fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57 #include <linux/iversion.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_BITMASK_SZ         3
  75
  76 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  77 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  78
  79 /* file attributes which can be mapped to nfs attributes */
  80 #define NFS4_VALID_ATTRS (ATTR_MODE \
  81         | ATTR_UID \
  82         | ATTR_GID \
  83         | ATTR_SIZE \
  84         | ATTR_ATIME \
  85         | ATTR_MTIME \
  86         | ATTR_CTIME \
  87         | ATTR_ATIME_SET \
  88         | ATTR_MTIME_SET)
  89
  90 struct nfs4_opendata;
  91 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  92 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  93 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  95 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  96                             struct nfs_fattr *fattr, struct iattr *sattr,
  97                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  98                             struct nfs4_label *olabel);
  99 #ifdef CONFIG_NFS_V4_1
 100 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 101                 const struct cred *cred,
 102                 struct nfs4_slot *slot,
 103                 bool is_privileged);
 104 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 105                 const struct cred *);
 106 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 107                 const struct cred *, bool);
 108 #endif
 109
 110 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 111 static inline struct nfs4_label *
 112 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 113         struct iattr *sattr, struct nfs4_label *label)
 114 {
 115         int err;
 116
 117         if (label == NULL)
 118                 return NULL;
 119
 120         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 121                 return NULL;
 122
 123         err = security_dentry_init_security(dentry, sattr->ia_mode,
 124                                 &dentry->d_name, (void **)&label->label, &label->len);
 125         if (err == 0)
 126                 return label;
 127
 128         return NULL;
 129 }
 130 static inline void
 131 nfs4_label_release_security(struct nfs4_label *label)
 132 {
 133         if (label)
 134                 security_release_secctx(label->label, label->len);
 135 }
 136 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 137 {
 138         if (label)
 139                 return server->attr_bitmask;
 140
 141         return server->attr_bitmask_nl;
 142 }
 143 #else
 144 static inline struct nfs4_label *
 145 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 146         struct iattr *sattr, struct nfs4_label *l)
 147 { return NULL; }
 148 static inline void
 149 nfs4_label_release_security(struct nfs4_label *label)
 150 { return; }
 151 static inline u32 *
 152 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 153 { return server->attr_bitmask; }
 154 #endif
 155
 156 /* Prevent leaks of NFSv4 errors into userland */
 157 static int nfs4_map_errors(int err)
 158 {
 159         if (err >= -1000)
 160                 return err;
 161         switch (err) {
 162         case -NFS4ERR_RESOURCE:
 163         case -NFS4ERR_LAYOUTTRYLATER:
 164         case -NFS4ERR_RECALLCONFLICT:
 165                 return -EREMOTEIO;
 166         case -NFS4ERR_WRONGSEC:
 167         case -NFS4ERR_WRONG_CRED:
 168                 return -EPERM;
 169         case -NFS4ERR_BADOWNER:
 170         case -NFS4ERR_BADNAME:
 171                 return -EINVAL;
 172         case -NFS4ERR_SHARE_DENIED:
 173                 return -EACCES;
 174         case -NFS4ERR_MINOR_VERS_MISMATCH:
 175                 return -EPROTONOSUPPORT;
 176         case -NFS4ERR_FILE_OPEN:
 177                 return -EBUSY;
 178         default:
 179                 dprintk("%s could not handle NFSv4 error %d\n",
 180                                 __func__, -err);
 181                 break;
 182         }
 183         return -EIO;
 184 }
 185
 186 /*
 187  * This is our standard bitmap for GETATTR requests.
 188  */
 189 const u32 nfs4_fattr_bitmap[3] = {
 190         FATTR4_WORD0_TYPE
 191         | FATTR4_WORD0_CHANGE
 192         | FATTR4_WORD0_SIZE
 193         | FATTR4_WORD0_FSID
 194         | FATTR4_WORD0_FILEID,
 195         FATTR4_WORD1_MODE
 196         | FATTR4_WORD1_NUMLINKS
 197         | FATTR4_WORD1_OWNER
 198         | FATTR4_WORD1_OWNER_GROUP
 199         | FATTR4_WORD1_RAWDEV
 200         | FATTR4_WORD1_SPACE_USED
 201         | FATTR4_WORD1_TIME_ACCESS
 202         | FATTR4_WORD1_TIME_METADATA
 203         | FATTR4_WORD1_TIME_MODIFY
 204         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 205 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 206         FATTR4_WORD2_SECURITY_LABEL
 207 #endif
 208 };
 209
 210 static const u32 nfs4_pnfs_open_bitmap[3] = {
 211         FATTR4_WORD0_TYPE
 212         | FATTR4_WORD0_CHANGE
 213         | FATTR4_WORD0_SIZE
 214         | FATTR4_WORD0_FSID
 215         | FATTR4_WORD0_FILEID,
 216         FATTR4_WORD1_MODE
 217         | FATTR4_WORD1_NUMLINKS
 218         | FATTR4_WORD1_OWNER
 219         | FATTR4_WORD1_OWNER_GROUP
 220         | FATTR4_WORD1_RAWDEV
 221         | FATTR4_WORD1_SPACE_USED
 222         | FATTR4_WORD1_TIME_ACCESS
 223         | FATTR4_WORD1_TIME_METADATA
 224         | FATTR4_WORD1_TIME_MODIFY,
 225         FATTR4_WORD2_MDSTHRESHOLD
 226 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 227         | FATTR4_WORD2_SECURITY_LABEL
 228 #endif
 229 };
 230
 231 static const u32 nfs4_open_noattr_bitmap[3] = {
 232         FATTR4_WORD0_TYPE
 233         | FATTR4_WORD0_FILEID,
 234 };
 235
 236 const u32 nfs4_statfs_bitmap[3] = {
 237         FATTR4_WORD0_FILES_AVAIL
 238         | FATTR4_WORD0_FILES_FREE
 239         | FATTR4_WORD0_FILES_TOTAL,
 240         FATTR4_WORD1_SPACE_AVAIL
 241         | FATTR4_WORD1_SPACE_FREE
 242         | FATTR4_WORD1_SPACE_TOTAL
 243 };
 244
 245 const u32 nfs4_pathconf_bitmap[3] = {
 246         FATTR4_WORD0_MAXLINK
 247         | FATTR4_WORD0_MAXNAME,
 248         0
 249 };
 250
 251 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 252                         | FATTR4_WORD0_MAXREAD
 253                         | FATTR4_WORD0_MAXWRITE
 254                         | FATTR4_WORD0_LEASE_TIME,
 255                         FATTR4_WORD1_TIME_DELTA
 256                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 257                         FATTR4_WORD2_LAYOUT_BLKSIZE
 258                         | FATTR4_WORD2_CLONE_BLKSIZE
 259 };
 260
 261 const u32 nfs4_fs_locations_bitmap[3] = {
 262         FATTR4_WORD0_CHANGE
 263         | FATTR4_WORD0_SIZE
 264         | FATTR4_WORD0_FSID
 265         | FATTR4_WORD0_FILEID
 266         | FATTR4_WORD0_FS_LOCATIONS,
 267         FATTR4_WORD1_OWNER
 268         | FATTR4_WORD1_OWNER_GROUP
 269         | FATTR4_WORD1_RAWDEV
 270         | FATTR4_WORD1_SPACE_USED
 271         | FATTR4_WORD1_TIME_ACCESS
 272         | FATTR4_WORD1_TIME_METADATA
 273         | FATTR4_WORD1_TIME_MODIFY
 274         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 275 };
 276
 277 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 278                 struct inode *inode)
 279 {
 280         unsigned long cache_validity;
 281
 282         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 283         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 284                 return;
 285
 286         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 287         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 288                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 289                                 | NFS_INO_INVALID_SIZE);
 290
 291         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 292                 dst[0] &= ~FATTR4_WORD0_SIZE;
 293
 294         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 295                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 296 }
 297
 298 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 299                 const __u32 *src, struct inode *inode)
 300 {
 301         nfs4_bitmap_copy_adjust(dst, src, inode);
 302 }
 303
 304 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 305                 struct nfs4_readdir_arg *readdir)
 306 {
 307         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 308         __be32 *start, *p;
 309
 310         if (cookie > 2) {
 311                 readdir->cookie = cookie;
 312                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 313                 return;
 314         }
 315
 316         readdir->cookie = 0;
 317         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 318         if (cookie == 2)
 319                 return;
 320
 321         /*
 322          * NFSv4 servers do not return entries for '.' and '..'
 323          * Therefore, we fake these entries here.  We let '.'
 324          * have cookie 0 and '..' have cookie 1.  Note that
 325          * when talking to the server, we always send cookie 0
 326          * instead of 1 or 2.
 327          */
 328         start = p = kmap_atomic(*readdir->pages);
 329
 330         if (cookie == 0) {
 331                 *p++ = xdr_one;                                  /* next */
 332                 *p++ = xdr_zero;                   /* cookie, first word */
 333                 *p++ = xdr_one;                   /* cookie, second word */
 334                 *p++ = xdr_one;                             /* entry len */
 335                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 336                 p++;
 337                 *p++ = xdr_one;                         /* bitmap length */
 338                 *p++ = htonl(attrs);                           /* bitmap */
 339                 *p++ = htonl(12);             /* attribute buffer length */
 340                 *p++ = htonl(NF4DIR);
 341                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 342         }
 343
 344         *p++ = xdr_one;                                  /* next */
 345         *p++ = xdr_zero;                   /* cookie, first word */
 346         *p++ = xdr_two;                   /* cookie, second word */
 347         *p++ = xdr_two;                             /* entry len */
 348         memcpy(p, "..\0\0", 4);                         /* entry */
 349         p++;
 350         *p++ = xdr_one;                         /* bitmap length */
 351         *p++ = htonl(attrs);                           /* bitmap */
 352         *p++ = htonl(12);             /* attribute buffer length */
 353         *p++ = htonl(NF4DIR);
 354         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 355
 356         readdir->pgbase = (char *)p - (char *)start;
 357         readdir->count -= readdir->pgbase;
 358         kunmap_atomic(start);
 359 }
 360
 361 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 362                 nfs4_stateid *stateid,
 363                 const struct cred *cred)
 364 {
 365         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 366
 367         ops->test_and_free_expired(server, stateid, cred);
 368 }
 369
 370 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 371                 nfs4_stateid *stateid,
 372                 const struct cred *cred)
 373 {
 374         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 375         nfs4_test_and_free_stateid(server, stateid, cred);
 376 }
 377
 378 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 379                 const nfs4_stateid *stateid,
 380                 const struct cred *cred)
 381 {
 382         nfs4_stateid tmp;
 383
 384         nfs4_stateid_copy(&tmp, stateid);
 385         __nfs4_free_revoked_stateid(server, &tmp, cred);
 386 }
 387
 388 static long nfs4_update_delay(long *timeout)
 389 {
 390         long ret;
 391         if (!timeout)
 392                 return NFS4_POLL_RETRY_MAX;
 393         if (*timeout <= 0)
 394                 *timeout = NFS4_POLL_RETRY_MIN;
 395         if (*timeout > NFS4_POLL_RETRY_MAX)
 396                 *timeout = NFS4_POLL_RETRY_MAX;
 397         ret = *timeout;
 398         *timeout <<= 1;
 399         return ret;
 400 }
 401
 402 static int nfs4_delay_killable(long *timeout)
 403 {
 404         might_sleep();
 405
 406         freezable_schedule_timeout_killable_unsafe(
 407                 nfs4_update_delay(timeout));
 408         if (!__fatal_signal_pending(current))
 409                 return 0;
 410         return -EINTR;
 411 }
 412
 413 static int nfs4_delay_interruptible(long *timeout)
 414 {
 415         might_sleep();
 416
 417         freezable_schedule_timeout_interruptible(nfs4_update_delay(timeout));
 418         if (!signal_pending(current))
 419                 return 0;
 420         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 421 }
 422
 423 static int nfs4_delay(long *timeout, bool interruptible)
 424 {
 425         if (interruptible)
 426                 return nfs4_delay_interruptible(timeout);
 427         return nfs4_delay_killable(timeout);
 428 }
 429
 430 static const nfs4_stateid *
 431 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 432 {
 433         if (!stateid)
 434                 return NULL;
 435         switch (stateid->type) {
 436         case NFS4_OPEN_STATEID_TYPE:
 437         case NFS4_LOCK_STATEID_TYPE:
 438         case NFS4_DELEGATION_STATEID_TYPE:
 439                 return stateid;
 440         default:
 441                 break;
 442         }
 443         return NULL;
 444 }
 445
 446 /* This is the error handling routine for processes that are allowed
 447  * to sleep.
 448  */
 449 static int nfs4_do_handle_exception(struct nfs_server *server,
 450                 int errorcode, struct nfs4_exception *exception)
 451 {
 452         struct nfs_client *clp = server->nfs_client;
 453         struct nfs4_state *state = exception->state;
 454         const nfs4_stateid *stateid;
 455         struct inode *inode = exception->inode;
 456         int ret = errorcode;
 457
 458         exception->delay = 0;
 459         exception->recovering = 0;
 460         exception->retry = 0;
 461
 462         stateid = nfs4_recoverable_stateid(exception->stateid);
 463         if (stateid == NULL && state != NULL)
 464                 stateid = nfs4_recoverable_stateid(&state->stateid);
 465
 466         switch(errorcode) {
 467                 case 0:
 468                         return 0;
 469                 case -NFS4ERR_BADHANDLE:
 470                 case -ESTALE:
 471                         if (inode != NULL && S_ISREG(inode->i_mode))
 472                                 pnfs_destroy_layout(NFS_I(inode));
 473                         break;
 474                 case -NFS4ERR_DELEG_REVOKED:
 475                 case -NFS4ERR_ADMIN_REVOKED:
 476                 case -NFS4ERR_EXPIRED:
 477                 case -NFS4ERR_BAD_STATEID:
 478                 case -NFS4ERR_PARTNER_NO_AUTH:
 479                         if (inode != NULL && stateid != NULL) {
 480                                 nfs_inode_find_state_and_recover(inode,
 481                                                 stateid);
 482                                 goto wait_on_recovery;
 483                         }
 484                         /* Fall through */
 485                 case -NFS4ERR_OPENMODE:
 486                         if (inode) {
 487                                 int err;
 488
 489                                 err = nfs_async_inode_return_delegation(inode,
 490                                                 stateid);
 491                                 if (err == 0)
 492                                         goto wait_on_recovery;
 493                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 494                                         exception->retry = 1;
 495                                         break;
 496                                 }
 497                         }
 498                         if (state == NULL)
 499                                 break;
 500                         ret = nfs4_schedule_stateid_recovery(server, state);
 501                         if (ret < 0)
 502                                 break;
 503                         goto wait_on_recovery;
 504                 case -NFS4ERR_STALE_STATEID:
 505                 case -NFS4ERR_STALE_CLIENTID:
 506                         nfs4_schedule_lease_recovery(clp);
 507                         goto wait_on_recovery;
 508                 case -NFS4ERR_MOVED:
 509                         ret = nfs4_schedule_migration_recovery(server);
 510                         if (ret < 0)
 511                                 break;
 512                         goto wait_on_recovery;
 513                 case -NFS4ERR_LEASE_MOVED:
 514                         nfs4_schedule_lease_moved_recovery(clp);
 515                         goto wait_on_recovery;
 516 #if defined(CONFIG_NFS_V4_1)
 517                 case -NFS4ERR_BADSESSION:
 518                 case -NFS4ERR_BADSLOT:
 519                 case -NFS4ERR_BAD_HIGH_SLOT:
 520                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 521                 case -NFS4ERR_DEADSESSION:
 522                 case -NFS4ERR_SEQ_FALSE_RETRY:
 523                 case -NFS4ERR_SEQ_MISORDERED:
 524                         /* Handled in nfs41_sequence_process() */
 525                         goto wait_on_recovery;
 526 #endif /* defined(CONFIG_NFS_V4_1) */
 527                 case -NFS4ERR_FILE_OPEN:
 528                         if (exception->timeout > HZ) {
 529                                 /* We have retried a decent amount, time to
 530                                  * fail
 531                                  */
 532                                 ret = -EBUSY;
 533                                 break;
 534                         }
 535                         /* Fall through */
 536                 case -NFS4ERR_DELAY:
 537                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 538                         /* Fall through */
 539                 case -NFS4ERR_GRACE:
 540                 case -NFS4ERR_LAYOUTTRYLATER:
 541                 case -NFS4ERR_RECALLCONFLICT:
 542                         exception->delay = 1;
 543                         return 0;
 544
 545                 case -NFS4ERR_RETRY_UNCACHED_REP:
 546                 case -NFS4ERR_OLD_STATEID:
 547                         exception->retry = 1;
 548                         break;
 549                 case -NFS4ERR_BADOWNER:
 550                         /* The following works around a Linux server bug! */
 551                 case -NFS4ERR_BADNAME:
 552                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 553                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 554                                 exception->retry = 1;
 555                                 printk(KERN_WARNING "NFS: v4 server %s "
 556                                                 "does not accept raw "
 557                                                 "uid/gids. "
 558                                                 "Reenabling the idmapper.\n",
 559                                                 server->nfs_client->cl_hostname);
 560                         }
 561         }
 562         /* We failed to handle the error */
 563         return nfs4_map_errors(ret);
 564 wait_on_recovery:
 565         exception->recovering = 1;
 566         return 0;
 567 }
 568
 569 /* This is the error handling routine for processes that are allowed
 570  * to sleep.
 571  */
 572 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 573 {
 574         struct nfs_client *clp = server->nfs_client;
 575         int ret;
 576
 577         ret = nfs4_do_handle_exception(server, errorcode, exception);
 578         if (exception->delay) {
 579                 ret = nfs4_delay(&exception->timeout,
 580                                 exception->interruptible);
 581                 goto out_retry;
 582         }
 583         if (exception->recovering) {
 584                 ret = nfs4_wait_clnt_recover(clp);
 585                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 586                         return -EIO;
 587                 goto out_retry;
 588         }
 589         return ret;
 590 out_retry:
 591         if (ret == 0)
 592                 exception->retry = 1;
 593         return ret;
 594 }
 595
 596 static int
 597 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 598                 int errorcode, struct nfs4_exception *exception)
 599 {
 600         struct nfs_client *clp = server->nfs_client;
 601         int ret;
 602
 603         ret = nfs4_do_handle_exception(server, errorcode, exception);
 604         if (exception->delay) {
 605                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 606                 goto out_retry;
 607         }
 608         if (exception->recovering) {
 609                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 610                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 611                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 612                 goto out_retry;
 613         }
 614         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 615                 ret = -EIO;
 616         return ret;
 617 out_retry:
 618         if (ret == 0) {
 619                 exception->retry = 1;
 620                 /*
 621                  * For NFS4ERR_MOVED, the client transport will need to
 622                  * be recomputed after migration recovery has completed.
 623                  */
 624                 if (errorcode == -NFS4ERR_MOVED)
 625                         rpc_task_release_transport(task);
 626         }
 627         return ret;
 628 }
 629
 630 int
 631 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 632                         struct nfs4_state *state, long *timeout)
 633 {
 634         struct nfs4_exception exception = {
 635                 .state = state,
 636         };
 637
 638         if (task->tk_status >= 0)
 639                 return 0;
 640         if (timeout)
 641                 exception.timeout = *timeout;
 642         task->tk_status = nfs4_async_handle_exception(task, server,
 643                         task->tk_status,
 644                         &exception);
 645         if (exception.delay && timeout)
 646                 *timeout = exception.timeout;
 647         if (exception.retry)
 648                 return -EAGAIN;
 649         return 0;
 650 }
 651
 652 /*
 653  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 654  * or 'false' otherwise.
 655  */
 656 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 657 {
 658         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 659         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 660 }
 661
 662 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 663 {
 664         spin_lock(&clp->cl_lock);
 665         if (time_before(clp->cl_last_renewal,timestamp))
 666                 clp->cl_last_renewal = timestamp;
 667         spin_unlock(&clp->cl_lock);
 668 }
 669
 670 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 671 {
 672         struct nfs_client *clp = server->nfs_client;
 673
 674         if (!nfs4_has_session(clp))
 675                 do_renew_lease(clp, timestamp);
 676 }
 677
 678 struct nfs4_call_sync_data {
 679         const struct nfs_server *seq_server;
 680         struct nfs4_sequence_args *seq_args;
 681         struct nfs4_sequence_res *seq_res;
 682 };
 683
 684 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 685                         struct nfs4_sequence_res *res, int cache_reply,
 686                         int privileged)
 687 {
 688         args->sa_slot = NULL;
 689         args->sa_cache_this = cache_reply;
 690         args->sa_privileged = privileged;
 691
 692         res->sr_slot = NULL;
 693 }
 694
 695 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 696 {
 697         struct nfs4_slot *slot = res->sr_slot;
 698         struct nfs4_slot_table *tbl;
 699
 700         tbl = slot->table;
 701         spin_lock(&tbl->slot_tbl_lock);
 702         if (!nfs41_wake_and_assign_slot(tbl, slot))
 703                 nfs4_free_slot(tbl, slot);
 704         spin_unlock(&tbl->slot_tbl_lock);
 705
 706         res->sr_slot = NULL;
 707 }
 708
 709 static int nfs40_sequence_done(struct rpc_task *task,
 710                                struct nfs4_sequence_res *res)
 711 {
 712         if (res->sr_slot != NULL)
 713                 nfs40_sequence_free_slot(res);
 714         return 1;
 715 }
 716
 717 #if defined(CONFIG_NFS_V4_1)
 718
 719 static void nfs41_release_slot(struct nfs4_slot *slot)
 720 {
 721         struct nfs4_session *session;
 722         struct nfs4_slot_table *tbl;
 723         bool send_new_highest_used_slotid = false;
 724
 725         if (!slot)
 726                 return;
 727         tbl = slot->table;
 728         session = tbl->session;
 729
 730         /* Bump the slot sequence number */
 731         if (slot->seq_done)
 732                 slot->seq_nr++;
 733         slot->seq_done = 0;
 734
 735         spin_lock(&tbl->slot_tbl_lock);
 736         /* Be nice to the server: try to ensure that the last transmitted
 737          * value for highest_user_slotid <= target_highest_slotid
 738          */
 739         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 740                 send_new_highest_used_slotid = true;
 741
 742         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 743                 send_new_highest_used_slotid = false;
 744                 goto out_unlock;
 745         }
 746         nfs4_free_slot(tbl, slot);
 747
 748         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 749                 send_new_highest_used_slotid = false;
 750 out_unlock:
 751         spin_unlock(&tbl->slot_tbl_lock);
 752         if (send_new_highest_used_slotid)
 753                 nfs41_notify_server(session->clp);
 754         if (waitqueue_active(&tbl->slot_waitq))
 755                 wake_up_all(&tbl->slot_waitq);
 756 }
 757
 758 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 759 {
 760         nfs41_release_slot(res->sr_slot);
 761         res->sr_slot = NULL;
 762 }
 763
 764 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 765                 u32 seqnr)
 766 {
 767         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 768                 slot->seq_nr_highest_sent = seqnr;
 769 }
 770 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 771                 u32 seqnr)
 772 {
 773         slot->seq_nr_highest_sent = seqnr;
 774         slot->seq_nr_last_acked = seqnr;
 775 }
 776
 777 static int nfs41_sequence_process(struct rpc_task *task,
 778                 struct nfs4_sequence_res *res)
 779 {
 780         struct nfs4_session *session;
 781         struct nfs4_slot *slot = res->sr_slot;
 782         struct nfs_client *clp;
 783         int status;
 784         int ret = 1;
 785
 786         if (slot == NULL)
 787                 goto out_noaction;
 788         /* don't increment the sequence number if the task wasn't sent */
 789         if (!RPC_WAS_SENT(task) || slot->seq_done)
 790                 goto out;
 791
 792         session = slot->table->session;
 793
 794         trace_nfs4_sequence_done(session, res);
 795
 796         status = res->sr_status;
 797         if (task->tk_status == -NFS4ERR_DEADSESSION)
 798                 status = -NFS4ERR_DEADSESSION;
 799
 800         /* Check the SEQUENCE operation status */
 801         switch (status) {
 802         case 0:
 803                 /* Mark this sequence number as having been acked */
 804                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 805                 /* Update the slot's sequence and clientid lease timer */
 806                 slot->seq_done = 1;
 807                 clp = session->clp;
 808                 do_renew_lease(clp, res->sr_timestamp);
 809                 /* Check sequence flags */
 810                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 811                                 !!slot->privileged);
 812                 nfs41_update_target_slotid(slot->table, slot, res);
 813                 break;
 814         case 1:
 815                 /*
 816                  * sr_status remains 1 if an RPC level error occurred.
 817                  * The server may or may not have processed the sequence
 818                  * operation..
 819                  */
 820                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 821                 slot->seq_done = 1;
 822                 goto out;
 823         case -NFS4ERR_DELAY:
 824                 /* The server detected a resend of the RPC call and
 825                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 826                  * of RFC5661.
 827                  */
 828                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 829                         __func__,
 830                         slot->slot_nr,
 831                         slot->seq_nr);
 832                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 833                 goto out_retry;
 834         case -NFS4ERR_RETRY_UNCACHED_REP:
 835         case -NFS4ERR_SEQ_FALSE_RETRY:
 836                 /*
 837                  * The server thinks we tried to replay a request.
 838                  * Retry the call after bumping the sequence ID.
 839                  */
 840                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 841                 goto retry_new_seq;
 842         case -NFS4ERR_BADSLOT:
 843                 /*
 844                  * The slot id we used was probably retired. Try again
 845                  * using a different slot id.
 846                  */
 847                 if (slot->slot_nr < slot->table->target_highest_slotid)
 848                         goto session_recover;
 849                 goto retry_nowait;
 850         case -NFS4ERR_SEQ_MISORDERED:
 851                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 852                 /*
 853                  * Were one or more calls using this slot interrupted?
 854                  * If the server never received the request, then our
 855                  * transmitted slot sequence number may be too high.
 856                  */
 857                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 858                         slot->seq_nr--;
 859                         goto retry_nowait;
 860                 }
 861                 /*
 862                  * RFC5661:
 863                  * A retry might be sent while the original request is
 864                  * still in progress on the replier. The replier SHOULD
 865                  * deal with the issue by returning NFS4ERR_DELAY as the
 866                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 867                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 868                  *
 869                  * Restart the search after a delay.
 870                  */
 871                 slot->seq_nr = slot->seq_nr_highest_sent;
 872                 goto out_retry;
 873         case -NFS4ERR_BADSESSION:
 874         case -NFS4ERR_DEADSESSION:
 875         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 876                 goto session_recover;
 877         default:
 878                 /* Just update the slot sequence no. */
 879                 slot->seq_done = 1;
 880         }
 881 out:
 882         /* The session may be reset by one of the error handlers. */
 883         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 884 out_noaction:
 885         return ret;
 886 session_recover:
 887         nfs4_schedule_session_recovery(session, status);
 888         dprintk("%s ERROR: %d Reset session\n", __func__, status);
 889         nfs41_sequence_free_slot(res);
 890         goto out;
 891 retry_new_seq:
 892         ++slot->seq_nr;
 893 retry_nowait:
 894         if (rpc_restart_call_prepare(task)) {
 895                 nfs41_sequence_free_slot(res);
 896                 task->tk_status = 0;
 897                 ret = 0;
 898         }
 899         goto out;
 900 out_retry:
 901         if (!rpc_restart_call(task))
 902                 goto out;
 903         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 904         return 0;
 905 }
 906
 907 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 908 {
 909         if (!nfs41_sequence_process(task, res))
 910                 return 0;
 911         if (res->sr_slot != NULL)
 912                 nfs41_sequence_free_slot(res);
 913         return 1;
 914
 915 }
 916 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 917
 918 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 919 {
 920         if (res->sr_slot == NULL)
 921                 return 1;
 922         if (res->sr_slot->table->session != NULL)
 923                 return nfs41_sequence_process(task, res);
 924         return nfs40_sequence_done(task, res);
 925 }
 926
 927 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 928 {
 929         if (res->sr_slot != NULL) {
 930                 if (res->sr_slot->table->session != NULL)
 931                         nfs41_sequence_free_slot(res);
 932                 else
 933                         nfs40_sequence_free_slot(res);
 934         }
 935 }
 936
 937 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 938 {
 939         if (res->sr_slot == NULL)
 940                 return 1;
 941         if (!res->sr_slot->table->session)
 942                 return nfs40_sequence_done(task, res);
 943         return nfs41_sequence_done(task, res);
 944 }
 945 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 946
 947 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 948 {
 949         struct nfs4_call_sync_data *data = calldata;
 950
 951         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 952
 953         nfs4_setup_sequence(data->seq_server->nfs_client,
 954                             data->seq_args, data->seq_res, task);
 955 }
 956
 957 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 958 {
 959         struct nfs4_call_sync_data *data = calldata;
 960
 961         nfs41_sequence_done(task, data->seq_res);
 962 }
 963
 964 static const struct rpc_call_ops nfs41_call_sync_ops = {
 965         .rpc_call_prepare = nfs41_call_sync_prepare,
 966         .rpc_call_done = nfs41_call_sync_done,
 967 };
 968
 969 #else   /* !CONFIG_NFS_V4_1 */
 970
 971 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 972 {
 973         return nfs40_sequence_done(task, res);
 974 }
 975
 976 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 977 {
 978         if (res->sr_slot != NULL)
 979                 nfs40_sequence_free_slot(res);
 980 }
 981
 982 int nfs4_sequence_done(struct rpc_task *task,
 983                        struct nfs4_sequence_res *res)
 984 {
 985         return nfs40_sequence_done(task, res);
 986 }
 987 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 988
 989 #endif  /* !CONFIG_NFS_V4_1 */
 990
 991 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
 992 {
 993         res->sr_timestamp = jiffies;
 994         res->sr_status_flags = 0;
 995         res->sr_status = 1;
 996 }
 997
 998 static
 999 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1000                 struct nfs4_sequence_res *res,
1001                 struct nfs4_slot *slot)
1002 {
1003         if (!slot)
1004                 return;
1005         slot->privileged = args->sa_privileged ? 1 : 0;
1006         args->sa_slot = slot;
1007
1008         res->sr_slot = slot;
1009 }
1010
1011 int nfs4_setup_sequence(struct nfs_client *client,
1012                         struct nfs4_sequence_args *args,
1013                         struct nfs4_sequence_res *res,
1014                         struct rpc_task *task)
1015 {
1016         struct nfs4_session *session = nfs4_get_session(client);
1017         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1018         struct nfs4_slot *slot;
1019
1020         /* slot already allocated? */
1021         if (res->sr_slot != NULL)
1022                 goto out_start;
1023
1024         if (session)
1025                 tbl = &session->fc_slot_table;
1026
1027         spin_lock(&tbl->slot_tbl_lock);
1028         /* The state manager will wait until the slot table is empty */
1029         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1030                 goto out_sleep;
1031
1032         slot = nfs4_alloc_slot(tbl);
1033         if (IS_ERR(slot)) {
1034                 if (slot == ERR_PTR(-ENOMEM))
1035                         goto out_sleep_timeout;
1036                 goto out_sleep;
1037         }
1038         spin_unlock(&tbl->slot_tbl_lock);
1039
1040         nfs4_sequence_attach_slot(args, res, slot);
1041
1042         trace_nfs4_setup_sequence(session, args);
1043 out_start:
1044         nfs41_sequence_res_init(res);
1045         rpc_call_start(task);
1046         return 0;
1047 out_sleep_timeout:
1048         /* Try again in 1/4 second */
1049         if (args->sa_privileged)
1050                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1051                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1052         else
1053                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1054                                 NULL, jiffies + (HZ >> 2));
1055         spin_unlock(&tbl->slot_tbl_lock);
1056         return -EAGAIN;
1057 out_sleep:
1058         if (args->sa_privileged)
1059                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1060                                 RPC_PRIORITY_PRIVILEGED);
1061         else
1062                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1063         spin_unlock(&tbl->slot_tbl_lock);
1064         return -EAGAIN;
1065 }
1066 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1067
1068 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1069 {
1070         struct nfs4_call_sync_data *data = calldata;
1071         nfs4_setup_sequence(data->seq_server->nfs_client,
1072                                 data->seq_args, data->seq_res, task);
1073 }
1074
1075 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1076 {
1077         struct nfs4_call_sync_data *data = calldata;
1078         nfs4_sequence_done(task, data->seq_res);
1079 }
1080
1081 static const struct rpc_call_ops nfs40_call_sync_ops = {
1082         .rpc_call_prepare = nfs40_call_sync_prepare,
1083         .rpc_call_done = nfs40_call_sync_done,
1084 };
1085
1086 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1087 {
1088         int ret;
1089         struct rpc_task *task;
1090
1091         task = rpc_run_task(task_setup);
1092         if (IS_ERR(task))
1093                 return PTR_ERR(task);
1094
1095         ret = task->tk_status;
1096         rpc_put_task(task);
1097         return ret;
1098 }
1099
1100 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1101                              struct nfs_server *server,
1102                              struct rpc_message *msg,
1103                              struct nfs4_sequence_args *args,
1104                              struct nfs4_sequence_res *res,
1105                              unsigned short task_flags)
1106 {
1107         struct nfs_client *clp = server->nfs_client;
1108         struct nfs4_call_sync_data data = {
1109                 .seq_server = server,
1110                 .seq_args = args,
1111                 .seq_res = res,
1112         };
1113         struct rpc_task_setup task_setup = {
1114                 .rpc_client = clnt,
1115                 .rpc_message = msg,
1116                 .callback_ops = clp->cl_mvops->call_sync_ops,
1117                 .callback_data = &data,
1118                 .flags = task_flags,
1119         };
1120
1121         return nfs4_call_sync_custom(&task_setup);
1122 }
1123
1124 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1125                                    struct nfs_server *server,
1126                                    struct rpc_message *msg,
1127                                    struct nfs4_sequence_args *args,
1128                                    struct nfs4_sequence_res *res)
1129 {
1130         return nfs4_do_call_sync(clnt, server, msg, args, res, 0);
1131 }
1132
1133
1134 int nfs4_call_sync(struct rpc_clnt *clnt,
1135                    struct nfs_server *server,
1136                    struct rpc_message *msg,
1137                    struct nfs4_sequence_args *args,
1138                    struct nfs4_sequence_res *res,
1139                    int cache_reply)
1140 {
1141         nfs4_init_sequence(args, res, cache_reply, 0);
1142         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1143 }
1144
1145 static void
1146 nfs4_inc_nlink_locked(struct inode *inode)
1147 {
1148         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1149         inc_nlink(inode);
1150 }
1151
1152 static void
1153 nfs4_dec_nlink_locked(struct inode *inode)
1154 {
1155         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1156         drop_nlink(inode);
1157 }
1158
1159 static void
1160 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1161                 unsigned long timestamp, unsigned long cache_validity)
1162 {
1163         struct nfs_inode *nfsi = NFS_I(dir);
1164
1165         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1166                 | NFS_INO_INVALID_MTIME
1167                 | NFS_INO_INVALID_DATA
1168                 | cache_validity;
1169         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1170                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1171                 nfsi->attrtimeo_timestamp = jiffies;
1172         } else {
1173                 nfs_force_lookup_revalidate(dir);
1174                 if (cinfo->before != inode_peek_iversion_raw(dir))
1175                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1176                                 NFS_INO_INVALID_ACL;
1177         }
1178         inode_set_iversion_raw(dir, cinfo->after);
1179         nfsi->read_cache_jiffies = timestamp;
1180         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1181         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1182         nfs_fscache_invalidate(dir);
1183 }
1184
1185 static void
1186 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1187                 unsigned long timestamp, unsigned long cache_validity)
1188 {
1189         spin_lock(&dir->i_lock);
1190         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1191         spin_unlock(&dir->i_lock);
1192 }
1193
1194 struct nfs4_open_createattrs {
1195         struct nfs4_label *label;
1196         struct iattr *sattr;
1197         const __u32 verf[2];
1198 };
1199
1200 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1201                 int err, struct nfs4_exception *exception)
1202 {
1203         if (err != -EINVAL)
1204                 return false;
1205         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1206                 return false;
1207         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1208         exception->retry = 1;
1209         return true;
1210 }
1211
1212 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1213 {
1214          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1215 }
1216
1217 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1218 {
1219         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1220
1221         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1222 }
1223
1224 static u32
1225 nfs4_map_atomic_open_share(struct nfs_server *server,
1226                 fmode_t fmode, int openflags)
1227 {
1228         u32 res = 0;
1229
1230         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1231         case FMODE_READ:
1232                 res = NFS4_SHARE_ACCESS_READ;
1233                 break;
1234         case FMODE_WRITE:
1235                 res = NFS4_SHARE_ACCESS_WRITE;
1236                 break;
1237         case FMODE_READ|FMODE_WRITE:
1238                 res = NFS4_SHARE_ACCESS_BOTH;
1239         }
1240         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1241                 goto out;
1242         /* Want no delegation if we're using O_DIRECT */
1243         if (openflags & O_DIRECT)
1244                 res |= NFS4_SHARE_WANT_NO_DELEG;
1245 out:
1246         return res;
1247 }
1248
1249 static enum open_claim_type4
1250 nfs4_map_atomic_open_claim(struct nfs_server *server,
1251                 enum open_claim_type4 claim)
1252 {
1253         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1254                 return claim;
1255         switch (claim) {
1256         default:
1257                 return claim;
1258         case NFS4_OPEN_CLAIM_FH:
1259                 return NFS4_OPEN_CLAIM_NULL;
1260         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1261                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1262         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1263                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1264         }
1265 }
1266
1267 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1268 {
1269         p->o_res.f_attr = &p->f_attr;
1270         p->o_res.f_label = p->f_label;
1271         p->o_res.seqid = p->o_arg.seqid;
1272         p->c_res.seqid = p->c_arg.seqid;
1273         p->o_res.server = p->o_arg.server;
1274         p->o_res.access_request = p->o_arg.access;
1275         nfs_fattr_init(&p->f_attr);
1276         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1277 }
1278
1279 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1280                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1281                 const struct nfs4_open_createattrs *c,
1282                 enum open_claim_type4 claim,
1283                 gfp_t gfp_mask)
1284 {
1285         struct dentry *parent = dget_parent(dentry);
1286         struct inode *dir = d_inode(parent);
1287         struct nfs_server *server = NFS_SERVER(dir);
1288         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1289         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1290         struct nfs4_opendata *p;
1291
1292         p = kzalloc(sizeof(*p), gfp_mask);
1293         if (p == NULL)
1294                 goto err;
1295
1296         p->f_label = nfs4_label_alloc(server, gfp_mask);
1297         if (IS_ERR(p->f_label))
1298                 goto err_free_p;
1299
1300         p->a_label = nfs4_label_alloc(server, gfp_mask);
1301         if (IS_ERR(p->a_label))
1302                 goto err_free_f;
1303
1304         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1305         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1306         if (IS_ERR(p->o_arg.seqid))
1307                 goto err_free_label;
1308         nfs_sb_active(dentry->d_sb);
1309         p->dentry = dget(dentry);
1310         p->dir = parent;
1311         p->owner = sp;
1312         atomic_inc(&sp->so_count);
1313         p->o_arg.open_flags = flags;
1314         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1315         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1316         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1317                         fmode, flags);
1318         if (flags & O_CREAT) {
1319                 p->o_arg.umask = current_umask();
1320                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1321                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1322                         p->o_arg.u.attrs = &p->attrs;
1323                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1324
1325                         memcpy(p->o_arg.u.verifier.data, c->verf,
1326                                         sizeof(p->o_arg.u.verifier.data));
1327                 }
1328         }
1329         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1330          * will return permission denied for all bits until close */
1331         if (!(flags & O_EXCL)) {
1332                 /* ask server to check for all possible rights as results
1333                  * are cached */
1334                 switch (p->o_arg.claim) {
1335                 default:
1336                         break;
1337                 case NFS4_OPEN_CLAIM_NULL:
1338                 case NFS4_OPEN_CLAIM_FH:
1339                         p->o_arg.access = NFS4_ACCESS_READ |
1340                                 NFS4_ACCESS_MODIFY |
1341                                 NFS4_ACCESS_EXTEND |
1342                                 NFS4_ACCESS_EXECUTE;
1343                 }
1344         }
1345         p->o_arg.clientid = server->nfs_client->cl_clientid;
1346         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1347         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1348         p->o_arg.name = &dentry->d_name;
1349         p->o_arg.server = server;
1350         p->o_arg.bitmask = nfs4_bitmask(server, label);
1351         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1352         switch (p->o_arg.claim) {
1353         case NFS4_OPEN_CLAIM_NULL:
1354         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1355         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1356                 p->o_arg.fh = NFS_FH(dir);
1357                 break;
1358         case NFS4_OPEN_CLAIM_PREVIOUS:
1359         case NFS4_OPEN_CLAIM_FH:
1360         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1361         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1362                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1363         }
1364         p->c_arg.fh = &p->o_res.fh;
1365         p->c_arg.stateid = &p->o_res.stateid;
1366         p->c_arg.seqid = p->o_arg.seqid;
1367         nfs4_init_opendata_res(p);
1368         kref_init(&p->kref);
1369         return p;
1370
1371 err_free_label:
1372         nfs4_label_free(p->a_label);
1373 err_free_f:
1374         nfs4_label_free(p->f_label);
1375 err_free_p:
1376         kfree(p);
1377 err:
1378         dput(parent);
1379         return NULL;
1380 }
1381
1382 static void nfs4_opendata_free(struct kref *kref)
1383 {
1384         struct nfs4_opendata *p = container_of(kref,
1385                         struct nfs4_opendata, kref);
1386         struct super_block *sb = p->dentry->d_sb;
1387
1388         nfs4_lgopen_release(p->lgp);
1389         nfs_free_seqid(p->o_arg.seqid);
1390         nfs4_sequence_free_slot(&p->o_res.seq_res);
1391         if (p->state != NULL)
1392                 nfs4_put_open_state(p->state);
1393         nfs4_put_state_owner(p->owner);
1394
1395         nfs4_label_free(p->a_label);
1396         nfs4_label_free(p->f_label);
1397
1398         dput(p->dir);
1399         dput(p->dentry);
1400         nfs_sb_deactive(sb);
1401         nfs_fattr_free_names(&p->f_attr);
1402         kfree(p->f_attr.mdsthreshold);
1403         kfree(p);
1404 }
1405
1406 static void nfs4_opendata_put(struct nfs4_opendata *p)
1407 {
1408         if (p != NULL)
1409                 kref_put(&p->kref, nfs4_opendata_free);
1410 }
1411
1412 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1413                 fmode_t fmode)
1414 {
1415         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1416         case FMODE_READ|FMODE_WRITE:
1417                 return state->n_rdwr != 0;
1418         case FMODE_WRITE:
1419                 return state->n_wronly != 0;
1420         case FMODE_READ:
1421                 return state->n_rdonly != 0;
1422         }
1423         WARN_ON_ONCE(1);
1424         return false;
1425 }
1426
1427 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1428                 int open_mode, enum open_claim_type4 claim)
1429 {
1430         int ret = 0;
1431
1432         if (open_mode & (O_EXCL|O_TRUNC))
1433                 goto out;
1434         switch (claim) {
1435         case NFS4_OPEN_CLAIM_NULL:
1436         case NFS4_OPEN_CLAIM_FH:
1437                 goto out;
1438         default:
1439                 break;
1440         }
1441         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1442                 case FMODE_READ:
1443                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1444                                 && state->n_rdonly != 0;
1445                         break;
1446                 case FMODE_WRITE:
1447                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1448                                 && state->n_wronly != 0;
1449                         break;
1450                 case FMODE_READ|FMODE_WRITE:
1451                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1452                                 && state->n_rdwr != 0;
1453         }
1454 out:
1455         return ret;
1456 }
1457
1458 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1459                 enum open_claim_type4 claim)
1460 {
1461         if (delegation == NULL)
1462                 return 0;
1463         if ((delegation->type & fmode) != fmode)
1464                 return 0;
1465         switch (claim) {
1466         case NFS4_OPEN_CLAIM_NULL:
1467         case NFS4_OPEN_CLAIM_FH:
1468                 break;
1469         case NFS4_OPEN_CLAIM_PREVIOUS:
1470                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1471                         break;
1472                 /* Fall through */
1473         default:
1474                 return 0;
1475         }
1476         nfs_mark_delegation_referenced(delegation);
1477         return 1;
1478 }
1479
1480 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1481 {
1482         switch (fmode) {
1483                 case FMODE_WRITE:
1484                         state->n_wronly++;
1485                         break;
1486                 case FMODE_READ:
1487                         state->n_rdonly++;
1488                         break;
1489                 case FMODE_READ|FMODE_WRITE:
1490                         state->n_rdwr++;
1491         }
1492         nfs4_state_set_mode_locked(state, state->state | fmode);
1493 }
1494
1495 #ifdef CONFIG_NFS_V4_1
1496 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1497 {
1498         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1499                 return true;
1500         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1501                 return true;
1502         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1503                 return true;
1504         return false;
1505 }
1506 #endif /* CONFIG_NFS_V4_1 */
1507
1508 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1509 {
1510         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1511                 wake_up_all(&state->waitq);
1512 }
1513
1514 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1515                 const nfs4_stateid *stateid)
1516 {
1517         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1518         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1519
1520         if (stateid_seqid == state_seqid + 1U ||
1521             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1522                 nfs_state_log_update_open_stateid(state);
1523         else
1524                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1525 }
1526
1527 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1528 {
1529         struct nfs_client *clp = state->owner->so_server->nfs_client;
1530         bool need_recover = false;
1531
1532         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1533                 need_recover = true;
1534         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1535                 need_recover = true;
1536         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1537                 need_recover = true;
1538         if (need_recover)
1539                 nfs4_state_mark_reclaim_nograce(clp, state);
1540 }
1541
1542 /*
1543  * Check for whether or not the caller may update the open stateid
1544  * to the value passed in by stateid.
1545  *
1546  * Note: This function relies heavily on the server implementing
1547  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1548  * correctly.
1549  * i.e. The stateid seqids have to be initialised to 1, and
1550  * are then incremented on every state transition.
1551  */
1552 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1553                 const nfs4_stateid *stateid)
1554 {
1555         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1556             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1557                 if (stateid->seqid == cpu_to_be32(1))
1558                         nfs_state_log_update_open_stateid(state);
1559                 else
1560                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1561                 return true;
1562         }
1563
1564         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1565                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1566                 return true;
1567         }
1568         return false;
1569 }
1570
1571 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1572 {
1573         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1574                 return;
1575         if (state->n_wronly)
1576                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1577         if (state->n_rdonly)
1578                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1579         if (state->n_rdwr)
1580                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1581         set_bit(NFS_OPEN_STATE, &state->flags);
1582 }
1583
1584 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1585                 nfs4_stateid *stateid, fmode_t fmode)
1586 {
1587         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1588         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1589         case FMODE_WRITE:
1590                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1591                 break;
1592         case FMODE_READ:
1593                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1594                 break;
1595         case 0:
1596                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1597                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1598                 clear_bit(NFS_OPEN_STATE, &state->flags);
1599         }
1600         if (stateid == NULL)
1601                 return;
1602         /* Handle OPEN+OPEN_DOWNGRADE races */
1603         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1604             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1605                 nfs_resync_open_stateid_locked(state);
1606                 goto out;
1607         }
1608         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1609                 nfs4_stateid_copy(&state->stateid, stateid);
1610         nfs4_stateid_copy(&state->open_stateid, stateid);
1611         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1612 out:
1613         nfs_state_log_update_open_stateid(state);
1614 }
1615
1616 static void nfs_clear_open_stateid(struct nfs4_state *state,
1617         nfs4_stateid *arg_stateid,
1618         nfs4_stateid *stateid, fmode_t fmode)
1619 {
1620         write_seqlock(&state->seqlock);
1621         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1622         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1623                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1624         write_sequnlock(&state->seqlock);
1625         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1626                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1627 }
1628
1629 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1630                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1631         __must_hold(&state->owner->so_lock)
1632         __must_hold(&state->seqlock)
1633         __must_hold(RCU)
1634
1635 {
1636         DEFINE_WAIT(wait);
1637         int status = 0;
1638         for (;;) {
1639
1640                 if (!nfs_need_update_open_stateid(state, stateid))
1641                         return;
1642                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1643                         break;
1644                 if (status)
1645                         break;
1646                 /* Rely on seqids for serialisation with NFSv4.0 */
1647                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1648                         break;
1649
1650                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1651                 /*
1652                  * Ensure we process the state changes in the same order
1653                  * in which the server processed them by delaying the
1654                  * update of the stateid until we are in sequence.
1655                  */
1656                 write_sequnlock(&state->seqlock);
1657                 spin_unlock(&state->owner->so_lock);
1658                 rcu_read_unlock();
1659                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1660                 if (!signal_pending(current)) {
1661                         if (schedule_timeout(5*HZ) == 0)
1662                                 status = -EAGAIN;
1663                         else
1664                                 status = 0;
1665                 } else
1666                         status = -EINTR;
1667                 finish_wait(&state->waitq, &wait);
1668                 rcu_read_lock();
1669                 spin_lock(&state->owner->so_lock);
1670                 write_seqlock(&state->seqlock);
1671         }
1672
1673         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1674             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1675                 nfs4_stateid_copy(freeme, &state->open_stateid);
1676                 nfs_test_and_clear_all_open_stateid(state);
1677         }
1678
1679         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1680                 nfs4_stateid_copy(&state->stateid, stateid);
1681         nfs4_stateid_copy(&state->open_stateid, stateid);
1682         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1683         nfs_state_log_update_open_stateid(state);
1684 }
1685
1686 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1687                 const nfs4_stateid *open_stateid,
1688                 fmode_t fmode,
1689                 nfs4_stateid *freeme)
1690 {
1691         /*
1692          * Protect the call to nfs4_state_set_mode_locked and
1693          * serialise the stateid update
1694          */
1695         write_seqlock(&state->seqlock);
1696         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1697         switch (fmode) {
1698         case FMODE_READ:
1699                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1700                 break;
1701         case FMODE_WRITE:
1702                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1703                 break;
1704         case FMODE_READ|FMODE_WRITE:
1705                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1706         }
1707         set_bit(NFS_OPEN_STATE, &state->flags);
1708         write_sequnlock(&state->seqlock);
1709 }
1710
1711 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1712 {
1713         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1714         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1715         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1716         clear_bit(NFS_OPEN_STATE, &state->flags);
1717 }
1718
1719 static void nfs_state_set_delegation(struct nfs4_state *state,
1720                 const nfs4_stateid *deleg_stateid,
1721                 fmode_t fmode)
1722 {
1723         /*
1724          * Protect the call to nfs4_state_set_mode_locked and
1725          * serialise the stateid update
1726          */
1727         write_seqlock(&state->seqlock);
1728         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1729         set_bit(NFS_DELEGATED_STATE, &state->flags);
1730         write_sequnlock(&state->seqlock);
1731 }
1732
1733 static void nfs_state_clear_delegation(struct nfs4_state *state)
1734 {
1735         write_seqlock(&state->seqlock);
1736         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1737         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1738         write_sequnlock(&state->seqlock);
1739 }
1740
1741 int update_open_stateid(struct nfs4_state *state,
1742                 const nfs4_stateid *open_stateid,
1743                 const nfs4_stateid *delegation,
1744                 fmode_t fmode)
1745 {
1746         struct nfs_server *server = NFS_SERVER(state->inode);
1747         struct nfs_client *clp = server->nfs_client;
1748         struct nfs_inode *nfsi = NFS_I(state->inode);
1749         struct nfs_delegation *deleg_cur;
1750         nfs4_stateid freeme = { };
1751         int ret = 0;
1752
1753         fmode &= (FMODE_READ|FMODE_WRITE);
1754
1755         rcu_read_lock();
1756         spin_lock(&state->owner->so_lock);
1757         if (open_stateid != NULL) {
1758                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1759                 ret = 1;
1760         }
1761
1762         deleg_cur = nfs4_get_valid_delegation(state->inode);
1763         if (deleg_cur == NULL)
1764                 goto no_delegation;
1765
1766         spin_lock(&deleg_cur->lock);
1767         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1768            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1769             (deleg_cur->type & fmode) != fmode)
1770                 goto no_delegation_unlock;
1771
1772         if (delegation == NULL)
1773                 delegation = &deleg_cur->stateid;
1774         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1775                 goto no_delegation_unlock;
1776
1777         nfs_mark_delegation_referenced(deleg_cur);
1778         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1779         ret = 1;
1780 no_delegation_unlock:
1781         spin_unlock(&deleg_cur->lock);
1782 no_delegation:
1783         if (ret)
1784                 update_open_stateflags(state, fmode);
1785         spin_unlock(&state->owner->so_lock);
1786         rcu_read_unlock();
1787
1788         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1789                 nfs4_schedule_state_manager(clp);
1790         if (freeme.type != 0)
1791                 nfs4_test_and_free_stateid(server, &freeme,
1792                                 state->owner->so_cred);
1793
1794         return ret;
1795 }
1796
1797 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1798                 const nfs4_stateid *stateid)
1799 {
1800         struct nfs4_state *state = lsp->ls_state;
1801         bool ret = false;
1802
1803         spin_lock(&state->state_lock);
1804         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1805                 goto out_noupdate;
1806         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1807                 goto out_noupdate;
1808         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1809         ret = true;
1810 out_noupdate:
1811         spin_unlock(&state->state_lock);
1812         return ret;
1813 }
1814
1815 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1816 {
1817         struct nfs_delegation *delegation;
1818
1819         fmode &= FMODE_READ|FMODE_WRITE;
1820         rcu_read_lock();
1821         delegation = nfs4_get_valid_delegation(inode);
1822         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1823                 rcu_read_unlock();
1824                 return;
1825         }
1826         rcu_read_unlock();
1827         nfs4_inode_return_delegation(inode);
1828 }
1829
1830 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1831 {
1832         struct nfs4_state *state = opendata->state;
1833         struct nfs_delegation *delegation;
1834         int open_mode = opendata->o_arg.open_flags;
1835         fmode_t fmode = opendata->o_arg.fmode;
1836         enum open_claim_type4 claim = opendata->o_arg.claim;
1837         nfs4_stateid stateid;
1838         int ret = -EAGAIN;
1839
1840         for (;;) {
1841                 spin_lock(&state->owner->so_lock);
1842                 if (can_open_cached(state, fmode, open_mode, claim)) {
1843                         update_open_stateflags(state, fmode);
1844                         spin_unlock(&state->owner->so_lock);
1845                         goto out_return_state;
1846                 }
1847                 spin_unlock(&state->owner->so_lock);
1848                 rcu_read_lock();
1849                 delegation = nfs4_get_valid_delegation(state->inode);
1850                 if (!can_open_delegated(delegation, fmode, claim)) {
1851                         rcu_read_unlock();
1852                         break;
1853                 }
1854                 /* Save the delegation */
1855                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1856                 rcu_read_unlock();
1857                 nfs_release_seqid(opendata->o_arg.seqid);
1858                 if (!opendata->is_recover) {
1859                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1860                         if (ret != 0)
1861                                 goto out;
1862                 }
1863                 ret = -EAGAIN;
1864
1865                 /* Try to update the stateid using the delegation */
1866                 if (update_open_stateid(state, NULL, &stateid, fmode))
1867                         goto out_return_state;
1868         }
1869 out:
1870         return ERR_PTR(ret);
1871 out_return_state:
1872         refcount_inc(&state->count);
1873         return state;
1874 }
1875
1876 static void
1877 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1878 {
1879         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1880         struct nfs_delegation *delegation;
1881         int delegation_flags = 0;
1882
1883         rcu_read_lock();
1884         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1885         if (delegation)
1886                 delegation_flags = delegation->flags;
1887         rcu_read_unlock();
1888         switch (data->o_arg.claim) {
1889         default:
1890                 break;
1891         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1892         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1893                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1894                                    "returning a delegation for "
1895                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1896                                    clp->cl_hostname);
1897                 return;
1898         }
1899         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1900                 nfs_inode_set_delegation(state->inode,
1901                                 data->owner->so_cred,
1902                                 data->o_res.delegation_type,
1903                                 &data->o_res.delegation,
1904                                 data->o_res.pagemod_limit);
1905         else
1906                 nfs_inode_reclaim_delegation(state->inode,
1907                                 data->owner->so_cred,
1908                                 data->o_res.delegation_type,
1909                                 &data->o_res.delegation,
1910                                 data->o_res.pagemod_limit);
1911
1912         if (data->o_res.do_recall)
1913                 nfs_async_inode_return_delegation(state->inode,
1914                                                   &data->o_res.delegation);
1915 }
1916
1917 /*
1918  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1919  * and update the nfs4_state.
1920  */
1921 static struct nfs4_state *
1922 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1923 {
1924         struct inode *inode = data->state->inode;
1925         struct nfs4_state *state = data->state;
1926         int ret;
1927
1928         if (!data->rpc_done) {
1929                 if (data->rpc_status)
1930                         return ERR_PTR(data->rpc_status);
1931                 /* cached opens have already been processed */
1932                 goto update;
1933         }
1934
1935         ret = nfs_refresh_inode(inode, &data->f_attr);
1936         if (ret)
1937                 return ERR_PTR(ret);
1938
1939         if (data->o_res.delegation_type != 0)
1940                 nfs4_opendata_check_deleg(data, state);
1941 update:
1942         if (!update_open_stateid(state, &data->o_res.stateid,
1943                                 NULL, data->o_arg.fmode))
1944                 return ERR_PTR(-EAGAIN);
1945         refcount_inc(&state->count);
1946
1947         return state;
1948 }
1949
1950 static struct inode *
1951 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1952 {
1953         struct inode *inode;
1954
1955         switch (data->o_arg.claim) {
1956         case NFS4_OPEN_CLAIM_NULL:
1957         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1958         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1959                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1960                         return ERR_PTR(-EAGAIN);
1961                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1962                                 &data->f_attr, data->f_label);
1963                 break;
1964         default:
1965                 inode = d_inode(data->dentry);
1966                 ihold(inode);
1967                 nfs_refresh_inode(inode, &data->f_attr);
1968         }
1969         return inode;
1970 }
1971
1972 static struct nfs4_state *
1973 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1974 {
1975         struct nfs4_state *state;
1976         struct inode *inode;
1977
1978         inode = nfs4_opendata_get_inode(data);
1979         if (IS_ERR(inode))
1980                 return ERR_CAST(inode);
1981         if (data->state != NULL && data->state->inode == inode) {
1982                 state = data->state;
1983                 refcount_inc(&state->count);
1984         } else
1985                 state = nfs4_get_open_state(inode, data->owner);
1986         iput(inode);
1987         if (state == NULL)
1988                 state = ERR_PTR(-ENOMEM);
1989         return state;
1990 }
1991
1992 static struct nfs4_state *
1993 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1994 {
1995         struct nfs4_state *state;
1996
1997         if (!data->rpc_done) {
1998                 state = nfs4_try_open_cached(data);
1999                 trace_nfs4_cached_open(data->state);
2000                 goto out;
2001         }
2002
2003         state = nfs4_opendata_find_nfs4_state(data);
2004         if (IS_ERR(state))
2005                 goto out;
2006
2007         if (data->o_res.delegation_type != 0)
2008                 nfs4_opendata_check_deleg(data, state);
2009         if (!update_open_stateid(state, &data->o_res.stateid,
2010                                 NULL, data->o_arg.fmode)) {
2011                 nfs4_put_open_state(state);
2012                 state = ERR_PTR(-EAGAIN);
2013         }
2014 out:
2015         nfs_release_seqid(data->o_arg.seqid);
2016         return state;
2017 }
2018
2019 static struct nfs4_state *
2020 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2021 {
2022         struct nfs4_state *ret;
2023
2024         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2025                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2026         else
2027                 ret = _nfs4_opendata_to_nfs4_state(data);
2028         nfs4_sequence_free_slot(&data->o_res.seq_res);
2029         return ret;
2030 }
2031
2032 static struct nfs_open_context *
2033 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2034 {
2035         struct nfs_inode *nfsi = NFS_I(state->inode);
2036         struct nfs_open_context *ctx;
2037
2038         rcu_read_lock();
2039         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2040                 if (ctx->state != state)
2041                         continue;
2042                 if ((ctx->mode & mode) != mode)
2043                         continue;
2044                 if (!get_nfs_open_context(ctx))
2045                         continue;
2046                 rcu_read_unlock();
2047                 return ctx;
2048         }
2049         rcu_read_unlock();
2050         return ERR_PTR(-ENOENT);
2051 }
2052
2053 static struct nfs_open_context *
2054 nfs4_state_find_open_context(struct nfs4_state *state)
2055 {
2056         struct nfs_open_context *ctx;
2057
2058         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2059         if (!IS_ERR(ctx))
2060                 return ctx;
2061         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2062         if (!IS_ERR(ctx))
2063                 return ctx;
2064         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2065 }
2066
2067 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2068                 struct nfs4_state *state, enum open_claim_type4 claim)
2069 {
2070         struct nfs4_opendata *opendata;
2071
2072         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2073                         NULL, claim, GFP_NOFS);
2074         if (opendata == NULL)
2075                 return ERR_PTR(-ENOMEM);
2076         opendata->state = state;
2077         refcount_inc(&state->count);
2078         return opendata;
2079 }
2080
2081 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2082                 fmode_t fmode)
2083 {
2084         struct nfs4_state *newstate;
2085         int ret;
2086
2087         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2088                 return 0;
2089         opendata->o_arg.open_flags = 0;
2090         opendata->o_arg.fmode = fmode;
2091         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2092                         NFS_SB(opendata->dentry->d_sb),
2093                         fmode, 0);
2094         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2095         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2096         nfs4_init_opendata_res(opendata);
2097         ret = _nfs4_recover_proc_open(opendata);
2098         if (ret != 0)
2099                 return ret;
2100         newstate = nfs4_opendata_to_nfs4_state(opendata);
2101         if (IS_ERR(newstate))
2102                 return PTR_ERR(newstate);
2103         if (newstate != opendata->state)
2104                 ret = -ESTALE;
2105         nfs4_close_state(newstate, fmode);
2106         return ret;
2107 }
2108
2109 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2110 {
2111         int ret;
2112
2113         /* memory barrier prior to reading state->n_* */
2114         smp_rmb();
2115         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2116         if (ret != 0)
2117                 return ret;
2118         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2119         if (ret != 0)
2120                 return ret;
2121         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2122         if (ret != 0)
2123                 return ret;
2124         /*
2125          * We may have performed cached opens for all three recoveries.
2126          * Check if we need to update the current stateid.
2127          */
2128         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2129             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2130                 write_seqlock(&state->seqlock);
2131                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2132                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2133                 write_sequnlock(&state->seqlock);
2134         }
2135         return 0;
2136 }
2137
2138 /*
2139  * OPEN_RECLAIM:
2140  *      reclaim state on the server after a reboot.
2141  */
2142 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2143 {
2144         struct nfs_delegation *delegation;
2145         struct nfs4_opendata *opendata;
2146         fmode_t delegation_type = 0;
2147         int status;
2148
2149         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2150                         NFS4_OPEN_CLAIM_PREVIOUS);
2151         if (IS_ERR(opendata))
2152                 return PTR_ERR(opendata);
2153         rcu_read_lock();
2154         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2155         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2156                 delegation_type = delegation->type;
2157         rcu_read_unlock();
2158         opendata->o_arg.u.delegation_type = delegation_type;
2159         status = nfs4_open_recover(opendata, state);
2160         nfs4_opendata_put(opendata);
2161         return status;
2162 }
2163
2164 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2165 {
2166         struct nfs_server *server = NFS_SERVER(state->inode);
2167         struct nfs4_exception exception = { };
2168         int err;
2169         do {
2170                 err = _nfs4_do_open_reclaim(ctx, state);
2171                 trace_nfs4_open_reclaim(ctx, 0, err);
2172                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2173                         continue;
2174                 if (err != -NFS4ERR_DELAY)
2175                         break;
2176                 nfs4_handle_exception(server, err, &exception);
2177         } while (exception.retry);
2178         return err;
2179 }
2180
2181 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2182 {
2183         struct nfs_open_context *ctx;
2184         int ret;
2185
2186         ctx = nfs4_state_find_open_context(state);
2187         if (IS_ERR(ctx))
2188                 return -EAGAIN;
2189         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2190         nfs_state_clear_open_state_flags(state);
2191         ret = nfs4_do_open_reclaim(ctx, state);
2192         put_nfs_open_context(ctx);
2193         return ret;
2194 }
2195
2196 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)
2197 {
2198         switch (err) {
2199                 default:
2200                         printk(KERN_ERR "NFS: %s: unhandled error "
2201                                         "%d.\n", __func__, err);
2202                 case 0:
2203                 case -ENOENT:
2204                 case -EAGAIN:
2205                 case -ESTALE:
2206                 case -ETIMEDOUT:
2207                         break;
2208                 case -NFS4ERR_BADSESSION:
2209                 case -NFS4ERR_BADSLOT:
2210                 case -NFS4ERR_BAD_HIGH_SLOT:
2211                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2212                 case -NFS4ERR_DEADSESSION:
2213                         return -EAGAIN;
2214                 case -NFS4ERR_STALE_CLIENTID:
2215                 case -NFS4ERR_STALE_STATEID:
2216                         /* Don't recall a delegation if it was lost */
2217                         nfs4_schedule_lease_recovery(server->nfs_client);
2218                         return -EAGAIN;
2219                 case -NFS4ERR_MOVED:
2220                         nfs4_schedule_migration_recovery(server);
2221                         return -EAGAIN;
2222                 case -NFS4ERR_LEASE_MOVED:
2223                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2224                         return -EAGAIN;
2225                 case -NFS4ERR_DELEG_REVOKED:
2226                 case -NFS4ERR_ADMIN_REVOKED:
2227                 case -NFS4ERR_EXPIRED:
2228                 case -NFS4ERR_BAD_STATEID:
2229                 case -NFS4ERR_OPENMODE:
2230                         nfs_inode_find_state_and_recover(state->inode,
2231                                         stateid);
2232                         nfs4_schedule_stateid_recovery(server, state);
2233                         return -EAGAIN;
2234                 case -NFS4ERR_DELAY:
2235                 case -NFS4ERR_GRACE:
2236                         ssleep(1);
2237                         return -EAGAIN;
2238                 case -ENOMEM:
2239                 case -NFS4ERR_DENIED:
2240                         if (fl) {
2241                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2242                                 if (lsp)
2243                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2244                         }
2245                         return 0;
2246         }
2247         return err;
2248 }
2249
2250 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2251                 struct nfs4_state *state, const nfs4_stateid *stateid)
2252 {
2253         struct nfs_server *server = NFS_SERVER(state->inode);
2254         struct nfs4_opendata *opendata;
2255         int err = 0;
2256
2257         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2258                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2259         if (IS_ERR(opendata))
2260                 return PTR_ERR(opendata);
2261         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2262         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2263                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2264                 if (err)
2265                         goto out;
2266         }
2267         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2268                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2269                 if (err)
2270                         goto out;
2271         }
2272         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2273                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2274                 if (err)
2275                         goto out;
2276         }
2277         nfs_state_clear_delegation(state);
2278 out:
2279         nfs4_opendata_put(opendata);
2280         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2281 }
2282
2283 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2284 {
2285         struct nfs4_opendata *data = calldata;
2286
2287         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2288                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2289 }
2290
2291 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2292 {
2293         struct nfs4_opendata *data = calldata;
2294
2295         nfs40_sequence_done(task, &data->c_res.seq_res);
2296
2297         data->rpc_status = task->tk_status;
2298         if (data->rpc_status == 0) {
2299                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2300                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2301                 renew_lease(data->o_res.server, data->timestamp);
2302                 data->rpc_done = true;
2303         }
2304 }
2305
2306 static void nfs4_open_confirm_release(void *calldata)
2307 {
2308         struct nfs4_opendata *data = calldata;
2309         struct nfs4_state *state = NULL;
2310
2311         /* If this request hasn't been cancelled, do nothing */
2312         if (!data->cancelled)
2313                 goto out_free;
2314         /* In case of error, no cleanup! */
2315         if (!data->rpc_done)
2316                 goto out_free;
2317         state = nfs4_opendata_to_nfs4_state(data);
2318         if (!IS_ERR(state))
2319                 nfs4_close_state(state, data->o_arg.fmode);
2320 out_free:
2321         nfs4_opendata_put(data);
2322 }
2323
2324 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2325         .rpc_call_prepare = nfs4_open_confirm_prepare,
2326         .rpc_call_done = nfs4_open_confirm_done,
2327         .rpc_release = nfs4_open_confirm_release,
2328 };
2329
2330 /*
2331  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2332  */
2333 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2334 {
2335         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2336         struct rpc_task *task;
2337         struct  rpc_message msg = {
2338                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2339                 .rpc_argp = &data->c_arg,
2340                 .rpc_resp = &data->c_res,
2341                 .rpc_cred = data->owner->so_cred,
2342         };
2343         struct rpc_task_setup task_setup_data = {
2344                 .rpc_client = server->client,
2345                 .rpc_message = &msg,
2346                 .callback_ops = &nfs4_open_confirm_ops,
2347                 .callback_data = data,
2348                 .workqueue = nfsiod_workqueue,
2349                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2350         };
2351         int status;
2352
2353         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2354                                 data->is_recover);
2355         kref_get(&data->kref);
2356         data->rpc_done = false;
2357         data->rpc_status = 0;
2358         data->timestamp = jiffies;
2359         task = rpc_run_task(&task_setup_data);
2360         if (IS_ERR(task))
2361                 return PTR_ERR(task);
2362         status = rpc_wait_for_completion_task(task);
2363         if (status != 0) {
2364                 data->cancelled = true;
2365                 smp_wmb();
2366         } else
2367                 status = data->rpc_status;
2368         rpc_put_task(task);
2369         return status;
2370 }
2371
2372 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2373 {
2374         struct nfs4_opendata *data = calldata;
2375         struct nfs4_state_owner *sp = data->owner;
2376         struct nfs_client *clp = sp->so_server->nfs_client;
2377         enum open_claim_type4 claim = data->o_arg.claim;
2378
2379         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2380                 goto out_wait;
2381         /*
2382          * Check if we still need to send an OPEN call, or if we can use
2383          * a delegation instead.
2384          */
2385         if (data->state != NULL) {
2386                 struct nfs_delegation *delegation;
2387
2388                 if (can_open_cached(data->state, data->o_arg.fmode,
2389                                         data->o_arg.open_flags, claim))
2390                         goto out_no_action;
2391                 rcu_read_lock();
2392                 delegation = nfs4_get_valid_delegation(data->state->inode);
2393                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2394                         goto unlock_no_action;
2395                 rcu_read_unlock();
2396         }
2397         /* Update client id. */
2398         data->o_arg.clientid = clp->cl_clientid;
2399         switch (claim) {
2400         default:
2401                 break;
2402         case NFS4_OPEN_CLAIM_PREVIOUS:
2403         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2404         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2405                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2406                 /* Fall through */
2407         case NFS4_OPEN_CLAIM_FH:
2408                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2409         }
2410         data->timestamp = jiffies;
2411         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2412                                 &data->o_arg.seq_args,
2413                                 &data->o_res.seq_res,
2414                                 task) != 0)
2415                 nfs_release_seqid(data->o_arg.seqid);
2416
2417         /* Set the create mode (note dependency on the session type) */
2418         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2419         if (data->o_arg.open_flags & O_EXCL) {
2420                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2421                 if (nfs4_has_persistent_session(clp))
2422                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2423                 else if (clp->cl_mvops->minor_version > 0)
2424                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2425         }
2426         return;
2427 unlock_no_action:
2428         trace_nfs4_cached_open(data->state);
2429         rcu_read_unlock();
2430 out_no_action:
2431         task->tk_action = NULL;
2432 out_wait:
2433         nfs4_sequence_done(task, &data->o_res.seq_res);
2434 }
2435
2436 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2437 {
2438         struct nfs4_opendata *data = calldata;
2439
2440         data->rpc_status = task->tk_status;
2441
2442         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2443                 return;
2444
2445         if (task->tk_status == 0) {
2446                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2447                         switch (data->o_res.f_attr->mode & S_IFMT) {
2448                         case S_IFREG:
2449                                 break;
2450                         case S_IFLNK:
2451                                 data->rpc_status = -ELOOP;
2452                                 break;
2453                         case S_IFDIR:
2454                                 data->rpc_status = -EISDIR;
2455                                 break;
2456                         default:
2457                                 data->rpc_status = -ENOTDIR;
2458                         }
2459                 }
2460                 renew_lease(data->o_res.server, data->timestamp);
2461                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2462                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2463         }
2464         data->rpc_done = true;
2465 }
2466
2467 static void nfs4_open_release(void *calldata)
2468 {
2469         struct nfs4_opendata *data = calldata;
2470         struct nfs4_state *state = NULL;
2471
2472         /* If this request hasn't been cancelled, do nothing */
2473         if (!data->cancelled)
2474                 goto out_free;
2475         /* In case of error, no cleanup! */
2476         if (data->rpc_status != 0 || !data->rpc_done)
2477                 goto out_free;
2478         /* In case we need an open_confirm, no cleanup! */
2479         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2480                 goto out_free;
2481         state = nfs4_opendata_to_nfs4_state(data);
2482         if (!IS_ERR(state))
2483                 nfs4_close_state(state, data->o_arg.fmode);
2484 out_free:
2485         nfs4_opendata_put(data);
2486 }
2487
2488 static const struct rpc_call_ops nfs4_open_ops = {
2489         .rpc_call_prepare = nfs4_open_prepare,
2490         .rpc_call_done = nfs4_open_done,
2491         .rpc_release = nfs4_open_release,
2492 };
2493
2494 static int nfs4_run_open_task(struct nfs4_opendata *data,
2495                               struct nfs_open_context *ctx)
2496 {
2497         struct inode *dir = d_inode(data->dir);
2498         struct nfs_server *server = NFS_SERVER(dir);
2499         struct nfs_openargs *o_arg = &data->o_arg;
2500         struct nfs_openres *o_res = &data->o_res;
2501         struct rpc_task *task;
2502         struct rpc_message msg = {
2503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2504                 .rpc_argp = o_arg,
2505                 .rpc_resp = o_res,
2506                 .rpc_cred = data->owner->so_cred,
2507         };
2508         struct rpc_task_setup task_setup_data = {
2509                 .rpc_client = server->client,
2510                 .rpc_message = &msg,
2511                 .callback_ops = &nfs4_open_ops,
2512                 .callback_data = data,
2513                 .workqueue = nfsiod_workqueue,
2514                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2515         };
2516         int status;
2517
2518         kref_get(&data->kref);
2519         data->rpc_done = false;
2520         data->rpc_status = 0;
2521         data->cancelled = false;
2522         data->is_recover = false;
2523         if (!ctx) {
2524                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2525                 data->is_recover = true;
2526                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2527         } else {
2528                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2529                 pnfs_lgopen_prepare(data, ctx);
2530         }
2531         task = rpc_run_task(&task_setup_data);
2532         if (IS_ERR(task))
2533                 return PTR_ERR(task);
2534         status = rpc_wait_for_completion_task(task);
2535         if (status != 0) {
2536                 data->cancelled = true;
2537                 smp_wmb();
2538         } else
2539                 status = data->rpc_status;
2540         rpc_put_task(task);
2541
2542         return status;
2543 }
2544
2545 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2546 {
2547         struct inode *dir = d_inode(data->dir);
2548         struct nfs_openres *o_res = &data->o_res;
2549         int status;
2550
2551         status = nfs4_run_open_task(data, NULL);
2552         if (status != 0 || !data->rpc_done)
2553                 return status;
2554
2555         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2556
2557         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2558                 status = _nfs4_proc_open_confirm(data);
2559
2560         return status;
2561 }
2562
2563 /*
2564  * Additional permission checks in order to distinguish between an
2565  * open for read, and an open for execute. This works around the
2566  * fact that NFSv4 OPEN treats read and execute permissions as being
2567  * the same.
2568  * Note that in the non-execute case, we want to turn off permission
2569  * checking if we just created a new file (POSIX open() semantics).
2570  */
2571 static int nfs4_opendata_access(const struct cred *cred,
2572                                 struct nfs4_opendata *opendata,
2573                                 struct nfs4_state *state, fmode_t fmode,
2574                                 int openflags)
2575 {
2576         struct nfs_access_entry cache;
2577         u32 mask, flags;
2578
2579         /* access call failed or for some reason the server doesn't
2580          * support any access modes -- defer access call until later */
2581         if (opendata->o_res.access_supported == 0)
2582                 return 0;
2583
2584         mask = 0;
2585         /*
2586          * Use openflags to check for exec, because fmode won't
2587          * always have FMODE_EXEC set when file open for exec.
2588          */
2589         if (openflags & __FMODE_EXEC) {
2590                 /* ONLY check for exec rights */
2591                 if (S_ISDIR(state->inode->i_mode))
2592                         mask = NFS4_ACCESS_LOOKUP;
2593                 else
2594                         mask = NFS4_ACCESS_EXECUTE;
2595         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2596                 mask = NFS4_ACCESS_READ;
2597
2598         cache.cred = cred;
2599         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2600         nfs_access_add_cache(state->inode, &cache);
2601
2602         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2603         if ((mask & ~cache.mask & flags) == 0)
2604                 return 0;
2605
2606         return -EACCES;
2607 }
2608
2609 /*
2610  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2611  */
2612 static int _nfs4_proc_open(struct nfs4_opendata *data,
2613                            struct nfs_open_context *ctx)
2614 {
2615         struct inode *dir = d_inode(data->dir);
2616         struct nfs_server *server = NFS_SERVER(dir);
2617         struct nfs_openargs *o_arg = &data->o_arg;
2618         struct nfs_openres *o_res = &data->o_res;
2619         int status;
2620
2621         status = nfs4_run_open_task(data, ctx);
2622         if (!data->rpc_done)
2623                 return status;
2624         if (status != 0) {
2625                 if (status == -NFS4ERR_BADNAME &&
2626                                 !(o_arg->open_flags & O_CREAT))
2627                         return -ENOENT;
2628                 return status;
2629         }
2630
2631         nfs_fattr_map_and_free_names(server, &data->f_attr);
2632
2633         if (o_arg->open_flags & O_CREAT) {
2634                 if (o_arg->open_flags & O_EXCL)
2635                         data->file_created = true;
2636                 else if (o_res->cinfo.before != o_res->cinfo.after)
2637                         data->file_created = true;
2638                 if (data->file_created ||
2639                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2640                         update_changeattr(dir, &o_res->cinfo,
2641                                         o_res->f_attr->time_start, 0);
2642         }
2643         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2644                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2645         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2646                 status = _nfs4_proc_open_confirm(data);
2647                 if (status != 0)
2648                         return status;
2649         }
2650         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2651                 nfs4_sequence_free_slot(&o_res->seq_res);
2652                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2653                                 o_res->f_label, NULL);
2654         }
2655         return 0;
2656 }
2657
2658 /*
2659  * OPEN_EXPIRED:
2660  *      reclaim state on the server after a network partition.
2661  *      Assumes caller holds the appropriate lock
2662  */
2663 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2664 {
2665         struct nfs4_opendata *opendata;
2666         int ret;
2667
2668         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2669                         NFS4_OPEN_CLAIM_FH);
2670         if (IS_ERR(opendata))
2671                 return PTR_ERR(opendata);
2672         ret = nfs4_open_recover(opendata, state);
2673         if (ret == -ESTALE)
2674                 d_drop(ctx->dentry);
2675         nfs4_opendata_put(opendata);
2676         return ret;
2677 }
2678
2679 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2680 {
2681         struct nfs_server *server = NFS_SERVER(state->inode);
2682         struct nfs4_exception exception = { };
2683         int err;
2684
2685         do {
2686                 err = _nfs4_open_expired(ctx, state);
2687                 trace_nfs4_open_expired(ctx, 0, err);
2688                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2689                         continue;
2690                 switch (err) {
2691                 default:
2692                         goto out;
2693                 case -NFS4ERR_GRACE:
2694                 case -NFS4ERR_DELAY:
2695                         nfs4_handle_exception(server, err, &exception);
2696                         err = 0;
2697                 }
2698         } while (exception.retry);
2699 out:
2700         return err;
2701 }
2702
2703 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2704 {
2705         struct nfs_open_context *ctx;
2706         int ret;
2707
2708         ctx = nfs4_state_find_open_context(state);
2709         if (IS_ERR(ctx))
2710                 return -EAGAIN;
2711         ret = nfs4_do_open_expired(ctx, state);
2712         put_nfs_open_context(ctx);
2713         return ret;
2714 }
2715
2716 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2717                 const nfs4_stateid *stateid)
2718 {
2719         nfs_remove_bad_delegation(state->inode, stateid);
2720         nfs_state_clear_delegation(state);
2721 }
2722
2723 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2724 {
2725         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2726                 nfs_finish_clear_delegation_stateid(state, NULL);
2727 }
2728
2729 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2730 {
2731         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2732         nfs40_clear_delegation_stateid(state);
2733         nfs_state_clear_open_state_flags(state);
2734         return nfs4_open_expired(sp, state);
2735 }
2736
2737 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2738                 nfs4_stateid *stateid,
2739                 const struct cred *cred)
2740 {
2741         return -NFS4ERR_BAD_STATEID;
2742 }
2743
2744 #if defined(CONFIG_NFS_V4_1)
2745 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2746                 nfs4_stateid *stateid,
2747                 const struct cred *cred)
2748 {
2749         int status;
2750
2751         switch (stateid->type) {
2752         default:
2753                 break;
2754         case NFS4_INVALID_STATEID_TYPE:
2755         case NFS4_SPECIAL_STATEID_TYPE:
2756                 return -NFS4ERR_BAD_STATEID;
2757         case NFS4_REVOKED_STATEID_TYPE:
2758                 goto out_free;
2759         }
2760
2761         status = nfs41_test_stateid(server, stateid, cred);
2762         switch (status) {
2763         case -NFS4ERR_EXPIRED:
2764         case -NFS4ERR_ADMIN_REVOKED:
2765         case -NFS4ERR_DELEG_REVOKED:
2766                 break;
2767         default:
2768                 return status;
2769         }
2770 out_free:
2771         /* Ack the revoked state to the server */
2772         nfs41_free_stateid(server, stateid, cred, true);
2773         return -NFS4ERR_EXPIRED;
2774 }
2775
2776 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2777 {
2778         struct nfs_server *server = NFS_SERVER(state->inode);
2779         nfs4_stateid stateid;
2780         struct nfs_delegation *delegation;
2781         const struct cred *cred = NULL;
2782         int status, ret = NFS_OK;
2783
2784         /* Get the delegation credential for use by test/free_stateid */
2785         rcu_read_lock();
2786         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2787         if (delegation == NULL) {
2788                 rcu_read_unlock();
2789                 nfs_state_clear_delegation(state);
2790                 return NFS_OK;
2791         }
2792
2793         spin_lock(&delegation->lock);
2794         nfs4_stateid_copy(&stateid, &delegation->stateid);
2795
2796         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2797                                 &delegation->flags)) {
2798                 spin_unlock(&delegation->lock);
2799                 rcu_read_unlock();
2800                 return NFS_OK;
2801         }
2802
2803         if (delegation->cred)
2804                 cred = get_cred(delegation->cred);
2805         spin_unlock(&delegation->lock);
2806         rcu_read_unlock();
2807         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2808         trace_nfs4_test_delegation_stateid(state, NULL, status);
2809         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2810                 nfs_finish_clear_delegation_stateid(state, &stateid);
2811         else
2812                 ret = status;
2813
2814         put_cred(cred);
2815         return ret;
2816 }
2817
2818 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2819 {
2820         nfs4_stateid tmp;
2821
2822         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2823             nfs4_copy_delegation_stateid(state->inode, state->state,
2824                                 &tmp, NULL) &&
2825             nfs4_stateid_match_other(&state->stateid, &tmp))
2826                 nfs_state_set_delegation(state, &tmp, state->state);
2827         else
2828                 nfs_state_clear_delegation(state);
2829 }
2830
2831 /**
2832  * nfs41_check_expired_locks - possibly free a lock stateid
2833  *
2834  * @state: NFSv4 state for an inode
2835  *
2836  * Returns NFS_OK if recovery for this stateid is now finished.
2837  * Otherwise a negative NFS4ERR value is returned.
2838  */
2839 static int nfs41_check_expired_locks(struct nfs4_state *state)
2840 {
2841         int status, ret = NFS_OK;
2842         struct nfs4_lock_state *lsp, *prev = NULL;
2843         struct nfs_server *server = NFS_SERVER(state->inode);
2844
2845         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2846                 goto out;
2847
2848         spin_lock(&state->state_lock);
2849         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2850                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2851                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2852
2853                         refcount_inc(&lsp->ls_count);
2854                         spin_unlock(&state->state_lock);
2855
2856                         nfs4_put_lock_state(prev);
2857                         prev = lsp;
2858
2859                         status = nfs41_test_and_free_expired_stateid(server,
2860                                         &lsp->ls_stateid,
2861                                         cred);
2862                         trace_nfs4_test_lock_stateid(state, lsp, status);
2863                         if (status == -NFS4ERR_EXPIRED ||
2864                             status == -NFS4ERR_BAD_STATEID) {
2865                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2866                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2867                                 if (!recover_lost_locks)
2868                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2869                         } else if (status != NFS_OK) {
2870                                 ret = status;
2871                                 nfs4_put_lock_state(prev);
2872                                 goto out;
2873                         }
2874                         spin_lock(&state->state_lock);
2875                 }
2876         }
2877         spin_unlock(&state->state_lock);
2878         nfs4_put_lock_state(prev);
2879 out:
2880         return ret;
2881 }
2882
2883 /**
2884  * nfs41_check_open_stateid - possibly free an open stateid
2885  *
2886  * @state: NFSv4 state for an inode
2887  *
2888  * Returns NFS_OK if recovery for this stateid is now finished.
2889  * Otherwise a negative NFS4ERR value is returned.
2890  */
2891 static int nfs41_check_open_stateid(struct nfs4_state *state)
2892 {
2893         struct nfs_server *server = NFS_SERVER(state->inode);
2894         nfs4_stateid *stateid = &state->open_stateid;
2895         const struct cred *cred = state->owner->so_cred;
2896         int status;
2897
2898         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2899                 return -NFS4ERR_BAD_STATEID;
2900         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2901         trace_nfs4_test_open_stateid(state, NULL, status);
2902         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2903                 nfs_state_clear_open_state_flags(state);
2904                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2905                 return status;
2906         }
2907         if (nfs_open_stateid_recover_openmode(state))
2908                 return -NFS4ERR_OPENMODE;
2909         return NFS_OK;
2910 }
2911
2912 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2913 {
2914         int status;
2915
2916         status = nfs41_check_delegation_stateid(state);
2917         if (status != NFS_OK)
2918                 return status;
2919         nfs41_delegation_recover_stateid(state);
2920
2921         status = nfs41_check_expired_locks(state);
2922         if (status != NFS_OK)
2923                 return status;
2924         status = nfs41_check_open_stateid(state);
2925         if (status != NFS_OK)
2926                 status = nfs4_open_expired(sp, state);
2927         return status;
2928 }
2929 #endif
2930
2931 /*
2932  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2933  * fields corresponding to attributes that were used to store the verifier.
2934  * Make sure we clobber those fields in the later setattr call
2935  */
2936 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2937                                 struct iattr *sattr, struct nfs4_label **label)
2938 {
2939         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2940         __u32 attrset[3];
2941         unsigned ret;
2942         unsigned i;
2943
2944         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2945                 attrset[i] = opendata->o_res.attrset[i];
2946                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2947                         attrset[i] &= ~bitmask[i];
2948         }
2949
2950         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2951                 sattr->ia_valid : 0;
2952
2953         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2954                 if (sattr->ia_valid & ATTR_ATIME_SET)
2955                         ret |= ATTR_ATIME_SET;
2956                 else
2957                         ret |= ATTR_ATIME;
2958         }
2959
2960         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2961                 if (sattr->ia_valid & ATTR_MTIME_SET)
2962                         ret |= ATTR_MTIME_SET;
2963                 else
2964                         ret |= ATTR_MTIME;
2965         }
2966
2967         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2968                 *label = NULL;
2969         return ret;
2970 }
2971
2972 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2973                 int flags, struct nfs_open_context *ctx)
2974 {
2975         struct nfs4_state_owner *sp = opendata->owner;
2976         struct nfs_server *server = sp->so_server;
2977         struct dentry *dentry;
2978         struct nfs4_state *state;
2979         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
2980         struct inode *dir = d_inode(opendata->dir);
2981         unsigned long dir_verifier;
2982         unsigned int seq;
2983         int ret;
2984
2985         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2986         dir_verifier = nfs_save_change_attribute(dir);
2987
2988         ret = _nfs4_proc_open(opendata, ctx);
2989         if (ret != 0)
2990                 goto out;
2991
2992         state = _nfs4_opendata_to_nfs4_state(opendata);
2993         ret = PTR_ERR(state);
2994         if (IS_ERR(state))
2995                 goto out;
2996         ctx->state = state;
2997         if (server->caps & NFS_CAP_POSIX_LOCK)
2998                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2999         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3000                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3001
3002         dentry = opendata->dentry;
3003         if (d_really_is_negative(dentry)) {
3004                 struct dentry *alias;
3005                 d_drop(dentry);
3006                 alias = d_exact_alias(dentry, state->inode);
3007                 if (!alias)
3008                         alias = d_splice_alias(igrab(state->inode), dentry);
3009                 /* d_splice_alias() can't fail here - it's a non-directory */
3010                 if (alias) {
3011                         dput(ctx->dentry);
3012                         ctx->dentry = dentry = alias;
3013                 }
3014         }
3015
3016         switch(opendata->o_arg.claim) {
3017         default:
3018                 break;
3019         case NFS4_OPEN_CLAIM_NULL:
3020         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3021         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3022                 if (!opendata->rpc_done)
3023                         break;
3024                 if (opendata->o_res.delegation_type != 0)
3025                         dir_verifier = nfs_save_change_attribute(dir);
3026                 nfs_set_verifier(dentry, dir_verifier);
3027         }
3028
3029         /* Parse layoutget results before we check for access */
3030         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3031
3032         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3033                         acc_mode, flags);
3034         if (ret != 0)
3035                 goto out;
3036
3037         if (d_inode(dentry) == state->inode) {
3038                 nfs_inode_attach_open_context(ctx);
3039                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3040                         nfs4_schedule_stateid_recovery(server, state);
3041         }
3042
3043 out:
3044         if (!opendata->cancelled)
3045                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3046         return ret;
3047 }
3048
3049 /*
3050  * Returns a referenced nfs4_state
3051  */
3052 static int _nfs4_do_open(struct inode *dir,
3053                         struct nfs_open_context *ctx,
3054                         int flags,
3055                         const struct nfs4_open_createattrs *c,
3056                         int *opened)
3057 {
3058         struct nfs4_state_owner  *sp;
3059         struct nfs4_state     *state = NULL;
3060         struct nfs_server       *server = NFS_SERVER(dir);
3061         struct nfs4_opendata *opendata;
3062         struct dentry *dentry = ctx->dentry;
3063         const struct cred *cred = ctx->cred;
3064         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3065         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3066         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3067         struct iattr *sattr = c->sattr;
3068         struct nfs4_label *label = c->label;
3069         struct nfs4_label *olabel = NULL;
3070         int status;
3071
3072         /* Protect against reboot recovery conflicts */
3073         status = -ENOMEM;
3074         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3075         if (sp == NULL) {
3076                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3077                 goto out_err;
3078         }
3079         status = nfs4_client_recover_expired_lease(server->nfs_client);
3080         if (status != 0)
3081                 goto err_put_state_owner;
3082         if (d_really_is_positive(dentry))
3083                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3084         status = -ENOMEM;
3085         if (d_really_is_positive(dentry))
3086                 claim = NFS4_OPEN_CLAIM_FH;
3087         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3088                         c, claim, GFP_KERNEL);
3089         if (opendata == NULL)
3090                 goto err_put_state_owner;
3091
3092         if (label) {
3093                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3094                 if (IS_ERR(olabel)) {
3095                         status = PTR_ERR(olabel);
3096                         goto err_opendata_put;
3097                 }
3098         }
3099
3100         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3101                 if (!opendata->f_attr.mdsthreshold) {
3102                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3103                         if (!opendata->f_attr.mdsthreshold)
3104                                 goto err_free_label;
3105                 }
3106                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3107         }
3108         if (d_really_is_positive(dentry))
3109                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3110
3111         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3112         if (status != 0)
3113                 goto err_free_label;
3114         state = ctx->state;
3115
3116         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3117             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3118                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3119                 /*
3120                  * send create attributes which was not set by open
3121                  * with an extra setattr.
3122                  */
3123                 if (attrs || label) {
3124                         unsigned ia_old = sattr->ia_valid;
3125
3126                         sattr->ia_valid = attrs;
3127                         nfs_fattr_init(opendata->o_res.f_attr);
3128                         status = nfs4_do_setattr(state->inode, cred,
3129                                         opendata->o_res.f_attr, sattr,
3130                                         ctx, label, olabel);
3131                         if (status == 0) {
3132                                 nfs_setattr_update_inode(state->inode, sattr,
3133                                                 opendata->o_res.f_attr);
3134                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3135                         }
3136                         sattr->ia_valid = ia_old;
3137                 }
3138         }
3139         if (opened && opendata->file_created)
3140                 *opened = 1;
3141
3142         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3143                 *ctx_th = opendata->f_attr.mdsthreshold;
3144                 opendata->f_attr.mdsthreshold = NULL;
3145         }
3146
3147         nfs4_label_free(olabel);
3148
3149         nfs4_opendata_put(opendata);
3150         nfs4_put_state_owner(sp);
3151         return 0;
3152 err_free_label:
3153         nfs4_label_free(olabel);
3154 err_opendata_put:
3155         nfs4_opendata_put(opendata);
3156 err_put_state_owner:
3157         nfs4_put_state_owner(sp);
3158 out_err:
3159         return status;
3160 }
3161
3162
3163 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3164                                         struct nfs_open_context *ctx,
3165                                         int flags,
3166                                         struct iattr *sattr,
3167                                         struct nfs4_label *label,
3168                                         int *opened)
3169 {
3170         struct nfs_server *server = NFS_SERVER(dir);
3171         struct nfs4_exception exception = {
3172                 .interruptible = true,
3173         };
3174         struct nfs4_state *res;
3175         struct nfs4_open_createattrs c = {
3176                 .label = label,
3177                 .sattr = sattr,
3178                 .verf = {
3179                         [0] = (__u32)jiffies,
3180                         [1] = (__u32)current->pid,
3181                 },
3182         };
3183         int status;
3184
3185         do {
3186                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3187                 res = ctx->state;
3188                 trace_nfs4_open_file(ctx, flags, status);
3189                 if (status == 0)
3190                         break;
3191                 /* NOTE: BAD_SEQID means the server and client disagree about the
3192                  * book-keeping w.r.t. state-changing operations
3193                  * (OPEN/CLOSE/LOCK/LOCKU...)
3194                  * It is actually a sign of a bug on the client or on the server.
3195                  *
3196                  * If we receive a BAD_SEQID error in the particular case of
3197                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3198                  * have unhashed the old state_owner for us, and that we can
3199                  * therefore safely retry using a new one. We should still warn
3200                  * the user though...
3201                  */
3202                 if (status == -NFS4ERR_BAD_SEQID) {
3203                         pr_warn_ratelimited("NFS: v4 server %s "
3204                                         " returned a bad sequence-id error!\n",
3205                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3206                         exception.retry = 1;
3207                         continue;
3208                 }
3209                 /*
3210                  * BAD_STATEID on OPEN means that the server cancelled our
3211                  * state before it received the OPEN_CONFIRM.
3212                  * Recover by retrying the request as per the discussion
3213                  * on Page 181 of RFC3530.
3214                  */
3215                 if (status == -NFS4ERR_BAD_STATEID) {
3216                         exception.retry = 1;
3217                         continue;
3218                 }
3219                 if (status == -NFS4ERR_EXPIRED) {
3220                         nfs4_schedule_lease_recovery(server->nfs_client);
3221                         exception.retry = 1;
3222                         continue;
3223                 }
3224                 if (status == -EAGAIN) {
3225                         /* We must have found a delegation */
3226                         exception.retry = 1;
3227                         continue;
3228                 }
3229                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3230                         continue;
3231                 res = ERR_PTR(nfs4_handle_exception(server,
3232                                         status, &exception));
3233         } while (exception.retry);
3234         return res;
3235 }
3236
3237 static int _nfs4_do_setattr(struct inode *inode,
3238                             struct nfs_setattrargs *arg,
3239                             struct nfs_setattrres *res,
3240                             const struct cred *cred,
3241                             struct nfs_open_context *ctx)
3242 {
3243         struct nfs_server *server = NFS_SERVER(inode);
3244         struct rpc_message msg = {
3245                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3246                 .rpc_argp       = arg,
3247                 .rpc_resp       = res,
3248                 .rpc_cred       = cred,
3249         };
3250         const struct cred *delegation_cred = NULL;
3251         unsigned long timestamp = jiffies;
3252         bool truncate;
3253         int status;
3254
3255         nfs_fattr_init(res->fattr);
3256
3257         /* Servers should only apply open mode checks for file size changes */
3258         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3259         if (!truncate)
3260                 goto zero_stateid;
3261
3262         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3263                 /* Use that stateid */
3264         } else if (ctx != NULL && ctx->state) {
3265                 struct nfs_lock_context *l_ctx;
3266                 if (!nfs4_valid_open_stateid(ctx->state))
3267                         return -EBADF;
3268                 l_ctx = nfs_get_lock_context(ctx);
3269                 if (IS_ERR(l_ctx))
3270                         return PTR_ERR(l_ctx);
3271                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3272                                                 &arg->stateid, &delegation_cred);
3273                 nfs_put_lock_context(l_ctx);
3274                 if (status == -EIO)
3275                         return -EBADF;
3276                 else if (status == -EAGAIN)
3277                         goto zero_stateid;
3278         } else {
3279 zero_stateid:
3280                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3281         }
3282         if (delegation_cred)
3283                 msg.rpc_cred = delegation_cred;
3284
3285         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3286
3287         put_cred(delegation_cred);
3288         if (status == 0 && ctx != NULL)
3289                 renew_lease(server, timestamp);
3290         trace_nfs4_setattr(inode, &arg->stateid, status);
3291         return status;
3292 }
3293
3294 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3295                            struct nfs_fattr *fattr, struct iattr *sattr,
3296                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3297                            struct nfs4_label *olabel)
3298 {
3299         struct nfs_server *server = NFS_SERVER(inode);
3300         __u32 bitmask[NFS4_BITMASK_SZ];
3301         struct nfs4_state *state = ctx ? ctx->state : NULL;
3302         struct nfs_setattrargs  arg = {
3303                 .fh             = NFS_FH(inode),
3304                 .iap            = sattr,
3305                 .server         = server,
3306                 .bitmask = bitmask,
3307                 .label          = ilabel,
3308         };
3309         struct nfs_setattrres  res = {
3310                 .fattr          = fattr,
3311                 .label          = olabel,
3312                 .server         = server,
3313         };
3314         struct nfs4_exception exception = {
3315                 .state = state,
3316                 .inode = inode,
3317                 .stateid = &arg.stateid,
3318         };
3319         int err;
3320
3321         do {
3322                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3323                                 nfs4_bitmask(server, olabel),
3324                                 inode);
3325
3326                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3327                 switch (err) {
3328                 case -NFS4ERR_OPENMODE:
3329                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3330                                 pr_warn_once("NFSv4: server %s is incorrectly "
3331                                                 "applying open mode checks to "
3332                                                 "a SETATTR that is not "
3333                                                 "changing file size.\n",
3334                                                 server->nfs_client->cl_hostname);
3335                         }
3336                         if (state && !(state->state & FMODE_WRITE)) {
3337                                 err = -EBADF;
3338                                 if (sattr->ia_valid & ATTR_OPEN)
3339                                         err = -EACCES;
3340                                 goto out;
3341                         }
3342                 }
3343                 err = nfs4_handle_exception(server, err, &exception);
3344         } while (exception.retry);
3345 out:
3346         return err;
3347 }
3348
3349 static bool
3350 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3351 {
3352         if (inode == NULL || !nfs_have_layout(inode))
3353                 return false;
3354
3355         return pnfs_wait_on_layoutreturn(inode, task);
3356 }
3357
3358 /*
3359  * Update the seqid of an open stateid
3360  */
3361 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3362                 struct nfs4_state *state)
3363 {
3364         __be32 seqid_open;
3365         u32 dst_seqid;
3366         int seq;
3367
3368         for (;;) {
3369                 if (!nfs4_valid_open_stateid(state))
3370                         break;
3371                 seq = read_seqbegin(&state->seqlock);
3372                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3373                         nfs4_stateid_copy(dst, &state->open_stateid);
3374                         if (read_seqretry(&state->seqlock, seq))
3375                                 continue;
3376                         break;
3377                 }
3378                 seqid_open = state->open_stateid.seqid;
3379                 if (read_seqretry(&state->seqlock, seq))
3380                         continue;
3381
3382                 dst_seqid = be32_to_cpu(dst->seqid);
3383                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3384                         dst->seqid = seqid_open;
3385                 break;
3386         }
3387 }
3388
3389 /*
3390  * Update the seqid of an open stateid after receiving
3391  * NFS4ERR_OLD_STATEID
3392  */
3393 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3394                 struct nfs4_state *state)
3395 {
3396         __be32 seqid_open;
3397         u32 dst_seqid;
3398         bool ret;
3399         int seq;
3400
3401         for (;;) {
3402                 ret = false;
3403                 if (!nfs4_valid_open_stateid(state))
3404                         break;
3405                 seq = read_seqbegin(&state->seqlock);
3406                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3407                         if (read_seqretry(&state->seqlock, seq))
3408                                 continue;
3409                         break;
3410                 }
3411                 seqid_open = state->open_stateid.seqid;
3412                 if (read_seqretry(&state->seqlock, seq))
3413                         continue;
3414
3415                 dst_seqid = be32_to_cpu(dst->seqid);
3416                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) >= 0)
3417                         dst->seqid = cpu_to_be32(dst_seqid + 1);
3418                 else
3419                         dst->seqid = seqid_open;
3420                 ret = true;
3421                 break;
3422         }
3423
3424         return ret;
3425 }
3426
3427 struct nfs4_closedata {
3428         struct inode *inode;
3429         struct nfs4_state *state;
3430         struct nfs_closeargs arg;
3431         struct nfs_closeres res;
3432         struct {
3433                 struct nfs4_layoutreturn_args arg;
3434                 struct nfs4_layoutreturn_res res;
3435                 struct nfs4_xdr_opaque_data ld_private;
3436                 u32 roc_barrier;
3437                 bool roc;
3438         } lr;
3439         struct nfs_fattr fattr;
3440         unsigned long timestamp;
3441 };
3442
3443 static void nfs4_free_closedata(void *data)
3444 {
3445         struct nfs4_closedata *calldata = data;
3446         struct nfs4_state_owner *sp = calldata->state->owner;
3447         struct super_block *sb = calldata->state->inode->i_sb;
3448
3449         if (calldata->lr.roc)
3450                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3451                                 calldata->res.lr_ret);
3452         nfs4_put_open_state(calldata->state);
3453         nfs_free_seqid(calldata->arg.seqid);
3454         nfs4_put_state_owner(sp);
3455         nfs_sb_deactive(sb);
3456         kfree(calldata);
3457 }
3458
3459 static void nfs4_close_done(struct rpc_task *task, void *data)
3460 {
3461         struct nfs4_closedata *calldata = data;
3462         struct nfs4_state *state = calldata->state;
3463         struct nfs_server *server = NFS_SERVER(calldata->inode);
3464         nfs4_stateid *res_stateid = NULL;
3465         struct nfs4_exception exception = {
3466                 .state = state,
3467                 .inode = calldata->inode,
3468                 .stateid = &calldata->arg.stateid,
3469         };
3470
3471         dprintk("%s: begin!\n", __func__);
3472         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3473                 return;
3474         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3475
3476         /* Handle Layoutreturn errors */
3477         if (pnfs_roc_done(task, calldata->inode,
3478                                 &calldata->arg.lr_args,
3479                                 &calldata->res.lr_res,
3480                                 &calldata->res.lr_ret) == -EAGAIN)
3481                 goto out_restart;
3482
3483         /* hmm. we are done with the inode, and in the process of freeing
3484          * the state_owner. we keep this around to process errors
3485          */
3486         switch (task->tk_status) {
3487                 case 0:
3488                         res_stateid = &calldata->res.stateid;
3489                         renew_lease(server, calldata->timestamp);
3490                         break;
3491                 case -NFS4ERR_ACCESS:
3492                         if (calldata->arg.bitmask != NULL) {
3493                                 calldata->arg.bitmask = NULL;
3494                                 calldata->res.fattr = NULL;
3495                                 goto out_restart;
3496
3497                         }
3498                         break;
3499                 case -NFS4ERR_OLD_STATEID:
3500                         /* Did we race with OPEN? */
3501                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3502                                                 state))
3503                                 goto out_restart;
3504                         goto out_release;
3505                 case -NFS4ERR_ADMIN_REVOKED:
3506                 case -NFS4ERR_STALE_STATEID:
3507                 case -NFS4ERR_EXPIRED:
3508                         nfs4_free_revoked_stateid(server,
3509                                         &calldata->arg.stateid,
3510                                         task->tk_msg.rpc_cred);
3511                         /* Fallthrough */
3512                 case -NFS4ERR_BAD_STATEID:
3513                         if (calldata->arg.fmode == 0)
3514                                 break;
3515                         /* Fallthrough */
3516                 default:
3517                         task->tk_status = nfs4_async_handle_exception(task,
3518                                         server, task->tk_status, &exception);
3519                         if (exception.retry)
3520                                 goto out_restart;
3521         }
3522         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3523                         res_stateid, calldata->arg.fmode);
3524 out_release:
3525         task->tk_status = 0;
3526         nfs_release_seqid(calldata->arg.seqid);
3527         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3528         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3529         return;
3530 out_restart:
3531         task->tk_status = 0;
3532         rpc_restart_call_prepare(task);
3533         goto out_release;
3534 }
3535
3536 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3537 {
3538         struct nfs4_closedata *calldata = data;
3539         struct nfs4_state *state = calldata->state;
3540         struct inode *inode = calldata->inode;
3541         struct pnfs_layout_hdr *lo;
3542         bool is_rdonly, is_wronly, is_rdwr;
3543         int call_close = 0;
3544
3545         dprintk("%s: begin!\n", __func__);
3546         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3547                 goto out_wait;
3548
3549         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3550         spin_lock(&state->owner->so_lock);
3551         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3552         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3553         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3554         /* Calculate the change in open mode */
3555         calldata->arg.fmode = 0;
3556         if (state->n_rdwr == 0) {
3557                 if (state->n_rdonly == 0)
3558                         call_close |= is_rdonly;
3559                 else if (is_rdonly)
3560                         calldata->arg.fmode |= FMODE_READ;
3561                 if (state->n_wronly == 0)
3562                         call_close |= is_wronly;
3563                 else if (is_wronly)
3564                         calldata->arg.fmode |= FMODE_WRITE;
3565                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3566                         call_close |= is_rdwr;
3567         } else if (is_rdwr)
3568                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3569
3570         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3571         if (!nfs4_valid_open_stateid(state))
3572                 call_close = 0;
3573         spin_unlock(&state->owner->so_lock);
3574
3575         if (!call_close) {
3576                 /* Note: exit _without_ calling nfs4_close_done */
3577                 goto out_no_action;
3578         }
3579
3580         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3581                 nfs_release_seqid(calldata->arg.seqid);
3582                 goto out_wait;
3583         }
3584
3585         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3586         if (lo && !pnfs_layout_is_valid(lo)) {
3587                 calldata->arg.lr_args = NULL;
3588                 calldata->res.lr_res = NULL;
3589         }
3590
3591         if (calldata->arg.fmode == 0)
3592                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3593
3594         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3595                 /* Close-to-open cache consistency revalidation */
3596                 if (!nfs4_have_delegation(inode, FMODE_READ))
3597                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3598                 else
3599                         calldata->arg.bitmask = NULL;
3600         }
3601
3602         calldata->arg.share_access =
3603                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3604                                 calldata->arg.fmode, 0);
3605
3606         if (calldata->res.fattr == NULL)
3607                 calldata->arg.bitmask = NULL;
3608         else if (calldata->arg.bitmask == NULL)
3609                 calldata->res.fattr = NULL;
3610         calldata->timestamp = jiffies;
3611         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3612                                 &calldata->arg.seq_args,
3613                                 &calldata->res.seq_res,
3614                                 task) != 0)
3615                 nfs_release_seqid(calldata->arg.seqid);
3616         dprintk("%s: done!\n", __func__);
3617         return;
3618 out_no_action:
3619         task->tk_action = NULL;
3620 out_wait:
3621         nfs4_sequence_done(task, &calldata->res.seq_res);
3622 }
3623
3624 static const struct rpc_call_ops nfs4_close_ops = {
3625         .rpc_call_prepare = nfs4_close_prepare,
3626         .rpc_call_done = nfs4_close_done,
3627         .rpc_release = nfs4_free_closedata,
3628 };
3629
3630 /*
3631  * It is possible for data to be read/written from a mem-mapped file
3632  * after the sys_close call (which hits the vfs layer as a flush).
3633  * This means that we can't safely call nfsv4 close on a file until
3634  * the inode is cleared. This in turn means that we are not good
3635  * NFSv4 citizens - we do not indicate to the server to update the file's
3636  * share state even when we are done with one of the three share
3637  * stateid's in the inode.
3638  *
3639  * NOTE: Caller must be holding the sp->so_owner semaphore!
3640  */
3641 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3642 {
3643         struct nfs_server *server = NFS_SERVER(state->inode);
3644         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3645         struct nfs4_closedata *calldata;
3646         struct nfs4_state_owner *sp = state->owner;
3647         struct rpc_task *task;
3648         struct rpc_message msg = {
3649                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3650                 .rpc_cred = state->owner->so_cred,
3651         };
3652         struct rpc_task_setup task_setup_data = {
3653                 .rpc_client = server->client,
3654                 .rpc_message = &msg,
3655                 .callback_ops = &nfs4_close_ops,
3656                 .workqueue = nfsiod_workqueue,
3657                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3658         };
3659         int status = -ENOMEM;
3660
3661         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3662                 &task_setup_data.rpc_client, &msg);
3663
3664         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3665         if (calldata == NULL)
3666                 goto out;
3667         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3668         calldata->inode = state->inode;
3669         calldata->state = state;
3670         calldata->arg.fh = NFS_FH(state->inode);
3671         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3672                 goto out_free_calldata;
3673         /* Serialization for the sequence id */
3674         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3675         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3676         if (IS_ERR(calldata->arg.seqid))
3677                 goto out_free_calldata;
3678         nfs_fattr_init(&calldata->fattr);
3679         calldata->arg.fmode = 0;
3680         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3681         calldata->res.fattr = &calldata->fattr;
3682         calldata->res.seqid = calldata->arg.seqid;
3683         calldata->res.server = server;
3684         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3685         calldata->lr.roc = pnfs_roc(state->inode,
3686                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3687         if (calldata->lr.roc) {
3688                 calldata->arg.lr_args = &calldata->lr.arg;
3689                 calldata->res.lr_res = &calldata->lr.res;
3690         }
3691         nfs_sb_active(calldata->inode->i_sb);
3692
3693         msg.rpc_argp = &calldata->arg;
3694         msg.rpc_resp = &calldata->res;
3695         task_setup_data.callback_data = calldata;
3696         task = rpc_run_task(&task_setup_data);
3697         if (IS_ERR(task))
3698                 return PTR_ERR(task);
3699         status = 0;
3700         if (wait)
3701                 status = rpc_wait_for_completion_task(task);
3702         rpc_put_task(task);
3703         return status;
3704 out_free_calldata:
3705         kfree(calldata);
3706 out:
3707         nfs4_put_open_state(state);
3708         nfs4_put_state_owner(sp);
3709         return status;
3710 }
3711
3712 static struct inode *
3713 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3714                 int open_flags, struct iattr *attr, int *opened)
3715 {
3716         struct nfs4_state *state;
3717         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3718
3719         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3720
3721         /* Protect against concurrent sillydeletes */
3722         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3723
3724         nfs4_label_release_security(label);
3725
3726         if (IS_ERR(state))
3727                 return ERR_CAST(state);
3728         return state->inode;
3729 }
3730
3731 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3732 {
3733         if (ctx->state == NULL)
3734                 return;
3735         if (is_sync)
3736                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3737         else
3738                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3739 }
3740
3741 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3742 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3743 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3744
3745 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3746 {
3747         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3748         struct nfs4_server_caps_arg args = {
3749                 .fhandle = fhandle,
3750                 .bitmask = bitmask,
3751         };
3752         struct nfs4_server_caps_res res = {};
3753         struct rpc_message msg = {
3754                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3755                 .rpc_argp = &args,
3756                 .rpc_resp = &res,
3757         };
3758         int status;
3759         int i;
3760
3761         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3762                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3763                      FATTR4_WORD0_LINK_SUPPORT |
3764                      FATTR4_WORD0_SYMLINK_SUPPORT |
3765                      FATTR4_WORD0_ACLSUPPORT;
3766         if (minorversion)
3767                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3768
3769         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3770         if (status == 0) {
3771                 /* Sanity check the server answers */
3772                 switch (minorversion) {
3773                 case 0:
3774                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3775                         res.attr_bitmask[2] = 0;
3776                         break;
3777                 case 1:
3778                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3779                         break;
3780                 case 2:
3781                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3782                 }
3783                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3784                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3785                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3786                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3787                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3788                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3789                                 NFS_CAP_SECURITY_LABEL);
3790                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3791                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3792                         server->caps |= NFS_CAP_ACLS;
3793                 if (res.has_links != 0)
3794                         server->caps |= NFS_CAP_HARDLINKS;
3795                 if (res.has_symlinks != 0)
3796                         server->caps |= NFS_CAP_SYMLINKS;
3797                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3798                         server->caps |= NFS_CAP_FILEID;
3799                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3800                         server->caps |= NFS_CAP_MODE;
3801                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3802                         server->caps |= NFS_CAP_NLINK;
3803                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3804                         server->caps |= NFS_CAP_OWNER;
3805                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3806                         server->caps |= NFS_CAP_OWNER_GROUP;
3807                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3808                         server->caps |= NFS_CAP_ATIME;
3809                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3810                         server->caps |= NFS_CAP_CTIME;
3811                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3812                         server->caps |= NFS_CAP_MTIME;
3813 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3814                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3815                         server->caps |= NFS_CAP_SECURITY_LABEL;
3816 #endif
3817                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3818                                 sizeof(server->attr_bitmask));
3819                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3820
3821                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3822                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3823                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3824                 server->cache_consistency_bitmask[2] = 0;
3825
3826                 /* Avoid a regression due to buggy server */
3827                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3828                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3829                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3830                         sizeof(server->exclcreat_bitmask));
3831
3832                 server->acl_bitmask = res.acl_bitmask;
3833                 server->fh_expire_type = res.fh_expire_type;
3834         }
3835
3836         return status;
3837 }
3838
3839 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3840 {
3841         struct nfs4_exception exception = {
3842                 .interruptible = true,
3843         };
3844         int err;
3845         do {
3846                 err = nfs4_handle_exception(server,
3847                                 _nfs4_server_capabilities(server, fhandle),
3848                                 &exception);
3849         } while (exception.retry);
3850         return err;
3851 }
3852
3853 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3854                 struct nfs_fsinfo *info)
3855 {
3856         u32 bitmask[3];
3857         struct nfs4_lookup_root_arg args = {
3858                 .bitmask = bitmask,
3859         };
3860         struct nfs4_lookup_res res = {
3861                 .server = server,
3862                 .fattr = info->fattr,
3863                 .fh = fhandle,
3864         };
3865         struct rpc_message msg = {
3866                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3867                 .rpc_argp = &args,
3868                 .rpc_resp = &res,
3869         };
3870
3871         bitmask[0] = nfs4_fattr_bitmap[0];
3872         bitmask[1] = nfs4_fattr_bitmap[1];
3873         /*
3874          * Process the label in the upcoming getfattr
3875          */
3876         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3877
3878         nfs_fattr_init(info->fattr);
3879         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3880 }
3881
3882 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3883                 struct nfs_fsinfo *info)
3884 {
3885         struct nfs4_exception exception = {
3886                 .interruptible = true,
3887         };
3888         int err;
3889         do {
3890                 err = _nfs4_lookup_root(server, fhandle, info);
3891                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3892                 switch (err) {
3893                 case 0:
3894                 case -NFS4ERR_WRONGSEC:
3895                         goto out;
3896                 default:
3897                         err = nfs4_handle_exception(server, err, &exception);
3898                 }
3899         } while (exception.retry);
3900 out:
3901         return err;
3902 }
3903
3904 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3905                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3906 {
3907         struct rpc_auth_create_args auth_args = {
3908                 .pseudoflavor = flavor,
3909         };
3910         struct rpc_auth *auth;
3911
3912         auth = rpcauth_create(&auth_args, server->client);
3913         if (IS_ERR(auth))
3914                 return -EACCES;
3915         return nfs4_lookup_root(server, fhandle, info);
3916 }
3917
3918 /*
3919  * Retry pseudoroot lookup with various security flavors.  We do this when:
3920  *
3921  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3922  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3923  *
3924  * Returns zero on success, or a negative NFS4ERR value, or a
3925  * negative errno value.
3926  */
3927 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3928                               struct nfs_fsinfo *info)
3929 {
3930         /* Per 3530bis 15.33.5 */
3931         static const rpc_authflavor_t flav_array[] = {
3932                 RPC_AUTH_GSS_KRB5P,
3933                 RPC_AUTH_GSS_KRB5I,
3934                 RPC_AUTH_GSS_KRB5,
3935                 RPC_AUTH_UNIX,                  /* courtesy */
3936                 RPC_AUTH_NULL,
3937         };
3938         int status = -EPERM;
3939         size_t i;
3940
3941         if (server->auth_info.flavor_len > 0) {
3942                 /* try each flavor specified by user */
3943                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3944                         status = nfs4_lookup_root_sec(server, fhandle, info,
3945                                                 server->auth_info.flavors[i]);
3946                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3947                                 continue;
3948                         break;
3949                 }
3950         } else {
3951                 /* no flavors specified by user, try default list */
3952                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3953                         status = nfs4_lookup_root_sec(server, fhandle, info,
3954                                                       flav_array[i]);
3955                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3956                                 continue;
3957                         break;
3958                 }
3959         }
3960
3961         /*
3962          * -EACCES could mean that the user doesn't have correct permissions
3963          * to access the mount.  It could also mean that we tried to mount
3964          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3965          * existing mount programs don't handle -EACCES very well so it should
3966          * be mapped to -EPERM instead.
3967          */
3968         if (status == -EACCES)
3969                 status = -EPERM;
3970         return status;
3971 }
3972
3973 /**
3974  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3975  * @server: initialized nfs_server handle
3976  * @fhandle: we fill in the pseudo-fs root file handle
3977  * @info: we fill in an FSINFO struct
3978  * @auth_probe: probe the auth flavours
3979  *
3980  * Returns zero on success, or a negative errno.
3981  */
3982 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3983                          struct nfs_fsinfo *info,
3984                          bool auth_probe)
3985 {
3986         int status = 0;
3987
3988         if (!auth_probe)
3989                 status = nfs4_lookup_root(server, fhandle, info);
3990
3991         if (auth_probe || status == NFS4ERR_WRONGSEC)
3992                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3993                                 fhandle, info);
3994
3995         if (status == 0)
3996                 status = nfs4_server_capabilities(server, fhandle);
3997         if (status == 0)
3998                 status = nfs4_do_fsinfo(server, fhandle, info);
3999
4000         return nfs4_map_errors(status);
4001 }
4002
4003 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4004                               struct nfs_fsinfo *info)
4005 {
4006         int error;
4007         struct nfs_fattr *fattr = info->fattr;
4008         struct nfs4_label *label = fattr->label;
4009
4010         error = nfs4_server_capabilities(server, mntfh);
4011         if (error < 0) {
4012                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4013                 return error;
4014         }
4015
4016         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4017         if (error < 0) {
4018                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4019                 goto out;
4020         }
4021
4022         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4023             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4024                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4025
4026 out:
4027         return error;
4028 }
4029
4030 /*
4031  * Get locations and (maybe) other attributes of a referral.
4032  * Note that we'll actually follow the referral later when
4033  * we detect fsid mismatch in inode revalidation
4034  */
4035 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4036                              const struct qstr *name, struct nfs_fattr *fattr,
4037                              struct nfs_fh *fhandle)
4038 {
4039         int status = -ENOMEM;
4040         struct page *page = NULL;
4041         struct nfs4_fs_locations *locations = NULL;
4042
4043         page = alloc_page(GFP_KERNEL);
4044         if (page == NULL)
4045                 goto out;
4046         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4047         if (locations == NULL)
4048                 goto out;
4049
4050         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4051         if (status != 0)
4052                 goto out;
4053
4054         /*
4055          * If the fsid didn't change, this is a migration event, not a
4056          * referral.  Cause us to drop into the exception handler, which
4057          * will kick off migration recovery.
4058          */
4059         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4060                 dprintk("%s: server did not return a different fsid for"
4061                         " a referral at %s\n", __func__, name->name);
4062                 status = -NFS4ERR_MOVED;
4063                 goto out;
4064         }
4065         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4066         nfs_fixup_referral_attributes(&locations->fattr);
4067
4068         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4069         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4070         memset(fhandle, 0, sizeof(struct nfs_fh));
4071 out:
4072         if (page)
4073                 __free_page(page);
4074         kfree(locations);
4075         return status;
4076 }
4077
4078 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4079                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4080                                 struct inode *inode)
4081 {
4082         __u32 bitmask[NFS4_BITMASK_SZ];
4083         struct nfs4_getattr_arg args = {
4084                 .fh = fhandle,
4085                 .bitmask = bitmask,
4086         };
4087         struct nfs4_getattr_res res = {
4088                 .fattr = fattr,
4089                 .label = label,
4090                 .server = server,
4091         };
4092         struct rpc_message msg = {
4093                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4094                 .rpc_argp = &args,
4095                 .rpc_resp = &res,
4096         };
4097         unsigned short task_flags = 0;
4098
4099         /* Is this is an attribute revalidation, subject to softreval? */
4100         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4101                 task_flags |= RPC_TASK_TIMEOUT;
4102
4103         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4104
4105         nfs_fattr_init(fattr);
4106         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4107         return nfs4_do_call_sync(server->client, server, &msg,
4108                         &args.seq_args, &res.seq_res, task_flags);
4109 }
4110
4111 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4112                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4113                                 struct inode *inode)
4114 {
4115         struct nfs4_exception exception = {
4116                 .interruptible = true,
4117         };
4118         int err;
4119         do {
4120                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4121                 trace_nfs4_getattr(server, fhandle, fattr, err);
4122                 err = nfs4_handle_exception(server, err,
4123                                 &exception);
4124         } while (exception.retry);
4125         return err;
4126 }
4127
4128 /*
4129  * The file is not closed if it is opened due to the a request to change
4130  * the size of the file. The open call will not be needed once the
4131  * VFS layer lookup-intents are implemented.
4132  *
4133  * Close is called when the inode is destroyed.
4134  * If we haven't opened the file for O_WRONLY, we
4135  * need to in the size_change case to obtain a stateid.
4136  *
4137  * Got race?
4138  * Because OPEN is always done by name in nfsv4, it is
4139  * possible that we opened a different file by the same
4140  * name.  We can recognize this race condition, but we
4141  * can't do anything about it besides returning an error.
4142  *
4143  * This will be fixed with VFS changes (lookup-intent).
4144  */
4145 static int
4146 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4147                   struct iattr *sattr)
4148 {
4149         struct inode *inode = d_inode(dentry);
4150         const struct cred *cred = NULL;
4151         struct nfs_open_context *ctx = NULL;
4152         struct nfs4_label *label = NULL;
4153         int status;
4154
4155         if (pnfs_ld_layoutret_on_setattr(inode) &&
4156             sattr->ia_valid & ATTR_SIZE &&
4157             sattr->ia_size < i_size_read(inode))
4158                 pnfs_commit_and_return_layout(inode);
4159
4160         nfs_fattr_init(fattr);
4161
4162         /* Deal with open(O_TRUNC) */
4163         if (sattr->ia_valid & ATTR_OPEN)
4164                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4165
4166         /* Optimization: if the end result is no change, don't RPC */
4167         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4168                 return 0;
4169
4170         /* Search for an existing open(O_WRITE) file */
4171         if (sattr->ia_valid & ATTR_FILE) {
4172
4173                 ctx = nfs_file_open_context(sattr->ia_file);
4174                 if (ctx)
4175                         cred = ctx->cred;
4176         }
4177
4178         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4179         if (IS_ERR(label))
4180                 return PTR_ERR(label);
4181
4182         /* Return any delegations if we're going to change ACLs */
4183         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4184                 nfs4_inode_make_writeable(inode);
4185
4186         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4187         if (status == 0) {
4188                 nfs_setattr_update_inode(inode, sattr, fattr);
4189                 nfs_setsecurity(inode, fattr, label);
4190         }
4191         nfs4_label_free(label);
4192         return status;
4193 }
4194
4195 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4196                 struct dentry *dentry, struct nfs_fh *fhandle,
4197                 struct nfs_fattr *fattr, struct nfs4_label *label)
4198 {
4199         struct nfs_server *server = NFS_SERVER(dir);
4200         int                    status;
4201         struct nfs4_lookup_arg args = {
4202                 .bitmask = server->attr_bitmask,
4203                 .dir_fh = NFS_FH(dir),
4204                 .name = &dentry->d_name,
4205         };
4206         struct nfs4_lookup_res res = {
4207                 .server = server,
4208                 .fattr = fattr,
4209                 .label = label,
4210                 .fh = fhandle,
4211         };
4212         struct rpc_message msg = {
4213                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4214                 .rpc_argp = &args,
4215                 .rpc_resp = &res,
4216         };
4217         unsigned short task_flags = 0;
4218
4219         /* Is this is an attribute revalidation, subject to softreval? */
4220         if (nfs_lookup_is_soft_revalidate(dentry))
4221                 task_flags |= RPC_TASK_TIMEOUT;
4222
4223         args.bitmask = nfs4_bitmask(server, label);
4224
4225         nfs_fattr_init(fattr);
4226
4227         dprintk("NFS call  lookup %pd2\n", dentry);
4228         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4229         status = nfs4_do_call_sync(clnt, server, &msg,
4230                         &args.seq_args, &res.seq_res, task_flags);
4231         dprintk("NFS reply lookup: %d\n", status);
4232         return status;
4233 }
4234
4235 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4236 {
4237         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4238                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4239         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4240         fattr->nlink = 2;
4241 }
4242
4243 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4244                                    struct dentry *dentry, struct nfs_fh *fhandle,
4245                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4246 {
4247         struct nfs4_exception exception = {
4248                 .interruptible = true,
4249         };
4250         struct rpc_clnt *client = *clnt;
4251         const struct qstr *name = &dentry->d_name;
4252         int err;
4253         do {
4254                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4255                 trace_nfs4_lookup(dir, name, err);
4256                 switch (err) {
4257                 case -NFS4ERR_BADNAME:
4258                         err = -ENOENT;
4259                         goto out;
4260                 case -NFS4ERR_MOVED:
4261                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4262                         if (err == -NFS4ERR_MOVED)
4263                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4264                         goto out;
4265                 case -NFS4ERR_WRONGSEC:
4266                         err = -EPERM;
4267                         if (client != *clnt)
4268                                 goto out;
4269                         client = nfs4_negotiate_security(client, dir, name);
4270                         if (IS_ERR(client))
4271                                 return PTR_ERR(client);
4272
4273                         exception.retry = 1;
4274                         break;
4275                 default:
4276                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4277                 }
4278         } while (exception.retry);
4279
4280 out:
4281         if (err == 0)
4282                 *clnt = client;
4283         else if (client != *clnt)
4284                 rpc_shutdown_client(client);
4285
4286         return err;
4287 }
4288
4289 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4290                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4291                             struct nfs4_label *label)
4292 {
4293         int status;
4294         struct rpc_clnt *client = NFS_CLIENT(dir);
4295
4296         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4297         if (client != NFS_CLIENT(dir)) {
4298                 rpc_shutdown_client(client);
4299                 nfs_fixup_secinfo_attributes(fattr);
4300         }
4301         return status;
4302 }
4303
4304 struct rpc_clnt *
4305 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4306                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4307 {
4308         struct rpc_clnt *client = NFS_CLIENT(dir);
4309         int status;
4310
4311         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4312         if (status < 0)
4313                 return ERR_PTR(status);
4314         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4315 }
4316
4317 static int _nfs4_proc_lookupp(struct inode *inode,
4318                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4319                 struct nfs4_label *label)
4320 {
4321         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4322         struct nfs_server *server = NFS_SERVER(inode);
4323         int                    status;
4324         struct nfs4_lookupp_arg args = {
4325                 .bitmask = server->attr_bitmask,
4326                 .fh = NFS_FH(inode),
4327         };
4328         struct nfs4_lookupp_res res = {
4329                 .server = server,
4330                 .fattr = fattr,
4331                 .label = label,
4332                 .fh = fhandle,
4333         };
4334         struct rpc_message msg = {
4335                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4336                 .rpc_argp = &args,
4337                 .rpc_resp = &res,
4338         };
4339
4340         args.bitmask = nfs4_bitmask(server, label);
4341
4342         nfs_fattr_init(fattr);
4343
4344         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4345         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4346                                 &res.seq_res, 0);
4347         dprintk("NFS reply lookupp: %d\n", status);
4348         return status;
4349 }
4350
4351 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4352                              struct nfs_fattr *fattr, struct nfs4_label *label)
4353 {
4354         struct nfs4_exception exception = {
4355                 .interruptible = true,
4356         };
4357         int err;
4358         do {
4359                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4360                 trace_nfs4_lookupp(inode, err);
4361                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4362                                 &exception);
4363         } while (exception.retry);
4364         return err;
4365 }
4366
4367 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4368 {
4369         struct nfs_server *server = NFS_SERVER(inode);
4370         struct nfs4_accessargs args = {
4371                 .fh = NFS_FH(inode),
4372                 .access = entry->mask,
4373         };
4374         struct nfs4_accessres res = {
4375                 .server = server,
4376         };
4377         struct rpc_message msg = {
4378                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4379                 .rpc_argp = &args,
4380                 .rpc_resp = &res,
4381                 .rpc_cred = entry->cred,
4382         };
4383         int status = 0;
4384
4385         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4386                 res.fattr = nfs_alloc_fattr();
4387                 if (res.fattr == NULL)
4388                         return -ENOMEM;
4389                 args.bitmask = server->cache_consistency_bitmask;
4390         }
4391         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4392         if (!status) {
4393                 nfs_access_set_mask(entry, res.access);
4394                 if (res.fattr)
4395                         nfs_refresh_inode(inode, res.fattr);
4396         }
4397         nfs_free_fattr(res.fattr);
4398         return status;
4399 }
4400
4401 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4402 {
4403         struct nfs4_exception exception = {
4404                 .interruptible = true,
4405         };
4406         int err;
4407         do {
4408                 err = _nfs4_proc_access(inode, entry);
4409                 trace_nfs4_access(inode, err);
4410                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4411                                 &exception);
4412         } while (exception.retry);
4413         return err;
4414 }
4415
4416 /*
4417  * TODO: For the time being, we don't try to get any attributes
4418  * along with any of the zero-copy operations READ, READDIR,
4419  * READLINK, WRITE.
4420  *
4421  * In the case of the first three, we want to put the GETATTR
4422  * after the read-type operation -- this is because it is hard
4423  * to predict the length of a GETATTR response in v4, and thus
4424  * align the READ data correctly.  This means that the GETATTR
4425  * may end up partially falling into the page cache, and we should
4426  * shift it into the 'tail' of the xdr_buf before processing.
4427  * To do this efficiently, we need to know the total length
4428  * of data received, which doesn't seem to be available outside
4429  * of the RPC layer.
4430  *
4431  * In the case of WRITE, we also want to put the GETATTR after
4432  * the operation -- in this case because we want to make sure
4433  * we get the post-operation mtime and size.
4434  *
4435  * Both of these changes to the XDR layer would in fact be quite
4436  * minor, but I decided to leave them for a subsequent patch.
4437  */
4438 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4439                 unsigned int pgbase, unsigned int pglen)
4440 {
4441         struct nfs4_readlink args = {
4442                 .fh       = NFS_FH(inode),
4443                 .pgbase   = pgbase,
4444                 .pglen    = pglen,
4445                 .pages    = &page,
4446         };
4447         struct nfs4_readlink_res res;
4448         struct rpc_message msg = {
4449                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4450                 .rpc_argp = &args,
4451                 .rpc_resp = &res,
4452         };
4453
4454         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4455 }
4456
4457 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4458                 unsigned int pgbase, unsigned int pglen)
4459 {
4460         struct nfs4_exception exception = {
4461                 .interruptible = true,
4462         };
4463         int err;
4464         do {
4465                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4466                 trace_nfs4_readlink(inode, err);
4467                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4468                                 &exception);
4469         } while (exception.retry);
4470         return err;
4471 }
4472
4473 /*
4474  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4475  */
4476 static int
4477 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4478                  int flags)
4479 {
4480         struct nfs_server *server = NFS_SERVER(dir);
4481         struct nfs4_label l, *ilabel = NULL;
4482         struct nfs_open_context *ctx;
4483         struct nfs4_state *state;
4484         int status = 0;
4485
4486         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4487         if (IS_ERR(ctx))
4488                 return PTR_ERR(ctx);
4489
4490         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4491
4492         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4493                 sattr->ia_mode &= ~current_umask();
4494         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4495         if (IS_ERR(state)) {
4496                 status = PTR_ERR(state);
4497                 goto out;
4498         }
4499 out:
4500         nfs4_label_release_security(ilabel);
4501         put_nfs_open_context(ctx);
4502         return status;
4503 }
4504
4505 static int
4506 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4507 {
4508         struct nfs_server *server = NFS_SERVER(dir);
4509         struct nfs_removeargs args = {
4510                 .fh = NFS_FH(dir),
4511                 .name = *name,
4512         };
4513         struct nfs_removeres res = {
4514                 .server = server,
4515         };
4516         struct rpc_message msg = {
4517                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4518                 .rpc_argp = &args,
4519                 .rpc_resp = &res,
4520         };
4521         unsigned long timestamp = jiffies;
4522         int status;
4523
4524         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4525         if (status == 0) {
4526                 spin_lock(&dir->i_lock);
4527                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4528                 /* Removing a directory decrements nlink in the parent */
4529                 if (ftype == NF4DIR && dir->i_nlink > 2)
4530                         nfs4_dec_nlink_locked(dir);
4531                 spin_unlock(&dir->i_lock);
4532         }
4533         return status;
4534 }
4535
4536 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4537 {
4538         struct nfs4_exception exception = {
4539                 .interruptible = true,
4540         };
4541         struct inode *inode = d_inode(dentry);
4542         int err;
4543
4544         if (inode) {
4545                 if (inode->i_nlink == 1)
4546                         nfs4_inode_return_delegation(inode);
4547                 else
4548                         nfs4_inode_make_writeable(inode);
4549         }
4550         do {
4551                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4552                 trace_nfs4_remove(dir, &dentry->d_name, err);
4553                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4554                                 &exception);
4555         } while (exception.retry);
4556         return err;
4557 }
4558
4559 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4560 {
4561         struct nfs4_exception exception = {
4562                 .interruptible = true,
4563         };
4564         int err;
4565
4566         do {
4567                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4568                 trace_nfs4_remove(dir, name, err);
4569                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4570                                 &exception);
4571         } while (exception.retry);
4572         return err;
4573 }
4574
4575 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4576                 struct dentry *dentry,
4577                 struct inode *inode)
4578 {
4579         struct nfs_removeargs *args = msg->rpc_argp;
4580         struct nfs_removeres *res = msg->rpc_resp;
4581
4582         res->server = NFS_SB(dentry->d_sb);
4583         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4584         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4585
4586         nfs_fattr_init(res->dir_attr);
4587
4588         if (inode)
4589                 nfs4_inode_return_delegation(inode);
4590 }
4591
4592 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4593 {
4594         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4595                         &data->args.seq_args,
4596                         &data->res.seq_res,
4597                         task);
4598 }
4599
4600 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4601 {
4602         struct nfs_unlinkdata *data = task->tk_calldata;
4603         struct nfs_removeres *res = &data->res;
4604
4605         if (!nfs4_sequence_done(task, &res->seq_res))
4606                 return 0;
4607         if (nfs4_async_handle_error(task, res->server, NULL,
4608                                     &data->timeout) == -EAGAIN)
4609                 return 0;
4610         if (task->tk_status == 0)
4611                 update_changeattr(dir, &res->cinfo,
4612                                 res->dir_attr->time_start, 0);
4613         return 1;
4614 }
4615
4616 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4617                 struct dentry *old_dentry,
4618                 struct dentry *new_dentry)
4619 {
4620         struct nfs_renameargs *arg = msg->rpc_argp;
4621         struct nfs_renameres *res = msg->rpc_resp;
4622         struct inode *old_inode = d_inode(old_dentry);
4623         struct inode *new_inode = d_inode(new_dentry);
4624
4625         if (old_inode)
4626                 nfs4_inode_make_writeable(old_inode);
4627         if (new_inode)
4628                 nfs4_inode_return_delegation(new_inode);
4629         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4630         res->server = NFS_SB(old_dentry->d_sb);
4631         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4632 }
4633
4634 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4635 {
4636         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4637                         &data->args.seq_args,
4638                         &data->res.seq_res,
4639                         task);
4640 }
4641
4642 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4643                                  struct inode *new_dir)
4644 {
4645         struct nfs_renamedata *data = task->tk_calldata;
4646         struct nfs_renameres *res = &data->res;
4647
4648         if (!nfs4_sequence_done(task, &res->seq_res))
4649                 return 0;
4650         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4651                 return 0;
4652
4653         if (task->tk_status == 0) {
4654                 if (new_dir != old_dir) {
4655                         /* Note: If we moved a directory, nlink will change */
4656                         update_changeattr(old_dir, &res->old_cinfo,
4657                                         res->old_fattr->time_start,
4658                                         NFS_INO_INVALID_OTHER);
4659                         update_changeattr(new_dir, &res->new_cinfo,
4660                                         res->new_fattr->time_start,
4661                                         NFS_INO_INVALID_OTHER);
4662                 } else
4663                         update_changeattr(old_dir, &res->old_cinfo,
4664                                         res->old_fattr->time_start,
4665                                         0);
4666         }
4667         return 1;
4668 }
4669
4670 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4671 {
4672         struct nfs_server *server = NFS_SERVER(inode);
4673         __u32 bitmask[NFS4_BITMASK_SZ];
4674         struct nfs4_link_arg arg = {
4675                 .fh     = NFS_FH(inode),
4676                 .dir_fh = NFS_FH(dir),
4677                 .name   = name,
4678                 .bitmask = bitmask,
4679         };
4680         struct nfs4_link_res res = {
4681                 .server = server,
4682                 .label = NULL,
4683         };
4684         struct rpc_message msg = {
4685                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4686                 .rpc_argp = &arg,
4687                 .rpc_resp = &res,
4688         };
4689         int status = -ENOMEM;
4690
4691         res.fattr = nfs_alloc_fattr();
4692         if (res.fattr == NULL)
4693                 goto out;
4694
4695         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4696         if (IS_ERR(res.label)) {
4697                 status = PTR_ERR(res.label);
4698                 goto out;
4699         }
4700
4701         nfs4_inode_make_writeable(inode);
4702         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4703
4704         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4705         if (!status) {
4706                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4707                 status = nfs_post_op_update_inode(inode, res.fattr);
4708                 if (!status)
4709                         nfs_setsecurity(inode, res.fattr, res.label);
4710         }
4711
4712
4713         nfs4_label_free(res.label);
4714
4715 out:
4716         nfs_free_fattr(res.fattr);
4717         return status;
4718 }
4719
4720 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4721 {
4722         struct nfs4_exception exception = {
4723                 .interruptible = true,
4724         };
4725         int err;
4726         do {
4727                 err = nfs4_handle_exception(NFS_SERVER(inode),
4728                                 _nfs4_proc_link(inode, dir, name),
4729                                 &exception);
4730         } while (exception.retry);
4731         return err;
4732 }
4733
4734 struct nfs4_createdata {
4735         struct rpc_message msg;
4736         struct nfs4_create_arg arg;
4737         struct nfs4_create_res res;
4738         struct nfs_fh fh;
4739         struct nfs_fattr fattr;
4740         struct nfs4_label *label;
4741 };
4742
4743 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4744                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4745 {
4746         struct nfs4_createdata *data;
4747
4748         data = kzalloc(sizeof(*data), GFP_KERNEL);
4749         if (data != NULL) {
4750                 struct nfs_server *server = NFS_SERVER(dir);
4751
4752                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4753                 if (IS_ERR(data->label))
4754                         goto out_free;
4755
4756                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4757                 data->msg.rpc_argp = &data->arg;
4758                 data->msg.rpc_resp = &data->res;
4759                 data->arg.dir_fh = NFS_FH(dir);
4760                 data->arg.server = server;
4761                 data->arg.name = name;
4762                 data->arg.attrs = sattr;
4763                 data->arg.ftype = ftype;
4764                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4765                 data->arg.umask = current_umask();
4766                 data->res.server = server;
4767                 data->res.fh = &data->fh;
4768                 data->res.fattr = &data->fattr;
4769                 data->res.label = data->label;
4770                 nfs_fattr_init(data->res.fattr);
4771         }
4772         return data;
4773 out_free:
4774         kfree(data);
4775         return NULL;
4776 }
4777
4778 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4779 {
4780         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4781                                     &data->arg.seq_args, &data->res.seq_res, 1);
4782         if (status == 0) {
4783                 spin_lock(&dir->i_lock);
4784                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4785                                 data->res.fattr->time_start, 0);
4786                 /* Creating a directory bumps nlink in the parent */
4787                 if (data->arg.ftype == NF4DIR)
4788                         nfs4_inc_nlink_locked(dir);
4789                 spin_unlock(&dir->i_lock);
4790                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4791         }
4792         return status;
4793 }
4794
4795 static void nfs4_free_createdata(struct nfs4_createdata *data)
4796 {
4797         nfs4_label_free(data->label);
4798         kfree(data);
4799 }
4800
4801 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4802                 struct page *page, unsigned int len, struct iattr *sattr,
4803                 struct nfs4_label *label)
4804 {
4805         struct nfs4_createdata *data;
4806         int status = -ENAMETOOLONG;
4807
4808         if (len > NFS4_MAXPATHLEN)
4809                 goto out;
4810
4811         status = -ENOMEM;
4812         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4813         if (data == NULL)
4814                 goto out;
4815
4816         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4817         data->arg.u.symlink.pages = &page;
4818         data->arg.u.symlink.len = len;
4819         data->arg.label = label;
4820
4821         status = nfs4_do_create(dir, dentry, data);
4822
4823         nfs4_free_createdata(data);
4824 out:
4825         return status;
4826 }
4827
4828 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4829                 struct page *page, unsigned int len, struct iattr *sattr)
4830 {
4831         struct nfs4_exception exception = {
4832                 .interruptible = true,
4833         };
4834         struct nfs4_label l, *label = NULL;
4835         int err;
4836
4837         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4838
4839         do {
4840                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4841                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4842                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4843                                 &exception);
4844         } while (exception.retry);
4845
4846         nfs4_label_release_security(label);
4847         return err;
4848 }
4849
4850 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4851                 struct iattr *sattr, struct nfs4_label *label)
4852 {
4853         struct nfs4_createdata *data;
4854         int status = -ENOMEM;
4855
4856         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4857         if (data == NULL)
4858                 goto out;
4859
4860         data->arg.label = label;
4861         status = nfs4_do_create(dir, dentry, data);
4862
4863         nfs4_free_createdata(data);
4864 out:
4865         return status;
4866 }
4867
4868 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4869                 struct iattr *sattr)
4870 {
4871         struct nfs_server *server = NFS_SERVER(dir);
4872         struct nfs4_exception exception = {
4873                 .interruptible = true,
4874         };
4875         struct nfs4_label l, *label = NULL;
4876         int err;
4877
4878         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4879
4880         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4881                 sattr->ia_mode &= ~current_umask();
4882         do {
4883                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4884                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4885                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4886                                 &exception);
4887         } while (exception.retry);
4888         nfs4_label_release_security(label);
4889
4890         return err;
4891 }
4892
4893 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4894                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4895 {
4896         struct inode            *dir = d_inode(dentry);
4897         struct nfs4_readdir_arg args = {
4898                 .fh = NFS_FH(dir),
4899                 .pages = pages,
4900                 .pgbase = 0,
4901                 .count = count,
4902                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4903                 .plus = plus,
4904         };
4905         struct nfs4_readdir_res res;
4906         struct rpc_message msg = {
4907                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4908                 .rpc_argp = &args,
4909                 .rpc_resp = &res,
4910                 .rpc_cred = cred,
4911         };
4912         int                     status;
4913
4914         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4915                         dentry,
4916                         (unsigned long long)cookie);
4917         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4918         res.pgbase = args.pgbase;
4919         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4920         if (status >= 0) {
4921                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4922                 status += args.pgbase;
4923         }
4924
4925         nfs_invalidate_atime(dir);
4926
4927         dprintk("%s: returns %d\n", __func__, status);
4928         return status;
4929 }
4930
4931 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4932                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4933 {
4934         struct nfs4_exception exception = {
4935                 .interruptible = true,
4936         };
4937         int err;
4938         do {
4939                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4940                                 pages, count, plus);
4941                 trace_nfs4_readdir(d_inode(dentry), err);
4942                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4943                                 &exception);
4944         } while (exception.retry);
4945         return err;
4946 }
4947
4948 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4949                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4950 {
4951         struct nfs4_createdata *data;
4952         int mode = sattr->ia_mode;
4953         int status = -ENOMEM;
4954
4955         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4956         if (data == NULL)
4957                 goto out;
4958
4959         if (S_ISFIFO(mode))
4960                 data->arg.ftype = NF4FIFO;
4961         else if (S_ISBLK(mode)) {
4962                 data->arg.ftype = NF4BLK;
4963                 data->arg.u.device.specdata1 = MAJOR(rdev);
4964                 data->arg.u.device.specdata2 = MINOR(rdev);
4965         }
4966         else if (S_ISCHR(mode)) {
4967                 data->arg.ftype = NF4CHR;
4968                 data->arg.u.device.specdata1 = MAJOR(rdev);
4969                 data->arg.u.device.specdata2 = MINOR(rdev);
4970         } else if (!S_ISSOCK(mode)) {
4971                 status = -EINVAL;
4972                 goto out_free;
4973         }
4974
4975         data->arg.label = label;
4976         status = nfs4_do_create(dir, dentry, data);
4977 out_free:
4978         nfs4_free_createdata(data);
4979 out:
4980         return status;
4981 }
4982
4983 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4984                 struct iattr *sattr, dev_t rdev)
4985 {
4986         struct nfs_server *server = NFS_SERVER(dir);
4987         struct nfs4_exception exception = {
4988                 .interruptible = true,
4989         };
4990         struct nfs4_label l, *label = NULL;
4991         int err;
4992
4993         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4994
4995         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4996                 sattr->ia_mode &= ~current_umask();
4997         do {
4998                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4999                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5000                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5001                                 &exception);
5002         } while (exception.retry);
5003
5004         nfs4_label_release_security(label);
5005
5006         return err;
5007 }
5008
5009 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5010                  struct nfs_fsstat *fsstat)
5011 {
5012         struct nfs4_statfs_arg args = {
5013                 .fh = fhandle,
5014                 .bitmask = server->attr_bitmask,
5015         };
5016         struct nfs4_statfs_res res = {
5017                 .fsstat = fsstat,
5018         };
5019         struct rpc_message msg = {
5020                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5021                 .rpc_argp = &args,
5022                 .rpc_resp = &res,
5023         };
5024
5025         nfs_fattr_init(fsstat->fattr);
5026         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5027 }
5028
5029 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5030 {
5031         struct nfs4_exception exception = {
5032                 .interruptible = true,
5033         };
5034         int err;
5035         do {
5036                 err = nfs4_handle_exception(server,
5037                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5038                                 &exception);
5039         } while (exception.retry);
5040         return err;
5041 }
5042
5043 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5044                 struct nfs_fsinfo *fsinfo)
5045 {
5046         struct nfs4_fsinfo_arg args = {
5047                 .fh = fhandle,
5048                 .bitmask = server->attr_bitmask,
5049         };
5050         struct nfs4_fsinfo_res res = {
5051                 .fsinfo = fsinfo,
5052         };
5053         struct rpc_message msg = {
5054                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5055                 .rpc_argp = &args,
5056                 .rpc_resp = &res,
5057         };
5058
5059         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5060 }
5061
5062 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5063 {
5064         struct nfs4_exception exception = {
5065                 .interruptible = true,
5066         };
5067         int err;
5068
5069         do {
5070                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5071                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5072                 if (err == 0) {
5073                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5074                         break;
5075                 }
5076                 err = nfs4_handle_exception(server, err, &exception);
5077         } while (exception.retry);
5078         return err;
5079 }
5080
5081 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5082 {
5083         int error;
5084
5085         nfs_fattr_init(fsinfo->fattr);
5086         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5087         if (error == 0) {
5088                 /* block layout checks this! */
5089                 server->pnfs_blksize = fsinfo->blksize;
5090                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5091         }
5092
5093         return error;
5094 }
5095
5096 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5097                 struct nfs_pathconf *pathconf)
5098 {
5099         struct nfs4_pathconf_arg args = {
5100                 .fh = fhandle,
5101                 .bitmask = server->attr_bitmask,
5102         };
5103         struct nfs4_pathconf_res res = {
5104                 .pathconf = pathconf,
5105         };
5106         struct rpc_message msg = {
5107                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5108                 .rpc_argp = &args,
5109                 .rpc_resp = &res,
5110         };
5111
5112         /* None of the pathconf attributes are mandatory to implement */
5113         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5114                 memset(pathconf, 0, sizeof(*pathconf));
5115                 return 0;
5116         }
5117
5118         nfs_fattr_init(pathconf->fattr);
5119         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5120 }
5121
5122 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5123                 struct nfs_pathconf *pathconf)
5124 {
5125         struct nfs4_exception exception = {
5126                 .interruptible = true,
5127         };
5128         int err;
5129
5130         do {
5131                 err = nfs4_handle_exception(server,
5132                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5133                                 &exception);
5134         } while (exception.retry);
5135         return err;
5136 }
5137
5138 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5139                 const struct nfs_open_context *ctx,
5140                 const struct nfs_lock_context *l_ctx,
5141                 fmode_t fmode)
5142 {
5143         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5144 }
5145 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5146
5147 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5148                 const struct nfs_open_context *ctx,
5149                 const struct nfs_lock_context *l_ctx,
5150                 fmode_t fmode)
5151 {
5152         nfs4_stateid _current_stateid;
5153
5154         /* If the current stateid represents a lost lock, then exit */
5155         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5156                 return true;
5157         return nfs4_stateid_match(stateid, &_current_stateid);
5158 }
5159
5160 static bool nfs4_error_stateid_expired(int err)
5161 {
5162         switch (err) {
5163         case -NFS4ERR_DELEG_REVOKED:
5164         case -NFS4ERR_ADMIN_REVOKED:
5165         case -NFS4ERR_BAD_STATEID:
5166         case -NFS4ERR_STALE_STATEID:
5167         case -NFS4ERR_OLD_STATEID:
5168         case -NFS4ERR_OPENMODE:
5169         case -NFS4ERR_EXPIRED:
5170                 return true;
5171         }
5172         return false;
5173 }
5174
5175 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5176 {
5177         struct nfs_server *server = NFS_SERVER(hdr->inode);
5178
5179         trace_nfs4_read(hdr, task->tk_status);
5180         if (task->tk_status < 0) {
5181                 struct nfs4_exception exception = {
5182                         .inode = hdr->inode,
5183                         .state = hdr->args.context->state,
5184                         .stateid = &hdr->args.stateid,
5185                 };
5186                 task->tk_status = nfs4_async_handle_exception(task,
5187                                 server, task->tk_status, &exception);
5188                 if (exception.retry) {
5189                         rpc_restart_call_prepare(task);
5190                         return -EAGAIN;
5191                 }
5192         }
5193
5194         if (task->tk_status > 0)
5195                 renew_lease(server, hdr->timestamp);
5196         return 0;
5197 }
5198
5199 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5200                 struct nfs_pgio_args *args)
5201 {
5202
5203         if (!nfs4_error_stateid_expired(task->tk_status) ||
5204                 nfs4_stateid_is_current(&args->stateid,
5205                                 args->context,
5206                                 args->lock_context,
5207                                 FMODE_READ))
5208                 return false;
5209         rpc_restart_call_prepare(task);
5210         return true;
5211 }
5212
5213 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5214 {
5215
5216         dprintk("--> %s\n", __func__);
5217
5218         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5219                 return -EAGAIN;
5220         if (nfs4_read_stateid_changed(task, &hdr->args))
5221                 return -EAGAIN;
5222         if (task->tk_status > 0)
5223                 nfs_invalidate_atime(hdr->inode);
5224         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5225                                     nfs4_read_done_cb(task, hdr);
5226 }
5227
5228 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5229                                  struct rpc_message *msg)
5230 {
5231         hdr->timestamp   = jiffies;
5232         if (!hdr->pgio_done_cb)
5233                 hdr->pgio_done_cb = nfs4_read_done_cb;
5234         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5235         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5236 }
5237
5238 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5239                                       struct nfs_pgio_header *hdr)
5240 {
5241         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5242                         &hdr->args.seq_args,
5243                         &hdr->res.seq_res,
5244                         task))
5245                 return 0;
5246         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5247                                 hdr->args.lock_context,
5248                                 hdr->rw_mode) == -EIO)
5249                 return -EIO;
5250         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5251                 return -EIO;
5252         return 0;
5253 }
5254
5255 static int nfs4_write_done_cb(struct rpc_task *task,
5256                               struct nfs_pgio_header *hdr)
5257 {
5258         struct inode *inode = hdr->inode;
5259
5260         trace_nfs4_write(hdr, task->tk_status);
5261         if (task->tk_status < 0) {
5262                 struct nfs4_exception exception = {
5263                         .inode = hdr->inode,
5264                         .state = hdr->args.context->state,
5265                         .stateid = &hdr->args.stateid,
5266                 };
5267                 task->tk_status = nfs4_async_handle_exception(task,
5268                                 NFS_SERVER(inode), task->tk_status,
5269                                 &exception);
5270                 if (exception.retry) {
5271                         rpc_restart_call_prepare(task);
5272                         return -EAGAIN;
5273                 }
5274         }
5275         if (task->tk_status >= 0) {
5276                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5277                 nfs_writeback_update_inode(hdr);
5278         }
5279         return 0;
5280 }
5281
5282 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5283                 struct nfs_pgio_args *args)
5284 {
5285
5286         if (!nfs4_error_stateid_expired(task->tk_status) ||
5287                 nfs4_stateid_is_current(&args->stateid,
5288                                 args->context,
5289                                 args->lock_context,
5290                                 FMODE_WRITE))
5291                 return false;
5292         rpc_restart_call_prepare(task);
5293         return true;
5294 }
5295
5296 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5297 {
5298         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5299                 return -EAGAIN;
5300         if (nfs4_write_stateid_changed(task, &hdr->args))
5301                 return -EAGAIN;
5302         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5303                 nfs4_write_done_cb(task, hdr);
5304 }
5305
5306 static
5307 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5308 {
5309         /* Don't request attributes for pNFS or O_DIRECT writes */
5310         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5311                 return false;
5312         /* Otherwise, request attributes if and only if we don't hold
5313          * a delegation
5314          */
5315         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5316 }
5317
5318 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5319                                   struct rpc_message *msg,
5320                                   struct rpc_clnt **clnt)
5321 {
5322         struct nfs_server *server = NFS_SERVER(hdr->inode);
5323
5324         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5325                 hdr->args.bitmask = NULL;
5326                 hdr->res.fattr = NULL;
5327         } else
5328                 hdr->args.bitmask = server->cache_consistency_bitmask;
5329
5330         if (!hdr->pgio_done_cb)
5331                 hdr->pgio_done_cb = nfs4_write_done_cb;
5332         hdr->res.server = server;
5333         hdr->timestamp   = jiffies;
5334
5335         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5336         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5337         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5338 }
5339
5340 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5341 {
5342         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5343                         &data->args.seq_args,
5344                         &data->res.seq_res,
5345                         task);
5346 }
5347
5348 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5349 {
5350         struct inode *inode = data->inode;
5351
5352         trace_nfs4_commit(data, task->tk_status);
5353         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5354                                     NULL, NULL) == -EAGAIN) {
5355                 rpc_restart_call_prepare(task);
5356                 return -EAGAIN;
5357         }
5358         return 0;
5359 }
5360
5361 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5362 {
5363         if (!nfs4_sequence_done(task, &data->res.seq_res))
5364                 return -EAGAIN;
5365         return data->commit_done_cb(task, data);
5366 }
5367
5368 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5369                                    struct rpc_clnt **clnt)
5370 {
5371         struct nfs_server *server = NFS_SERVER(data->inode);
5372
5373         if (data->commit_done_cb == NULL)
5374                 data->commit_done_cb = nfs4_commit_done_cb;
5375         data->res.server = server;
5376         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5377         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5378         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5379 }
5380
5381 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5382                                 struct nfs_commitres *res)
5383 {
5384         struct inode *dst_inode = file_inode(dst);
5385         struct nfs_server *server = NFS_SERVER(dst_inode);
5386         struct rpc_message msg = {
5387                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5388                 .rpc_argp = args,
5389                 .rpc_resp = res,
5390         };
5391
5392         args->fh = NFS_FH(dst_inode);
5393         return nfs4_call_sync(server->client, server, &msg,
5394                         &args->seq_args, &res->seq_res, 1);
5395 }
5396
5397 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5398 {
5399         struct nfs_commitargs args = {
5400                 .offset = offset,
5401                 .count = count,
5402         };
5403         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5404         struct nfs4_exception exception = { };
5405         int status;
5406
5407         do {
5408                 status = _nfs4_proc_commit(dst, &args, res);
5409                 status = nfs4_handle_exception(dst_server, status, &exception);
5410         } while (exception.retry);
5411
5412         return status;
5413 }
5414
5415 struct nfs4_renewdata {
5416         struct nfs_client       *client;
5417         unsigned long           timestamp;
5418 };
5419
5420 /*
5421  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5422  * standalone procedure for queueing an asynchronous RENEW.
5423  */
5424 static void nfs4_renew_release(void *calldata)
5425 {
5426         struct nfs4_renewdata *data = calldata;
5427         struct nfs_client *clp = data->client;
5428
5429         if (refcount_read(&clp->cl_count) > 1)
5430                 nfs4_schedule_state_renewal(clp);
5431         nfs_put_client(clp);
5432         kfree(data);
5433 }
5434
5435 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5436 {
5437         struct nfs4_renewdata *data = calldata;
5438         struct nfs_client *clp = data->client;
5439         unsigned long timestamp = data->timestamp;
5440
5441         trace_nfs4_renew_async(clp, task->tk_status);
5442         switch (task->tk_status) {
5443         case 0:
5444                 break;
5445         case -NFS4ERR_LEASE_MOVED:
5446                 nfs4_schedule_lease_moved_recovery(clp);
5447                 break;
5448         default:
5449                 /* Unless we're shutting down, schedule state recovery! */
5450                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5451                         return;
5452                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5453                         nfs4_schedule_lease_recovery(clp);
5454                         return;
5455                 }
5456                 nfs4_schedule_path_down_recovery(clp);
5457         }
5458         do_renew_lease(clp, timestamp);
5459 }
5460
5461 static const struct rpc_call_ops nfs4_renew_ops = {
5462         .rpc_call_done = nfs4_renew_done,
5463         .rpc_release = nfs4_renew_release,
5464 };
5465
5466 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5467 {
5468         struct rpc_message msg = {
5469                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5470                 .rpc_argp       = clp,
5471                 .rpc_cred       = cred,
5472         };
5473         struct nfs4_renewdata *data;
5474
5475         if (renew_flags == 0)
5476                 return 0;
5477         if (!refcount_inc_not_zero(&clp->cl_count))
5478                 return -EIO;
5479         data = kmalloc(sizeof(*data), GFP_NOFS);
5480         if (data == NULL) {
5481                 nfs_put_client(clp);
5482                 return -ENOMEM;
5483         }
5484         data->client = clp;
5485         data->timestamp = jiffies;
5486         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5487                         &nfs4_renew_ops, data);
5488 }
5489
5490 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5491 {
5492         struct rpc_message msg = {
5493                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5494                 .rpc_argp       = clp,
5495                 .rpc_cred       = cred,
5496         };
5497         unsigned long now = jiffies;
5498         int status;
5499
5500         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5501         if (status < 0)
5502                 return status;
5503         do_renew_lease(clp, now);
5504         return 0;
5505 }
5506
5507 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5508 {
5509         return server->caps & NFS_CAP_ACLS;
5510 }
5511
5512 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5513  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5514  * the stack.
5515  */
5516 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5517
5518 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5519                 struct page **pages)
5520 {
5521         struct page *newpage, **spages;
5522         int rc = 0;
5523         size_t len;
5524         spages = pages;
5525
5526         do {
5527                 len = min_t(size_t, PAGE_SIZE, buflen);
5528                 newpage = alloc_page(GFP_KERNEL);
5529
5530                 if (newpage == NULL)
5531                         goto unwind;
5532                 memcpy(page_address(newpage), buf, len);
5533                 buf += len;
5534                 buflen -= len;
5535                 *pages++ = newpage;
5536                 rc++;
5537         } while (buflen != 0);
5538
5539         return rc;
5540
5541 unwind:
5542         for(; rc > 0; rc--)
5543                 __free_page(spages[rc-1]);
5544         return -ENOMEM;
5545 }
5546
5547 struct nfs4_cached_acl {
5548         int cached;
5549         size_t len;
5550         char data[];
5551 };
5552
5553 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5554 {
5555         struct nfs_inode *nfsi = NFS_I(inode);
5556
5557         spin_lock(&inode->i_lock);
5558         kfree(nfsi->nfs4_acl);
5559         nfsi->nfs4_acl = acl;
5560         spin_unlock(&inode->i_lock);
5561 }
5562
5563 static void nfs4_zap_acl_attr(struct inode *inode)
5564 {
5565         nfs4_set_cached_acl(inode, NULL);
5566 }
5567
5568 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5569 {
5570         struct nfs_inode *nfsi = NFS_I(inode);
5571         struct nfs4_cached_acl *acl;
5572         int ret = -ENOENT;
5573
5574         spin_lock(&inode->i_lock);
5575         acl = nfsi->nfs4_acl;
5576         if (acl == NULL)
5577                 goto out;
5578         if (buf == NULL) /* user is just asking for length */
5579                 goto out_len;
5580         if (acl->cached == 0)
5581                 goto out;
5582         ret = -ERANGE; /* see getxattr(2) man page */
5583         if (acl->len > buflen)
5584                 goto out;
5585         memcpy(buf, acl->data, acl->len);
5586 out_len:
5587         ret = acl->len;
5588 out:
5589         spin_unlock(&inode->i_lock);
5590         return ret;
5591 }
5592
5593 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5594 {
5595         struct nfs4_cached_acl *acl;
5596         size_t buflen = sizeof(*acl) + acl_len;
5597
5598         if (buflen <= PAGE_SIZE) {
5599                 acl = kmalloc(buflen, GFP_KERNEL);
5600                 if (acl == NULL)
5601                         goto out;
5602                 acl->cached = 1;
5603                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5604         } else {
5605                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5606                 if (acl == NULL)
5607                         goto out;
5608                 acl->cached = 0;
5609         }
5610         acl->len = acl_len;
5611 out:
5612         nfs4_set_cached_acl(inode, acl);
5613 }
5614
5615 /*
5616  * The getxattr API returns the required buffer length when called with a
5617  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5618  * the required buf.  On a NULL buf, we send a page of data to the server
5619  * guessing that the ACL request can be serviced by a page. If so, we cache
5620  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5621  * the cache. If not so, we throw away the page, and cache the required
5622  * length. The next getxattr call will then produce another round trip to
5623  * the server, this time with the input buf of the required size.
5624  */
5625 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5626 {
5627         struct page **pages;
5628         struct nfs_getaclargs args = {
5629                 .fh = NFS_FH(inode),
5630                 .acl_len = buflen,
5631         };
5632         struct nfs_getaclres res = {
5633                 .acl_len = buflen,
5634         };
5635         struct rpc_message msg = {
5636                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5637                 .rpc_argp = &args,
5638                 .rpc_resp = &res,
5639         };
5640         unsigned int npages;
5641         int ret = -ENOMEM, i;
5642         struct nfs_server *server = NFS_SERVER(inode);
5643
5644         if (buflen == 0)
5645                 buflen = server->rsize;
5646
5647         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5648         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5649         if (!pages)
5650                 return -ENOMEM;
5651
5652         args.acl_pages = pages;
5653
5654         for (i = 0; i < npages; i++) {
5655                 pages[i] = alloc_page(GFP_KERNEL);
5656                 if (!pages[i])
5657                         goto out_free;
5658         }
5659
5660         /* for decoding across pages */
5661         res.acl_scratch = alloc_page(GFP_KERNEL);
5662         if (!res.acl_scratch)
5663                 goto out_free;
5664
5665         args.acl_len = npages * PAGE_SIZE;
5666
5667         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5668                 __func__, buf, buflen, npages, args.acl_len);
5669         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5670                              &msg, &args.seq_args, &res.seq_res, 0);
5671         if (ret)
5672                 goto out_free;
5673
5674         /* Handle the case where the passed-in buffer is too short */
5675         if (res.acl_flags & NFS4_ACL_TRUNC) {
5676                 /* Did the user only issue a request for the acl length? */
5677                 if (buf == NULL)
5678                         goto out_ok;
5679                 ret = -ERANGE;
5680                 goto out_free;
5681         }
5682         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5683         if (buf) {
5684                 if (res.acl_len > buflen) {
5685                         ret = -ERANGE;
5686                         goto out_free;
5687                 }
5688                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5689         }
5690 out_ok:
5691         ret = res.acl_len;
5692 out_free:
5693         for (i = 0; i < npages; i++)
5694                 if (pages[i])
5695                         __free_page(pages[i]);
5696         if (res.acl_scratch)
5697                 __free_page(res.acl_scratch);
5698         kfree(pages);
5699         return ret;
5700 }
5701
5702 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5703 {
5704         struct nfs4_exception exception = {
5705                 .interruptible = true,
5706         };
5707         ssize_t ret;
5708         do {
5709                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5710                 trace_nfs4_get_acl(inode, ret);
5711                 if (ret >= 0)
5712                         break;
5713                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5714         } while (exception.retry);
5715         return ret;
5716 }
5717
5718 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5719 {
5720         struct nfs_server *server = NFS_SERVER(inode);
5721         int ret;
5722
5723         if (!nfs4_server_supports_acls(server))
5724                 return -EOPNOTSUPP;
5725         ret = nfs_revalidate_inode(server, inode);
5726         if (ret < 0)
5727                 return ret;
5728         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5729                 nfs_zap_acl_cache(inode);
5730         ret = nfs4_read_cached_acl(inode, buf, buflen);
5731         if (ret != -ENOENT)
5732                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5733                  * but no cached acl data, just the acl length */
5734                 return ret;
5735         return nfs4_get_acl_uncached(inode, buf, buflen);
5736 }
5737
5738 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5739 {
5740         struct nfs_server *server = NFS_SERVER(inode);
5741         struct page *pages[NFS4ACL_MAXPAGES];
5742         struct nfs_setaclargs arg = {
5743                 .fh             = NFS_FH(inode),
5744                 .acl_pages      = pages,
5745                 .acl_len        = buflen,
5746         };
5747         struct nfs_setaclres res;
5748         struct rpc_message msg = {
5749                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5750                 .rpc_argp       = &arg,
5751                 .rpc_resp       = &res,
5752         };
5753         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5754         int ret, i;
5755
5756         if (!nfs4_server_supports_acls(server))
5757                 return -EOPNOTSUPP;
5758         if (npages > ARRAY_SIZE(pages))
5759                 return -ERANGE;
5760         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5761         if (i < 0)
5762                 return i;
5763         nfs4_inode_make_writeable(inode);
5764         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5765
5766         /*
5767          * Free each page after tx, so the only ref left is
5768          * held by the network stack
5769          */
5770         for (; i > 0; i--)
5771                 put_page(pages[i-1]);
5772
5773         /*
5774          * Acl update can result in inode attribute update.
5775          * so mark the attribute cache invalid.
5776          */
5777         spin_lock(&inode->i_lock);
5778         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5779                 | NFS_INO_INVALID_CTIME
5780                 | NFS_INO_REVAL_FORCED;
5781         spin_unlock(&inode->i_lock);
5782         nfs_access_zap_cache(inode);
5783         nfs_zap_acl_cache(inode);
5784         return ret;
5785 }
5786
5787 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5788 {
5789         struct nfs4_exception exception = { };
5790         int err;
5791         do {
5792                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5793                 trace_nfs4_set_acl(inode, err);
5794                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5795                                 &exception);
5796         } while (exception.retry);
5797         return err;
5798 }
5799
5800 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5801 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5802                                         size_t buflen)
5803 {
5804         struct nfs_server *server = NFS_SERVER(inode);
5805         struct nfs_fattr fattr;
5806         struct nfs4_label label = {0, 0, buflen, buf};
5807
5808         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5809         struct nfs4_getattr_arg arg = {
5810                 .fh             = NFS_FH(inode),
5811                 .bitmask        = bitmask,
5812         };
5813         struct nfs4_getattr_res res = {
5814                 .fattr          = &fattr,
5815                 .label          = &label,
5816                 .server         = server,
5817         };
5818         struct rpc_message msg = {
5819                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5820                 .rpc_argp       = &arg,
5821                 .rpc_resp       = &res,
5822         };
5823         int ret;
5824
5825         nfs_fattr_init(&fattr);
5826
5827         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5828         if (ret)
5829                 return ret;
5830         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5831                 return -ENOENT;
5832         if (buflen < label.len)
5833                 return -ERANGE;
5834         return 0;
5835 }
5836
5837 static int nfs4_get_security_label(struct inode *inode, void *buf,
5838                                         size_t buflen)
5839 {
5840         struct nfs4_exception exception = {
5841                 .interruptible = true,
5842         };
5843         int err;
5844
5845         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5846                 return -EOPNOTSUPP;
5847
5848         do {
5849                 err = _nfs4_get_security_label(inode, buf, buflen);
5850                 trace_nfs4_get_security_label(inode, err);
5851                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5852                                 &exception);
5853         } while (exception.retry);
5854         return err;
5855 }
5856
5857 static int _nfs4_do_set_security_label(struct inode *inode,
5858                 struct nfs4_label *ilabel,
5859                 struct nfs_fattr *fattr,
5860                 struct nfs4_label *olabel)
5861 {
5862
5863         struct iattr sattr = {0};
5864         struct nfs_server *server = NFS_SERVER(inode);
5865         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5866         struct nfs_setattrargs arg = {
5867                 .fh             = NFS_FH(inode),
5868                 .iap            = &sattr,
5869                 .server         = server,
5870                 .bitmask        = bitmask,
5871                 .label          = ilabel,
5872         };
5873         struct nfs_setattrres res = {
5874                 .fattr          = fattr,
5875                 .label          = olabel,
5876                 .server         = server,
5877         };
5878         struct rpc_message msg = {
5879                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5880                 .rpc_argp       = &arg,
5881                 .rpc_resp       = &res,
5882         };
5883         int status;
5884
5885         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5886
5887         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5888         if (status)
5889                 dprintk("%s failed: %d\n", __func__, status);
5890
5891         return status;
5892 }
5893
5894 static int nfs4_do_set_security_label(struct inode *inode,
5895                 struct nfs4_label *ilabel,
5896                 struct nfs_fattr *fattr,
5897                 struct nfs4_label *olabel)
5898 {
5899         struct nfs4_exception exception = { };
5900         int err;
5901
5902         do {
5903                 err = _nfs4_do_set_security_label(inode, ilabel,
5904                                 fattr, olabel);
5905                 trace_nfs4_set_security_label(inode, err);
5906                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5907                                 &exception);
5908         } while (exception.retry);
5909         return err;
5910 }
5911
5912 static int
5913 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5914 {
5915         struct nfs4_label ilabel, *olabel = NULL;
5916         struct nfs_fattr fattr;
5917         int status;
5918
5919         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5920                 return -EOPNOTSUPP;
5921
5922         nfs_fattr_init(&fattr);
5923
5924         ilabel.pi = 0;
5925         ilabel.lfs = 0;
5926         ilabel.label = (char *)buf;
5927         ilabel.len = buflen;
5928
5929         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5930         if (IS_ERR(olabel)) {
5931                 status = -PTR_ERR(olabel);
5932                 goto out;
5933         }
5934
5935         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5936         if (status == 0)
5937                 nfs_setsecurity(inode, &fattr, olabel);
5938
5939         nfs4_label_free(olabel);
5940 out:
5941         return status;
5942 }
5943 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5944
5945
5946 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5947                                     nfs4_verifier *bootverf)
5948 {
5949         __be32 verf[2];
5950
5951         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5952                 /* An impossible timestamp guarantees this value
5953                  * will never match a generated boot time. */
5954                 verf[0] = cpu_to_be32(U32_MAX);
5955                 verf[1] = cpu_to_be32(U32_MAX);
5956         } else {
5957                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5958                 u64 ns = ktime_to_ns(nn->boot_time);
5959
5960                 verf[0] = cpu_to_be32(ns >> 32);
5961                 verf[1] = cpu_to_be32(ns);
5962         }
5963         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5964 }
5965
5966 static int
5967 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5968 {
5969         size_t len;
5970         char *str;
5971
5972         if (clp->cl_owner_id != NULL)
5973                 return 0;
5974
5975         rcu_read_lock();
5976         len = 14 +
5977                 strlen(clp->cl_rpcclient->cl_nodename) +
5978                 1 +
5979                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5980                 1;
5981         rcu_read_unlock();
5982         if (nfs4_client_id_uniquifier[0] != '\0')
5983                 len += strlen(nfs4_client_id_uniquifier) + 1;
5984         if (len > NFS4_OPAQUE_LIMIT + 1)
5985                 return -EINVAL;
5986
5987         /*
5988          * Since this string is allocated at mount time, and held until the
5989          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5990          * about a memory-reclaim deadlock.
5991          */
5992         str = kmalloc(len, GFP_KERNEL);
5993         if (!str)
5994                 return -ENOMEM;
5995
5996         rcu_read_lock();
5997         if (nfs4_client_id_uniquifier[0] != '\0')
5998                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5999                           clp->cl_rpcclient->cl_nodename,
6000                           nfs4_client_id_uniquifier,
6001                           rpc_peeraddr2str(clp->cl_rpcclient,
6002                                            RPC_DISPLAY_ADDR));
6003         else
6004                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6005                           clp->cl_rpcclient->cl_nodename,
6006                           rpc_peeraddr2str(clp->cl_rpcclient,
6007                                            RPC_DISPLAY_ADDR));
6008         rcu_read_unlock();
6009
6010         clp->cl_owner_id = str;
6011         return 0;
6012 }
6013
6014 static int
6015 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
6016 {
6017         size_t len;
6018         char *str;
6019
6020         len = 10 + 10 + 1 + 10 + 1 +
6021                 strlen(nfs4_client_id_uniquifier) + 1 +
6022                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6023
6024         if (len > NFS4_OPAQUE_LIMIT + 1)
6025                 return -EINVAL;
6026
6027         /*
6028          * Since this string is allocated at mount time, and held until the
6029          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6030          * about a memory-reclaim deadlock.
6031          */
6032         str = kmalloc(len, GFP_KERNEL);
6033         if (!str)
6034                 return -ENOMEM;
6035
6036         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6037                         clp->rpc_ops->version, clp->cl_minorversion,
6038                         nfs4_client_id_uniquifier,
6039                         clp->cl_rpcclient->cl_nodename);
6040         clp->cl_owner_id = str;
6041         return 0;
6042 }
6043
6044 static int
6045 nfs4_init_uniform_client_string(struct nfs_client *clp)
6046 {
6047         size_t len;
6048         char *str;
6049
6050         if (clp->cl_owner_id != NULL)
6051                 return 0;
6052
6053         if (nfs4_client_id_uniquifier[0] != '\0')
6054                 return nfs4_init_uniquifier_client_string(clp);
6055
6056         len = 10 + 10 + 1 + 10 + 1 +
6057                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6058
6059         if (len > NFS4_OPAQUE_LIMIT + 1)
6060                 return -EINVAL;
6061
6062         /*
6063          * Since this string is allocated at mount time, and held until the
6064          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6065          * about a memory-reclaim deadlock.
6066          */
6067         str = kmalloc(len, GFP_KERNEL);
6068         if (!str)
6069                 return -ENOMEM;
6070
6071         scnprintf(str, len, "Linux NFSv%u.%u %s",
6072                         clp->rpc_ops->version, clp->cl_minorversion,
6073                         clp->cl_rpcclient->cl_nodename);
6074         clp->cl_owner_id = str;
6075         return 0;
6076 }
6077
6078 /*
6079  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6080  * services.  Advertise one based on the address family of the
6081  * clientaddr.
6082  */
6083 static unsigned int
6084 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6085 {
6086         if (strchr(clp->cl_ipaddr, ':') != NULL)
6087                 return scnprintf(buf, len, "tcp6");
6088         else
6089                 return scnprintf(buf, len, "tcp");
6090 }
6091
6092 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6093 {
6094         struct nfs4_setclientid *sc = calldata;
6095
6096         if (task->tk_status == 0)
6097                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6098 }
6099
6100 static const struct rpc_call_ops nfs4_setclientid_ops = {
6101         .rpc_call_done = nfs4_setclientid_done,
6102 };
6103
6104 /**
6105  * nfs4_proc_setclientid - Negotiate client ID
6106  * @clp: state data structure
6107  * @program: RPC program for NFSv4 callback service
6108  * @port: IP port number for NFS4 callback service
6109  * @cred: credential to use for this call
6110  * @res: where to place the result
6111  *
6112  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6113  */
6114 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6115                 unsigned short port, const struct cred *cred,
6116                 struct nfs4_setclientid_res *res)
6117 {
6118         nfs4_verifier sc_verifier;
6119         struct nfs4_setclientid setclientid = {
6120                 .sc_verifier = &sc_verifier,
6121                 .sc_prog = program,
6122                 .sc_clnt = clp,
6123         };
6124         struct rpc_message msg = {
6125                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6126                 .rpc_argp = &setclientid,
6127                 .rpc_resp = res,
6128                 .rpc_cred = cred,
6129         };
6130         struct rpc_task_setup task_setup_data = {
6131                 .rpc_client = clp->cl_rpcclient,
6132                 .rpc_message = &msg,
6133                 .callback_ops = &nfs4_setclientid_ops,
6134                 .callback_data = &setclientid,
6135                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6136         };
6137         unsigned long now = jiffies;
6138         int status;
6139
6140         /* nfs_client_id4 */
6141         nfs4_init_boot_verifier(clp, &sc_verifier);
6142
6143         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6144                 status = nfs4_init_uniform_client_string(clp);
6145         else
6146                 status = nfs4_init_nonuniform_client_string(clp);
6147
6148         if (status)
6149                 goto out;
6150
6151         /* cb_client4 */
6152         setclientid.sc_netid_len =
6153                                 nfs4_init_callback_netid(clp,
6154                                                 setclientid.sc_netid,
6155                                                 sizeof(setclientid.sc_netid));
6156         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6157                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6158                                 clp->cl_ipaddr, port >> 8, port & 255);
6159
6160         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6161                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6162                 clp->cl_owner_id);
6163
6164         status = nfs4_call_sync_custom(&task_setup_data);
6165         if (setclientid.sc_cred) {
6166                 kfree(clp->cl_acceptor);
6167                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6168                 put_rpccred(setclientid.sc_cred);
6169         }
6170
6171         if (status == 0)
6172                 do_renew_lease(clp, now);
6173 out:
6174         trace_nfs4_setclientid(clp, status);
6175         dprintk("NFS reply setclientid: %d\n", status);
6176         return status;
6177 }
6178
6179 /**
6180  * nfs4_proc_setclientid_confirm - Confirm client ID
6181  * @clp: state data structure
6182  * @arg: result of a previous SETCLIENTID
6183  * @cred: credential to use for this call
6184  *
6185  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6186  */
6187 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6188                 struct nfs4_setclientid_res *arg,
6189                 const struct cred *cred)
6190 {
6191         struct rpc_message msg = {
6192                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6193                 .rpc_argp = arg,
6194                 .rpc_cred = cred,
6195         };
6196         int status;
6197
6198         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6199                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6200                 clp->cl_clientid);
6201         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6202                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6203         trace_nfs4_setclientid_confirm(clp, status);
6204         dprintk("NFS reply setclientid_confirm: %d\n", status);
6205         return status;
6206 }
6207
6208 struct nfs4_delegreturndata {
6209         struct nfs4_delegreturnargs args;
6210         struct nfs4_delegreturnres res;
6211         struct nfs_fh fh;
6212         nfs4_stateid stateid;
6213         unsigned long timestamp;
6214         struct {
6215                 struct nfs4_layoutreturn_args arg;
6216                 struct nfs4_layoutreturn_res res;
6217                 struct nfs4_xdr_opaque_data ld_private;
6218                 u32 roc_barrier;
6219                 bool roc;
6220         } lr;
6221         struct nfs_fattr fattr;
6222         int rpc_status;
6223         struct inode *inode;
6224 };
6225
6226 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6227 {
6228         struct nfs4_delegreturndata *data = calldata;
6229         struct nfs4_exception exception = {
6230                 .inode = data->inode,
6231                 .stateid = &data->stateid,
6232         };
6233
6234         if (!nfs4_sequence_done(task, &data->res.seq_res))
6235                 return;
6236
6237         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6238
6239         /* Handle Layoutreturn errors */
6240         if (pnfs_roc_done(task, data->inode,
6241                                 &data->args.lr_args,
6242                                 &data->res.lr_res,
6243                                 &data->res.lr_ret) == -EAGAIN)
6244                 goto out_restart;
6245
6246         switch (task->tk_status) {
6247         case 0:
6248                 renew_lease(data->res.server, data->timestamp);
6249                 break;
6250         case -NFS4ERR_ADMIN_REVOKED:
6251         case -NFS4ERR_DELEG_REVOKED:
6252         case -NFS4ERR_EXPIRED:
6253                 nfs4_free_revoked_stateid(data->res.server,
6254                                 data->args.stateid,
6255                                 task->tk_msg.rpc_cred);
6256                 /* Fallthrough */
6257         case -NFS4ERR_BAD_STATEID:
6258         case -NFS4ERR_STALE_STATEID:
6259         case -ETIMEDOUT:
6260                 task->tk_status = 0;
6261                 break;
6262         case -NFS4ERR_OLD_STATEID:
6263                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6264                         nfs4_stateid_seqid_inc(&data->stateid);
6265                 if (data->args.bitmask) {
6266                         data->args.bitmask = NULL;
6267                         data->res.fattr = NULL;
6268                 }
6269                 goto out_restart;
6270         case -NFS4ERR_ACCESS:
6271                 if (data->args.bitmask) {
6272                         data->args.bitmask = NULL;
6273                         data->res.fattr = NULL;
6274                         goto out_restart;
6275                 }
6276                 /* Fallthrough */
6277         default:
6278                 task->tk_status = nfs4_async_handle_exception(task,
6279                                 data->res.server, task->tk_status,
6280                                 &exception);
6281                 if (exception.retry)
6282                         goto out_restart;
6283         }
6284         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6285         data->rpc_status = task->tk_status;
6286         return;
6287 out_restart:
6288         task->tk_status = 0;
6289         rpc_restart_call_prepare(task);
6290 }
6291
6292 static void nfs4_delegreturn_release(void *calldata)
6293 {
6294         struct nfs4_delegreturndata *data = calldata;
6295         struct inode *inode = data->inode;
6296
6297         if (inode) {
6298                 if (data->lr.roc)
6299                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6300                                         data->res.lr_ret);
6301                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6302                 nfs_iput_and_deactive(inode);
6303         }
6304         kfree(calldata);
6305 }
6306
6307 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6308 {
6309         struct nfs4_delegreturndata *d_data;
6310         struct pnfs_layout_hdr *lo;
6311
6312         d_data = (struct nfs4_delegreturndata *)data;
6313
6314         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6315                 nfs4_sequence_done(task, &d_data->res.seq_res);
6316                 return;
6317         }
6318
6319         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6320         if (lo && !pnfs_layout_is_valid(lo)) {
6321                 d_data->args.lr_args = NULL;
6322                 d_data->res.lr_res = NULL;
6323         }
6324
6325         nfs4_setup_sequence(d_data->res.server->nfs_client,
6326                         &d_data->args.seq_args,
6327                         &d_data->res.seq_res,
6328                         task);
6329 }
6330
6331 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6332         .rpc_call_prepare = nfs4_delegreturn_prepare,
6333         .rpc_call_done = nfs4_delegreturn_done,
6334         .rpc_release = nfs4_delegreturn_release,
6335 };
6336
6337 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6338 {
6339         struct nfs4_delegreturndata *data;
6340         struct nfs_server *server = NFS_SERVER(inode);
6341         struct rpc_task *task;
6342         struct rpc_message msg = {
6343                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6344                 .rpc_cred = cred,
6345         };
6346         struct rpc_task_setup task_setup_data = {
6347                 .rpc_client = server->client,
6348                 .rpc_message = &msg,
6349                 .callback_ops = &nfs4_delegreturn_ops,
6350                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF | RPC_TASK_TIMEOUT,
6351         };
6352         int status = 0;
6353
6354         data = kzalloc(sizeof(*data), GFP_NOFS);
6355         if (data == NULL)
6356                 return -ENOMEM;
6357         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6358
6359         nfs4_state_protect(server->nfs_client,
6360                         NFS_SP4_MACH_CRED_CLEANUP,
6361                         &task_setup_data.rpc_client, &msg);
6362
6363         data->args.fhandle = &data->fh;
6364         data->args.stateid = &data->stateid;
6365         data->args.bitmask = server->cache_consistency_bitmask;
6366         nfs_copy_fh(&data->fh, NFS_FH(inode));
6367         nfs4_stateid_copy(&data->stateid, stateid);
6368         data->res.fattr = &data->fattr;
6369         data->res.server = server;
6370         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6371         data->lr.arg.ld_private = &data->lr.ld_private;
6372         nfs_fattr_init(data->res.fattr);
6373         data->timestamp = jiffies;
6374         data->rpc_status = 0;
6375         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6376         data->inode = nfs_igrab_and_active(inode);
6377         if (data->inode) {
6378                 if (data->lr.roc) {
6379                         data->args.lr_args = &data->lr.arg;
6380                         data->res.lr_res = &data->lr.res;
6381                 }
6382         } else if (data->lr.roc) {
6383                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6384                 data->lr.roc = false;
6385         }
6386
6387         task_setup_data.callback_data = data;
6388         msg.rpc_argp = &data->args;
6389         msg.rpc_resp = &data->res;
6390         task = rpc_run_task(&task_setup_data);
6391         if (IS_ERR(task))
6392                 return PTR_ERR(task);
6393         if (!issync)
6394                 goto out;
6395         status = rpc_wait_for_completion_task(task);
6396         if (status != 0)
6397                 goto out;
6398         status = data->rpc_status;
6399 out:
6400         rpc_put_task(task);
6401         return status;
6402 }
6403
6404 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6405 {
6406         struct nfs_server *server = NFS_SERVER(inode);
6407         struct nfs4_exception exception = { };
6408         int err;
6409         do {
6410                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6411                 trace_nfs4_delegreturn(inode, stateid, err);
6412                 switch (err) {
6413                         case -NFS4ERR_STALE_STATEID:
6414                         case -NFS4ERR_EXPIRED:
6415                         case 0:
6416                                 return 0;
6417                 }
6418                 err = nfs4_handle_exception(server, err, &exception);
6419         } while (exception.retry);
6420         return err;
6421 }
6422
6423 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6424 {
6425         struct inode *inode = state->inode;
6426         struct nfs_server *server = NFS_SERVER(inode);
6427         struct nfs_client *clp = server->nfs_client;
6428         struct nfs_lockt_args arg = {
6429                 .fh = NFS_FH(inode),
6430                 .fl = request,
6431         };
6432         struct nfs_lockt_res res = {
6433                 .denied = request,
6434         };
6435         struct rpc_message msg = {
6436                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6437                 .rpc_argp       = &arg,
6438                 .rpc_resp       = &res,
6439                 .rpc_cred       = state->owner->so_cred,
6440         };
6441         struct nfs4_lock_state *lsp;
6442         int status;
6443
6444         arg.lock_owner.clientid = clp->cl_clientid;
6445         status = nfs4_set_lock_state(state, request);
6446         if (status != 0)
6447                 goto out;
6448         lsp = request->fl_u.nfs4_fl.owner;
6449         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6450         arg.lock_owner.s_dev = server->s_dev;
6451         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6452         switch (status) {
6453                 case 0:
6454                         request->fl_type = F_UNLCK;
6455                         break;
6456                 case -NFS4ERR_DENIED:
6457                         status = 0;
6458         }
6459         request->fl_ops->fl_release_private(request);
6460         request->fl_ops = NULL;
6461 out:
6462         return status;
6463 }
6464
6465 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6466 {
6467         struct nfs4_exception exception = {
6468                 .interruptible = true,
6469         };
6470         int err;
6471
6472         do {
6473                 err = _nfs4_proc_getlk(state, cmd, request);
6474                 trace_nfs4_get_lock(request, state, cmd, err);
6475                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6476                                 &exception);
6477         } while (exception.retry);
6478         return err;
6479 }
6480
6481 /*
6482  * Update the seqid of a lock stateid after receiving
6483  * NFS4ERR_OLD_STATEID
6484  */
6485 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6486                 struct nfs4_lock_state *lsp)
6487 {
6488         struct nfs4_state *state = lsp->ls_state;
6489         bool ret = false;
6490
6491         spin_lock(&state->state_lock);
6492         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6493                 goto out;
6494         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6495                 nfs4_stateid_seqid_inc(dst);
6496         else
6497                 dst->seqid = lsp->ls_stateid.seqid;
6498         ret = true;
6499 out:
6500         spin_unlock(&state->state_lock);
6501         return ret;
6502 }
6503
6504 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6505                 struct nfs4_lock_state *lsp)
6506 {
6507         struct nfs4_state *state = lsp->ls_state;
6508         bool ret;
6509
6510         spin_lock(&state->state_lock);
6511         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6512         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6513         spin_unlock(&state->state_lock);
6514         return ret;
6515 }
6516
6517 struct nfs4_unlockdata {
6518         struct nfs_locku_args arg;
6519         struct nfs_locku_res res;
6520         struct nfs4_lock_state *lsp;
6521         struct nfs_open_context *ctx;
6522         struct nfs_lock_context *l_ctx;
6523         struct file_lock fl;
6524         struct nfs_server *server;
6525         unsigned long timestamp;
6526 };
6527
6528 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6529                 struct nfs_open_context *ctx,
6530                 struct nfs4_lock_state *lsp,
6531                 struct nfs_seqid *seqid)
6532 {
6533         struct nfs4_unlockdata *p;
6534         struct nfs4_state *state = lsp->ls_state;
6535         struct inode *inode = state->inode;
6536
6537         p = kzalloc(sizeof(*p), GFP_NOFS);
6538         if (p == NULL)
6539                 return NULL;
6540         p->arg.fh = NFS_FH(inode);
6541         p->arg.fl = &p->fl;
6542         p->arg.seqid = seqid;
6543         p->res.seqid = seqid;
6544         p->lsp = lsp;
6545         /* Ensure we don't close file until we're done freeing locks! */
6546         p->ctx = get_nfs_open_context(ctx);
6547         p->l_ctx = nfs_get_lock_context(ctx);
6548         locks_init_lock(&p->fl);
6549         locks_copy_lock(&p->fl, fl);
6550         p->server = NFS_SERVER(inode);
6551         spin_lock(&state->state_lock);
6552         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6553         spin_unlock(&state->state_lock);
6554         return p;
6555 }
6556
6557 static void nfs4_locku_release_calldata(void *data)
6558 {
6559         struct nfs4_unlockdata *calldata = data;
6560         nfs_free_seqid(calldata->arg.seqid);
6561         nfs4_put_lock_state(calldata->lsp);
6562         nfs_put_lock_context(calldata->l_ctx);
6563         put_nfs_open_context(calldata->ctx);
6564         kfree(calldata);
6565 }
6566
6567 static void nfs4_locku_done(struct rpc_task *task, void *data)
6568 {
6569         struct nfs4_unlockdata *calldata = data;
6570         struct nfs4_exception exception = {
6571                 .inode = calldata->lsp->ls_state->inode,
6572                 .stateid = &calldata->arg.stateid,
6573         };
6574
6575         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6576                 return;
6577         switch (task->tk_status) {
6578                 case 0:
6579                         renew_lease(calldata->server, calldata->timestamp);
6580                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6581                         if (nfs4_update_lock_stateid(calldata->lsp,
6582                                         &calldata->res.stateid))
6583                                 break;
6584                         /* Fall through */
6585                 case -NFS4ERR_ADMIN_REVOKED:
6586                 case -NFS4ERR_EXPIRED:
6587                         nfs4_free_revoked_stateid(calldata->server,
6588                                         &calldata->arg.stateid,
6589                                         task->tk_msg.rpc_cred);
6590                         /* Fall through */
6591                 case -NFS4ERR_BAD_STATEID:
6592                 case -NFS4ERR_STALE_STATEID:
6593                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6594                                                 calldata->lsp))
6595                                 rpc_restart_call_prepare(task);
6596                         break;
6597                 case -NFS4ERR_OLD_STATEID:
6598                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6599                                                 calldata->lsp))
6600                                 rpc_restart_call_prepare(task);
6601                         break;
6602                 default:
6603                         task->tk_status = nfs4_async_handle_exception(task,
6604                                         calldata->server, task->tk_status,
6605                                         &exception);
6606                         if (exception.retry)
6607                                 rpc_restart_call_prepare(task);
6608         }
6609         nfs_release_seqid(calldata->arg.seqid);
6610 }
6611
6612 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6613 {
6614         struct nfs4_unlockdata *calldata = data;
6615
6616         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6617                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6618                 return;
6619
6620         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6621                 goto out_wait;
6622         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6623                 /* Note: exit _without_ running nfs4_locku_done */
6624                 goto out_no_action;
6625         }
6626         calldata->timestamp = jiffies;
6627         if (nfs4_setup_sequence(calldata->server->nfs_client,
6628                                 &calldata->arg.seq_args,
6629                                 &calldata->res.seq_res,
6630                                 task) != 0)
6631                 nfs_release_seqid(calldata->arg.seqid);
6632         return;
6633 out_no_action:
6634         task->tk_action = NULL;
6635 out_wait:
6636         nfs4_sequence_done(task, &calldata->res.seq_res);
6637 }
6638
6639 static const struct rpc_call_ops nfs4_locku_ops = {
6640         .rpc_call_prepare = nfs4_locku_prepare,
6641         .rpc_call_done = nfs4_locku_done,
6642         .rpc_release = nfs4_locku_release_calldata,
6643 };
6644
6645 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6646                 struct nfs_open_context *ctx,
6647                 struct nfs4_lock_state *lsp,
6648                 struct nfs_seqid *seqid)
6649 {
6650         struct nfs4_unlockdata *data;
6651         struct rpc_message msg = {
6652                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6653                 .rpc_cred = ctx->cred,
6654         };
6655         struct rpc_task_setup task_setup_data = {
6656                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6657                 .rpc_message = &msg,
6658                 .callback_ops = &nfs4_locku_ops,
6659                 .workqueue = nfsiod_workqueue,
6660                 .flags = RPC_TASK_ASYNC,
6661         };
6662
6663         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6664                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6665
6666         /* Ensure this is an unlock - when canceling a lock, the
6667          * canceled lock is passed in, and it won't be an unlock.
6668          */
6669         fl->fl_type = F_UNLCK;
6670         if (fl->fl_flags & FL_CLOSE)
6671                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6672
6673         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6674         if (data == NULL) {
6675                 nfs_free_seqid(seqid);
6676                 return ERR_PTR(-ENOMEM);
6677         }
6678
6679         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6680         msg.rpc_argp = &data->arg;
6681         msg.rpc_resp = &data->res;
6682         task_setup_data.callback_data = data;
6683         return rpc_run_task(&task_setup_data);
6684 }
6685
6686 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6687 {
6688         struct inode *inode = state->inode;
6689         struct nfs4_state_owner *sp = state->owner;
6690         struct nfs_inode *nfsi = NFS_I(inode);
6691         struct nfs_seqid *seqid;
6692         struct nfs4_lock_state *lsp;
6693         struct rpc_task *task;
6694         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6695         int status = 0;
6696         unsigned char fl_flags = request->fl_flags;
6697
6698         status = nfs4_set_lock_state(state, request);
6699         /* Unlock _before_ we do the RPC call */
6700         request->fl_flags |= FL_EXISTS;
6701         /* Exclude nfs_delegation_claim_locks() */
6702         mutex_lock(&sp->so_delegreturn_mutex);
6703         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6704         down_read(&nfsi->rwsem);
6705         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6706                 up_read(&nfsi->rwsem);
6707                 mutex_unlock(&sp->so_delegreturn_mutex);
6708                 goto out;
6709         }
6710         up_read(&nfsi->rwsem);
6711         mutex_unlock(&sp->so_delegreturn_mutex);
6712         if (status != 0)
6713                 goto out;
6714         /* Is this a delegated lock? */
6715         lsp = request->fl_u.nfs4_fl.owner;
6716         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6717                 goto out;
6718         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6719         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6720         status = -ENOMEM;
6721         if (IS_ERR(seqid))
6722                 goto out;
6723         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6724         status = PTR_ERR(task);
6725         if (IS_ERR(task))
6726                 goto out;
6727         status = rpc_wait_for_completion_task(task);
6728         rpc_put_task(task);
6729 out:
6730         request->fl_flags = fl_flags;
6731         trace_nfs4_unlock(request, state, F_SETLK, status);
6732         return status;
6733 }
6734
6735 struct nfs4_lockdata {
6736         struct nfs_lock_args arg;
6737         struct nfs_lock_res res;
6738         struct nfs4_lock_state *lsp;
6739         struct nfs_open_context *ctx;
6740         struct file_lock fl;
6741         unsigned long timestamp;
6742         int rpc_status;
6743         int cancelled;
6744         struct nfs_server *server;
6745 };
6746
6747 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6748                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6749                 gfp_t gfp_mask)
6750 {
6751         struct nfs4_lockdata *p;
6752         struct inode *inode = lsp->ls_state->inode;
6753         struct nfs_server *server = NFS_SERVER(inode);
6754         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6755
6756         p = kzalloc(sizeof(*p), gfp_mask);
6757         if (p == NULL)
6758                 return NULL;
6759
6760         p->arg.fh = NFS_FH(inode);
6761         p->arg.fl = &p->fl;
6762         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6763         if (IS_ERR(p->arg.open_seqid))
6764                 goto out_free;
6765         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6766         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6767         if (IS_ERR(p->arg.lock_seqid))
6768                 goto out_free_seqid;
6769         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6770         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6771         p->arg.lock_owner.s_dev = server->s_dev;
6772         p->res.lock_seqid = p->arg.lock_seqid;
6773         p->lsp = lsp;
6774         p->server = server;
6775         p->ctx = get_nfs_open_context(ctx);
6776         locks_init_lock(&p->fl);
6777         locks_copy_lock(&p->fl, fl);
6778         return p;
6779 out_free_seqid:
6780         nfs_free_seqid(p->arg.open_seqid);
6781 out_free:
6782         kfree(p);
6783         return NULL;
6784 }
6785
6786 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6787 {
6788         struct nfs4_lockdata *data = calldata;
6789         struct nfs4_state *state = data->lsp->ls_state;
6790
6791         dprintk("%s: begin!\n", __func__);
6792         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6793                 goto out_wait;
6794         /* Do we need to do an open_to_lock_owner? */
6795         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6796                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6797                         goto out_release_lock_seqid;
6798                 }
6799                 nfs4_stateid_copy(&data->arg.open_stateid,
6800                                 &state->open_stateid);
6801                 data->arg.new_lock_owner = 1;
6802                 data->res.open_seqid = data->arg.open_seqid;
6803         } else {
6804                 data->arg.new_lock_owner = 0;
6805                 nfs4_stateid_copy(&data->arg.lock_stateid,
6806                                 &data->lsp->ls_stateid);
6807         }
6808         if (!nfs4_valid_open_stateid(state)) {
6809                 data->rpc_status = -EBADF;
6810                 task->tk_action = NULL;
6811                 goto out_release_open_seqid;
6812         }
6813         data->timestamp = jiffies;
6814         if (nfs4_setup_sequence(data->server->nfs_client,
6815                                 &data->arg.seq_args,
6816                                 &data->res.seq_res,
6817                                 task) == 0)
6818                 return;
6819 out_release_open_seqid:
6820         nfs_release_seqid(data->arg.open_seqid);
6821 out_release_lock_seqid:
6822         nfs_release_seqid(data->arg.lock_seqid);
6823 out_wait:
6824         nfs4_sequence_done(task, &data->res.seq_res);
6825         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6826 }
6827
6828 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6829 {
6830         struct nfs4_lockdata *data = calldata;
6831         struct nfs4_lock_state *lsp = data->lsp;
6832
6833         dprintk("%s: begin!\n", __func__);
6834
6835         if (!nfs4_sequence_done(task, &data->res.seq_res))
6836                 return;
6837
6838         data->rpc_status = task->tk_status;
6839         switch (task->tk_status) {
6840         case 0:
6841                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6842                                 data->timestamp);
6843                 if (data->arg.new_lock && !data->cancelled) {
6844                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6845                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6846                                 goto out_restart;
6847                 }
6848                 if (data->arg.new_lock_owner != 0) {
6849                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6850                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6851                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6852                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6853                         goto out_restart;
6854                 break;
6855         case -NFS4ERR_BAD_STATEID:
6856         case -NFS4ERR_OLD_STATEID:
6857         case -NFS4ERR_STALE_STATEID:
6858         case -NFS4ERR_EXPIRED:
6859                 if (data->arg.new_lock_owner != 0) {
6860                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6861                                                 &lsp->ls_state->open_stateid))
6862                                 goto out_restart;
6863                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6864                                                 &lsp->ls_stateid))
6865                                 goto out_restart;
6866         }
6867 out_done:
6868         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6869         return;
6870 out_restart:
6871         if (!data->cancelled)
6872                 rpc_restart_call_prepare(task);
6873         goto out_done;
6874 }
6875
6876 static void nfs4_lock_release(void *calldata)
6877 {
6878         struct nfs4_lockdata *data = calldata;
6879
6880         dprintk("%s: begin!\n", __func__);
6881         nfs_free_seqid(data->arg.open_seqid);
6882         if (data->cancelled && data->rpc_status == 0) {
6883                 struct rpc_task *task;
6884                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6885                                 data->arg.lock_seqid);
6886                 if (!IS_ERR(task))
6887                         rpc_put_task_async(task);
6888                 dprintk("%s: cancelling lock!\n", __func__);
6889         } else
6890                 nfs_free_seqid(data->arg.lock_seqid);
6891         nfs4_put_lock_state(data->lsp);
6892         put_nfs_open_context(data->ctx);
6893         kfree(data);
6894         dprintk("%s: done!\n", __func__);
6895 }
6896
6897 static const struct rpc_call_ops nfs4_lock_ops = {
6898         .rpc_call_prepare = nfs4_lock_prepare,
6899         .rpc_call_done = nfs4_lock_done,
6900         .rpc_release = nfs4_lock_release,
6901 };
6902
6903 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6904 {
6905         switch (error) {
6906         case -NFS4ERR_ADMIN_REVOKED:
6907         case -NFS4ERR_EXPIRED:
6908         case -NFS4ERR_BAD_STATEID:
6909                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6910                 if (new_lock_owner != 0 ||
6911                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6912                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6913                 break;
6914         case -NFS4ERR_STALE_STATEID:
6915                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6916                 nfs4_schedule_lease_recovery(server->nfs_client);
6917         }
6918 }
6919
6920 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6921 {
6922         struct nfs4_lockdata *data;
6923         struct rpc_task *task;
6924         struct rpc_message msg = {
6925                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6926                 .rpc_cred = state->owner->so_cred,
6927         };
6928         struct rpc_task_setup task_setup_data = {
6929                 .rpc_client = NFS_CLIENT(state->inode),
6930                 .rpc_message = &msg,
6931                 .callback_ops = &nfs4_lock_ops,
6932                 .workqueue = nfsiod_workqueue,
6933                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
6934         };
6935         int ret;
6936
6937         dprintk("%s: begin!\n", __func__);
6938         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6939                         fl->fl_u.nfs4_fl.owner,
6940                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6941         if (data == NULL)
6942                 return -ENOMEM;
6943         if (IS_SETLKW(cmd))
6944                 data->arg.block = 1;
6945         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6946                                 recovery_type > NFS_LOCK_NEW);
6947         msg.rpc_argp = &data->arg;
6948         msg.rpc_resp = &data->res;
6949         task_setup_data.callback_data = data;
6950         if (recovery_type > NFS_LOCK_NEW) {
6951                 if (recovery_type == NFS_LOCK_RECLAIM)
6952                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6953         } else
6954                 data->arg.new_lock = 1;
6955         task = rpc_run_task(&task_setup_data);
6956         if (IS_ERR(task))
6957                 return PTR_ERR(task);
6958         ret = rpc_wait_for_completion_task(task);
6959         if (ret == 0) {
6960                 ret = data->rpc_status;
6961                 if (ret)
6962                         nfs4_handle_setlk_error(data->server, data->lsp,
6963                                         data->arg.new_lock_owner, ret);
6964         } else
6965                 data->cancelled = true;
6966         rpc_put_task(task);
6967         dprintk("%s: done, ret = %d!\n", __func__, ret);
6968         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6969         return ret;
6970 }
6971
6972 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6973 {
6974         struct nfs_server *server = NFS_SERVER(state->inode);
6975         struct nfs4_exception exception = {
6976                 .inode = state->inode,
6977         };
6978         int err;
6979
6980         do {
6981                 /* Cache the lock if possible... */
6982                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6983                         return 0;
6984                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6985                 if (err != -NFS4ERR_DELAY)
6986                         break;
6987                 nfs4_handle_exception(server, err, &exception);
6988         } while (exception.retry);
6989         return err;
6990 }
6991
6992 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6993 {
6994         struct nfs_server *server = NFS_SERVER(state->inode);
6995         struct nfs4_exception exception = {
6996                 .inode = state->inode,
6997         };
6998         int err;
6999
7000         err = nfs4_set_lock_state(state, request);
7001         if (err != 0)
7002                 return err;
7003         if (!recover_lost_locks) {
7004                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7005                 return 0;
7006         }
7007         do {
7008                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7009                         return 0;
7010                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7011                 switch (err) {
7012                 default:
7013                         goto out;
7014                 case -NFS4ERR_GRACE:
7015                 case -NFS4ERR_DELAY:
7016                         nfs4_handle_exception(server, err, &exception);
7017                         err = 0;
7018                 }
7019         } while (exception.retry);
7020 out:
7021         return err;
7022 }
7023
7024 #if defined(CONFIG_NFS_V4_1)
7025 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7026 {
7027         struct nfs4_lock_state *lsp;
7028         int status;
7029
7030         status = nfs4_set_lock_state(state, request);
7031         if (status != 0)
7032                 return status;
7033         lsp = request->fl_u.nfs4_fl.owner;
7034         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7035             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7036                 return 0;
7037         return nfs4_lock_expired(state, request);
7038 }
7039 #endif
7040
7041 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7042 {
7043         struct nfs_inode *nfsi = NFS_I(state->inode);
7044         struct nfs4_state_owner *sp = state->owner;
7045         unsigned char fl_flags = request->fl_flags;
7046         int status;
7047
7048         request->fl_flags |= FL_ACCESS;
7049         status = locks_lock_inode_wait(state->inode, request);
7050         if (status < 0)
7051                 goto out;
7052         mutex_lock(&sp->so_delegreturn_mutex);
7053         down_read(&nfsi->rwsem);
7054         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7055                 /* Yes: cache locks! */
7056                 /* ...but avoid races with delegation recall... */
7057                 request->fl_flags = fl_flags & ~FL_SLEEP;
7058                 status = locks_lock_inode_wait(state->inode, request);
7059                 up_read(&nfsi->rwsem);
7060                 mutex_unlock(&sp->so_delegreturn_mutex);
7061                 goto out;
7062         }
7063         up_read(&nfsi->rwsem);
7064         mutex_unlock(&sp->so_delegreturn_mutex);
7065         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7066 out:
7067         request->fl_flags = fl_flags;
7068         return status;
7069 }
7070
7071 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7072 {
7073         struct nfs4_exception exception = {
7074                 .state = state,
7075                 .inode = state->inode,
7076                 .interruptible = true,
7077         };
7078         int err;
7079
7080         do {
7081                 err = _nfs4_proc_setlk(state, cmd, request);
7082                 if (err == -NFS4ERR_DENIED)
7083                         err = -EAGAIN;
7084                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7085                                 err, &exception);
7086         } while (exception.retry);
7087         return err;
7088 }
7089
7090 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7091 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7092
7093 static int
7094 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7095                         struct file_lock *request)
7096 {
7097         int             status = -ERESTARTSYS;
7098         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7099
7100         while(!signalled()) {
7101                 status = nfs4_proc_setlk(state, cmd, request);
7102                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7103                         break;
7104                 freezable_schedule_timeout_interruptible(timeout);
7105                 timeout *= 2;
7106                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7107                 status = -ERESTARTSYS;
7108         }
7109         return status;
7110 }
7111
7112 #ifdef CONFIG_NFS_V4_1
7113 struct nfs4_lock_waiter {
7114         struct task_struct      *task;
7115         struct inode            *inode;
7116         struct nfs_lowner       *owner;
7117 };
7118
7119 static int
7120 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7121 {
7122         int ret;
7123         struct nfs4_lock_waiter *waiter = wait->private;
7124
7125         /* NULL key means to wake up everyone */
7126         if (key) {
7127                 struct cb_notify_lock_args      *cbnl = key;
7128                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7129                                                 *wowner = waiter->owner;
7130
7131                 /* Only wake if the callback was for the same owner. */
7132                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7133                         return 0;
7134
7135                 /* Make sure it's for the right inode */
7136                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7137                         return 0;
7138         }
7139
7140         /* override "private" so we can use default_wake_function */
7141         wait->private = waiter->task;
7142         ret = woken_wake_function(wait, mode, flags, key);
7143         if (ret)
7144                 list_del_init(&wait->entry);
7145         wait->private = waiter;
7146         return ret;
7147 }
7148
7149 static int
7150 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7151 {
7152         int status = -ERESTARTSYS;
7153         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7154         struct nfs_server *server = NFS_SERVER(state->inode);
7155         struct nfs_client *clp = server->nfs_client;
7156         wait_queue_head_t *q = &clp->cl_lock_waitq;
7157         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7158                                     .id = lsp->ls_seqid.owner_id,
7159                                     .s_dev = server->s_dev };
7160         struct nfs4_lock_waiter waiter = { .task  = current,
7161                                            .inode = state->inode,
7162                                            .owner = &owner};
7163         wait_queue_entry_t wait;
7164
7165         /* Don't bother with waitqueue if we don't expect a callback */
7166         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7167                 return nfs4_retry_setlk_simple(state, cmd, request);
7168
7169         init_wait(&wait);
7170         wait.private = &waiter;
7171         wait.func = nfs4_wake_lock_waiter;
7172
7173         while(!signalled()) {
7174                 add_wait_queue(q, &wait);
7175                 status = nfs4_proc_setlk(state, cmd, request);
7176                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7177                         finish_wait(q, &wait);
7178                         break;
7179                 }
7180
7181                 status = -ERESTARTSYS;
7182                 freezer_do_not_count();
7183                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7184                 freezer_count();
7185                 finish_wait(q, &wait);
7186         }
7187
7188         return status;
7189 }
7190 #else /* !CONFIG_NFS_V4_1 */
7191 static inline int
7192 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7193 {
7194         return nfs4_retry_setlk_simple(state, cmd, request);
7195 }
7196 #endif
7197
7198 static int
7199 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7200 {
7201         struct nfs_open_context *ctx;
7202         struct nfs4_state *state;
7203         int status;
7204
7205         /* verify open state */
7206         ctx = nfs_file_open_context(filp);
7207         state = ctx->state;
7208
7209         if (IS_GETLK(cmd)) {
7210                 if (state != NULL)
7211                         return nfs4_proc_getlk(state, F_GETLK, request);
7212                 return 0;
7213         }
7214
7215         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7216                 return -EINVAL;
7217
7218         if (request->fl_type == F_UNLCK) {
7219                 if (state != NULL)
7220                         return nfs4_proc_unlck(state, cmd, request);
7221                 return 0;
7222         }
7223
7224         if (state == NULL)
7225                 return -ENOLCK;
7226
7227         if ((request->fl_flags & FL_POSIX) &&
7228             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7229                 return -ENOLCK;
7230
7231         /*
7232          * Don't rely on the VFS having checked the file open mode,
7233          * since it won't do this for flock() locks.
7234          */
7235         switch (request->fl_type) {
7236         case F_RDLCK:
7237                 if (!(filp->f_mode & FMODE_READ))
7238                         return -EBADF;
7239                 break;
7240         case F_WRLCK:
7241                 if (!(filp->f_mode & FMODE_WRITE))
7242                         return -EBADF;
7243         }
7244
7245         status = nfs4_set_lock_state(state, request);
7246         if (status != 0)
7247                 return status;
7248
7249         return nfs4_retry_setlk(state, cmd, request);
7250 }
7251
7252 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7253 {
7254         struct nfs_server *server = NFS_SERVER(state->inode);
7255         int err;
7256
7257         err = nfs4_set_lock_state(state, fl);
7258         if (err != 0)
7259                 return err;
7260         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7261         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7262 }
7263
7264 struct nfs_release_lockowner_data {
7265         struct nfs4_lock_state *lsp;
7266         struct nfs_server *server;
7267         struct nfs_release_lockowner_args args;
7268         struct nfs_release_lockowner_res res;
7269         unsigned long timestamp;
7270 };
7271
7272 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7273 {
7274         struct nfs_release_lockowner_data *data = calldata;
7275         struct nfs_server *server = data->server;
7276         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7277                            &data->res.seq_res, task);
7278         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7279         data->timestamp = jiffies;
7280 }
7281
7282 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7283 {
7284         struct nfs_release_lockowner_data *data = calldata;
7285         struct nfs_server *server = data->server;
7286
7287         nfs40_sequence_done(task, &data->res.seq_res);
7288
7289         switch (task->tk_status) {
7290         case 0:
7291                 renew_lease(server, data->timestamp);
7292                 break;
7293         case -NFS4ERR_STALE_CLIENTID:
7294         case -NFS4ERR_EXPIRED:
7295                 nfs4_schedule_lease_recovery(server->nfs_client);
7296                 break;
7297         case -NFS4ERR_LEASE_MOVED:
7298         case -NFS4ERR_DELAY:
7299                 if (nfs4_async_handle_error(task, server,
7300                                             NULL, NULL) == -EAGAIN)
7301                         rpc_restart_call_prepare(task);
7302         }
7303 }
7304
7305 static void nfs4_release_lockowner_release(void *calldata)
7306 {
7307         struct nfs_release_lockowner_data *data = calldata;
7308         nfs4_free_lock_state(data->server, data->lsp);
7309         kfree(calldata);
7310 }
7311
7312 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7313         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7314         .rpc_call_done = nfs4_release_lockowner_done,
7315         .rpc_release = nfs4_release_lockowner_release,
7316 };
7317
7318 static void
7319 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7320 {
7321         struct nfs_release_lockowner_data *data;
7322         struct rpc_message msg = {
7323                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7324         };
7325
7326         if (server->nfs_client->cl_mvops->minor_version != 0)
7327                 return;
7328
7329         data = kmalloc(sizeof(*data), GFP_NOFS);
7330         if (!data)
7331                 return;
7332         data->lsp = lsp;
7333         data->server = server;
7334         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7335         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7336         data->args.lock_owner.s_dev = server->s_dev;
7337
7338         msg.rpc_argp = &data->args;
7339         msg.rpc_resp = &data->res;
7340         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7341         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7342 }
7343
7344 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7345
7346 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7347                                    struct dentry *unused, struct inode *inode,
7348                                    const char *key, const void *buf,
7349                                    size_t buflen, int flags)
7350 {
7351         return nfs4_proc_set_acl(inode, buf, buflen);
7352 }
7353
7354 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7355                                    struct dentry *unused, struct inode *inode,
7356                                    const char *key, void *buf, size_t buflen)
7357 {
7358         return nfs4_proc_get_acl(inode, buf, buflen);
7359 }
7360
7361 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7362 {
7363         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7364 }
7365
7366 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7367
7368 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7369                                      struct dentry *unused, struct inode *inode,
7370                                      const char *key, const void *buf,
7371                                      size_t buflen, int flags)
7372 {
7373         if (security_ismaclabel(key))
7374                 return nfs4_set_security_label(inode, buf, buflen);
7375
7376         return -EOPNOTSUPP;
7377 }
7378
7379 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7380                                      struct dentry *unused, struct inode *inode,
7381                                      const char *key, void *buf, size_t buflen)
7382 {
7383         if (security_ismaclabel(key))
7384                 return nfs4_get_security_label(inode, buf, buflen);
7385         return -EOPNOTSUPP;
7386 }
7387
7388 static ssize_t
7389 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7390 {
7391         int len = 0;
7392
7393         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7394                 len = security_inode_listsecurity(inode, list, list_len);
7395                 if (list_len && len > list_len)
7396                         return -ERANGE;
7397         }
7398         return len;
7399 }
7400
7401 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7402         .prefix = XATTR_SECURITY_PREFIX,
7403         .get    = nfs4_xattr_get_nfs4_label,
7404         .set    = nfs4_xattr_set_nfs4_label,
7405 };
7406
7407 #else
7408
7409 static ssize_t
7410 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7411 {
7412         return 0;
7413 }
7414
7415 #endif
7416
7417 /*
7418  * nfs_fhget will use either the mounted_on_fileid or the fileid
7419  */
7420 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7421 {
7422         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7423                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7424               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7425               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7426                 return;
7427
7428         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7429                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7430         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7431         fattr->nlink = 2;
7432 }
7433
7434 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7435                                    const struct qstr *name,
7436                                    struct nfs4_fs_locations *fs_locations,
7437                                    struct page *page)
7438 {
7439         struct nfs_server *server = NFS_SERVER(dir);
7440         u32 bitmask[3];
7441         struct nfs4_fs_locations_arg args = {
7442                 .dir_fh = NFS_FH(dir),
7443                 .name = name,
7444                 .page = page,
7445                 .bitmask = bitmask,
7446         };
7447         struct nfs4_fs_locations_res res = {
7448                 .fs_locations = fs_locations,
7449         };
7450         struct rpc_message msg = {
7451                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7452                 .rpc_argp = &args,
7453                 .rpc_resp = &res,
7454         };
7455         int status;
7456
7457         dprintk("%s: start\n", __func__);
7458
7459         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7460         bitmask[1] = nfs4_fattr_bitmap[1];
7461
7462         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7463          * is not supported */
7464         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7465                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7466         else
7467                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7468
7469         nfs_fattr_init(&fs_locations->fattr);
7470         fs_locations->server = server;
7471         fs_locations->nlocations = 0;
7472         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7473         dprintk("%s: returned status = %d\n", __func__, status);
7474         return status;
7475 }
7476
7477 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7478                            const struct qstr *name,
7479                            struct nfs4_fs_locations *fs_locations,
7480                            struct page *page)
7481 {
7482         struct nfs4_exception exception = {
7483                 .interruptible = true,
7484         };
7485         int err;
7486         do {
7487                 err = _nfs4_proc_fs_locations(client, dir, name,
7488                                 fs_locations, page);
7489                 trace_nfs4_get_fs_locations(dir, name, err);
7490                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7491                                 &exception);
7492         } while (exception.retry);
7493         return err;
7494 }
7495
7496 /*
7497  * This operation also signals the server that this client is
7498  * performing migration recovery.  The server can stop returning
7499  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7500  * appended to this compound to identify the client ID which is
7501  * performing recovery.
7502  */
7503 static int _nfs40_proc_get_locations(struct inode *inode,
7504                                      struct nfs4_fs_locations *locations,
7505                                      struct page *page, const struct cred *cred)
7506 {
7507         struct nfs_server *server = NFS_SERVER(inode);
7508         struct rpc_clnt *clnt = server->client;
7509         u32 bitmask[2] = {
7510                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7511         };
7512         struct nfs4_fs_locations_arg args = {
7513                 .clientid       = server->nfs_client->cl_clientid,
7514                 .fh             = NFS_FH(inode),
7515                 .page           = page,
7516                 .bitmask        = bitmask,
7517                 .migration      = 1,            /* skip LOOKUP */
7518                 .renew          = 1,            /* append RENEW */
7519         };
7520         struct nfs4_fs_locations_res res = {
7521                 .fs_locations   = locations,
7522                 .migration      = 1,
7523                 .renew          = 1,
7524         };
7525         struct rpc_message msg = {
7526                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7527                 .rpc_argp       = &args,
7528                 .rpc_resp       = &res,
7529                 .rpc_cred       = cred,
7530         };
7531         unsigned long now = jiffies;
7532         int status;
7533
7534         nfs_fattr_init(&locations->fattr);
7535         locations->server = server;
7536         locations->nlocations = 0;
7537
7538         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7539         status = nfs4_call_sync_sequence(clnt, server, &msg,
7540                                         &args.seq_args, &res.seq_res);
7541         if (status)
7542                 return status;
7543
7544         renew_lease(server, now);
7545         return 0;
7546 }
7547
7548 #ifdef CONFIG_NFS_V4_1
7549
7550 /*
7551  * This operation also signals the server that this client is
7552  * performing migration recovery.  The server can stop asserting
7553  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7554  * performing this operation is identified in the SEQUENCE
7555  * operation in this compound.
7556  *
7557  * When the client supports GETATTR(fs_locations_info), it can
7558  * be plumbed in here.
7559  */
7560 static int _nfs41_proc_get_locations(struct inode *inode,
7561                                      struct nfs4_fs_locations *locations,
7562                                      struct page *page, const struct cred *cred)
7563 {
7564         struct nfs_server *server = NFS_SERVER(inode);
7565         struct rpc_clnt *clnt = server->client;
7566         u32 bitmask[2] = {
7567                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7568         };
7569         struct nfs4_fs_locations_arg args = {
7570                 .fh             = NFS_FH(inode),
7571                 .page           = page,
7572                 .bitmask        = bitmask,
7573                 .migration      = 1,            /* skip LOOKUP */
7574         };
7575         struct nfs4_fs_locations_res res = {
7576                 .fs_locations   = locations,
7577                 .migration      = 1,
7578         };
7579         struct rpc_message msg = {
7580                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7581                 .rpc_argp       = &args,
7582                 .rpc_resp       = &res,
7583                 .rpc_cred       = cred,
7584         };
7585         int status;
7586
7587         nfs_fattr_init(&locations->fattr);
7588         locations->server = server;
7589         locations->nlocations = 0;
7590
7591         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7592         status = nfs4_call_sync_sequence(clnt, server, &msg,
7593                                         &args.seq_args, &res.seq_res);
7594         if (status == NFS4_OK &&
7595             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7596                 status = -NFS4ERR_LEASE_MOVED;
7597         return status;
7598 }
7599
7600 #endif  /* CONFIG_NFS_V4_1 */
7601
7602 /**
7603  * nfs4_proc_get_locations - discover locations for a migrated FSID
7604  * @inode: inode on FSID that is migrating
7605  * @locations: result of query
7606  * @page: buffer
7607  * @cred: credential to use for this operation
7608  *
7609  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7610  * operation failed, or a negative errno if a local error occurred.
7611  *
7612  * On success, "locations" is filled in, but if the server has
7613  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7614  * asserted.
7615  *
7616  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7617  * from this client that require migration recovery.
7618  */
7619 int nfs4_proc_get_locations(struct inode *inode,
7620                             struct nfs4_fs_locations *locations,
7621                             struct page *page, const struct cred *cred)
7622 {
7623         struct nfs_server *server = NFS_SERVER(inode);
7624         struct nfs_client *clp = server->nfs_client;
7625         const struct nfs4_mig_recovery_ops *ops =
7626                                         clp->cl_mvops->mig_recovery_ops;
7627         struct nfs4_exception exception = {
7628                 .interruptible = true,
7629         };
7630         int status;
7631
7632         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7633                 (unsigned long long)server->fsid.major,
7634                 (unsigned long long)server->fsid.minor,
7635                 clp->cl_hostname);
7636         nfs_display_fhandle(NFS_FH(inode), __func__);
7637
7638         do {
7639                 status = ops->get_locations(inode, locations, page, cred);
7640                 if (status != -NFS4ERR_DELAY)
7641                         break;
7642                 nfs4_handle_exception(server, status, &exception);
7643         } while (exception.retry);
7644         return status;
7645 }
7646
7647 /*
7648  * This operation also signals the server that this client is
7649  * performing "lease moved" recovery.  The server can stop
7650  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7651  * is appended to this compound to identify the client ID which is
7652  * performing recovery.
7653  */
7654 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7655 {
7656         struct nfs_server *server = NFS_SERVER(inode);
7657         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7658         struct rpc_clnt *clnt = server->client;
7659         struct nfs4_fsid_present_arg args = {
7660                 .fh             = NFS_FH(inode),
7661                 .clientid       = clp->cl_clientid,
7662                 .renew          = 1,            /* append RENEW */
7663         };
7664         struct nfs4_fsid_present_res res = {
7665                 .renew          = 1,
7666         };
7667         struct rpc_message msg = {
7668                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7669                 .rpc_argp       = &args,
7670                 .rpc_resp       = &res,
7671                 .rpc_cred       = cred,
7672         };
7673         unsigned long now = jiffies;
7674         int status;
7675
7676         res.fh = nfs_alloc_fhandle();
7677         if (res.fh == NULL)
7678                 return -ENOMEM;
7679
7680         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7681         status = nfs4_call_sync_sequence(clnt, server, &msg,
7682                                                 &args.seq_args, &res.seq_res);
7683         nfs_free_fhandle(res.fh);
7684         if (status)
7685                 return status;
7686
7687         do_renew_lease(clp, now);
7688         return 0;
7689 }
7690
7691 #ifdef CONFIG_NFS_V4_1
7692
7693 /*
7694  * This operation also signals the server that this client is
7695  * performing "lease moved" recovery.  The server can stop asserting
7696  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7697  * this operation is identified in the SEQUENCE operation in this
7698  * compound.
7699  */
7700 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7701 {
7702         struct nfs_server *server = NFS_SERVER(inode);
7703         struct rpc_clnt *clnt = server->client;
7704         struct nfs4_fsid_present_arg args = {
7705                 .fh             = NFS_FH(inode),
7706         };
7707         struct nfs4_fsid_present_res res = {
7708         };
7709         struct rpc_message msg = {
7710                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7711                 .rpc_argp       = &args,
7712                 .rpc_resp       = &res,
7713                 .rpc_cred       = cred,
7714         };
7715         int status;
7716
7717         res.fh = nfs_alloc_fhandle();
7718         if (res.fh == NULL)
7719                 return -ENOMEM;
7720
7721         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7722         status = nfs4_call_sync_sequence(clnt, server, &msg,
7723                                                 &args.seq_args, &res.seq_res);
7724         nfs_free_fhandle(res.fh);
7725         if (status == NFS4_OK &&
7726             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7727                 status = -NFS4ERR_LEASE_MOVED;
7728         return status;
7729 }
7730
7731 #endif  /* CONFIG_NFS_V4_1 */
7732
7733 /**
7734  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7735  * @inode: inode on FSID to check
7736  * @cred: credential to use for this operation
7737  *
7738  * Server indicates whether the FSID is present, moved, or not
7739  * recognized.  This operation is necessary to clear a LEASE_MOVED
7740  * condition for this client ID.
7741  *
7742  * Returns NFS4_OK if the FSID is present on this server,
7743  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7744  *  NFS4ERR code if some error occurred on the server, or a
7745  *  negative errno if a local failure occurred.
7746  */
7747 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
7748 {
7749         struct nfs_server *server = NFS_SERVER(inode);
7750         struct nfs_client *clp = server->nfs_client;
7751         const struct nfs4_mig_recovery_ops *ops =
7752                                         clp->cl_mvops->mig_recovery_ops;
7753         struct nfs4_exception exception = {
7754                 .interruptible = true,
7755         };
7756         int status;
7757
7758         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7759                 (unsigned long long)server->fsid.major,
7760                 (unsigned long long)server->fsid.minor,
7761                 clp->cl_hostname);
7762         nfs_display_fhandle(NFS_FH(inode), __func__);
7763
7764         do {
7765                 status = ops->fsid_present(inode, cred);
7766                 if (status != -NFS4ERR_DELAY)
7767                         break;
7768                 nfs4_handle_exception(server, status, &exception);
7769         } while (exception.retry);
7770         return status;
7771 }
7772
7773 /*
7774  * If 'use_integrity' is true and the state managment nfs_client
7775  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7776  * and the machine credential as per RFC3530bis and RFC5661 Security
7777  * Considerations sections. Otherwise, just use the user cred with the
7778  * filesystem's rpc_client.
7779  */
7780 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7781 {
7782         int status;
7783         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7784         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
7785         struct nfs4_secinfo_arg args = {
7786                 .dir_fh = NFS_FH(dir),
7787                 .name   = name,
7788         };
7789         struct nfs4_secinfo_res res = {
7790                 .flavors     = flavors,
7791         };
7792         struct rpc_message msg = {
7793                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7794                 .rpc_argp = &args,
7795                 .rpc_resp = &res,
7796         };
7797         struct nfs4_call_sync_data data = {
7798                 .seq_server = NFS_SERVER(dir),
7799                 .seq_args = &args.seq_args,
7800                 .seq_res = &res.seq_res,
7801         };
7802         struct rpc_task_setup task_setup = {
7803                 .rpc_client = clnt,
7804                 .rpc_message = &msg,
7805                 .callback_ops = clp->cl_mvops->call_sync_ops,
7806                 .callback_data = &data,
7807                 .flags = RPC_TASK_NO_ROUND_ROBIN,
7808         };
7809         const struct cred *cred = NULL;
7810
7811         if (use_integrity) {
7812                 clnt = clp->cl_rpcclient;
7813                 task_setup.rpc_client = clnt;
7814
7815                 cred = nfs4_get_clid_cred(clp);
7816                 msg.rpc_cred = cred;
7817         }
7818
7819         dprintk("NFS call  secinfo %s\n", name->name);
7820
7821         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7822         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
7823         status = nfs4_call_sync_custom(&task_setup);
7824
7825         dprintk("NFS reply  secinfo: %d\n", status);
7826
7827         put_cred(cred);
7828         return status;
7829 }
7830
7831 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7832                       struct nfs4_secinfo_flavors *flavors)
7833 {
7834         struct nfs4_exception exception = {
7835                 .interruptible = true,
7836         };
7837         int err;
7838         do {
7839                 err = -NFS4ERR_WRONGSEC;
7840
7841                 /* try to use integrity protection with machine cred */
7842                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7843                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7844
7845                 /*
7846                  * if unable to use integrity protection, or SECINFO with
7847                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7848                  * disallowed by spec, but exists in deployed servers) use
7849                  * the current filesystem's rpc_client and the user cred.
7850                  */
7851                 if (err == -NFS4ERR_WRONGSEC)
7852                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7853
7854                 trace_nfs4_secinfo(dir, name, err);
7855                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7856                                 &exception);
7857         } while (exception.retry);
7858         return err;
7859 }
7860
7861 #ifdef CONFIG_NFS_V4_1
7862 /*
7863  * Check the exchange flags returned by the server for invalid flags, having
7864  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7865  * DS flags set.
7866  */
7867 static int nfs4_check_cl_exchange_flags(u32 flags)
7868 {
7869         if (flags & ~EXCHGID4_FLAG_MASK_R)
7870                 goto out_inval;
7871         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7872             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7873                 goto out_inval;
7874         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7875                 goto out_inval;
7876         return NFS_OK;
7877 out_inval:
7878         return -NFS4ERR_INVAL;
7879 }
7880
7881 static bool
7882 nfs41_same_server_scope(struct nfs41_server_scope *a,
7883                         struct nfs41_server_scope *b)
7884 {
7885         if (a->server_scope_sz != b->server_scope_sz)
7886                 return false;
7887         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7888 }
7889
7890 static void
7891 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7892 {
7893         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
7894         struct nfs_client *clp = args->client;
7895
7896         switch (task->tk_status) {
7897         case -NFS4ERR_BADSESSION:
7898         case -NFS4ERR_DEADSESSION:
7899                 nfs4_schedule_session_recovery(clp->cl_session,
7900                                 task->tk_status);
7901         }
7902 }
7903
7904 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7905         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7906 };
7907
7908 /*
7909  * nfs4_proc_bind_one_conn_to_session()
7910  *
7911  * The 4.1 client currently uses the same TCP connection for the
7912  * fore and backchannel.
7913  */
7914 static
7915 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7916                 struct rpc_xprt *xprt,
7917                 struct nfs_client *clp,
7918                 const struct cred *cred)
7919 {
7920         int status;
7921         struct nfs41_bind_conn_to_session_args args = {
7922                 .client = clp,
7923                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7924         };
7925         struct nfs41_bind_conn_to_session_res res;
7926         struct rpc_message msg = {
7927                 .rpc_proc =
7928                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7929                 .rpc_argp = &args,
7930                 .rpc_resp = &res,
7931                 .rpc_cred = cred,
7932         };
7933         struct rpc_task_setup task_setup_data = {
7934                 .rpc_client = clnt,
7935                 .rpc_xprt = xprt,
7936                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7937                 .rpc_message = &msg,
7938                 .flags = RPC_TASK_TIMEOUT,
7939         };
7940         struct rpc_task *task;
7941
7942         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7943         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7944                 args.dir = NFS4_CDFC4_FORE;
7945
7946         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7947         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7948                 args.dir = NFS4_CDFC4_FORE;
7949
7950         task = rpc_run_task(&task_setup_data);
7951         if (!IS_ERR(task)) {
7952                 status = task->tk_status;
7953                 rpc_put_task(task);
7954         } else
7955                 status = PTR_ERR(task);
7956         trace_nfs4_bind_conn_to_session(clp, status);
7957         if (status == 0) {
7958                 if (memcmp(res.sessionid.data,
7959                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7960                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7961                         return -EIO;
7962                 }
7963                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7964                         dprintk("NFS: %s: Unexpected direction from server\n",
7965                                 __func__);
7966                         return -EIO;
7967                 }
7968                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7969                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7970                                 __func__);
7971                         return -EIO;
7972                 }
7973         }
7974
7975         return status;
7976 }
7977
7978 struct rpc_bind_conn_calldata {
7979         struct nfs_client *clp;
7980         const struct cred *cred;
7981 };
7982
7983 static int
7984 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7985                 struct rpc_xprt *xprt,
7986                 void *calldata)
7987 {
7988         struct rpc_bind_conn_calldata *p = calldata;
7989
7990         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7991 }
7992
7993 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
7994 {
7995         struct rpc_bind_conn_calldata data = {
7996                 .clp = clp,
7997                 .cred = cred,
7998         };
7999         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8000                         nfs4_proc_bind_conn_to_session_callback, &data);
8001 }
8002
8003 /*
8004  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8005  * and operations we'd like to see to enable certain features in the allow map
8006  */
8007 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8008         .how = SP4_MACH_CRED,
8009         .enforce.u.words = {
8010                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8011                       1 << (OP_EXCHANGE_ID - 32) |
8012                       1 << (OP_CREATE_SESSION - 32) |
8013                       1 << (OP_DESTROY_SESSION - 32) |
8014                       1 << (OP_DESTROY_CLIENTID - 32)
8015         },
8016         .allow.u.words = {
8017                 [0] = 1 << (OP_CLOSE) |
8018                       1 << (OP_OPEN_DOWNGRADE) |
8019                       1 << (OP_LOCKU) |
8020                       1 << (OP_DELEGRETURN) |
8021                       1 << (OP_COMMIT),
8022                 [1] = 1 << (OP_SECINFO - 32) |
8023                       1 << (OP_SECINFO_NO_NAME - 32) |
8024                       1 << (OP_LAYOUTRETURN - 32) |
8025                       1 << (OP_TEST_STATEID - 32) |
8026                       1 << (OP_FREE_STATEID - 32) |
8027                       1 << (OP_WRITE - 32)
8028         }
8029 };
8030
8031 /*
8032  * Select the state protection mode for client `clp' given the server results
8033  * from exchange_id in `sp'.
8034  *
8035  * Returns 0 on success, negative errno otherwise.
8036  */
8037 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8038                                  struct nfs41_state_protection *sp)
8039 {
8040         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8041                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8042                       1 << (OP_EXCHANGE_ID - 32) |
8043                       1 << (OP_CREATE_SESSION - 32) |
8044                       1 << (OP_DESTROY_SESSION - 32) |
8045                       1 << (OP_DESTROY_CLIENTID - 32)
8046         };
8047         unsigned long flags = 0;
8048         unsigned int i;
8049         int ret = 0;
8050
8051         if (sp->how == SP4_MACH_CRED) {
8052                 /* Print state protect result */
8053                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8054                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8055                         if (test_bit(i, sp->enforce.u.longs))
8056                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8057                         if (test_bit(i, sp->allow.u.longs))
8058                                 dfprintk(MOUNT, "  allow op %d\n", i);
8059                 }
8060
8061                 /* make sure nothing is on enforce list that isn't supported */
8062                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8063                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8064                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8065                                 ret = -EINVAL;
8066                                 goto out;
8067                         }
8068                 }
8069
8070                 /*
8071                  * Minimal mode - state operations are allowed to use machine
8072                  * credential.  Note this already happens by default, so the
8073                  * client doesn't have to do anything more than the negotiation.
8074                  *
8075                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8076                  *       we're already using the machine cred for exchange_id
8077                  *       and will never use a different cred.
8078                  */
8079                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8080                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8081                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8082                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8083                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8084                         dfprintk(MOUNT, "  minimal mode enabled\n");
8085                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8086                 } else {
8087                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8088                         ret = -EINVAL;
8089                         goto out;
8090                 }
8091
8092                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8093                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8094                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8095                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8096                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8097                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8098                 }
8099
8100                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8101                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8102                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8103                 }
8104
8105                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8106                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8107                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8108                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8109                 }
8110
8111                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8112                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8113                         dfprintk(MOUNT, "  stateid mode enabled\n");
8114                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8115                 }
8116
8117                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8118                         dfprintk(MOUNT, "  write mode enabled\n");
8119                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8120                 }
8121
8122                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8123                         dfprintk(MOUNT, "  commit mode enabled\n");
8124                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8125                 }
8126         }
8127 out:
8128         clp->cl_sp4_flags = flags;
8129         return ret;
8130 }
8131
8132 struct nfs41_exchange_id_data {
8133         struct nfs41_exchange_id_res res;
8134         struct nfs41_exchange_id_args args;
8135 };
8136
8137 static void nfs4_exchange_id_release(void *data)
8138 {
8139         struct nfs41_exchange_id_data *cdata =
8140                                         (struct nfs41_exchange_id_data *)data;
8141
8142         nfs_put_client(cdata->args.client);
8143         kfree(cdata->res.impl_id);
8144         kfree(cdata->res.server_scope);
8145         kfree(cdata->res.server_owner);
8146         kfree(cdata);
8147 }
8148
8149 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8150         .rpc_release = nfs4_exchange_id_release,
8151 };
8152
8153 /*
8154  * _nfs4_proc_exchange_id()
8155  *
8156  * Wrapper for EXCHANGE_ID operation.
8157  */
8158 static struct rpc_task *
8159 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8160                         u32 sp4_how, struct rpc_xprt *xprt)
8161 {
8162         struct rpc_message msg = {
8163                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8164                 .rpc_cred = cred,
8165         };
8166         struct rpc_task_setup task_setup_data = {
8167                 .rpc_client = clp->cl_rpcclient,
8168                 .callback_ops = &nfs4_exchange_id_call_ops,
8169                 .rpc_message = &msg,
8170                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8171         };
8172         struct nfs41_exchange_id_data *calldata;
8173         int status;
8174
8175         if (!refcount_inc_not_zero(&clp->cl_count))
8176                 return ERR_PTR(-EIO);
8177
8178         status = -ENOMEM;
8179         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8180         if (!calldata)
8181                 goto out;
8182
8183         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8184
8185         status = nfs4_init_uniform_client_string(clp);
8186         if (status)
8187                 goto out_calldata;
8188
8189         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8190                                                 GFP_NOFS);
8191         status = -ENOMEM;
8192         if (unlikely(calldata->res.server_owner == NULL))
8193                 goto out_calldata;
8194
8195         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8196                                         GFP_NOFS);
8197         if (unlikely(calldata->res.server_scope == NULL))
8198                 goto out_server_owner;
8199
8200         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8201         if (unlikely(calldata->res.impl_id == NULL))
8202                 goto out_server_scope;
8203
8204         switch (sp4_how) {
8205         case SP4_NONE:
8206                 calldata->args.state_protect.how = SP4_NONE;
8207                 break;
8208
8209         case SP4_MACH_CRED:
8210                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8211                 break;
8212
8213         default:
8214                 /* unsupported! */
8215                 WARN_ON_ONCE(1);
8216                 status = -EINVAL;
8217                 goto out_impl_id;
8218         }
8219         if (xprt) {
8220                 task_setup_data.rpc_xprt = xprt;
8221                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8222                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8223                                 sizeof(calldata->args.verifier.data));
8224         }
8225         calldata->args.client = clp;
8226         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8227         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8228 #ifdef CONFIG_NFS_V4_1_MIGRATION
8229         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8230 #endif
8231         msg.rpc_argp = &calldata->args;
8232         msg.rpc_resp = &calldata->res;
8233         task_setup_data.callback_data = calldata;
8234
8235         return rpc_run_task(&task_setup_data);
8236
8237 out_impl_id:
8238         kfree(calldata->res.impl_id);
8239 out_server_scope:
8240         kfree(calldata->res.server_scope);
8241 out_server_owner:
8242         kfree(calldata->res.server_owner);
8243 out_calldata:
8244         kfree(calldata);
8245 out:
8246         nfs_put_client(clp);
8247         return ERR_PTR(status);
8248 }
8249
8250 /*
8251  * _nfs4_proc_exchange_id()
8252  *
8253  * Wrapper for EXCHANGE_ID operation.
8254  */
8255 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8256                         u32 sp4_how)
8257 {
8258         struct rpc_task *task;
8259         struct nfs41_exchange_id_args *argp;
8260         struct nfs41_exchange_id_res *resp;
8261         unsigned long now = jiffies;
8262         int status;
8263
8264         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8265         if (IS_ERR(task))
8266                 return PTR_ERR(task);
8267
8268         argp = task->tk_msg.rpc_argp;
8269         resp = task->tk_msg.rpc_resp;
8270         status = task->tk_status;
8271         if (status  != 0)
8272                 goto out;
8273
8274         status = nfs4_check_cl_exchange_flags(resp->flags);
8275         if (status  != 0)
8276                 goto out;
8277
8278         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8279         if (status != 0)
8280                 goto out;
8281
8282         do_renew_lease(clp, now);
8283
8284         clp->cl_clientid = resp->clientid;
8285         clp->cl_exchange_flags = resp->flags;
8286         clp->cl_seqid = resp->seqid;
8287         /* Client ID is not confirmed */
8288         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8289                 clear_bit(NFS4_SESSION_ESTABLISHED,
8290                           &clp->cl_session->session_state);
8291
8292         if (clp->cl_serverscope != NULL &&
8293             !nfs41_same_server_scope(clp->cl_serverscope,
8294                                 resp->server_scope)) {
8295                 dprintk("%s: server_scope mismatch detected\n",
8296                         __func__);
8297                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8298         }
8299
8300         swap(clp->cl_serverowner, resp->server_owner);
8301         swap(clp->cl_serverscope, resp->server_scope);
8302         swap(clp->cl_implid, resp->impl_id);
8303
8304         /* Save the EXCHANGE_ID verifier session trunk tests */
8305         memcpy(clp->cl_confirm.data, argp->verifier.data,
8306                sizeof(clp->cl_confirm.data));
8307 out:
8308         trace_nfs4_exchange_id(clp, status);
8309         rpc_put_task(task);
8310         return status;
8311 }
8312
8313 /*
8314  * nfs4_proc_exchange_id()
8315  *
8316  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8317  *
8318  * Since the clientid has expired, all compounds using sessions
8319  * associated with the stale clientid will be returning
8320  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8321  * be in some phase of session reset.
8322  *
8323  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8324  */
8325 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8326 {
8327         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8328         int status;
8329
8330         /* try SP4_MACH_CRED if krb5i/p */
8331         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8332             authflavor == RPC_AUTH_GSS_KRB5P) {
8333                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8334                 if (!status)
8335                         return 0;
8336         }
8337
8338         /* try SP4_NONE */
8339         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8340 }
8341
8342 /**
8343  * nfs4_test_session_trunk
8344  *
8345  * This is an add_xprt_test() test function called from
8346  * rpc_clnt_setup_test_and_add_xprt.
8347  *
8348  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8349  * and is dereferrenced in nfs4_exchange_id_release
8350  *
8351  * Upon success, add the new transport to the rpc_clnt
8352  *
8353  * @clnt: struct rpc_clnt to get new transport
8354  * @xprt: the rpc_xprt to test
8355  * @data: call data for _nfs4_proc_exchange_id.
8356  */
8357 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8358                             void *data)
8359 {
8360         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8361         struct rpc_task *task;
8362         int status;
8363
8364         u32 sp4_how;
8365
8366         dprintk("--> %s try %s\n", __func__,
8367                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8368
8369         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8370
8371         /* Test connection for session trunking. Async exchange_id call */
8372         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8373         if (IS_ERR(task))
8374                 return;
8375
8376         status = task->tk_status;
8377         if (status == 0)
8378                 status = nfs4_detect_session_trunking(adata->clp,
8379                                 task->tk_msg.rpc_resp, xprt);
8380
8381         if (status == 0)
8382                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8383
8384         rpc_put_task(task);
8385 }
8386 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8387
8388 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8389                 const struct cred *cred)
8390 {
8391         struct rpc_message msg = {
8392                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8393                 .rpc_argp = clp,
8394                 .rpc_cred = cred,
8395         };
8396         int status;
8397
8398         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8399                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8400         trace_nfs4_destroy_clientid(clp, status);
8401         if (status)
8402                 dprintk("NFS: Got error %d from the server %s on "
8403                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8404         return status;
8405 }
8406
8407 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8408                 const struct cred *cred)
8409 {
8410         unsigned int loop;
8411         int ret;
8412
8413         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8414                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8415                 switch (ret) {
8416                 case -NFS4ERR_DELAY:
8417                 case -NFS4ERR_CLIENTID_BUSY:
8418                         ssleep(1);
8419                         break;
8420                 default:
8421                         return ret;
8422                 }
8423         }
8424         return 0;
8425 }
8426
8427 int nfs4_destroy_clientid(struct nfs_client *clp)
8428 {
8429         const struct cred *cred;
8430         int ret = 0;
8431
8432         if (clp->cl_mvops->minor_version < 1)
8433                 goto out;
8434         if (clp->cl_exchange_flags == 0)
8435                 goto out;
8436         if (clp->cl_preserve_clid)
8437                 goto out;
8438         cred = nfs4_get_clid_cred(clp);
8439         ret = nfs4_proc_destroy_clientid(clp, cred);
8440         put_cred(cred);
8441         switch (ret) {
8442         case 0:
8443         case -NFS4ERR_STALE_CLIENTID:
8444                 clp->cl_exchange_flags = 0;
8445         }
8446 out:
8447         return ret;
8448 }
8449
8450 #endif /* CONFIG_NFS_V4_1 */
8451
8452 struct nfs4_get_lease_time_data {
8453         struct nfs4_get_lease_time_args *args;
8454         struct nfs4_get_lease_time_res *res;
8455         struct nfs_client *clp;
8456 };
8457
8458 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8459                                         void *calldata)
8460 {
8461         struct nfs4_get_lease_time_data *data =
8462                         (struct nfs4_get_lease_time_data *)calldata;
8463
8464         dprintk("--> %s\n", __func__);
8465         /* just setup sequence, do not trigger session recovery
8466            since we're invoked within one */
8467         nfs4_setup_sequence(data->clp,
8468                         &data->args->la_seq_args,
8469                         &data->res->lr_seq_res,
8470                         task);
8471         dprintk("<-- %s\n", __func__);
8472 }
8473
8474 /*
8475  * Called from nfs4_state_manager thread for session setup, so don't recover
8476  * from sequence operation or clientid errors.
8477  */
8478 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8479 {
8480         struct nfs4_get_lease_time_data *data =
8481                         (struct nfs4_get_lease_time_data *)calldata;
8482
8483         dprintk("--> %s\n", __func__);
8484         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8485                 return;
8486         switch (task->tk_status) {
8487         case -NFS4ERR_DELAY:
8488         case -NFS4ERR_GRACE:
8489                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8490                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8491                 task->tk_status = 0;
8492                 /* fall through */
8493         case -NFS4ERR_RETRY_UNCACHED_REP:
8494                 rpc_restart_call_prepare(task);
8495                 return;
8496         }
8497         dprintk("<-- %s\n", __func__);
8498 }
8499
8500 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8501         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8502         .rpc_call_done = nfs4_get_lease_time_done,
8503 };
8504
8505 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8506 {
8507         struct nfs4_get_lease_time_args args;
8508         struct nfs4_get_lease_time_res res = {
8509                 .lr_fsinfo = fsinfo,
8510         };
8511         struct nfs4_get_lease_time_data data = {
8512                 .args = &args,
8513                 .res = &res,
8514                 .clp = clp,
8515         };
8516         struct rpc_message msg = {
8517                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8518                 .rpc_argp = &args,
8519                 .rpc_resp = &res,
8520         };
8521         struct rpc_task_setup task_setup = {
8522                 .rpc_client = clp->cl_rpcclient,
8523                 .rpc_message = &msg,
8524                 .callback_ops = &nfs4_get_lease_time_ops,
8525                 .callback_data = &data,
8526                 .flags = RPC_TASK_TIMEOUT,
8527         };
8528
8529         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8530         return nfs4_call_sync_custom(&task_setup);
8531 }
8532
8533 #ifdef CONFIG_NFS_V4_1
8534
8535 /*
8536  * Initialize the values to be used by the client in CREATE_SESSION
8537  * If nfs4_init_session set the fore channel request and response sizes,
8538  * use them.
8539  *
8540  * Set the back channel max_resp_sz_cached to zero to force the client to
8541  * always set csa_cachethis to FALSE because the current implementation
8542  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8543  */
8544 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8545                                     struct rpc_clnt *clnt)
8546 {
8547         unsigned int max_rqst_sz, max_resp_sz;
8548         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8549         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8550
8551         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8552         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8553
8554         /* Fore channel attributes */
8555         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8556         args->fc_attrs.max_resp_sz = max_resp_sz;
8557         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8558         args->fc_attrs.max_reqs = max_session_slots;
8559
8560         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8561                 "max_ops=%u max_reqs=%u\n",
8562                 __func__,
8563                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8564                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8565
8566         /* Back channel attributes */
8567         args->bc_attrs.max_rqst_sz = max_bc_payload;
8568         args->bc_attrs.max_resp_sz = max_bc_payload;
8569         args->bc_attrs.max_resp_sz_cached = 0;
8570         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8571         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8572         if (args->bc_attrs.max_reqs > max_bc_slots)
8573                 args->bc_attrs.max_reqs = max_bc_slots;
8574
8575         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8576                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8577                 __func__,
8578                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8579                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8580                 args->bc_attrs.max_reqs);
8581 }
8582
8583 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8584                 struct nfs41_create_session_res *res)
8585 {
8586         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8587         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8588
8589         if (rcvd->max_resp_sz > sent->max_resp_sz)
8590                 return -EINVAL;
8591         /*
8592          * Our requested max_ops is the minimum we need; we're not
8593          * prepared to break up compounds into smaller pieces than that.
8594          * So, no point even trying to continue if the server won't
8595          * cooperate:
8596          */
8597         if (rcvd->max_ops < sent->max_ops)
8598                 return -EINVAL;
8599         if (rcvd->max_reqs == 0)
8600                 return -EINVAL;
8601         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8602                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8603         return 0;
8604 }
8605
8606 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8607                 struct nfs41_create_session_res *res)
8608 {
8609         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8610         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8611
8612         if (!(res->flags & SESSION4_BACK_CHAN))
8613                 goto out;
8614         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8615                 return -EINVAL;
8616         if (rcvd->max_resp_sz < sent->max_resp_sz)
8617                 return -EINVAL;
8618         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8619                 return -EINVAL;
8620         if (rcvd->max_ops > sent->max_ops)
8621                 return -EINVAL;
8622         if (rcvd->max_reqs > sent->max_reqs)
8623                 return -EINVAL;
8624 out:
8625         return 0;
8626 }
8627
8628 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8629                                      struct nfs41_create_session_res *res)
8630 {
8631         int ret;
8632
8633         ret = nfs4_verify_fore_channel_attrs(args, res);
8634         if (ret)
8635                 return ret;
8636         return nfs4_verify_back_channel_attrs(args, res);
8637 }
8638
8639 static void nfs4_update_session(struct nfs4_session *session,
8640                 struct nfs41_create_session_res *res)
8641 {
8642         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8643         /* Mark client id and session as being confirmed */
8644         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8645         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8646         session->flags = res->flags;
8647         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8648         if (res->flags & SESSION4_BACK_CHAN)
8649                 memcpy(&session->bc_attrs, &res->bc_attrs,
8650                                 sizeof(session->bc_attrs));
8651 }
8652
8653 static int _nfs4_proc_create_session(struct nfs_client *clp,
8654                 const struct cred *cred)
8655 {
8656         struct nfs4_session *session = clp->cl_session;
8657         struct nfs41_create_session_args args = {
8658                 .client = clp,
8659                 .clientid = clp->cl_clientid,
8660                 .seqid = clp->cl_seqid,
8661                 .cb_program = NFS4_CALLBACK,
8662         };
8663         struct nfs41_create_session_res res;
8664
8665         struct rpc_message msg = {
8666                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8667                 .rpc_argp = &args,
8668                 .rpc_resp = &res,
8669                 .rpc_cred = cred,
8670         };
8671         int status;
8672
8673         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8674         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8675
8676         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8677                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8678         trace_nfs4_create_session(clp, status);
8679
8680         switch (status) {
8681         case -NFS4ERR_STALE_CLIENTID:
8682         case -NFS4ERR_DELAY:
8683         case -ETIMEDOUT:
8684         case -EACCES:
8685         case -EAGAIN:
8686                 goto out;
8687         }
8688
8689         clp->cl_seqid++;
8690         if (!status) {
8691                 /* Verify the session's negotiated channel_attrs values */
8692                 status = nfs4_verify_channel_attrs(&args, &res);
8693                 /* Increment the clientid slot sequence id */
8694                 if (status)
8695                         goto out;
8696                 nfs4_update_session(session, &res);
8697         }
8698 out:
8699         return status;
8700 }
8701
8702 /*
8703  * Issues a CREATE_SESSION operation to the server.
8704  * It is the responsibility of the caller to verify the session is
8705  * expired before calling this routine.
8706  */
8707 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8708 {
8709         int status;
8710         unsigned *ptr;
8711         struct nfs4_session *session = clp->cl_session;
8712
8713         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8714
8715         status = _nfs4_proc_create_session(clp, cred);
8716         if (status)
8717                 goto out;
8718
8719         /* Init or reset the session slot tables */
8720         status = nfs4_setup_session_slot_tables(session);
8721         dprintk("slot table setup returned %d\n", status);
8722         if (status)
8723                 goto out;
8724
8725         ptr = (unsigned *)&session->sess_id.data[0];
8726         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8727                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8728 out:
8729         dprintk("<-- %s\n", __func__);
8730         return status;
8731 }
8732
8733 /*
8734  * Issue the over-the-wire RPC DESTROY_SESSION.
8735  * The caller must serialize access to this routine.
8736  */
8737 int nfs4_proc_destroy_session(struct nfs4_session *session,
8738                 const struct cred *cred)
8739 {
8740         struct rpc_message msg = {
8741                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8742                 .rpc_argp = session,
8743                 .rpc_cred = cred,
8744         };
8745         int status = 0;
8746
8747         dprintk("--> nfs4_proc_destroy_session\n");
8748
8749         /* session is still being setup */
8750         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8751                 return 0;
8752
8753         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8754                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8755         trace_nfs4_destroy_session(session->clp, status);
8756
8757         if (status)
8758                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8759                         "Session has been destroyed regardless...\n", status);
8760
8761         dprintk("<-- nfs4_proc_destroy_session\n");
8762         return status;
8763 }
8764
8765 /*
8766  * Renew the cl_session lease.
8767  */
8768 struct nfs4_sequence_data {
8769         struct nfs_client *clp;
8770         struct nfs4_sequence_args args;
8771         struct nfs4_sequence_res res;
8772 };
8773
8774 static void nfs41_sequence_release(void *data)
8775 {
8776         struct nfs4_sequence_data *calldata = data;
8777         struct nfs_client *clp = calldata->clp;
8778
8779         if (refcount_read(&clp->cl_count) > 1)
8780                 nfs4_schedule_state_renewal(clp);
8781         nfs_put_client(clp);
8782         kfree(calldata);
8783 }
8784
8785 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8786 {
8787         switch(task->tk_status) {
8788         case -NFS4ERR_DELAY:
8789                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8790                 return -EAGAIN;
8791         default:
8792                 nfs4_schedule_lease_recovery(clp);
8793         }
8794         return 0;
8795 }
8796
8797 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8798 {
8799         struct nfs4_sequence_data *calldata = data;
8800         struct nfs_client *clp = calldata->clp;
8801
8802         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8803                 return;
8804
8805         trace_nfs4_sequence(clp, task->tk_status);
8806         if (task->tk_status < 0) {
8807                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8808                 if (refcount_read(&clp->cl_count) == 1)
8809                         goto out;
8810
8811                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8812                         rpc_restart_call_prepare(task);
8813                         return;
8814                 }
8815         }
8816         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8817 out:
8818         dprintk("<-- %s\n", __func__);
8819 }
8820
8821 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8822 {
8823         struct nfs4_sequence_data *calldata = data;
8824         struct nfs_client *clp = calldata->clp;
8825         struct nfs4_sequence_args *args;
8826         struct nfs4_sequence_res *res;
8827
8828         args = task->tk_msg.rpc_argp;
8829         res = task->tk_msg.rpc_resp;
8830
8831         nfs4_setup_sequence(clp, args, res, task);
8832 }
8833
8834 static const struct rpc_call_ops nfs41_sequence_ops = {
8835         .rpc_call_done = nfs41_sequence_call_done,
8836         .rpc_call_prepare = nfs41_sequence_prepare,
8837         .rpc_release = nfs41_sequence_release,
8838 };
8839
8840 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8841                 const struct cred *cred,
8842                 struct nfs4_slot *slot,
8843                 bool is_privileged)
8844 {
8845         struct nfs4_sequence_data *calldata;
8846         struct rpc_message msg = {
8847                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8848                 .rpc_cred = cred,
8849         };
8850         struct rpc_task_setup task_setup_data = {
8851                 .rpc_client = clp->cl_rpcclient,
8852                 .rpc_message = &msg,
8853                 .callback_ops = &nfs41_sequence_ops,
8854                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8855         };
8856         struct rpc_task *ret;
8857
8858         ret = ERR_PTR(-EIO);
8859         if (!refcount_inc_not_zero(&clp->cl_count))
8860                 goto out_err;
8861
8862         ret = ERR_PTR(-ENOMEM);
8863         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8864         if (calldata == NULL)
8865                 goto out_put_clp;
8866         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8867         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8868         msg.rpc_argp = &calldata->args;
8869         msg.rpc_resp = &calldata->res;
8870         calldata->clp = clp;
8871         task_setup_data.callback_data = calldata;
8872
8873         ret = rpc_run_task(&task_setup_data);
8874         if (IS_ERR(ret))
8875                 goto out_err;
8876         return ret;
8877 out_put_clp:
8878         nfs_put_client(clp);
8879 out_err:
8880         nfs41_release_slot(slot);
8881         return ret;
8882 }
8883
8884 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
8885 {
8886         struct rpc_task *task;
8887         int ret = 0;
8888
8889         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8890                 return -EAGAIN;
8891         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8892         if (IS_ERR(task))
8893                 ret = PTR_ERR(task);
8894         else
8895                 rpc_put_task_async(task);
8896         dprintk("<-- %s status=%d\n", __func__, ret);
8897         return ret;
8898 }
8899
8900 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
8901 {
8902         struct rpc_task *task;
8903         int ret;
8904
8905         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8906         if (IS_ERR(task)) {
8907                 ret = PTR_ERR(task);
8908                 goto out;
8909         }
8910         ret = rpc_wait_for_completion_task(task);
8911         if (!ret)
8912                 ret = task->tk_status;
8913         rpc_put_task(task);
8914 out:
8915         dprintk("<-- %s status=%d\n", __func__, ret);
8916         return ret;
8917 }
8918
8919 struct nfs4_reclaim_complete_data {
8920         struct nfs_client *clp;
8921         struct nfs41_reclaim_complete_args arg;
8922         struct nfs41_reclaim_complete_res res;
8923 };
8924
8925 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8926 {
8927         struct nfs4_reclaim_complete_data *calldata = data;
8928
8929         nfs4_setup_sequence(calldata->clp,
8930                         &calldata->arg.seq_args,
8931                         &calldata->res.seq_res,
8932                         task);
8933 }
8934
8935 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8936 {
8937         switch(task->tk_status) {
8938         case 0:
8939                 wake_up_all(&clp->cl_lock_waitq);
8940                 /* Fallthrough */
8941         case -NFS4ERR_COMPLETE_ALREADY:
8942         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8943                 break;
8944         case -NFS4ERR_DELAY:
8945                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8946                 /* fall through */
8947         case -NFS4ERR_RETRY_UNCACHED_REP:
8948                 return -EAGAIN;
8949         case -NFS4ERR_BADSESSION:
8950         case -NFS4ERR_DEADSESSION:
8951         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8952                 break;
8953         default:
8954                 nfs4_schedule_lease_recovery(clp);
8955         }
8956         return 0;
8957 }
8958
8959 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8960 {
8961         struct nfs4_reclaim_complete_data *calldata = data;
8962         struct nfs_client *clp = calldata->clp;
8963         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8964
8965         dprintk("--> %s\n", __func__);
8966         if (!nfs41_sequence_done(task, res))
8967                 return;
8968
8969         trace_nfs4_reclaim_complete(clp, task->tk_status);
8970         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8971                 rpc_restart_call_prepare(task);
8972                 return;
8973         }
8974         dprintk("<-- %s\n", __func__);
8975 }
8976
8977 static void nfs4_free_reclaim_complete_data(void *data)
8978 {
8979         struct nfs4_reclaim_complete_data *calldata = data;
8980
8981         kfree(calldata);
8982 }
8983
8984 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8985         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8986         .rpc_call_done = nfs4_reclaim_complete_done,
8987         .rpc_release = nfs4_free_reclaim_complete_data,
8988 };
8989
8990 /*
8991  * Issue a global reclaim complete.
8992  */
8993 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8994                 const struct cred *cred)
8995 {
8996         struct nfs4_reclaim_complete_data *calldata;
8997         struct rpc_message msg = {
8998                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8999                 .rpc_cred = cred,
9000         };
9001         struct rpc_task_setup task_setup_data = {
9002                 .rpc_client = clp->cl_rpcclient,
9003                 .rpc_message = &msg,
9004                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9005                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9006         };
9007         int status = -ENOMEM;
9008
9009         dprintk("--> %s\n", __func__);
9010         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9011         if (calldata == NULL)
9012                 goto out;
9013         calldata->clp = clp;
9014         calldata->arg.one_fs = 0;
9015
9016         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9017         msg.rpc_argp = &calldata->arg;
9018         msg.rpc_resp = &calldata->res;
9019         task_setup_data.callback_data = calldata;
9020         status = nfs4_call_sync_custom(&task_setup_data);
9021 out:
9022         dprintk("<-- %s status=%d\n", __func__, status);
9023         return status;
9024 }
9025
9026 static void
9027 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9028 {
9029         struct nfs4_layoutget *lgp = calldata;
9030         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9031
9032         dprintk("--> %s\n", __func__);
9033         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9034                                 &lgp->res.seq_res, task);
9035         dprintk("<-- %s\n", __func__);
9036 }
9037
9038 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9039 {
9040         struct nfs4_layoutget *lgp = calldata;
9041
9042         dprintk("--> %s\n", __func__);
9043         nfs41_sequence_process(task, &lgp->res.seq_res);
9044         dprintk("<-- %s\n", __func__);
9045 }
9046
9047 static int
9048 nfs4_layoutget_handle_exception(struct rpc_task *task,
9049                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9050 {
9051         struct inode *inode = lgp->args.inode;
9052         struct nfs_server *server = NFS_SERVER(inode);
9053         struct pnfs_layout_hdr *lo;
9054         int nfs4err = task->tk_status;
9055         int err, status = 0;
9056         LIST_HEAD(head);
9057
9058         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9059
9060         nfs4_sequence_free_slot(&lgp->res.seq_res);
9061
9062         switch (nfs4err) {
9063         case 0:
9064                 goto out;
9065
9066         /*
9067          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9068          * on the file. set tk_status to -ENODATA to tell upper layer to
9069          * retry go inband.
9070          */
9071         case -NFS4ERR_LAYOUTUNAVAILABLE:
9072                 status = -ENODATA;
9073                 goto out;
9074         /*
9075          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9076          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9077          */
9078         case -NFS4ERR_BADLAYOUT:
9079                 status = -EOVERFLOW;
9080                 goto out;
9081         /*
9082          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9083          * (or clients) writing to the same RAID stripe except when
9084          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9085          *
9086          * Treat it like we would RECALLCONFLICT -- we retry for a little
9087          * while, and then eventually give up.
9088          */
9089         case -NFS4ERR_LAYOUTTRYLATER:
9090                 if (lgp->args.minlength == 0) {
9091                         status = -EOVERFLOW;
9092                         goto out;
9093                 }
9094                 status = -EBUSY;
9095                 break;
9096         case -NFS4ERR_RECALLCONFLICT:
9097                 status = -ERECALLCONFLICT;
9098                 break;
9099         case -NFS4ERR_DELEG_REVOKED:
9100         case -NFS4ERR_ADMIN_REVOKED:
9101         case -NFS4ERR_EXPIRED:
9102         case -NFS4ERR_BAD_STATEID:
9103                 exception->timeout = 0;
9104                 spin_lock(&inode->i_lock);
9105                 lo = NFS_I(inode)->layout;
9106                 /* If the open stateid was bad, then recover it. */
9107                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9108                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9109                         spin_unlock(&inode->i_lock);
9110                         exception->state = lgp->args.ctx->state;
9111                         exception->stateid = &lgp->args.stateid;
9112                         break;
9113                 }
9114
9115                 /*
9116                  * Mark the bad layout state as invalid, then retry
9117                  */
9118                 pnfs_mark_layout_stateid_invalid(lo, &head);
9119                 spin_unlock(&inode->i_lock);
9120                 nfs_commit_inode(inode, 0);
9121                 pnfs_free_lseg_list(&head);
9122                 status = -EAGAIN;
9123                 goto out;
9124         }
9125
9126         err = nfs4_handle_exception(server, nfs4err, exception);
9127         if (!status) {
9128                 if (exception->retry)
9129                         status = -EAGAIN;
9130                 else
9131                         status = err;
9132         }
9133 out:
9134         dprintk("<-- %s\n", __func__);
9135         return status;
9136 }
9137
9138 size_t max_response_pages(struct nfs_server *server)
9139 {
9140         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9141         return nfs_page_array_len(0, max_resp_sz);
9142 }
9143
9144 static void nfs4_layoutget_release(void *calldata)
9145 {
9146         struct nfs4_layoutget *lgp = calldata;
9147
9148         dprintk("--> %s\n", __func__);
9149         nfs4_sequence_free_slot(&lgp->res.seq_res);
9150         pnfs_layoutget_free(lgp);
9151         dprintk("<-- %s\n", __func__);
9152 }
9153
9154 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9155         .rpc_call_prepare = nfs4_layoutget_prepare,
9156         .rpc_call_done = nfs4_layoutget_done,
9157         .rpc_release = nfs4_layoutget_release,
9158 };
9159
9160 struct pnfs_layout_segment *
9161 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9162 {
9163         struct inode *inode = lgp->args.inode;
9164         struct nfs_server *server = NFS_SERVER(inode);
9165         struct rpc_task *task;
9166         struct rpc_message msg = {
9167                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9168                 .rpc_argp = &lgp->args,
9169                 .rpc_resp = &lgp->res,
9170                 .rpc_cred = lgp->cred,
9171         };
9172         struct rpc_task_setup task_setup_data = {
9173                 .rpc_client = server->client,
9174                 .rpc_message = &msg,
9175                 .callback_ops = &nfs4_layoutget_call_ops,
9176                 .callback_data = lgp,
9177                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
9178         };
9179         struct pnfs_layout_segment *lseg = NULL;
9180         struct nfs4_exception exception = {
9181                 .inode = inode,
9182                 .timeout = *timeout,
9183         };
9184         int status = 0;
9185
9186         dprintk("--> %s\n", __func__);
9187
9188         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9189         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9190
9191         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9192
9193         task = rpc_run_task(&task_setup_data);
9194
9195         status = rpc_wait_for_completion_task(task);
9196         if (status != 0)
9197                 goto out;
9198
9199         if (task->tk_status < 0) {
9200                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9201                 *timeout = exception.timeout;
9202         } else if (lgp->res.layoutp->len == 0) {
9203                 status = -EAGAIN;
9204                 *timeout = nfs4_update_delay(&exception.timeout);
9205         } else
9206                 lseg = pnfs_layout_process(lgp);
9207 out:
9208         trace_nfs4_layoutget(lgp->args.ctx,
9209                         &lgp->args.range,
9210                         &lgp->res.range,
9211                         &lgp->res.stateid,
9212                         status);
9213
9214         rpc_put_task(task);
9215         dprintk("<-- %s status=%d\n", __func__, status);
9216         if (status)
9217                 return ERR_PTR(status);
9218         return lseg;
9219 }
9220
9221 static void
9222 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9223 {
9224         struct nfs4_layoutreturn *lrp = calldata;
9225
9226         dprintk("--> %s\n", __func__);
9227         nfs4_setup_sequence(lrp->clp,
9228                         &lrp->args.seq_args,
9229                         &lrp->res.seq_res,
9230                         task);
9231         if (!pnfs_layout_is_valid(lrp->args.layout))
9232                 rpc_exit(task, 0);
9233 }
9234
9235 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9236 {
9237         struct nfs4_layoutreturn *lrp = calldata;
9238         struct nfs_server *server;
9239
9240         dprintk("--> %s\n", __func__);
9241
9242         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9243                 return;
9244
9245         /*
9246          * Was there an RPC level error? Assume the call succeeded,
9247          * and that we need to release the layout
9248          */
9249         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9250                 lrp->res.lrs_present = 0;
9251                 return;
9252         }
9253
9254         server = NFS_SERVER(lrp->args.inode);
9255         switch (task->tk_status) {
9256         case -NFS4ERR_OLD_STATEID:
9257                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9258                                         &lrp->args.range,
9259                                         lrp->args.inode))
9260                         goto out_restart;
9261                 /* Fallthrough */
9262         default:
9263                 task->tk_status = 0;
9264                 /* Fallthrough */
9265         case 0:
9266                 break;
9267         case -NFS4ERR_DELAY:
9268                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9269                         break;
9270                 goto out_restart;
9271         }
9272         dprintk("<-- %s\n", __func__);
9273         return;
9274 out_restart:
9275         task->tk_status = 0;
9276         nfs4_sequence_free_slot(&lrp->res.seq_res);
9277         rpc_restart_call_prepare(task);
9278 }
9279
9280 static void nfs4_layoutreturn_release(void *calldata)
9281 {
9282         struct nfs4_layoutreturn *lrp = calldata;
9283         struct pnfs_layout_hdr *lo = lrp->args.layout;
9284
9285         dprintk("--> %s\n", __func__);
9286         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9287                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9288         nfs4_sequence_free_slot(&lrp->res.seq_res);
9289         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9290                 lrp->ld_private.ops->free(&lrp->ld_private);
9291         pnfs_put_layout_hdr(lrp->args.layout);
9292         nfs_iput_and_deactive(lrp->inode);
9293         put_cred(lrp->cred);
9294         kfree(calldata);
9295         dprintk("<-- %s\n", __func__);
9296 }
9297
9298 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9299         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9300         .rpc_call_done = nfs4_layoutreturn_done,
9301         .rpc_release = nfs4_layoutreturn_release,
9302 };
9303
9304 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9305 {
9306         struct rpc_task *task;
9307         struct rpc_message msg = {
9308                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9309                 .rpc_argp = &lrp->args,
9310                 .rpc_resp = &lrp->res,
9311                 .rpc_cred = lrp->cred,
9312         };
9313         struct rpc_task_setup task_setup_data = {
9314                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9315                 .rpc_message = &msg,
9316                 .callback_ops = &nfs4_layoutreturn_call_ops,
9317                 .callback_data = lrp,
9318         };
9319         int status = 0;
9320
9321         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9322                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9323                         &task_setup_data.rpc_client, &msg);
9324
9325         dprintk("--> %s\n", __func__);
9326         if (!sync) {
9327                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9328                 if (!lrp->inode) {
9329                         nfs4_layoutreturn_release(lrp);
9330                         return -EAGAIN;
9331                 }
9332                 task_setup_data.flags |= RPC_TASK_ASYNC;
9333         }
9334         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9335         task = rpc_run_task(&task_setup_data);
9336         if (IS_ERR(task))
9337                 return PTR_ERR(task);
9338         if (sync)
9339                 status = task->tk_status;
9340         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9341         dprintk("<-- %s status=%d\n", __func__, status);
9342         rpc_put_task(task);
9343         return status;
9344 }
9345
9346 static int
9347 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9348                 struct pnfs_device *pdev,
9349                 const struct cred *cred)
9350 {
9351         struct nfs4_getdeviceinfo_args args = {
9352                 .pdev = pdev,
9353                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9354                         NOTIFY_DEVICEID4_DELETE,
9355         };
9356         struct nfs4_getdeviceinfo_res res = {
9357                 .pdev = pdev,
9358         };
9359         struct rpc_message msg = {
9360                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9361                 .rpc_argp = &args,
9362                 .rpc_resp = &res,
9363                 .rpc_cred = cred,
9364         };
9365         int status;
9366
9367         dprintk("--> %s\n", __func__);
9368         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9369         if (res.notification & ~args.notify_types)
9370                 dprintk("%s: unsupported notification\n", __func__);
9371         if (res.notification != args.notify_types)
9372                 pdev->nocache = 1;
9373
9374         dprintk("<-- %s status=%d\n", __func__, status);
9375
9376         return status;
9377 }
9378
9379 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9380                 struct pnfs_device *pdev,
9381                 const struct cred *cred)
9382 {
9383         struct nfs4_exception exception = { };
9384         int err;
9385
9386         do {
9387                 err = nfs4_handle_exception(server,
9388                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9389                                         &exception);
9390         } while (exception.retry);
9391         return err;
9392 }
9393 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9394
9395 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9396 {
9397         struct nfs4_layoutcommit_data *data = calldata;
9398         struct nfs_server *server = NFS_SERVER(data->args.inode);
9399
9400         nfs4_setup_sequence(server->nfs_client,
9401                         &data->args.seq_args,
9402                         &data->res.seq_res,
9403                         task);
9404 }
9405
9406 static void
9407 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9408 {
9409         struct nfs4_layoutcommit_data *data = calldata;
9410         struct nfs_server *server = NFS_SERVER(data->args.inode);
9411
9412         if (!nfs41_sequence_done(task, &data->res.seq_res))
9413                 return;
9414
9415         switch (task->tk_status) { /* Just ignore these failures */
9416         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9417         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9418         case -NFS4ERR_BADLAYOUT:     /* no layout */
9419         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9420                 task->tk_status = 0;
9421         case 0:
9422                 break;
9423         default:
9424                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9425                         rpc_restart_call_prepare(task);
9426                         return;
9427                 }
9428         }
9429 }
9430
9431 static void nfs4_layoutcommit_release(void *calldata)
9432 {
9433         struct nfs4_layoutcommit_data *data = calldata;
9434
9435         pnfs_cleanup_layoutcommit(data);
9436         nfs_post_op_update_inode_force_wcc(data->args.inode,
9437                                            data->res.fattr);
9438         put_cred(data->cred);
9439         nfs_iput_and_deactive(data->inode);
9440         kfree(data);
9441 }
9442
9443 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9444         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9445         .rpc_call_done = nfs4_layoutcommit_done,
9446         .rpc_release = nfs4_layoutcommit_release,
9447 };
9448
9449 int
9450 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9451 {
9452         struct rpc_message msg = {
9453                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9454                 .rpc_argp = &data->args,
9455                 .rpc_resp = &data->res,
9456                 .rpc_cred = data->cred,
9457         };
9458         struct rpc_task_setup task_setup_data = {
9459                 .task = &data->task,
9460                 .rpc_client = NFS_CLIENT(data->args.inode),
9461                 .rpc_message = &msg,
9462                 .callback_ops = &nfs4_layoutcommit_ops,
9463                 .callback_data = data,
9464         };
9465         struct rpc_task *task;
9466         int status = 0;
9467
9468         dprintk("NFS: initiating layoutcommit call. sync %d "
9469                 "lbw: %llu inode %lu\n", sync,
9470                 data->args.lastbytewritten,
9471                 data->args.inode->i_ino);
9472
9473         if (!sync) {
9474                 data->inode = nfs_igrab_and_active(data->args.inode);
9475                 if (data->inode == NULL) {
9476                         nfs4_layoutcommit_release(data);
9477                         return -EAGAIN;
9478                 }
9479                 task_setup_data.flags = RPC_TASK_ASYNC;
9480         }
9481         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9482         task = rpc_run_task(&task_setup_data);
9483         if (IS_ERR(task))
9484                 return PTR_ERR(task);
9485         if (sync)
9486                 status = task->tk_status;
9487         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9488         dprintk("%s: status %d\n", __func__, status);
9489         rpc_put_task(task);
9490         return status;
9491 }
9492
9493 /*
9494  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9495  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9496  */
9497 static int
9498 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9499                     struct nfs_fsinfo *info,
9500                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9501 {
9502         struct nfs41_secinfo_no_name_args args = {
9503                 .style = SECINFO_STYLE_CURRENT_FH,
9504         };
9505         struct nfs4_secinfo_res res = {
9506                 .flavors = flavors,
9507         };
9508         struct rpc_message msg = {
9509                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9510                 .rpc_argp = &args,
9511                 .rpc_resp = &res,
9512         };
9513         struct rpc_clnt *clnt = server->client;
9514         struct nfs4_call_sync_data data = {
9515                 .seq_server = server,
9516                 .seq_args = &args.seq_args,
9517                 .seq_res = &res.seq_res,
9518         };
9519         struct rpc_task_setup task_setup = {
9520                 .rpc_client = server->client,
9521                 .rpc_message = &msg,
9522                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9523                 .callback_data = &data,
9524                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9525         };
9526         const struct cred *cred = NULL;
9527         int status;
9528
9529         if (use_integrity) {
9530                 clnt = server->nfs_client->cl_rpcclient;
9531                 task_setup.rpc_client = clnt;
9532
9533                 cred = nfs4_get_clid_cred(server->nfs_client);
9534                 msg.rpc_cred = cred;
9535         }
9536
9537         dprintk("--> %s\n", __func__);
9538         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9539         status = nfs4_call_sync_custom(&task_setup);
9540         dprintk("<-- %s status=%d\n", __func__, status);
9541
9542         put_cred(cred);
9543
9544         return status;
9545 }
9546
9547 static int
9548 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9549                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9550 {
9551         struct nfs4_exception exception = {
9552                 .interruptible = true,
9553         };
9554         int err;
9555         do {
9556                 /* first try using integrity protection */
9557                 err = -NFS4ERR_WRONGSEC;
9558
9559                 /* try to use integrity protection with machine cred */
9560                 if (_nfs4_is_integrity_protected(server->nfs_client))
9561                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9562                                                           flavors, true);
9563
9564                 /*
9565                  * if unable to use integrity protection, or SECINFO with
9566                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9567                  * disallowed by spec, but exists in deployed servers) use
9568                  * the current filesystem's rpc_client and the user cred.
9569                  */
9570                 if (err == -NFS4ERR_WRONGSEC)
9571                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9572                                                           flavors, false);
9573
9574                 switch (err) {
9575                 case 0:
9576                 case -NFS4ERR_WRONGSEC:
9577                 case -ENOTSUPP:
9578                         goto out;
9579                 default:
9580                         err = nfs4_handle_exception(server, err, &exception);
9581                 }
9582         } while (exception.retry);
9583 out:
9584         return err;
9585 }
9586
9587 static int
9588 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9589                     struct nfs_fsinfo *info)
9590 {
9591         int err;
9592         struct page *page;
9593         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9594         struct nfs4_secinfo_flavors *flavors;
9595         struct nfs4_secinfo4 *secinfo;
9596         int i;
9597
9598         page = alloc_page(GFP_KERNEL);
9599         if (!page) {
9600                 err = -ENOMEM;
9601                 goto out;
9602         }
9603
9604         flavors = page_address(page);
9605         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9606
9607         /*
9608          * Fall back on "guess and check" method if
9609          * the server doesn't support SECINFO_NO_NAME
9610          */
9611         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9612                 err = nfs4_find_root_sec(server, fhandle, info);
9613                 goto out_freepage;
9614         }
9615         if (err)
9616                 goto out_freepage;
9617
9618         for (i = 0; i < flavors->num_flavors; i++) {
9619                 secinfo = &flavors->flavors[i];
9620
9621                 switch (secinfo->flavor) {
9622                 case RPC_AUTH_NULL:
9623                 case RPC_AUTH_UNIX:
9624                 case RPC_AUTH_GSS:
9625                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9626                                         &secinfo->flavor_info);
9627                         break;
9628                 default:
9629                         flavor = RPC_AUTH_MAXFLAVOR;
9630                         break;
9631                 }
9632
9633                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9634                         flavor = RPC_AUTH_MAXFLAVOR;
9635
9636                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9637                         err = nfs4_lookup_root_sec(server, fhandle,
9638                                                    info, flavor);
9639                         if (!err)
9640                                 break;
9641                 }
9642         }
9643
9644         if (flavor == RPC_AUTH_MAXFLAVOR)
9645                 err = -EPERM;
9646
9647 out_freepage:
9648         put_page(page);
9649         if (err == -EACCES)
9650                 return -EPERM;
9651 out:
9652         return err;
9653 }
9654
9655 static int _nfs41_test_stateid(struct nfs_server *server,
9656                 nfs4_stateid *stateid,
9657                 const struct cred *cred)
9658 {
9659         int status;
9660         struct nfs41_test_stateid_args args = {
9661                 .stateid = stateid,
9662         };
9663         struct nfs41_test_stateid_res res;
9664         struct rpc_message msg = {
9665                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9666                 .rpc_argp = &args,
9667                 .rpc_resp = &res,
9668                 .rpc_cred = cred,
9669         };
9670         struct rpc_clnt *rpc_client = server->client;
9671
9672         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9673                 &rpc_client, &msg);
9674
9675         dprintk("NFS call  test_stateid %p\n", stateid);
9676         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9677         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9678                         &args.seq_args, &res.seq_res);
9679         if (status != NFS_OK) {
9680                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9681                 return status;
9682         }
9683         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9684         return -res.status;
9685 }
9686
9687 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9688                 int err, struct nfs4_exception *exception)
9689 {
9690         exception->retry = 0;
9691         switch(err) {
9692         case -NFS4ERR_DELAY:
9693         case -NFS4ERR_RETRY_UNCACHED_REP:
9694                 nfs4_handle_exception(server, err, exception);
9695                 break;
9696         case -NFS4ERR_BADSESSION:
9697         case -NFS4ERR_BADSLOT:
9698         case -NFS4ERR_BAD_HIGH_SLOT:
9699         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9700         case -NFS4ERR_DEADSESSION:
9701                 nfs4_do_handle_exception(server, err, exception);
9702         }
9703 }
9704
9705 /**
9706  * nfs41_test_stateid - perform a TEST_STATEID operation
9707  *
9708  * @server: server / transport on which to perform the operation
9709  * @stateid: state ID to test
9710  * @cred: credential
9711  *
9712  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9713  * Otherwise a negative NFS4ERR value is returned if the operation
9714  * failed or the state ID is not currently valid.
9715  */
9716 static int nfs41_test_stateid(struct nfs_server *server,
9717                 nfs4_stateid *stateid,
9718                 const struct cred *cred)
9719 {
9720         struct nfs4_exception exception = {
9721                 .interruptible = true,
9722         };
9723         int err;
9724         do {
9725                 err = _nfs41_test_stateid(server, stateid, cred);
9726                 nfs4_handle_delay_or_session_error(server, err, &exception);
9727         } while (exception.retry);
9728         return err;
9729 }
9730
9731 struct nfs_free_stateid_data {
9732         struct nfs_server *server;
9733         struct nfs41_free_stateid_args args;
9734         struct nfs41_free_stateid_res res;
9735 };
9736
9737 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9738 {
9739         struct nfs_free_stateid_data *data = calldata;
9740         nfs4_setup_sequence(data->server->nfs_client,
9741                         &data->args.seq_args,
9742                         &data->res.seq_res,
9743                         task);
9744 }
9745
9746 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9747 {
9748         struct nfs_free_stateid_data *data = calldata;
9749
9750         nfs41_sequence_done(task, &data->res.seq_res);
9751
9752         switch (task->tk_status) {
9753         case -NFS4ERR_DELAY:
9754                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9755                         rpc_restart_call_prepare(task);
9756         }
9757 }
9758
9759 static void nfs41_free_stateid_release(void *calldata)
9760 {
9761         kfree(calldata);
9762 }
9763
9764 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9765         .rpc_call_prepare = nfs41_free_stateid_prepare,
9766         .rpc_call_done = nfs41_free_stateid_done,
9767         .rpc_release = nfs41_free_stateid_release,
9768 };
9769
9770 /**
9771  * nfs41_free_stateid - perform a FREE_STATEID operation
9772  *
9773  * @server: server / transport on which to perform the operation
9774  * @stateid: state ID to release
9775  * @cred: credential
9776  * @privileged: set to true if this call needs to be privileged
9777  *
9778  * Note: this function is always asynchronous.
9779  */
9780 static int nfs41_free_stateid(struct nfs_server *server,
9781                 const nfs4_stateid *stateid,
9782                 const struct cred *cred,
9783                 bool privileged)
9784 {
9785         struct rpc_message msg = {
9786                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9787                 .rpc_cred = cred,
9788         };
9789         struct rpc_task_setup task_setup = {
9790                 .rpc_client = server->client,
9791                 .rpc_message = &msg,
9792                 .callback_ops = &nfs41_free_stateid_ops,
9793                 .flags = RPC_TASK_ASYNC,
9794         };
9795         struct nfs_free_stateid_data *data;
9796         struct rpc_task *task;
9797
9798         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9799                 &task_setup.rpc_client, &msg);
9800
9801         dprintk("NFS call  free_stateid %p\n", stateid);
9802         data = kmalloc(sizeof(*data), GFP_NOFS);
9803         if (!data)
9804                 return -ENOMEM;
9805         data->server = server;
9806         nfs4_stateid_copy(&data->args.stateid, stateid);
9807
9808         task_setup.callback_data = data;
9809
9810         msg.rpc_argp = &data->args;
9811         msg.rpc_resp = &data->res;
9812         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9813         task = rpc_run_task(&task_setup);
9814         if (IS_ERR(task))
9815                 return PTR_ERR(task);
9816         rpc_put_task(task);
9817         return 0;
9818 }
9819
9820 static void
9821 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9822 {
9823         const struct cred *cred = lsp->ls_state->owner->so_cred;
9824
9825         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9826         nfs4_free_lock_state(server, lsp);
9827 }
9828
9829 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9830                 const nfs4_stateid *s2)
9831 {
9832         if (s1->type != s2->type)
9833                 return false;
9834
9835         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9836                 return false;
9837
9838         if (s1->seqid == s2->seqid)
9839                 return true;
9840
9841         return s1->seqid == 0 || s2->seqid == 0;
9842 }
9843
9844 #endif /* CONFIG_NFS_V4_1 */
9845
9846 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9847                 const nfs4_stateid *s2)
9848 {
9849         return nfs4_stateid_match(s1, s2);
9850 }
9851
9852
9853 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9854         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9855         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9856         .recover_open   = nfs4_open_reclaim,
9857         .recover_lock   = nfs4_lock_reclaim,
9858         .establish_clid = nfs4_init_clientid,
9859         .detect_trunking = nfs40_discover_server_trunking,
9860 };
9861
9862 #if defined(CONFIG_NFS_V4_1)
9863 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9864         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9865         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9866         .recover_open   = nfs4_open_reclaim,
9867         .recover_lock   = nfs4_lock_reclaim,
9868         .establish_clid = nfs41_init_clientid,
9869         .reclaim_complete = nfs41_proc_reclaim_complete,
9870         .detect_trunking = nfs41_discover_server_trunking,
9871 };
9872 #endif /* CONFIG_NFS_V4_1 */
9873
9874 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9875         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9876         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9877         .recover_open   = nfs40_open_expired,
9878         .recover_lock   = nfs4_lock_expired,
9879         .establish_clid = nfs4_init_clientid,
9880 };
9881
9882 #if defined(CONFIG_NFS_V4_1)
9883 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9884         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9885         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9886         .recover_open   = nfs41_open_expired,
9887         .recover_lock   = nfs41_lock_expired,
9888         .establish_clid = nfs41_init_clientid,
9889 };
9890 #endif /* CONFIG_NFS_V4_1 */
9891
9892 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9893         .sched_state_renewal = nfs4_proc_async_renew,
9894         .get_state_renewal_cred = nfs4_get_renew_cred,
9895         .renew_lease = nfs4_proc_renew,
9896 };
9897
9898 #if defined(CONFIG_NFS_V4_1)
9899 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9900         .sched_state_renewal = nfs41_proc_async_sequence,
9901         .get_state_renewal_cred = nfs4_get_machine_cred,
9902         .renew_lease = nfs4_proc_sequence,
9903 };
9904 #endif
9905
9906 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9907         .get_locations = _nfs40_proc_get_locations,
9908         .fsid_present = _nfs40_proc_fsid_present,
9909 };
9910
9911 #if defined(CONFIG_NFS_V4_1)
9912 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9913         .get_locations = _nfs41_proc_get_locations,
9914         .fsid_present = _nfs41_proc_fsid_present,
9915 };
9916 #endif  /* CONFIG_NFS_V4_1 */
9917
9918 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9919         .minor_version = 0,
9920         .init_caps = NFS_CAP_READDIRPLUS
9921                 | NFS_CAP_ATOMIC_OPEN
9922                 | NFS_CAP_POSIX_LOCK,
9923         .init_client = nfs40_init_client,
9924         .shutdown_client = nfs40_shutdown_client,
9925         .match_stateid = nfs4_match_stateid,
9926         .find_root_sec = nfs4_find_root_sec,
9927         .free_lock_state = nfs4_release_lockowner,
9928         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9929         .alloc_seqid = nfs_alloc_seqid,
9930         .call_sync_ops = &nfs40_call_sync_ops,
9931         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9932         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9933         .state_renewal_ops = &nfs40_state_renewal_ops,
9934         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9935 };
9936
9937 #if defined(CONFIG_NFS_V4_1)
9938 static struct nfs_seqid *
9939 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9940 {
9941         return NULL;
9942 }
9943
9944 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9945         .minor_version = 1,
9946         .init_caps = NFS_CAP_READDIRPLUS
9947                 | NFS_CAP_ATOMIC_OPEN
9948                 | NFS_CAP_POSIX_LOCK
9949                 | NFS_CAP_STATEID_NFSV41
9950                 | NFS_CAP_ATOMIC_OPEN_V1
9951                 | NFS_CAP_LGOPEN,
9952         .init_client = nfs41_init_client,
9953         .shutdown_client = nfs41_shutdown_client,
9954         .match_stateid = nfs41_match_stateid,
9955         .find_root_sec = nfs41_find_root_sec,
9956         .free_lock_state = nfs41_free_lock_state,
9957         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9958         .alloc_seqid = nfs_alloc_no_seqid,
9959         .session_trunk = nfs4_test_session_trunk,
9960         .call_sync_ops = &nfs41_call_sync_ops,
9961         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9962         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9963         .state_renewal_ops = &nfs41_state_renewal_ops,
9964         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9965 };
9966 #endif
9967
9968 #if defined(CONFIG_NFS_V4_2)
9969 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9970         .minor_version = 2,
9971         .init_caps = NFS_CAP_READDIRPLUS
9972                 | NFS_CAP_ATOMIC_OPEN
9973                 | NFS_CAP_POSIX_LOCK
9974                 | NFS_CAP_STATEID_NFSV41
9975                 | NFS_CAP_ATOMIC_OPEN_V1
9976                 | NFS_CAP_LGOPEN
9977                 | NFS_CAP_ALLOCATE
9978                 | NFS_CAP_COPY
9979                 | NFS_CAP_OFFLOAD_CANCEL
9980                 | NFS_CAP_COPY_NOTIFY
9981                 | NFS_CAP_DEALLOCATE
9982                 | NFS_CAP_SEEK
9983                 | NFS_CAP_LAYOUTSTATS
9984                 | NFS_CAP_CLONE
9985                 | NFS_CAP_LAYOUTERROR,
9986         .init_client = nfs41_init_client,
9987         .shutdown_client = nfs41_shutdown_client,
9988         .match_stateid = nfs41_match_stateid,
9989         .find_root_sec = nfs41_find_root_sec,
9990         .free_lock_state = nfs41_free_lock_state,
9991         .call_sync_ops = &nfs41_call_sync_ops,
9992         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9993         .alloc_seqid = nfs_alloc_no_seqid,
9994         .session_trunk = nfs4_test_session_trunk,
9995         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9996         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9997         .state_renewal_ops = &nfs41_state_renewal_ops,
9998         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9999 };
10000 #endif
10001
10002 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10003         [0] = &nfs_v4_0_minor_ops,
10004 #if defined(CONFIG_NFS_V4_1)
10005         [1] = &nfs_v4_1_minor_ops,
10006 #endif
10007 #if defined(CONFIG_NFS_V4_2)
10008         [2] = &nfs_v4_2_minor_ops,
10009 #endif
10010 };
10011
10012 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10013 {
10014         ssize_t error, error2;
10015
10016         error = generic_listxattr(dentry, list, size);
10017         if (error < 0)
10018                 return error;
10019         if (list) {
10020                 list += error;
10021                 size -= error;
10022         }
10023
10024         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10025         if (error2 < 0)
10026                 return error2;
10027         return error + error2;
10028 }
10029
10030 static const struct inode_operations nfs4_dir_inode_operations = {
10031         .create         = nfs_create,
10032         .lookup         = nfs_lookup,
10033         .atomic_open    = nfs_atomic_open,
10034         .link           = nfs_link,
10035         .unlink         = nfs_unlink,
10036         .symlink        = nfs_symlink,
10037         .mkdir          = nfs_mkdir,
10038         .rmdir          = nfs_rmdir,
10039         .mknod          = nfs_mknod,
10040         .rename         = nfs_rename,
10041         .permission     = nfs_permission,
10042         .getattr        = nfs_getattr,
10043         .setattr        = nfs_setattr,
10044         .listxattr      = nfs4_listxattr,
10045 };
10046
10047 static const struct inode_operations nfs4_file_inode_operations = {
10048         .permission     = nfs_permission,
10049         .getattr        = nfs_getattr,
10050         .setattr        = nfs_setattr,
10051         .listxattr      = nfs4_listxattr,
10052 };
10053
10054 const struct nfs_rpc_ops nfs_v4_clientops = {
10055         .version        = 4,                    /* protocol version */
10056         .dentry_ops     = &nfs4_dentry_operations,
10057         .dir_inode_ops  = &nfs4_dir_inode_operations,
10058         .file_inode_ops = &nfs4_file_inode_operations,
10059         .file_ops       = &nfs4_file_operations,
10060         .getroot        = nfs4_proc_get_root,
10061         .submount       = nfs4_submount,
10062         .try_get_tree   = nfs4_try_get_tree,
10063         .getattr        = nfs4_proc_getattr,
10064         .setattr        = nfs4_proc_setattr,
10065         .lookup         = nfs4_proc_lookup,
10066         .lookupp        = nfs4_proc_lookupp,
10067         .access         = nfs4_proc_access,
10068         .readlink       = nfs4_proc_readlink,
10069         .create         = nfs4_proc_create,
10070         .remove         = nfs4_proc_remove,
10071         .unlink_setup   = nfs4_proc_unlink_setup,
10072         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10073         .unlink_done    = nfs4_proc_unlink_done,
10074         .rename_setup   = nfs4_proc_rename_setup,
10075         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10076         .rename_done    = nfs4_proc_rename_done,
10077         .link           = nfs4_proc_link,
10078         .symlink        = nfs4_proc_symlink,
10079         .mkdir          = nfs4_proc_mkdir,
10080         .rmdir          = nfs4_proc_rmdir,
10081         .readdir        = nfs4_proc_readdir,
10082         .mknod          = nfs4_proc_mknod,
10083         .statfs         = nfs4_proc_statfs,
10084         .fsinfo         = nfs4_proc_fsinfo,
10085         .pathconf       = nfs4_proc_pathconf,
10086         .set_capabilities = nfs4_server_capabilities,
10087         .decode_dirent  = nfs4_decode_dirent,
10088         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10089         .read_setup     = nfs4_proc_read_setup,
10090         .read_done      = nfs4_read_done,
10091         .write_setup    = nfs4_proc_write_setup,
10092         .write_done     = nfs4_write_done,
10093         .commit_setup   = nfs4_proc_commit_setup,
10094         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10095         .commit_done    = nfs4_commit_done,
10096         .lock           = nfs4_proc_lock,
10097         .clear_acl_cache = nfs4_zap_acl_attr,
10098         .close_context  = nfs4_close_context,
10099         .open_context   = nfs4_atomic_open,
10100         .have_delegation = nfs4_have_delegation,
10101         .alloc_client   = nfs4_alloc_client,
10102         .init_client    = nfs4_init_client,
10103         .free_client    = nfs4_free_client,
10104         .create_server  = nfs4_create_server,
10105         .clone_server   = nfs_clone_server,
10106 };
10107
10108 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10109         .name   = XATTR_NAME_NFSV4_ACL,
10110         .list   = nfs4_xattr_list_nfs4_acl,
10111         .get    = nfs4_xattr_get_nfs4_acl,
10112         .set    = nfs4_xattr_set_nfs4_acl,
10113 };
10114
10115 const struct xattr_handler *nfs4_xattr_handlers[] = {
10116         &nfs4_xattr_nfs4_acl_handler,
10117 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10118         &nfs4_xattr_nfs4_label_handler,
10119 #endif
10120         NULL
10121 };
10122
10123 /*
10124  * Local variables:
10125  *  c-basic-offset: 8
10126  * End:
10127  */