                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7257         };
7258
7259         if (server->nfs_client->cl_mvops->minor_version != 0)
7260                 return;
7261
7262         data = kmalloc(sizeof(*data), GFP_NOFS);
7263         if (!data)
7264                 return;
7265         data->lsp = lsp;
7266         data->server = server;
7267         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7268         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7269         data->args.lock_owner.s_dev = server->s_dev;
7270
7271         msg.rpc_argp = &data->args;
7272         msg.rpc_resp = &data->res;
7273         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7274         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7275 }
7276
7277 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7278
7279 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7280                                    struct dentry *unused, struct inode *inode,
7281                                    const char *key, const void *buf,
7282                                    size_t buflen, int flags)
7283 {
7284         return nfs4_proc_set_acl(inode, buf, buflen);
7285 }
7286
7287 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7288                                    struct dentry *unused, struct inode *inode,
7289                                    const char *key, void *buf, size_t buflen)
7290 {
7291         return nfs4_proc_get_acl(inode, buf, buflen);
7292 }
7293
7294 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7295 {
7296         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7297 }
7298
7299 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7300
7301 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7302                                      struct dentry *unused, struct inode *inode,
7303                                      const char *key, const void *buf,
7304                                      size_t buflen, int flags)
7305 {
7306         if (security_ismaclabel(key))
7307                 return nfs4_set_security_label(inode, buf, buflen);
7308
7309         return -EOPNOTSUPP;
7310 }
7311
7312 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7313                                      struct dentry *unused, struct inode *inode,
7314                                      const char *key, void *buf, size_t buflen)
7315 {
7316         if (security_ismaclabel(key))
7317                 return nfs4_get_security_label(inode, buf, buflen);
7318         return -EOPNOTSUPP;
7319 }
7320
7321 static ssize_t
7322 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7323 {
7324         int len = 0;
7325
7326         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7327                 len = security_inode_listsecurity(inode, list, list_len);
7328                 if (list_len && len > list_len)
7329                         return -ERANGE;
7330         }
7331         return len;
7332 }
7333
7334 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7335         .prefix = XATTR_SECURITY_PREFIX,
7336         .get    = nfs4_xattr_get_nfs4_label,
7337         .set    = nfs4_xattr_set_nfs4_label,
7338 };
7339
7340 #else
7341
7342 static ssize_t
7343 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7344 {
7345         return 0;
7346 }
7347
7348 #endif
7349
7350 /*
7351  * nfs_fhget will use either the mounted_on_fileid or the fileid
7352  */
7353 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7354 {
7355         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7356                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7357               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7358               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7359                 return;
7360
7361         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7362                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7363         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7364         fattr->nlink = 2;
7365 }
7366
7367 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7368                                    const struct qstr *name,
7369                                    struct nfs4_fs_locations *fs_locations,
7370                                    struct page *page)
7371 {
7372         struct nfs_server *server = NFS_SERVER(dir);
7373         u32 bitmask[3];
7374         struct nfs4_fs_locations_arg args = {
7375                 .dir_fh = NFS_FH(dir),
7376                 .name = name,
7377                 .page = page,
7378                 .bitmask = bitmask,
7379         };
7380         struct nfs4_fs_locations_res res = {
7381                 .fs_locations = fs_locations,
7382         };
7383         struct rpc_message msg = {
7384                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7385                 .rpc_argp = &args,
7386                 .rpc_resp = &res,
7387         };
7388         int status;
7389
7390         dprintk("%s: start\n", __func__);
7391
7392         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7393         bitmask[1] = nfs4_fattr_bitmap[1];
7394
7395         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7396          * is not supported */
7397         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7398                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7399         else
7400                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7401
7402         nfs_fattr_init(&fs_locations->fattr);
7403         fs_locations->server = server;
7404         fs_locations->nlocations = 0;
7405         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7406         dprintk("%s: returned status = %d\n", __func__, status);
7407         return status;
7408 }
7409
7410 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7411                            const struct qstr *name,
7412                            struct nfs4_fs_locations *fs_locations,
7413                            struct page *page)
7414 {
7415         struct nfs4_exception exception = {
7416                 .interruptible = true,
7417         };
7418         int err;
7419         do {
7420                 err = _nfs4_proc_fs_locations(client, dir, name,
7421                                 fs_locations, page);
7422                 trace_nfs4_get_fs_locations(dir, name, err);
7423                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7424                                 &exception);
7425         } while (exception.retry);
7426         return err;
7427 }
7428
7429 /*
7430  * This operation also signals the server that this client is
7431  * performing migration recovery.  The server can stop returning
7432  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7433  * appended to this compound to identify the client ID which is
7434  * performing recovery.
7435  */
7436 static int _nfs40_proc_get_locations(struct inode *inode,
7437                                      struct nfs4_fs_locations *locations,
7438                                      struct page *page, const struct cred *cred)
7439 {
7440         struct nfs_server *server = NFS_SERVER(inode);
7441         struct rpc_clnt *clnt = server->client;
7442         u32 bitmask[2] = {
7443                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7444         };
7445         struct nfs4_fs_locations_arg args = {
7446                 .clientid       = server->nfs_client->cl_clientid,
7447                 .fh             = NFS_FH(inode),
7448                 .page           = page,
7449                 .bitmask        = bitmask,
7450                 .migration      = 1,            /* skip LOOKUP */
7451                 .renew          = 1,            /* append RENEW */
7452         };
7453         struct nfs4_fs_locations_res res = {
7454                 .fs_locations   = locations,
7455                 .migration      = 1,
7456                 .renew          = 1,
7457         };
7458         struct rpc_message msg = {
7459                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7460                 .rpc_argp       = &args,
7461                 .rpc_resp       = &res,
7462                 .rpc_cred       = cred,
7463         };
7464         unsigned long now = jiffies;
7465         int status;
7466
7467         nfs_fattr_init(&locations->fattr);
7468         locations->server = server;
7469         locations->nlocations = 0;
7470
7471         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7472         status = nfs4_call_sync_sequence(clnt, server, &msg,
7473                                         &args.seq_args, &res.seq_res);
7474         if (status)
7475                 return status;
7476
7477         renew_lease(server, now);
7478         return 0;
7479 }
7480
7481 #ifdef CONFIG_NFS_V4_1
7482
7483 /*
7484  * This operation also signals the server that this client is
7485  * performing migration recovery.  The server can stop asserting
7486  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7487  * performing this operation is identified in the SEQUENCE
7488  * operation in this compound.
7489  *
7490  * When the client supports GETATTR(fs_locations_info), it can
7491  * be plumbed in here.
7492  */
7493 static int _nfs41_proc_get_locations(struct inode *inode,
7494                                      struct nfs4_fs_locations *locations,
7495                                      struct page *page, const struct cred *cred)
7496 {
7497         struct nfs_server *server = NFS_SERVER(inode);
7498         struct rpc_clnt *clnt = server->client;
7499         u32 bitmask[2] = {
7500                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7501         };
7502         struct nfs4_fs_locations_arg args = {
7503                 .fh             = NFS_FH(inode),
7504                 .page           = page,
7505                 .bitmask        = bitmask,
7506                 .migration      = 1,            /* skip LOOKUP */
7507         };
7508         struct nfs4_fs_locations_res res = {
7509                 .fs_locations   = locations,
7510                 .migration      = 1,
7511         };
7512         struct rpc_message msg = {
7513                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7514                 .rpc_argp       = &args,
7515                 .rpc_resp       = &res,
7516                 .rpc_cred       = cred,
7517         };
7518         int status;
7519
7520         nfs_fattr_init(&locations->fattr);
7521         locations->server = server;
7522         locations->nlocations = 0;
7523
7524         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7525         status = nfs4_call_sync_sequence(clnt, server, &msg,
7526                                         &args.seq_args, &res.seq_res);
7527         if (status == NFS4_OK &&
7528             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7529                 status = -NFS4ERR_LEASE_MOVED;
7530         return status;
7531 }
7532
7533 #endif  /* CONFIG_NFS_V4_1 */
7534
7535 /**
7536  * nfs4_proc_get_locations - discover locations for a migrated FSID
7537  * @inode: inode on FSID that is migrating
7538  * @locations: result of query
7539  * @page: buffer
7540  * @cred: credential to use for this operation
7541  *
7542  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7543  * operation failed, or a negative errno if a local error occurred.
7544  *
7545  * On success, "locations" is filled in, but if the server has
7546  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7547  * asserted.
7548  *
7549  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7550  * from this client that require migration recovery.
7551  */
7552 int nfs4_proc_get_locations(struct inode *inode,
7553                             struct nfs4_fs_locations *locations,
7554                             struct page *page, const struct cred *cred)
7555 {
7556         struct nfs_server *server = NFS_SERVER(inode);
7557         struct nfs_client *clp = server->nfs_client;
7558         const struct nfs4_mig_recovery_ops *ops =
7559                                         clp->cl_mvops->mig_recovery_ops;
7560         struct nfs4_exception exception = {
7561                 .interruptible = true,
7562         };
7563         int status;
7564
7565         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7566                 (unsigned long long)server->fsid.major,
7567                 (unsigned long long)server->fsid.minor,
7568                 clp->cl_hostname);
7569         nfs_display_fhandle(NFS_FH(inode), __func__);
7570
7571         do {
7572                 status = ops->get_locations(inode, locations, page, cred);
7573                 if (status != -NFS4ERR_DELAY)
7574                         break;
7575                 nfs4_handle_exception(server, status, &exception);
7576         } while (exception.retry);
7577         return status;
7578 }
7579
7580 /*
7581  * This operation also signals the server that this client is
7582  * performing "lease moved" recovery.  The server can stop
7583  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7584  * is appended to this compound to identify the client ID which is
7585  * performing recovery.
7586  */
7587 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7588 {
7589         struct nfs_server *server = NFS_SERVER(inode);
7590         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7591         struct rpc_clnt *clnt = server->client;
7592         struct nfs4_fsid_present_arg args = {
7593                 .fh             = NFS_FH(inode),
7594                 .clientid       = clp->cl_clientid,
7595                 .renew          = 1,            /* append RENEW */
7596         };
7597         struct nfs4_fsid_present_res res = {
7598                 .renew          = 1,
7599         };
7600         struct rpc_message msg = {
7601                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7602                 .rpc_argp       = &args,
7603                 .rpc_resp       = &res,
7604                 .rpc_cred       = cred,
7605         };
7606         unsigned long now = jiffies;
7607         int status;
7608
7609         res.fh = nfs_alloc_fhandle();
7610         if (res.fh == NULL)
7611                 return -ENOMEM;
7612
7613         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7614         status = nfs4_call_sync_sequence(clnt, server, &msg,
7615                                                 &args.seq_args, &res.seq_res);
7616         nfs_free_fhandle(res.fh);
7617         if (status)
7618                 return status;
7619
7620         do_renew_lease(clp, now);
7621         return 0;
7622 }
7623
7624 #ifdef CONFIG_NFS_V4_1
7625
7626 /*
7627  * This operation also signals the server that this client is
7628  * performing "lease moved" recovery.  The server can stop asserting
7629  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7630  * this operation is identified in the SEQUENCE operation in this
7631  * compound.
7632  */
7633 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7634 {
7635         struct nfs_server *server = NFS_SERVER(inode);
7636         struct rpc_clnt *clnt = server->client;
7637         struct nfs4_fsid_present_arg args = {
7638                 .fh             = NFS_FH(inode),
7639         };
7640         struct nfs4_fsid_present_res res = {
7641         };
7642         struct rpc_message msg = {
7643                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7644                 .rpc_argp       = &args,
7645                 .rpc_resp       = &res,
7646                 .rpc_cred       = cred,
7647         };
7648         int status;
7649
7650         res.fh = nfs_alloc_fhandle();
7651         if (res.fh == NULL)
7652                 return -ENOMEM;
7653
7654         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7655         status = nfs4_call_sync_sequence(clnt, server, &msg,
7656                                                 &args.seq_args, &res.seq_res);
7657         nfs_free_fhandle(res.fh);
7658         if (status == NFS4_OK &&
7659             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7660                 status = -NFS4ERR_LEASE_MOVED;
7661         return status;
7662 }
7663
7664 #endif  /* CONFIG_NFS_V4_1 */
7665
7666 /**
7667  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7668  * @inode: inode on FSID to check
7669  * @cred: credential to use for this operation
7670  *
7671  * Server indicates whether the FSID is present, moved, or not
7672  * recognized.  This operation is necessary to clear a LEASE_MOVED
7673  * condition for this client ID.
7674  *
7675  * Returns NFS4_OK if the FSID is present on this server,
7676  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7677  *  NFS4ERR code if some error occurred on the server, or a
7678  *  negative errno if a local failure occurred.
7679  */
7680 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
7681 {
7682         struct nfs_server *server = NFS_SERVER(inode);
7683         struct nfs_client *clp = server->nfs_client;
7684         const struct nfs4_mig_recovery_ops *ops =
7685                                         clp->cl_mvops->mig_recovery_ops;
7686         struct nfs4_exception exception = {
7687                 .interruptible = true,
7688         };
7689         int status;
7690
7691         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7692                 (unsigned long long)server->fsid.major,
7693                 (unsigned long long)server->fsid.minor,
7694                 clp->cl_hostname);
7695         nfs_display_fhandle(NFS_FH(inode), __func__);
7696
7697         do {
7698                 status = ops->fsid_present(inode, cred);
7699                 if (status != -NFS4ERR_DELAY)
7700                         break;
7701                 nfs4_handle_exception(server, status, &exception);
7702         } while (exception.retry);
7703         return status;
7704 }
7705
7706 /*
7707  * If 'use_integrity' is true and the state managment nfs_client
7708  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7709  * and the machine credential as per RFC3530bis and RFC5661 Security
7710  * Considerations sections. Otherwise, just use the user cred with the
7711  * filesystem's rpc_client.
7712  */
7713 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7714 {
7715         int status;
7716         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7717         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
7718         struct nfs4_secinfo_arg args = {
7719                 .dir_fh = NFS_FH(dir),
7720                 .name   = name,
7721         };
7722         struct nfs4_secinfo_res res = {
7723                 .flavors     = flavors,
7724         };
7725         struct rpc_message msg = {
7726                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7727                 .rpc_argp = &args,
7728                 .rpc_resp = &res,
7729         };
7730         struct nfs4_call_sync_data data = {
7731                 .seq_server = NFS_SERVER(dir),
7732                 .seq_args = &args.seq_args,
7733                 .seq_res = &res.seq_res,
7734         };
7735         struct rpc_task_setup task_setup = {
7736                 .rpc_client = clnt,
7737                 .rpc_message = &msg,
7738                 .callback_ops = clp->cl_mvops->call_sync_ops,
7739                 .callback_data = &data,
7740                 .flags = RPC_TASK_NO_ROUND_ROBIN,
7741         };
7742         const struct cred *cred = NULL;
7743
7744         if (use_integrity) {
7745                 clnt = clp->cl_rpcclient;
7746                 task_setup.rpc_client = clnt;
7747
7748                 cred = nfs4_get_clid_cred(clp);
7749                 msg.rpc_cred = cred;
7750         }
7751
7752         dprintk("NFS call  secinfo %s\n", name->name);
7753
7754         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7755         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
7756         status = nfs4_call_sync_custom(&task_setup);
7757
7758         dprintk("NFS reply  secinfo: %d\n", status);
7759
7760         put_cred(cred);
7761         return status;
7762 }
7763
7764 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7765                       struct nfs4_secinfo_flavors *flavors)
7766 {
7767         struct nfs4_exception exception = {
7768                 .interruptible = true,
7769         };
7770         int err;
7771         do {
7772                 err = -NFS4ERR_WRONGSEC;
7773
7774                 /* try to use integrity protection with machine cred */
7775                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7776                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7777
7778                 /*
7779                  * if unable to use integrity protection, or SECINFO with
7780                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7781                  * disallowed by spec, but exists in deployed servers) use
7782                  * the current filesystem's rpc_client and the user cred.
7783                  */
7784                 if (err == -NFS4ERR_WRONGSEC)
7785                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7786
7787                 trace_nfs4_secinfo(dir, name, err);
7788                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7789                                 &exception);
7790         } while (exception.retry);
7791         return err;
7792 }
7793
7794 #ifdef CONFIG_NFS_V4_1
7795 /*
7796  * Check the exchange flags returned by the server for invalid flags, having
7797  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7798  * DS flags set.
7799  */
7800 static int nfs4_check_cl_exchange_flags(u32 flags)
7801 {
7802         if (flags & ~EXCHGID4_FLAG_MASK_R)
7803                 goto out_inval;
7804         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7805             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7806                 goto out_inval;
7807         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7808                 goto out_inval;
7809         return NFS_OK;
7810 out_inval:
7811         return -NFS4ERR_INVAL;
7812 }
7813
7814 static bool
7815 nfs41_same_server_scope(struct nfs41_server_scope *a,
7816                         struct nfs41_server_scope *b)
7817 {
7818         if (a->server_scope_sz != b->server_scope_sz)
7819                 return false;
7820         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7821 }
7822
7823 static void
7824 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7825 {
7826 }
7827
7828 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7829         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7830 };
7831
7832 /*
7833  * nfs4_proc_bind_one_conn_to_session()
7834  *
7835  * The 4.1 client currently uses the same TCP connection for the
7836  * fore and backchannel.
7837  */
7838 static
7839 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7840                 struct rpc_xprt *xprt,
7841                 struct nfs_client *clp,
7842                 const struct cred *cred)
7843 {
7844         int status;
7845         struct nfs41_bind_conn_to_session_args args = {
7846                 .client = clp,
7847                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7848         };
7849         struct nfs41_bind_conn_to_session_res res;
7850         struct rpc_message msg = {
7851                 .rpc_proc =
7852                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7853                 .rpc_argp = &args,
7854                 .rpc_resp = &res,
7855                 .rpc_cred = cred,
7856         };
7857         struct rpc_task_setup task_setup_data = {
7858                 .rpc_client = clnt,
7859                 .rpc_xprt = xprt,
7860                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7861                 .rpc_message = &msg,
7862                 .flags = RPC_TASK_TIMEOUT,
7863         };
7864         struct rpc_task *task;
7865
7866         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7867         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7868                 args.dir = NFS4_CDFC4_FORE;
7869
7870         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7871         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7872                 args.dir = NFS4_CDFC4_FORE;
7873
7874         task = rpc_run_task(&task_setup_data);
7875         if (!IS_ERR(task)) {
7876                 status = task->tk_status;
7877                 rpc_put_task(task);
7878         } else
7879                 status = PTR_ERR(task);
7880         trace_nfs4_bind_conn_to_session(clp, status);
7881         if (status == 0) {
7882                 if (memcmp(res.sessionid.data,
7883                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7884                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7885                         return -EIO;
7886                 }
7887                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7888                         dprintk("NFS: %s: Unexpected direction from server\n",
7889                                 __func__);
7890                         return -EIO;
7891                 }
7892                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7893                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7894                                 __func__);
7895                         return -EIO;
7896                 }
7897         }
7898
7899         return status;
7900 }
7901
7902 struct rpc_bind_conn_calldata {
7903         struct nfs_client *clp;
7904         const struct cred *cred;
7905 };
7906
7907 static int
7908 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7909                 struct rpc_xprt *xprt,
7910                 void *calldata)
7911 {
7912         struct rpc_bind_conn_calldata *p = calldata;
7913
7914         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7915 }
7916
7917 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
7918 {
7919         struct rpc_bind_conn_calldata data = {
7920                 .clp = clp,
7921                 .cred = cred,
7922         };
7923         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7924                         nfs4_proc_bind_conn_to_session_callback, &data);
7925 }
7926
7927 /*
7928  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7929  * and operations we'd like to see to enable certain features in the allow map
7930  */
7931 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7932         .how = SP4_MACH_CRED,
7933         .enforce.u.words = {
7934                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7935                       1 << (OP_EXCHANGE_ID - 32) |
7936                       1 << (OP_CREATE_SESSION - 32) |
7937                       1 << (OP_DESTROY_SESSION - 32) |
7938                       1 << (OP_DESTROY_CLIENTID - 32)
7939         },
7940         .allow.u.words = {
7941                 [0] = 1 << (OP_CLOSE) |
7942                       1 << (OP_OPEN_DOWNGRADE) |
7943                       1 << (OP_LOCKU) |
7944                       1 << (OP_DELEGRETURN) |
7945                       1 << (OP_COMMIT),
7946                 [1] = 1 << (OP_SECINFO - 32) |
7947                       1 << (OP_SECINFO_NO_NAME - 32) |
7948                       1 << (OP_LAYOUTRETURN - 32) |
7949                       1 << (OP_TEST_STATEID - 32) |
7950                       1 << (OP_FREE_STATEID - 32) |
7951                       1 << (OP_WRITE - 32)
7952         }
7953 };
7954
7955 /*
7956  * Select the state protection mode for client `clp' given the server results
7957  * from exchange_id in `sp'.
7958  *
7959  * Returns 0 on success, negative errno otherwise.
7960  */
7961 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7962                                  struct nfs41_state_protection *sp)
7963 {
7964         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7965                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7966                       1 << (OP_EXCHANGE_ID - 32) |
7967                       1 << (OP_CREATE_SESSION - 32) |
7968                       1 << (OP_DESTROY_SESSION - 32) |
7969                       1 << (OP_DESTROY_CLIENTID - 32)
7970         };
7971         unsigned long flags = 0;
7972         unsigned int i;
7973         int ret = 0;
7974
7975         if (sp->how == SP4_MACH_CRED) {
7976                 /* Print state protect result */
7977                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7978                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7979                         if (test_bit(i, sp->enforce.u.longs))
7980                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7981                         if (test_bit(i, sp->allow.u.longs))
7982                                 dfprintk(MOUNT, "  allow op %d\n", i);
7983                 }
7984
7985                 /* make sure nothing is on enforce list that isn't supported */
7986                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7987                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7988                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7989                                 ret = -EINVAL;
7990                                 goto out;
7991                         }
7992                 }
7993
7994                 /*
7995                  * Minimal mode - state operations are allowed to use machine
7996                  * credential.  Note this already happens by default, so the
7997                  * client doesn't have to do anything more than the negotiation.
7998                  *
7999                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8000                  *       we're already using the machine cred for exchange_id
8001                  *       and will never use a different cred.
8002                  */
8003                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8004                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8005                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8006                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8007                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8008                         dfprintk(MOUNT, "  minimal mode enabled\n");
8009                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8010                 } else {
8011                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8012                         ret = -EINVAL;
8013                         goto out;
8014                 }
8015
8016                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8017                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8018                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8019                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8020                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8021                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8022                 }
8023
8024                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8025                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8026                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8027                 }
8028
8029                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8030                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8031                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8032                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8033                 }
8034
8035                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8036                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8037                         dfprintk(MOUNT, "  stateid mode enabled\n");
8038                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8039                 }
8040
8041                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8042                         dfprintk(MOUNT, "  write mode enabled\n");
8043                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8044                 }
8045
8046                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8047                         dfprintk(MOUNT, "  commit mode enabled\n");
8048                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8049                 }
8050         }
8051 out:
8052         clp->cl_sp4_flags = flags;
8053         return ret;
8054 }
8055
8056 struct nfs41_exchange_id_data {
8057         struct nfs41_exchange_id_res res;
8058         struct nfs41_exchange_id_args args;
8059 };
8060
8061 static void nfs4_exchange_id_release(void *data)
8062 {
8063         struct nfs41_exchange_id_data *cdata =
8064                                         (struct nfs41_exchange_id_data *)data;
8065
8066         nfs_put_client(cdata->args.client);
8067         kfree(cdata->res.impl_id);
8068         kfree(cdata->res.server_scope);
8069         kfree(cdata->res.server_owner);
8070         kfree(cdata);
8071 }
8072
8073 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8074         .rpc_release = nfs4_exchange_id_release,
8075 };
8076
8077 /*
8078  * _nfs4_proc_exchange_id()
8079  *
8080  * Wrapper for EXCHANGE_ID operation.
8081  */
8082 static struct rpc_task *
8083 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8084                         u32 sp4_how, struct rpc_xprt *xprt)
8085 {
8086         struct rpc_message msg = {
8087                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8088                 .rpc_cred = cred,
8089         };
8090         struct rpc_task_setup task_setup_data = {
8091                 .rpc_client = clp->cl_rpcclient,
8092                 .callback_ops = &nfs4_exchange_id_call_ops,
8093                 .rpc_message = &msg,
8094                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8095         };
8096         struct nfs41_exchange_id_data *calldata;
8097         int status;
8098
8099         if (!refcount_inc_not_zero(&clp->cl_count))
8100                 return ERR_PTR(-EIO);
8101
8102         status = -ENOMEM;
8103         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8104         if (!calldata)
8105                 goto out;
8106
8107         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8108
8109         status = nfs4_init_uniform_client_string(clp);
8110         if (status)
8111                 goto out_calldata;
8112
8113         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8114                                                 GFP_NOFS);
8115         status = -ENOMEM;
8116         if (unlikely(calldata->res.server_owner == NULL))
8117                 goto out_calldata;
8118
8119         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8120                                         GFP_NOFS);
8121         if (unlikely(calldata->res.server_scope == NULL))
8122                 goto out_server_owner;
8123
8124         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8125         if (unlikely(calldata->res.impl_id == NULL))
8126                 goto out_server_scope;
8127
8128         switch (sp4_how) {
8129         case SP4_NONE:
8130                 calldata->args.state_protect.how = SP4_NONE;
8131                 break;
8132
8133         case SP4_MACH_CRED:
8134                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8135                 break;
8136
8137         default:
8138                 /* unsupported! */
8139                 WARN_ON_ONCE(1);
8140                 status = -EINVAL;
8141                 goto out_impl_id;
8142         }
8143         if (xprt) {
8144                 task_setup_data.rpc_xprt = xprt;
8145                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8146                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8147                                 sizeof(calldata->args.verifier.data));
8148         }
8149         calldata->args.client = clp;
8150         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8151         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8152 #ifdef CONFIG_NFS_V4_1_MIGRATION
8153         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8154 #endif
8155         msg.rpc_argp = &calldata->args;
8156         msg.rpc_resp = &calldata->res;
8157         task_setup_data.callback_data = calldata;
8158
8159         return rpc_run_task(&task_setup_data);
8160
8161 out_impl_id:
8162         kfree(calldata->res.impl_id);
8163 out_server_scope:
8164         kfree(calldata->res.server_scope);
8165 out_server_owner:
8166         kfree(calldata->res.server_owner);
8167 out_calldata:
8168         kfree(calldata);
8169 out:
8170         nfs_put_client(clp);
8171         return ERR_PTR(status);
8172 }
8173
8174 /*
8175  * _nfs4_proc_exchange_id()
8176  *
8177  * Wrapper for EXCHANGE_ID operation.
8178  */
8179 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8180                         u32 sp4_how)
8181 {
8182         struct rpc_task *task;
8183         struct nfs41_exchange_id_args *argp;
8184         struct nfs41_exchange_id_res *resp;
8185         int status;
8186
8187         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8188         if (IS_ERR(task))
8189                 return PTR_ERR(task);
8190
8191         argp = task->tk_msg.rpc_argp;
8192         resp = task->tk_msg.rpc_resp;
8193         status = task->tk_status;
8194         if (status  != 0)
8195                 goto out;
8196
8197         status = nfs4_check_cl_exchange_flags(resp->flags);
8198         if (status  != 0)
8199                 goto out;
8200
8201         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8202         if (status != 0)
8203                 goto out;
8204
8205         clp->cl_clientid = resp->clientid;
8206         clp->cl_exchange_flags = resp->flags;
8207         clp->cl_seqid = resp->seqid;
8208         /* Client ID is not confirmed */
8209         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8210                 clear_bit(NFS4_SESSION_ESTABLISHED,
8211                           &clp->cl_session->session_state);
8212
8213         if (clp->cl_serverscope != NULL &&
8214             !nfs41_same_server_scope(clp->cl_serverscope,
8215                                 resp->server_scope)) {
8216                 dprintk("%s: server_scope mismatch detected\n",
8217                         __func__);
8218                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8219         }
8220
8221         swap(clp->cl_serverowner, resp->server_owner);
8222         swap(clp->cl_serverscope, resp->server_scope);
8223         swap(clp->cl_implid, resp->impl_id);
8224
8225         /* Save the EXCHANGE_ID verifier session trunk tests */
8226         memcpy(clp->cl_confirm.data, argp->verifier.data,
8227                sizeof(clp->cl_confirm.data));
8228 out:
8229         trace_nfs4_exchange_id(clp, status);
8230         rpc_put_task(task);
8231         return status;
8232 }
8233
8234 /*
8235  * nfs4_proc_exchange_id()
8236  *
8237  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8238  *
8239  * Since the clientid has expired, all compounds using sessions
8240  * associated with the stale clientid will be returning
8241  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8242  * be in some phase of session reset.
8243  *
8244  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8245  */
8246 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8247 {
8248         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8249         int status;
8250
8251         /* try SP4_MACH_CRED if krb5i/p */
8252         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8253             authflavor == RPC_AUTH_GSS_KRB5P) {
8254                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8255                 if (!status)
8256                         return 0;
8257         }
8258
8259         /* try SP4_NONE */
8260         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8261 }
8262
8263 /**
8264  * nfs4_test_session_trunk
8265  *
8266  * This is an add_xprt_test() test function called from
8267  * rpc_clnt_setup_test_and_add_xprt.
8268  *
8269  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8270  * and is dereferrenced in nfs4_exchange_id_release
8271  *
8272  * Upon success, add the new transport to the rpc_clnt
8273  *
8274  * @clnt: struct rpc_clnt to get new transport
8275  * @xprt: the rpc_xprt to test
8276  * @data: call data for _nfs4_proc_exchange_id.
8277  */
8278 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8279                             void *data)
8280 {
8281         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8282         struct rpc_task *task;
8283         int status;
8284
8285         u32 sp4_how;
8286
8287         dprintk("--> %s try %s\n", __func__,
8288                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8289
8290         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8291
8292         /* Test connection for session trunking. Async exchange_id call */
8293         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8294         if (IS_ERR(task))
8295                 return;
8296
8297         status = task->tk_status;
8298         if (status == 0)
8299                 status = nfs4_detect_session_trunking(adata->clp,
8300                                 task->tk_msg.rpc_resp, xprt);
8301
8302         if (status == 0)
8303                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8304
8305         rpc_put_task(task);
8306 }
8307 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8308
8309 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8310                 const struct cred *cred)
8311 {
8312         struct rpc_message msg = {
8313                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8314                 .rpc_argp = clp,
8315                 .rpc_cred = cred,
8316         };
8317         int status;
8318
8319         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8320                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8321         trace_nfs4_destroy_clientid(clp, status);
8322         if (status)
8323                 dprintk("NFS: Got error %d from the server %s on "
8324                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8325         return status;
8326 }
8327
8328 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8329                 const struct cred *cred)
8330 {
8331         unsigned int loop;
8332         int ret;
8333
8334         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8335                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8336                 switch (ret) {
8337                 case -NFS4ERR_DELAY:
8338                 case -NFS4ERR_CLIENTID_BUSY:
8339                         ssleep(1);
8340                         break;
8341                 default:
8342                         return ret;
8343                 }
8344         }
8345         return 0;
8346 }
8347
8348 int nfs4_destroy_clientid(struct nfs_client *clp)
8349 {
8350         const struct cred *cred;
8351         int ret = 0;
8352
8353         if (clp->cl_mvops->minor_version < 1)
8354                 goto out;
8355         if (clp->cl_exchange_flags == 0)
8356                 goto out;
8357         if (clp->cl_preserve_clid)
8358                 goto out;
8359         cred = nfs4_get_clid_cred(clp);
8360         ret = nfs4_proc_destroy_clientid(clp, cred);
8361         put_cred(cred);
8362         switch (ret) {
8363         case 0:
8364         case -NFS4ERR_STALE_CLIENTID:
8365                 clp->cl_exchange_flags = 0;
8366         }
8367 out:
8368         return ret;
8369 }
8370
8371 #endif /* CONFIG_NFS_V4_1 */
8372
8373 struct nfs4_get_lease_time_data {
8374         struct nfs4_get_lease_time_args *args;
8375         struct nfs4_get_lease_time_res *res;
8376         struct nfs_client *clp;
8377 };
8378
8379 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8380                                         void *calldata)
8381 {
8382         struct nfs4_get_lease_time_data *data =
8383                         (struct nfs4_get_lease_time_data *)calldata;
8384
8385         dprintk("--> %s\n", __func__);
8386         /* just setup sequence, do not trigger session recovery
8387            since we're invoked within one */
8388         nfs4_setup_sequence(data->clp,
8389                         &data->args->la_seq_args,
8390                         &data->res->lr_seq_res,
8391                         task);
8392         dprintk("<-- %s\n", __func__);
8393 }
8394
8395 /*
8396  * Called from nfs4_state_manager thread for session setup, so don't recover
8397  * from sequence operation or clientid errors.
8398  */
8399 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8400 {
8401         struct nfs4_get_lease_time_data *data =
8402                         (struct nfs4_get_lease_time_data *)calldata;
8403
8404         dprintk("--> %s\n", __func__);
8405         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8406                 return;
8407         switch (task->tk_status) {
8408         case -NFS4ERR_DELAY:
8409         case -NFS4ERR_GRACE:
8410                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8411                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8412                 task->tk_status = 0;
8413                 /* fall through */
8414         case -NFS4ERR_RETRY_UNCACHED_REP:
8415                 rpc_restart_call_prepare(task);
8416                 return;
8417         }
8418         dprintk("<-- %s\n", __func__);
8419 }
8420
8421 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8422         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8423         .rpc_call_done = nfs4_get_lease_time_done,
8424 };
8425
8426 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8427 {
8428         struct nfs4_get_lease_time_args args;
8429         struct nfs4_get_lease_time_res res = {
8430                 .lr_fsinfo = fsinfo,
8431         };
8432         struct nfs4_get_lease_time_data data = {
8433                 .args = &args,
8434                 .res = &res,
8435                 .clp = clp,
8436         };
8437         struct rpc_message msg = {
8438                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8439                 .rpc_argp = &args,
8440                 .rpc_resp = &res,
8441         };
8442         struct rpc_task_setup task_setup = {
8443                 .rpc_client = clp->cl_rpcclient,
8444                 .rpc_message = &msg,
8445                 .callback_ops = &nfs4_get_lease_time_ops,
8446                 .callback_data = &data,
8447                 .flags = RPC_TASK_TIMEOUT,
8448         };
8449
8450         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8451         return nfs4_call_sync_custom(&task_setup);
8452 }
8453
8454 #ifdef CONFIG_NFS_V4_1
8455
8456 /*
8457  * Initialize the values to be used by the client in CREATE_SESSION
8458  * If nfs4_init_session set the fore channel request and response sizes,
8459  * use them.
8460  *
8461  * Set the back channel max_resp_sz_cached to zero to force the client to
8462  * always set csa_cachethis to FALSE because the current implementation
8463  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8464  */
8465 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8466                                     struct rpc_clnt *clnt)
8467 {
8468         unsigned int max_rqst_sz, max_resp_sz;
8469         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8470         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8471
8472         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8473         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8474
8475         /* Fore channel attributes */
8476         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8477         args->fc_attrs.max_resp_sz = max_resp_sz;
8478         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8479         args->fc_attrs.max_reqs = max_session_slots;
8480
8481         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8482                 "max_ops=%u max_reqs=%u\n",
8483                 __func__,
8484                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8485                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8486
8487         /* Back channel attributes */
8488         args->bc_attrs.max_rqst_sz = max_bc_payload;
8489         args->bc_attrs.max_resp_sz = max_bc_payload;
8490         args->bc_attrs.max_resp_sz_cached = 0;
8491         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8492         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8493         if (args->bc_attrs.max_reqs > max_bc_slots)
8494                 args->bc_attrs.max_reqs = max_bc_slots;
8495
8496         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8497                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8498                 __func__,
8499                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8500                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8501                 args->bc_attrs.max_reqs);
8502 }
8503
8504 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8505                 struct nfs41_create_session_res *res)
8506 {
8507         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8508         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8509
8510         if (rcvd->max_resp_sz > sent->max_resp_sz)
8511                 return -EINVAL;
8512         /*
8513          * Our requested max_ops is the minimum we need; we're not
8514          * prepared to break up compounds into smaller pieces than that.
8515          * So, no point even trying to continue if the server won't
8516          * cooperate:
8517          */
8518         if (rcvd->max_ops < sent->max_ops)
8519                 return -EINVAL;
8520         if (rcvd->max_reqs == 0)
8521                 return -EINVAL;
8522         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8523                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8524         return 0;
8525 }
8526
8527 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8528                 struct nfs41_create_session_res *res)
8529 {
8530         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8531         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8532
8533         if (!(res->flags & SESSION4_BACK_CHAN))
8534                 goto out;
8535         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8536                 return -EINVAL;
8537         if (rcvd->max_resp_sz < sent->max_resp_sz)
8538                 return -EINVAL;
8539         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8540                 return -EINVAL;
8541         if (rcvd->max_ops > sent->max_ops)
8542                 return -EINVAL;
8543         if (rcvd->max_reqs > sent->max_reqs)
8544                 return -EINVAL;
8545 out:
8546         return 0;
8547 }
8548
8549 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8550                                      struct nfs41_create_session_res *res)
8551 {
8552         int ret;
8553
8554         ret = nfs4_verify_fore_channel_attrs(args, res);
8555         if (ret)
8556                 return ret;
8557         return nfs4_verify_back_channel_attrs(args, res);
8558 }
8559
8560 static void nfs4_update_session(struct nfs4_session *session,
8561                 struct nfs41_create_session_res *res)
8562 {
8563         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8564         /* Mark client id and session as being confirmed */
8565         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8566         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8567         session->flags = res->flags;
8568         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8569         if (res->flags & SESSION4_BACK_CHAN)
8570                 memcpy(&session->bc_attrs, &res->bc_attrs,
8571                                 sizeof(session->bc_attrs));
8572 }
8573
8574 static int _nfs4_proc_create_session(struct nfs_client *clp,
8575                 const struct cred *cred)
8576 {
8577         struct nfs4_session *session = clp->cl_session;
8578         struct nfs41_create_session_args args = {
8579                 .client = clp,
8580                 .clientid = clp->cl_clientid,
8581                 .seqid = clp->cl_seqid,
8582                 .cb_program = NFS4_CALLBACK,
8583         };
8584         struct nfs41_create_session_res res;
8585
8586         struct rpc_message msg = {
8587                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8588                 .rpc_argp = &args,
8589                 .rpc_resp = &res,
8590                 .rpc_cred = cred,
8591         };
8592         int status;
8593
8594         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8595         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8596
8597         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8598                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8599         trace_nfs4_create_session(clp, status);
8600
8601         switch (status) {
8602         case -NFS4ERR_STALE_CLIENTID:
8603         case -NFS4ERR_DELAY:
8604         case -ETIMEDOUT:
8605         case -EACCES:
8606         case -EAGAIN:
8607                 goto out;
8608         };
8609
8610         clp->cl_seqid++;
8611         if (!status) {
8612                 /* Verify the session's negotiated channel_attrs values */
8613                 status = nfs4_verify_channel_attrs(&args, &res);
8614                 /* Increment the clientid slot sequence id */
8615                 if (status)
8616                         goto out;
8617                 nfs4_update_session(session, &res);
8618         }
8619 out:
8620         return status;
8621 }
8622
8623 /*
8624  * Issues a CREATE_SESSION operation to the server.
8625  * It is the responsibility of the caller to verify the session is
8626  * expired before calling this routine.
8627  */
8628 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8629 {
8630         int status;
8631         unsigned *ptr;
8632         struct nfs4_session *session = clp->cl_session;
8633
8634         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8635
8636         status = _nfs4_proc_create_session(clp, cred);
8637         if (status)
8638                 goto out;
8639
8640         /* Init or reset the session slot tables */
8641         status = nfs4_setup_session_slot_tables(session);
8642         dprintk("slot table setup returned %d\n", status);
8643         if (status)
8644                 goto out;
8645
8646         ptr = (unsigned *)&session->sess_id.data[0];
8647         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8648                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8649 out:
8650         dprintk("<-- %s\n", __func__);
8651         return status;
8652 }
8653
8654 /*
8655  * Issue the over-the-wire RPC DESTROY_SESSION.
8656  * The caller must serialize access to this routine.
8657  */
8658 int nfs4_proc_destroy_session(struct nfs4_session *session,
8659                 const struct cred *cred)
8660 {
8661         struct rpc_message msg = {
8662                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8663                 .rpc_argp = session,
8664                 .rpc_cred = cred,
8665         };
8666         int status = 0;
8667
8668         dprintk("--> nfs4_proc_destroy_session\n");
8669
8670         /* session is still being setup */
8671         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8672                 return 0;
8673
8674         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8675                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8676         trace_nfs4_destroy_session(session->clp, status);
8677
8678         if (status)
8679                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8680                         "Session has been destroyed regardless...\n", status);
8681
8682         dprintk("<-- nfs4_proc_destroy_session\n");
8683         return status;
8684 }
8685
8686 /*
8687  * Renew the cl_session lease.
8688  */
8689 struct nfs4_sequence_data {
8690         struct nfs_client *clp;
8691         struct nfs4_sequence_args args;
8692         struct nfs4_sequence_res res;
8693 };
8694
8695 static void nfs41_sequence_release(void *data)
8696 {
8697         struct nfs4_sequence_data *calldata = data;
8698         struct nfs_client *clp = calldata->clp;
8699
8700         if (refcount_read(&clp->cl_count) > 1)
8701                 nfs4_schedule_state_renewal(clp);
8702         nfs_put_client(clp);
8703         kfree(calldata);
8704 }
8705
8706 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8707 {
8708         switch(task->tk_status) {
8709         case -NFS4ERR_DELAY:
8710                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8711                 return -EAGAIN;
8712         default:
8713                 nfs4_schedule_lease_recovery(clp);
8714         }
8715         return 0;
8716 }
8717
8718 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8719 {
8720         struct nfs4_sequence_data *calldata = data;
8721         struct nfs_client *clp = calldata->clp;
8722
8723         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8724                 return;
8725
8726         trace_nfs4_sequence(clp, task->tk_status);
8727         if (task->tk_status < 0) {
8728                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8729                 if (refcount_read(&clp->cl_count) == 1)
8730                         goto out;
8731
8732                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8733                         rpc_restart_call_prepare(task);
8734                         return;
8735                 }
8736         }
8737         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8738 out:
8739         dprintk("<-- %s\n", __func__);
8740 }
8741
8742 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8743 {
8744         struct nfs4_sequence_data *calldata = data;
8745         struct nfs_client *clp = calldata->clp;
8746         struct nfs4_sequence_args *args;
8747         struct nfs4_sequence_res *res;
8748
8749         args = task->tk_msg.rpc_argp;
8750         res = task->tk_msg.rpc_resp;
8751
8752         nfs4_setup_sequence(clp, args, res, task);
8753 }
8754
8755 static const struct rpc_call_ops nfs41_sequence_ops = {
8756         .rpc_call_done = nfs41_sequence_call_done,
8757         .rpc_call_prepare = nfs41_sequence_prepare,
8758         .rpc_release = nfs41_sequence_release,
8759 };
8760
8761 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8762                 const struct cred *cred,
8763                 struct nfs4_slot *slot,
8764                 bool is_privileged)
8765 {
8766         struct nfs4_sequence_data *calldata;
8767         struct rpc_message msg = {
8768                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8769                 .rpc_cred = cred,
8770         };
8771         struct rpc_task_setup task_setup_data = {
8772                 .rpc_client = clp->cl_rpcclient,
8773                 .rpc_message = &msg,
8774                 .callback_ops = &nfs41_sequence_ops,
8775                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8776         };
8777         struct rpc_task *ret;
8778
8779         ret = ERR_PTR(-EIO);
8780         if (!refcount_inc_not_zero(&clp->cl_count))
8781                 goto out_err;
8782
8783         ret = ERR_PTR(-ENOMEM);
8784         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8785         if (calldata == NULL)
8786                 goto out_put_clp;
8787         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8788         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8789         msg.rpc_argp = &calldata->args;
8790         msg.rpc_resp = &calldata->res;
8791         calldata->clp = clp;
8792         task_setup_data.callback_data = calldata;
8793
8794         ret = rpc_run_task(&task_setup_data);
8795         if (IS_ERR(ret))
8796                 goto out_err;
8797         return ret;
8798 out_put_clp:
8799         nfs_put_client(clp);
8800 out_err:
8801         nfs41_release_slot(slot);
8802         return ret;
8803 }
8804
8805 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
8806 {
8807         struct rpc_task *task;
8808         int ret = 0;
8809
8810         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8811                 return -EAGAIN;
8812         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8813         if (IS_ERR(task))
8814                 ret = PTR_ERR(task);
8815         else
8816                 rpc_put_task_async(task);
8817         dprintk("<-- %s status=%d\n", __func__, ret);
8818         return ret;
8819 }
8820
8821 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
8822 {
8823         struct rpc_task *task;
8824         int ret;
8825
8826         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8827         if (IS_ERR(task)) {
8828                 ret = PTR_ERR(task);
8829                 goto out;
8830         }
8831         ret = rpc_wait_for_completion_task(task);
8832         if (!ret)
8833                 ret = task->tk_status;
8834         rpc_put_task(task);
8835 out:
8836         dprintk("<-- %s status=%d\n", __func__, ret);
8837         return ret;
8838 }
8839
8840 struct nfs4_reclaim_complete_data {
8841         struct nfs_client *clp;
8842         struct nfs41_reclaim_complete_args arg;
8843         struct nfs41_reclaim_complete_res res;
8844 };
8845
8846 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8847 {
8848         struct nfs4_reclaim_complete_data *calldata = data;
8849
8850         nfs4_setup_sequence(calldata->clp,
8851                         &calldata->arg.seq_args,
8852                         &calldata->res.seq_res,
8853                         task);
8854 }
8855
8856 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8857 {
8858         switch(task->tk_status) {
8859         case 0:
8860                 wake_up_all(&clp->cl_lock_waitq);
8861                 /* Fallthrough */
8862         case -NFS4ERR_COMPLETE_ALREADY:
8863         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8864                 break;
8865         case -NFS4ERR_DELAY:
8866                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8867                 /* fall through */
8868         case -NFS4ERR_RETRY_UNCACHED_REP:
8869                 return -EAGAIN;
8870         case -NFS4ERR_BADSESSION:
8871         case -NFS4ERR_DEADSESSION:
8872         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8873                 nfs4_schedule_session_recovery(clp->cl_session,
8874                                 task->tk_status);
8875                 break;
8876         default:
8877                 nfs4_schedule_lease_recovery(clp);
8878         }
8879         return 0;
8880 }
8881
8882 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8883 {
8884         struct nfs4_reclaim_complete_data *calldata = data;
8885         struct nfs_client *clp = calldata->clp;
8886         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8887
8888         dprintk("--> %s\n", __func__);
8889         if (!nfs41_sequence_done(task, res))
8890                 return;
8891
8892         trace_nfs4_reclaim_complete(clp, task->tk_status);
8893         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8894                 rpc_restart_call_prepare(task);
8895                 return;
8896         }
8897         dprintk("<-- %s\n", __func__);
8898 }
8899
8900 static void nfs4_free_reclaim_complete_data(void *data)
8901 {
8902         struct nfs4_reclaim_complete_data *calldata = data;
8903
8904         kfree(calldata);
8905 }
8906
8907 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8908         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8909         .rpc_call_done = nfs4_reclaim_complete_done,
8910         .rpc_release = nfs4_free_reclaim_complete_data,
8911 };
8912
8913 /*
8914  * Issue a global reclaim complete.
8915  */
8916 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8917                 const struct cred *cred)
8918 {
8919         struct nfs4_reclaim_complete_data *calldata;
8920         struct rpc_message msg = {
8921                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8922                 .rpc_cred = cred,
8923         };
8924         struct rpc_task_setup task_setup_data = {
8925                 .rpc_client = clp->cl_rpcclient,
8926                 .rpc_message = &msg,
8927                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8928                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8929         };
8930         int status = -ENOMEM;
8931
8932         dprintk("--> %s\n", __func__);
8933         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8934         if (calldata == NULL)
8935                 goto out;
8936         calldata->clp = clp;
8937         calldata->arg.one_fs = 0;
8938
8939         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
8940         msg.rpc_argp = &calldata->arg;
8941         msg.rpc_resp = &calldata->res;
8942         task_setup_data.callback_data = calldata;
8943         status = nfs4_call_sync_custom(&task_setup_data);
8944 out:
8945         dprintk("<-- %s status=%d\n", __func__, status);
8946         return status;
8947 }
8948
8949 static void
8950 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8951 {
8952         struct nfs4_layoutget *lgp = calldata;
8953         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8954
8955         dprintk("--> %s\n", __func__);
8956         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8957                                 &lgp->res.seq_res, task);
8958         dprintk("<-- %s\n", __func__);
8959 }
8960
8961 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8962 {
8963         struct nfs4_layoutget *lgp = calldata;
8964
8965         dprintk("--> %s\n", __func__);
8966         nfs41_sequence_process(task, &lgp->res.seq_res);
8967         dprintk("<-- %s\n", __func__);
8968 }
8969
8970 static int
8971 nfs4_layoutget_handle_exception(struct rpc_task *task,
8972                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8973 {
8974         struct inode *inode = lgp->args.inode;
8975         struct nfs_server *server = NFS_SERVER(inode);
8976         struct pnfs_layout_hdr *lo;
8977         int nfs4err = task->tk_status;
8978         int err, status = 0;
8979         LIST_HEAD(head);
8980
8981         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8982
8983         nfs4_sequence_free_slot(&lgp->res.seq_res);
8984
8985         switch (nfs4err) {
8986         case 0:
8987                 goto out;
8988
8989         /*
8990          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8991          * on the file. set tk_status to -ENODATA to tell upper layer to
8992          * retry go inband.
8993          */
8994         case -NFS4ERR_LAYOUTUNAVAILABLE:
8995                 status = -ENODATA;
8996                 goto out;
8997         /*
8998          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8999          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9000          */
9001         case -NFS4ERR_BADLAYOUT:
9002                 status = -EOVERFLOW;
9003                 goto out;
9004         /*
9005          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9006          * (or clients) writing to the same RAID stripe except when
9007          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9008          *
9009          * Treat it like we would RECALLCONFLICT -- we retry for a little
9010          * while, and then eventually give up.
9011          */
9012         case -NFS4ERR_LAYOUTTRYLATER:
9013                 if (lgp->args.minlength == 0) {
9014                         status = -EOVERFLOW;
9015                         goto out;
9016                 }
9017                 status = -EBUSY;
9018                 break;
9019         case -NFS4ERR_RECALLCONFLICT:
9020                 status = -ERECALLCONFLICT;
9021                 break;
9022         case -NFS4ERR_DELEG_REVOKED:
9023         case -NFS4ERR_ADMIN_REVOKED:
9024         case -NFS4ERR_EXPIRED:
9025         case -NFS4ERR_BAD_STATEID:
9026                 exception->timeout = 0;
9027                 spin_lock(&inode->i_lock);
9028                 lo = NFS_I(inode)->layout;
9029                 /* If the open stateid was bad, then recover it. */
9030                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9031                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9032                         spin_unlock(&inode->i_lock);
9033                         exception->state = lgp->args.ctx->state;
9034                         exception->stateid = &lgp->args.stateid;
9035                         break;
9036                 }
9037
9038                 /*
9039                  * Mark the bad layout state as invalid, then retry
9040                  */
9041                 pnfs_mark_layout_stateid_invalid(lo, &head);
9042                 spin_unlock(&inode->i_lock);
9043                 nfs_commit_inode(inode, 0);
9044                 pnfs_free_lseg_list(&head);
9045                 status = -EAGAIN;
9046                 goto out;
9047         }
9048
9049         err = nfs4_handle_exception(server, nfs4err, exception);
9050         if (!status) {
9051                 if (exception->retry)
9052                         status = -EAGAIN;
9053                 else
9054                         status = err;
9055         }
9056 out:
9057         dprintk("<-- %s\n", __func__);
9058         return status;
9059 }
9060
9061 size_t max_response_pages(struct nfs_server *server)
9062 {
9063         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9064         return nfs_page_array_len(0, max_resp_sz);
9065 }
9066
9067 static void nfs4_layoutget_release(void *calldata)
9068 {
9069         struct nfs4_layoutget *lgp = calldata;
9070
9071         dprintk("--> %s\n", __func__);
9072         nfs4_sequence_free_slot(&lgp->res.seq_res);
9073         pnfs_layoutget_free(lgp);
9074         dprintk("<-- %s\n", __func__);
9075 }
9076
9077 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9078         .rpc_call_prepare = nfs4_layoutget_prepare,
9079         .rpc_call_done = nfs4_layoutget_done,
9080         .rpc_release = nfs4_layoutget_release,
9081 };
9082
9083 struct pnfs_layout_segment *
9084 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9085 {
9086         struct inode *inode = lgp->args.inode;
9087         struct nfs_server *server = NFS_SERVER(inode);
9088         struct rpc_task *task;
9089         struct rpc_message msg = {
9090                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9091                 .rpc_argp = &lgp->args,
9092                 .rpc_resp = &lgp->res,
9093                 .rpc_cred = lgp->cred,
9094         };
9095         struct rpc_task_setup task_setup_data = {
9096                 .rpc_client = server->client,
9097                 .rpc_message = &msg,
9098                 .callback_ops = &nfs4_layoutget_call_ops,
9099                 .callback_data = lgp,
9100                 .flags = RPC_TASK_ASYNC,
9101         };
9102         struct pnfs_layout_segment *lseg = NULL;
9103         struct nfs4_exception exception = {
9104                 .inode = inode,
9105                 .timeout = *timeout,
9106         };
9107         int status = 0;
9108
9109         dprintk("--> %s\n", __func__);
9110
9111         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9112         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9113
9114         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9115
9116         task = rpc_run_task(&task_setup_data);
9117         if (IS_ERR(task))
9118                 return ERR_CAST(task);
9119         status = rpc_wait_for_completion_task(task);
9120         if (status != 0)
9121                 goto out;
9122
9123         if (task->tk_status < 0) {
9124                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9125                 *timeout = exception.timeout;
9126         } else if (lgp->res.layoutp->len == 0) {
9127                 status = -EAGAIN;
9128                 *timeout = nfs4_update_delay(&exception.timeout);
9129         } else
9130                 lseg = pnfs_layout_process(lgp);
9131 out:
9132         trace_nfs4_layoutget(lgp->args.ctx,
9133                         &lgp->args.range,
9134                         &lgp->res.range,
9135                         &lgp->res.stateid,
9136                         status);
9137
9138         rpc_put_task(task);
9139         dprintk("<-- %s status=%d\n", __func__, status);
9140         if (status)
9141                 return ERR_PTR(status);
9142         return lseg;
9143 }
9144
9145 static void
9146 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9147 {
9148         struct nfs4_layoutreturn *lrp = calldata;
9149
9150         dprintk("--> %s\n", __func__);
9151         nfs4_setup_sequence(lrp->clp,
9152                         &lrp->args.seq_args,
9153                         &lrp->res.seq_res,
9154                         task);
9155         if (!pnfs_layout_is_valid(lrp->args.layout))
9156                 rpc_exit(task, 0);
9157 }
9158
9159 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9160 {
9161         struct nfs4_layoutreturn *lrp = calldata;
9162         struct nfs_server *server;
9163
9164         dprintk("--> %s\n", __func__);
9165
9166         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9167                 return;
9168
9169         /*
9170          * Was there an RPC level error? Assume the call succeeded,
9171          * and that we need to release the layout
9172          */
9173         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9174                 lrp->res.lrs_present = 0;
9175                 return;
9176         }
9177
9178         server = NFS_SERVER(lrp->args.inode);
9179         switch (task->tk_status) {
9180         case -NFS4ERR_OLD_STATEID:
9181                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9182                                         &lrp->args.range,
9183                                         lrp->args.inode))
9184                         goto out_restart;
9185                 /* Fallthrough */
9186         default:
9187                 task->tk_status = 0;
9188                 /* Fallthrough */
9189         case 0:
9190                 break;
9191         case -NFS4ERR_DELAY:
9192                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9193                         break;
9194                 goto out_restart;
9195         }
9196         dprintk("<-- %s\n", __func__);
9197         return;
9198 out_restart:
9199         task->tk_status = 0;
9200         nfs4_sequence_free_slot(&lrp->res.seq_res);
9201         rpc_restart_call_prepare(task);
9202 }
9203
9204 static void nfs4_layoutreturn_release(void *calldata)
9205 {
9206         struct nfs4_layoutreturn *lrp = calldata;
9207         struct pnfs_layout_hdr *lo = lrp->args.layout;
9208
9209         dprintk("--> %s\n", __func__);
9210         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9211                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9212         nfs4_sequence_free_slot(&lrp->res.seq_res);
9213         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9214                 lrp->ld_private.ops->free(&lrp->ld_private);
9215         pnfs_put_layout_hdr(lrp->args.layout);
9216         nfs_iput_and_deactive(lrp->inode);
9217         kfree(calldata);
9218         dprintk("<-- %s\n", __func__);
9219 }
9220
9221 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9222         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9223         .rpc_call_done = nfs4_layoutreturn_done,
9224         .rpc_release = nfs4_layoutreturn_release,
9225 };
9226
9227 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9228 {
9229         struct rpc_task *task;
9230         struct rpc_message msg = {
9231                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9232                 .rpc_argp = &lrp->args,
9233                 .rpc_resp = &lrp->res,
9234                 .rpc_cred = lrp->cred,
9235         };
9236         struct rpc_task_setup task_setup_data = {
9237                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9238                 .rpc_message = &msg,
9239                 .callback_ops = &nfs4_layoutreturn_call_ops,
9240                 .callback_data = lrp,
9241         };
9242         int status = 0;
9243
9244         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9245                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9246                         &task_setup_data.rpc_client, &msg);
9247
9248         dprintk("--> %s\n", __func__);
9249         if (!sync) {
9250                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9251                 if (!lrp->inode) {
9252                         nfs4_layoutreturn_release(lrp);
9253                         return -EAGAIN;
9254                 }
9255                 task_setup_data.flags |= RPC_TASK_ASYNC;
9256         }
9257         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9258         task = rpc_run_task(&task_setup_data);
9259         if (IS_ERR(task))
9260                 return PTR_ERR(task);
9261         if (sync)
9262                 status = task->tk_status;
9263         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9264         dprintk("<-- %s status=%d\n", __func__, status);
9265         rpc_put_task(task);
9266         return status;
9267 }
9268
9269 static int
9270 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9271                 struct pnfs_device *pdev,
9272                 const struct cred *cred)
9273 {
9274         struct nfs4_getdeviceinfo_args args = {
9275                 .pdev = pdev,
9276                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9277                         NOTIFY_DEVICEID4_DELETE,
9278         };
9279         struct nfs4_getdeviceinfo_res res = {
9280                 .pdev = pdev,
9281         };
9282         struct rpc_message msg = {
9283                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9284                 .rpc_argp = &args,
9285                 .rpc_resp = &res,
9286                 .rpc_cred = cred,
9287         };
9288         int status;
9289
9290         dprintk("--> %s\n", __func__);
9291         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9292         if (res.notification & ~args.notify_types)
9293                 dprintk("%s: unsupported notification\n", __func__);
9294         if (res.notification != args.notify_types)
9295                 pdev->nocache = 1;
9296
9297         dprintk("<-- %s status=%d\n", __func__, status);
9298
9299         return status;
9300 }
9301
9302 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9303                 struct pnfs_device *pdev,
9304                 const struct cred *cred)
9305 {
9306         struct nfs4_exception exception = { };
9307         int err;
9308
9309         do {
9310                 err = nfs4_handle_exception(server,
9311                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9312                                         &exception);
9313         } while (exception.retry);
9314         return err;
9315 }
9316 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9317
9318 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9319 {
9320         struct nfs4_layoutcommit_data *data = calldata;
9321         struct nfs_server *server = NFS_SERVER(data->args.inode);
9322
9323         nfs4_setup_sequence(server->nfs_client,
9324                         &data->args.seq_args,
9325                         &data->res.seq_res,
9326                         task);
9327 }
9328
9329 static void
9330 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9331 {
9332         struct nfs4_layoutcommit_data *data = calldata;
9333         struct nfs_server *server = NFS_SERVER(data->args.inode);
9334
9335         if (!nfs41_sequence_done(task, &data->res.seq_res))
9336                 return;
9337
9338         switch (task->tk_status) { /* Just ignore these failures */
9339         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9340         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9341         case -NFS4ERR_BADLAYOUT:     /* no layout */
9342         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9343                 task->tk_status = 0;
9344         case 0:
9345                 break;
9346         default:
9347                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9348                         rpc_restart_call_prepare(task);
9349                         return;
9350                 }
9351         }
9352 }
9353
9354 static void nfs4_layoutcommit_release(void *calldata)
9355 {
9356         struct nfs4_layoutcommit_data *data = calldata;
9357
9358         pnfs_cleanup_layoutcommit(data);
9359         nfs_post_op_update_inode_force_wcc(data->args.inode,
9360                                            data->res.fattr);
9361         put_cred(data->cred);
9362         nfs_iput_and_deactive(data->inode);
9363         kfree(data);
9364 }
9365
9366 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9367         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9368         .rpc_call_done = nfs4_layoutcommit_done,
9369         .rpc_release = nfs4_layoutcommit_release,
9370 };
9371
9372 int
9373 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9374 {
9375         struct rpc_message msg = {
9376                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9377                 .rpc_argp = &data->args,
9378                 .rpc_resp = &data->res,
9379                 .rpc_cred = data->cred,
9380         };
9381         struct rpc_task_setup task_setup_data = {
9382                 .task = &data->task,
9383                 .rpc_client = NFS_CLIENT(data->args.inode),
9384                 .rpc_message = &msg,
9385                 .callback_ops = &nfs4_layoutcommit_ops,
9386                 .callback_data = data,
9387         };
9388         struct rpc_task *task;
9389         int status = 0;
9390
9391         dprintk("NFS: initiating layoutcommit call. sync %d "
9392                 "lbw: %llu inode %lu\n", sync,
9393                 data->args.lastbytewritten,
9394                 data->args.inode->i_ino);
9395
9396         if (!sync) {
9397                 data->inode = nfs_igrab_and_active(data->args.inode);
9398                 if (data->inode == NULL) {
9399                         nfs4_layoutcommit_release(data);
9400                         return -EAGAIN;
9401                 }
9402                 task_setup_data.flags = RPC_TASK_ASYNC;
9403         }
9404         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9405         task = rpc_run_task(&task_setup_data);
9406         if (IS_ERR(task))
9407                 return PTR_ERR(task);
9408         if (sync)
9409                 status = task->tk_status;
9410         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9411         dprintk("%s: status %d\n", __func__, status);
9412         rpc_put_task(task);
9413         return status;
9414 }
9415
9416 /*
9417  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9418  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9419  */
9420 static int
9421 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9422                     struct nfs_fsinfo *info,
9423                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9424 {
9425         struct nfs41_secinfo_no_name_args args = {
9426                 .style = SECINFO_STYLE_CURRENT_FH,
9427         };
9428         struct nfs4_secinfo_res res = {
9429                 .flavors = flavors,
9430         };
9431         struct rpc_message msg = {
9432                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9433                 .rpc_argp = &args,
9434                 .rpc_resp = &res,
9435         };
9436         struct rpc_clnt *clnt = server->client;
9437         struct nfs4_call_sync_data data = {
9438                 .seq_server = server,
9439                 .seq_args = &args.seq_args,
9440                 .seq_res = &res.seq_res,
9441         };
9442         struct rpc_task_setup task_setup = {
9443                 .rpc_client = server->client,
9444                 .rpc_message = &msg,
9445                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9446                 .callback_data = &data,
9447                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9448         };
9449         const struct cred *cred = NULL;
9450         int status;
9451
9452         if (use_integrity) {
9453                 clnt = server->nfs_client->cl_rpcclient;
9454                 task_setup.rpc_client = clnt;
9455
9456                 cred = nfs4_get_clid_cred(server->nfs_client);
9457                 msg.rpc_cred = cred;
9458         }
9459
9460         dprintk("--> %s\n", __func__);
9461         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9462         status = nfs4_call_sync_custom(&task_setup);
9463         dprintk("<-- %s status=%d\n", __func__, status);
9464
9465         put_cred(cred);
9466
9467         return status;
9468 }
9469
9470 static int
9471 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9472                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9473 {
9474         struct nfs4_exception exception = {
9475                 .interruptible = true,
9476         };
9477         int err;
9478         do {
9479                 /* first try using integrity protection */
9480                 err = -NFS4ERR_WRONGSEC;
9481
9482                 /* try to use integrity protection with machine cred */
9483                 if (_nfs4_is_integrity_protected(server->nfs_client))
9484                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9485                                                           flavors, true);
9486
9487                 /*
9488                  * if unable to use integrity protection, or SECINFO with
9489                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9490                  * disallowed by spec, but exists in deployed servers) use
9491                  * the current filesystem's rpc_client and the user cred.
9492                  */
9493                 if (err == -NFS4ERR_WRONGSEC)
9494                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9495                                                           flavors, false);
9496
9497                 switch (err) {
9498                 case 0:
9499                 case -NFS4ERR_WRONGSEC:
9500                 case -ENOTSUPP:
9501                         goto out;
9502                 default:
9503                         err = nfs4_handle_exception(server, err, &exception);
9504                 }
9505         } while (exception.retry);
9506 out:
9507         return err;
9508 }
9509
9510 static int
9511 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9512                     struct nfs_fsinfo *info)
9513 {
9514         int err;
9515         struct page *page;
9516         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9517         struct nfs4_secinfo_flavors *flavors;
9518         struct nfs4_secinfo4 *secinfo;
9519         int i;
9520
9521         page = alloc_page(GFP_KERNEL);
9522         if (!page) {
9523                 err = -ENOMEM;
9524                 goto out;
9525         }
9526
9527         flavors = page_address(page);
9528         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9529
9530         /*
9531          * Fall back on "guess and check" method if
9532          * the server doesn't support SECINFO_NO_NAME
9533          */
9534         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9535                 err = nfs4_find_root_sec(server, fhandle, info);
9536                 goto out_freepage;
9537         }
9538         if (err)
9539                 goto out_freepage;
9540
9541         for (i = 0; i < flavors->num_flavors; i++) {
9542                 secinfo = &flavors->flavors[i];
9543
9544                 switch (secinfo->flavor) {
9545                 case RPC_AUTH_NULL:
9546                 case RPC_AUTH_UNIX:
9547                 case RPC_AUTH_GSS:
9548                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9549                                         &secinfo->flavor_info);
9550                         break;
9551                 default:
9552                         flavor = RPC_AUTH_MAXFLAVOR;
9553                         break;
9554                 }
9555
9556                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9557                         flavor = RPC_AUTH_MAXFLAVOR;
9558
9559                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9560                         err = nfs4_lookup_root_sec(server, fhandle,
9561                                                    info, flavor);
9562                         if (!err)
9563                                 break;
9564                 }
9565         }
9566
9567         if (flavor == RPC_AUTH_MAXFLAVOR)
9568                 err = -EPERM;
9569
9570 out_freepage:
9571         put_page(page);
9572         if (err == -EACCES)
9573                 return -EPERM;
9574 out:
9575         return err;
9576 }
9577
9578 static int _nfs41_test_stateid(struct nfs_server *server,
9579                 nfs4_stateid *stateid,
9580                 const struct cred *cred)
9581 {
9582         int status;
9583         struct nfs41_test_stateid_args args = {
9584                 .stateid = stateid,
9585         };
9586         struct nfs41_test_stateid_res res;
9587         struct rpc_message msg = {
9588                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9589                 .rpc_argp = &args,
9590                 .rpc_resp = &res,
9591                 .rpc_cred = cred,
9592         };
9593         struct rpc_clnt *rpc_client = server->client;
9594
9595         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9596                 &rpc_client, &msg);
9597
9598         dprintk("NFS call  test_stateid %p\n", stateid);
9599         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9600         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9601                         &args.seq_args, &res.seq_res);
9602         if (status != NFS_OK) {
9603                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9604                 return status;
9605         }
9606         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9607         return -res.status;
9608 }
9609
9610 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9611                 int err, struct nfs4_exception *exception)
9612 {
9613         exception->retry = 0;
9614         switch(err) {
9615         case -NFS4ERR_DELAY:
9616         case -NFS4ERR_RETRY_UNCACHED_REP:
9617                 nfs4_handle_exception(server, err, exception);
9618                 break;
9619         case -NFS4ERR_BADSESSION:
9620         case -NFS4ERR_BADSLOT:
9621         case -NFS4ERR_BAD_HIGH_SLOT:
9622         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9623         case -NFS4ERR_DEADSESSION:
9624                 nfs4_do_handle_exception(server, err, exception);
9625         }
9626 }
9627
9628 /**
9629  * nfs41_test_stateid - perform a TEST_STATEID operation
9630  *
9631  * @server: server / transport on which to perform the operation
9632  * @stateid: state ID to test
9633  * @cred: credential
9634  *
9635  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9636  * Otherwise a negative NFS4ERR value is returned if the operation
9637  * failed or the state ID is not currently valid.
9638  */
9639 static int nfs41_test_stateid(struct nfs_server *server,
9640                 nfs4_stateid *stateid,
9641                 const struct cred *cred)
9642 {
9643         struct nfs4_exception exception = {
9644                 .interruptible = true,
9645         };
9646         int err;
9647         do {
9648                 err = _nfs41_test_stateid(server, stateid, cred);
9649                 nfs4_handle_delay_or_session_error(server, err, &exception);
9650         } while (exception.retry);
9651         return err;
9652 }
9653
9654 struct nfs_free_stateid_data {
9655         struct nfs_server *server;
9656         struct nfs41_free_stateid_args args;
9657         struct nfs41_free_stateid_res res;
9658 };
9659
9660 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9661 {
9662         struct nfs_free_stateid_data *data = calldata;
9663         nfs4_setup_sequence(data->server->nfs_client,
9664                         &data->args.seq_args,
9665                         &data->res.seq_res,
9666                         task);
9667 }
9668
9669 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9670 {
9671         struct nfs_free_stateid_data *data = calldata;
9672
9673         nfs41_sequence_done(task, &data->res.seq_res);
9674
9675         switch (task->tk_status) {
9676         case -NFS4ERR_DELAY:
9677                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9678                         rpc_restart_call_prepare(task);
9679         }
9680 }
9681
9682 static void nfs41_free_stateid_release(void *calldata)
9683 {
9684         kfree(calldata);
9685 }
9686
9687 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9688         .rpc_call_prepare = nfs41_free_stateid_prepare,
9689         .rpc_call_done = nfs41_free_stateid_done,
9690         .rpc_release = nfs41_free_stateid_release,
9691 };
9692
9693 /**
9694  * nfs41_free_stateid - perform a FREE_STATEID operation
9695  *
9696  * @server: server / transport on which to perform the operation
9697  * @stateid: state ID to release
9698  * @cred: credential
9699  * @privileged: set to true if this call needs to be privileged
9700  *
9701  * Note: this function is always asynchronous.
9702  */
9703 static int nfs41_free_stateid(struct nfs_server *server,
9704                 const nfs4_stateid *stateid,
9705                 const struct cred *cred,
9706                 bool privileged)
9707 {
9708         struct rpc_message msg = {
9709                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9710                 .rpc_cred = cred,
9711         };
9712         struct rpc_task_setup task_setup = {
9713                 .rpc_client = server->client,
9714                 .rpc_message = &msg,
9715                 .callback_ops = &nfs41_free_stateid_ops,
9716                 .flags = RPC_TASK_ASYNC,
9717         };
9718         struct nfs_free_stateid_data *data;
9719         struct rpc_task *task;
9720
9721         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9722                 &task_setup.rpc_client, &msg);
9723
9724         dprintk("NFS call  free_stateid %p\n", stateid);
9725         data = kmalloc(sizeof(*data), GFP_NOFS);
9726         if (!data)
9727                 return -ENOMEM;
9728         data->server = server;
9729         nfs4_stateid_copy(&data->args.stateid, stateid);
9730
9731         task_setup.callback_data = data;
9732
9733         msg.rpc_argp = &data->args;
9734         msg.rpc_resp = &data->res;
9735         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9736         task = rpc_run_task(&task_setup);
9737         if (IS_ERR(task))
9738                 return PTR_ERR(task);
9739         rpc_put_task(task);
9740         return 0;
9741 }
9742
9743 static void
9744 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9745 {
9746         const struct cred *cred = lsp->ls_state->owner->so_cred;
9747
9748         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9749         nfs4_free_lock_state(server, lsp);
9750 }
9751
9752 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9753                 const nfs4_stateid *s2)
9754 {
9755         if (s1->type != s2->type)
9756                 return false;
9757
9758         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9759                 return false;
9760
9761         if (s1->seqid == s2->seqid)
9762                 return true;
9763
9764         return s1->seqid == 0 || s2->seqid == 0;
9765 }
9766
9767 #endif /* CONFIG_NFS_V4_1 */
9768
9769 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9770                 const nfs4_stateid *s2)
9771 {
9772         return nfs4_stateid_match(s1, s2);
9773 }
9774
9775
9776 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9777         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9778         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9779         .recover_open   = nfs4_open_reclaim,
9780         .recover_lock   = nfs4_lock_reclaim,
9781         .establish_clid = nfs4_init_clientid,
9782         .detect_trunking = nfs40_discover_server_trunking,
9783 };
9784
9785 #if defined(CONFIG_NFS_V4_1)
9786 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9787         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9788         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9789         .recover_open   = nfs4_open_reclaim,
9790         .recover_lock   = nfs4_lock_reclaim,
9791         .establish_clid = nfs41_init_clientid,
9792         .reclaim_complete = nfs41_proc_reclaim_complete,
9793         .detect_trunking = nfs41_discover_server_trunking,
9794 };
9795 #endif /* CONFIG_NFS_V4_1 */
9796
9797 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9798         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9799         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9800         .recover_open   = nfs40_open_expired,
9801         .recover_lock   = nfs4_lock_expired,
9802         .establish_clid = nfs4_init_clientid,
9803 };
9804
9805 #if defined(CONFIG_NFS_V4_1)
9806 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9807         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9808         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9809         .recover_open   = nfs41_open_expired,
9810         .recover_lock   = nfs41_lock_expired,
9811         .establish_clid = nfs41_init_clientid,
9812 };
9813 #endif /* CONFIG_NFS_V4_1 */
9814
9815 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9816         .sched_state_renewal = nfs4_proc_async_renew,
9817         .get_state_renewal_cred = nfs4_get_renew_cred,
9818         .renew_lease = nfs4_proc_renew,
9819 };
9820
9821 #if defined(CONFIG_NFS_V4_1)
9822 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9823         .sched_state_renewal = nfs41_proc_async_sequence,
9824         .get_state_renewal_cred = nfs4_get_machine_cred,
9825         .renew_lease = nfs4_proc_sequence,
9826 };
9827 #endif
9828
9829 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9830         .get_locations = _nfs40_proc_get_locations,
9831         .fsid_present = _nfs40_proc_fsid_present,
9832 };
9833
9834 #if defined(CONFIG_NFS_V4_1)
9835 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9836         .get_locations = _nfs41_proc_get_locations,
9837         .fsid_present = _nfs41_proc_fsid_present,
9838 };
9839 #endif  /* CONFIG_NFS_V4_1 */
9840
9841 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9842         .minor_version = 0,
9843         .init_caps = NFS_CAP_READDIRPLUS
9844                 | NFS_CAP_ATOMIC_OPEN
9845                 | NFS_CAP_POSIX_LOCK,
9846         .init_client = nfs40_init_client,
9847         .shutdown_client = nfs40_shutdown_client,
9848         .match_stateid = nfs4_match_stateid,
9849         .find_root_sec = nfs4_find_root_sec,
9850         .free_lock_state = nfs4_release_lockowner,
9851         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9852         .alloc_seqid = nfs_alloc_seqid,
9853         .call_sync_ops = &nfs40_call_sync_ops,
9854         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9855         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9856         .state_renewal_ops = &nfs40_state_renewal_ops,
9857         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9858 };
9859
9860 #if defined(CONFIG_NFS_V4_1)
9861 static struct nfs_seqid *
9862 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9863 {
9864         return NULL;
9865 }
9866
9867 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9868         .minor_version = 1,
9869         .init_caps = NFS_CAP_READDIRPLUS
9870                 | NFS_CAP_ATOMIC_OPEN
9871                 | NFS_CAP_POSIX_LOCK
9872                 | NFS_CAP_STATEID_NFSV41
9873                 | NFS_CAP_ATOMIC_OPEN_V1
9874                 | NFS_CAP_LGOPEN,
9875         .init_client = nfs41_init_client,
9876         .shutdown_client = nfs41_shutdown_client,
9877         .match_stateid = nfs41_match_stateid,
9878         .find_root_sec = nfs41_find_root_sec,
9879         .free_lock_state = nfs41_free_lock_state,
9880         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9881         .alloc_seqid = nfs_alloc_no_seqid,
9882         .session_trunk = nfs4_test_session_trunk,
9883         .call_sync_ops = &nfs41_call_sync_ops,
9884         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9885         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9886         .state_renewal_ops = &nfs41_state_renewal_ops,
9887         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9888 };
9889 #endif
9890
9891 #if defined(CONFIG_NFS_V4_2)
9892 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9893         .minor_version = 2,
9894         .init_caps = NFS_CAP_READDIRPLUS
9895                 | NFS_CAP_ATOMIC_OPEN
9896                 | NFS_CAP_POSIX_LOCK
9897                 | NFS_CAP_STATEID_NFSV41
9898                 | NFS_CAP_ATOMIC_OPEN_V1
9899                 | NFS_CAP_LGOPEN
9900                 | NFS_CAP_ALLOCATE
9901                 | NFS_CAP_COPY
9902                 | NFS_CAP_OFFLOAD_CANCEL
9903                 | NFS_CAP_DEALLOCATE
9904                 | NFS_CAP_SEEK
9905                 | NFS_CAP_LAYOUTSTATS
9906                 | NFS_CAP_CLONE
9907                 | NFS_CAP_LAYOUTERROR,
9908         .init_client = nfs41_init_client,
9909         .shutdown_client = nfs41_shutdown_client,
9910         .match_stateid = nfs41_match_stateid,
9911         .find_root_sec = nfs41_find_root_sec,
9912         .free_lock_state = nfs41_free_lock_state,
9913         .call_sync_ops = &nfs41_call_sync_ops,
9914         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9915         .alloc_seqid = nfs_alloc_no_seqid,
9916         .session_trunk = nfs4_test_session_trunk,
9917         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9918         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9919         .state_renewal_ops = &nfs41_state_renewal_ops,
9920         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9921 };
9922 #endif
9923
9924 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9925         [0] = &nfs_v4_0_minor_ops,
9926 #if defined(CONFIG_NFS_V4_1)
9927         [1] = &nfs_v4_1_minor_ops,
9928 #endif
9929 #if defined(CONFIG_NFS_V4_2)
9930         [2] = &nfs_v4_2_minor_ops,
9931 #endif
9932 };
9933
9934 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9935 {
9936         ssize_t error, error2;
9937
9938         error = generic_listxattr(dentry, list, size);
9939         if (error < 0)
9940                 return error;
9941         if (list) {
9942                 list += error;
9943                 size -= error;
9944         }
9945
9946         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9947         if (error2 < 0)
9948                 return error2;
9949         return error + error2;
9950 }
9951
9952 static const struct inode_operations nfs4_dir_inode_operations = {
9953         .create         = nfs_create,
9954         .lookup         = nfs_lookup,
9955         .atomic_open    = nfs_atomic_open,
9956         .link           = nfs_link,
9957         .unlink         = nfs_unlink,
9958         .symlink        = nfs_symlink,
9959         .mkdir          = nfs_mkdir,
9960         .rmdir          = nfs_rmdir,
9961         .mknod          = nfs_mknod,
9962         .rename         = nfs_rename,
9963         .permission     = nfs_permission,
9964         .getattr        = nfs_getattr,
9965         .setattr        = nfs_setattr,
9966         .listxattr      = nfs4_listxattr,
9967 };
9968
9969 static const struct inode_operations nfs4_file_inode_operations = {
9970         .permission     = nfs_permission,
9971         .getattr        = nfs_getattr,
9972         .setattr        = nfs_setattr,
9973         .listxattr      = nfs4_listxattr,
9974 };
9975
9976 const struct nfs_rpc_ops nfs_v4_clientops = {
9977         .version        = 4,                    /* protocol version */
9978         .dentry_ops     = &nfs4_dentry_operations,
9979         .dir_inode_ops  = &nfs4_dir_inode_operations,
9980         .file_inode_ops = &nfs4_file_inode_operations,
9981         .file_ops       = &nfs4_file_operations,
9982         .getroot        = nfs4_proc_get_root,
9983         .submount       = nfs4_submount,
9984         .try_mount      = nfs4_try_mount,
9985         .getattr        = nfs4_proc_getattr,
9986         .setattr        = nfs4_proc_setattr,
9987         .lookup         = nfs4_proc_lookup,
9988         .lookupp        = nfs4_proc_lookupp,
9989         .access         = nfs4_proc_access,
9990         .readlink       = nfs4_proc_readlink,
9991         .create         = nfs4_proc_create,
9992         .remove         = nfs4_proc_remove,
9993         .unlink_setup   = nfs4_proc_unlink_setup,
9994         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9995         .unlink_done    = nfs4_proc_unlink_done,
9996         .rename_setup   = nfs4_proc_rename_setup,
9997         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9998         .rename_done    = nfs4_proc_rename_done,
9999         .link           = nfs4_proc_link,
10000         .symlink        = nfs4_proc_symlink,
10001         .mkdir          = nfs4_proc_mkdir,
10002         .rmdir          = nfs4_proc_rmdir,
10003         .readdir        = nfs4_proc_readdir,
10004         .mknod          = nfs4_proc_mknod,
10005         .statfs         = nfs4_proc_statfs,
10006         .fsinfo         = nfs4_proc_fsinfo,
10007         .pathconf       = nfs4_proc_pathconf,
10008         .set_capabilities = nfs4_server_capabilities,
10009         .decode_dirent  = nfs4_decode_dirent,
10010         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10011         .read_setup     = nfs4_proc_read_setup,
10012         .read_done      = nfs4_read_done,
10013         .write_setup    = nfs4_proc_write_setup,
10014         .write_done     = nfs4_write_done,
10015         .commit_setup   = nfs4_proc_commit_setup,
10016         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10017         .commit_done    = nfs4_commit_done,
10018         .lock           = nfs4_proc_lock,
10019         .clear_acl_cache = nfs4_zap_acl_attr,
10020         .close_context  = nfs4_close_context,
10021         .open_context   = nfs4_atomic_open,
10022         .have_delegation = nfs4_have_delegation,
10023         .alloc_client   = nfs4_alloc_client,
10024         .init_client    = nfs4_init_client,
10025         .free_client    = nfs4_free_client,
10026         .create_server  = nfs4_create_server,
10027         .clone_server   = nfs_clone_server,
10028 };
10029
10030 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10031         .name   = XATTR_NAME_NFSV4_ACL,
10032         .list   = nfs4_xattr_list_nfs4_acl,
10033         .get    = nfs4_xattr_get_nfs4_acl,
10034         .set    = nfs4_xattr_set_nfs4_acl,
10035 };
10036
10037 const struct xattr_handler *nfs4_xattr_handlers[] = {
10038         &nfs4_xattr_nfs4_acl_handler,
10039 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10040         &nfs4_xattr_nfs4_label_handler,
10041 #endif
10042         NULL
10043 };
10044
10045 /*
10046  * Local variables:
10047  *  c-basic-offset: 8
10048  * End:
10049  */