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/nfs_idmap.h>
  55 #include <linux/xattr.h>
  56 #include <linux/utsname.h>
  57 #include <linux/freezer.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 "nfs4session.h"
  67 #include "fscache.h"
  68
  69 #include "nfs4trace.h"
  70
  71 #define NFSDBG_FACILITY         NFSDBG_PROC
  72
  73 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  74 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  75
  76 struct nfs4_opendata;
  77 static int _nfs4_proc_open(struct nfs4_opendata *data);
  78 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  79 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  80 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
  81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  85                             struct nfs_fattr *fattr, struct iattr *sattr,
  86                             struct nfs4_state *state, struct nfs4_label *ilabel,
  87                             struct nfs4_label *olabel);
  88 #ifdef CONFIG_NFS_V4_1
  89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
  90                 struct rpc_cred *);
  91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
  92                 struct rpc_cred *);
  93 #endif
  94
  95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
  96 static inline struct nfs4_label *
  97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
  98         struct iattr *sattr, struct nfs4_label *label)
  99 {
 100         int err;
 101
 102         if (label == NULL)
 103                 return NULL;
 104
 105         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 106                 return NULL;
 107
 108         err = security_dentry_init_security(dentry, sattr->ia_mode,
 109                                 &dentry->d_name, (void **)&label->label, &label->len);
 110         if (err == 0)
 111                 return label;
 112
 113         return NULL;
 114 }
 115 static inline void
 116 nfs4_label_release_security(struct nfs4_label *label)
 117 {
 118         if (label)
 119                 security_release_secctx(label->label, label->len);
 120 }
 121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 122 {
 123         if (label)
 124                 return server->attr_bitmask;
 125
 126         return server->attr_bitmask_nl;
 127 }
 128 #else
 129 static inline struct nfs4_label *
 130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 131         struct iattr *sattr, struct nfs4_label *l)
 132 { return NULL; }
 133 static inline void
 134 nfs4_label_release_security(struct nfs4_label *label)
 135 { return; }
 136 static inline u32 *
 137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 138 { return server->attr_bitmask; }
 139 #endif
 140
 141 /* Prevent leaks of NFSv4 errors into userland */
 142 static int nfs4_map_errors(int err)
 143 {
 144         if (err >= -1000)
 145                 return err;
 146         switch (err) {
 147         case -NFS4ERR_RESOURCE:
 148         case -NFS4ERR_LAYOUTTRYLATER:
 149         case -NFS4ERR_RECALLCONFLICT:
 150                 return -EREMOTEIO;
 151         case -NFS4ERR_WRONGSEC:
 152         case -NFS4ERR_WRONG_CRED:
 153                 return -EPERM;
 154         case -NFS4ERR_BADOWNER:
 155         case -NFS4ERR_BADNAME:
 156                 return -EINVAL;
 157         case -NFS4ERR_SHARE_DENIED:
 158                 return -EACCES;
 159         case -NFS4ERR_MINOR_VERS_MISMATCH:
 160                 return -EPROTONOSUPPORT;
 161         case -NFS4ERR_FILE_OPEN:
 162                 return -EBUSY;
 163         default:
 164                 dprintk("%s could not handle NFSv4 error %d\n",
 165                                 __func__, -err);
 166                 break;
 167         }
 168         return -EIO;
 169 }
 170
 171 /*
 172  * This is our standard bitmap for GETATTR requests.
 173  */
 174 const u32 nfs4_fattr_bitmap[3] = {
 175         FATTR4_WORD0_TYPE
 176         | FATTR4_WORD0_CHANGE
 177         | FATTR4_WORD0_SIZE
 178         | FATTR4_WORD0_FSID
 179         | FATTR4_WORD0_FILEID,
 180         FATTR4_WORD1_MODE
 181         | FATTR4_WORD1_NUMLINKS
 182         | FATTR4_WORD1_OWNER
 183         | FATTR4_WORD1_OWNER_GROUP
 184         | FATTR4_WORD1_RAWDEV
 185         | FATTR4_WORD1_SPACE_USED
 186         | FATTR4_WORD1_TIME_ACCESS
 187         | FATTR4_WORD1_TIME_METADATA
 188         | FATTR4_WORD1_TIME_MODIFY,
 189 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 190         FATTR4_WORD2_SECURITY_LABEL
 191 #endif
 192 };
 193
 194 static const u32 nfs4_pnfs_open_bitmap[3] = {
 195         FATTR4_WORD0_TYPE
 196         | FATTR4_WORD0_CHANGE
 197         | FATTR4_WORD0_SIZE
 198         | FATTR4_WORD0_FSID
 199         | FATTR4_WORD0_FILEID,
 200         FATTR4_WORD1_MODE
 201         | FATTR4_WORD1_NUMLINKS
 202         | FATTR4_WORD1_OWNER
 203         | FATTR4_WORD1_OWNER_GROUP
 204         | FATTR4_WORD1_RAWDEV
 205         | FATTR4_WORD1_SPACE_USED
 206         | FATTR4_WORD1_TIME_ACCESS
 207         | FATTR4_WORD1_TIME_METADATA
 208         | FATTR4_WORD1_TIME_MODIFY,
 209         FATTR4_WORD2_MDSTHRESHOLD
 210 };
 211
 212 static const u32 nfs4_open_noattr_bitmap[3] = {
 213         FATTR4_WORD0_TYPE
 214         | FATTR4_WORD0_CHANGE
 215         | FATTR4_WORD0_FILEID,
 216 };
 217
 218 const u32 nfs4_statfs_bitmap[3] = {
 219         FATTR4_WORD0_FILES_AVAIL
 220         | FATTR4_WORD0_FILES_FREE
 221         | FATTR4_WORD0_FILES_TOTAL,
 222         FATTR4_WORD1_SPACE_AVAIL
 223         | FATTR4_WORD1_SPACE_FREE
 224         | FATTR4_WORD1_SPACE_TOTAL
 225 };
 226
 227 const u32 nfs4_pathconf_bitmap[3] = {
 228         FATTR4_WORD0_MAXLINK
 229         | FATTR4_WORD0_MAXNAME,
 230         0
 231 };
 232
 233 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 234                         | FATTR4_WORD0_MAXREAD
 235                         | FATTR4_WORD0_MAXWRITE
 236                         | FATTR4_WORD0_LEASE_TIME,
 237                         FATTR4_WORD1_TIME_DELTA
 238                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 239                         FATTR4_WORD2_LAYOUT_BLKSIZE
 240 };
 241
 242 const u32 nfs4_fs_locations_bitmap[3] = {
 243         FATTR4_WORD0_TYPE
 244         | FATTR4_WORD0_CHANGE
 245         | FATTR4_WORD0_SIZE
 246         | FATTR4_WORD0_FSID
 247         | FATTR4_WORD0_FILEID
 248         | FATTR4_WORD0_FS_LOCATIONS,
 249         FATTR4_WORD1_MODE
 250         | FATTR4_WORD1_NUMLINKS
 251         | FATTR4_WORD1_OWNER
 252         | FATTR4_WORD1_OWNER_GROUP
 253         | FATTR4_WORD1_RAWDEV
 254         | FATTR4_WORD1_SPACE_USED
 255         | FATTR4_WORD1_TIME_ACCESS
 256         | FATTR4_WORD1_TIME_METADATA
 257         | FATTR4_WORD1_TIME_MODIFY
 258         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 259 };
 260
 261 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 262                 struct nfs4_readdir_arg *readdir)
 263 {
 264         __be32 *start, *p;
 265
 266         if (cookie > 2) {
 267                 readdir->cookie = cookie;
 268                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 269                 return;
 270         }
 271
 272         readdir->cookie = 0;
 273         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 274         if (cookie == 2)
 275                 return;
 276
 277         /*
 278          * NFSv4 servers do not return entries for '.' and '..'
 279          * Therefore, we fake these entries here.  We let '.'
 280          * have cookie 0 and '..' have cookie 1.  Note that
 281          * when talking to the server, we always send cookie 0
 282          * instead of 1 or 2.
 283          */
 284         start = p = kmap_atomic(*readdir->pages);
 285
 286         if (cookie == 0) {
 287                 *p++ = xdr_one;                                  /* next */
 288                 *p++ = xdr_zero;                   /* cookie, first word */
 289                 *p++ = xdr_one;                   /* cookie, second word */
 290                 *p++ = xdr_one;                             /* entry len */
 291                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 292                 p++;
 293                 *p++ = xdr_one;                         /* bitmap length */
 294                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 295                 *p++ = htonl(8);              /* attribute buffer length */
 296                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 297         }
 298
 299         *p++ = xdr_one;                                  /* next */
 300         *p++ = xdr_zero;                   /* cookie, first word */
 301         *p++ = xdr_two;                   /* cookie, second word */
 302         *p++ = xdr_two;                             /* entry len */
 303         memcpy(p, "..\0\0", 4);                         /* entry */
 304         p++;
 305         *p++ = xdr_one;                         /* bitmap length */
 306         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 307         *p++ = htonl(8);              /* attribute buffer length */
 308         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 309
 310         readdir->pgbase = (char *)p - (char *)start;
 311         readdir->count -= readdir->pgbase;
 312         kunmap_atomic(start);
 313 }
 314
 315 static long nfs4_update_delay(long *timeout)
 316 {
 317         long ret;
 318         if (!timeout)
 319                 return NFS4_POLL_RETRY_MAX;
 320         if (*timeout <= 0)
 321                 *timeout = NFS4_POLL_RETRY_MIN;
 322         if (*timeout > NFS4_POLL_RETRY_MAX)
 323                 *timeout = NFS4_POLL_RETRY_MAX;
 324         ret = *timeout;
 325         *timeout <<= 1;
 326         return ret;
 327 }
 328
 329 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 330 {
 331         int res = 0;
 332
 333         might_sleep();
 334
 335         freezable_schedule_timeout_killable_unsafe(
 336                 nfs4_update_delay(timeout));
 337         if (fatal_signal_pending(current))
 338                 res = -ERESTARTSYS;
 339         return res;
 340 }
 341
 342 /* This is the error handling routine for processes that are allowed
 343  * to sleep.
 344  */
 345 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 346 {
 347         struct nfs_client *clp = server->nfs_client;
 348         struct nfs4_state *state = exception->state;
 349         struct inode *inode = exception->inode;
 350         int ret = errorcode;
 351
 352         exception->retry = 0;
 353         switch(errorcode) {
 354                 case 0:
 355                         return 0;
 356                 case -NFS4ERR_OPENMODE:
 357                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 358                                 nfs4_inode_return_delegation(inode);
 359                                 exception->retry = 1;
 360                                 return 0;
 361                         }
 362                         if (state == NULL)
 363                                 break;
 364                         ret = nfs4_schedule_stateid_recovery(server, state);
 365                         if (ret < 0)
 366                                 break;
 367                         goto wait_on_recovery;
 368                 case -NFS4ERR_DELEG_REVOKED:
 369                 case -NFS4ERR_ADMIN_REVOKED:
 370                 case -NFS4ERR_BAD_STATEID:
 371                         if (state == NULL)
 372                                 break;
 373                         ret = nfs4_schedule_stateid_recovery(server, state);
 374                         if (ret < 0)
 375                                 break;
 376                         goto wait_on_recovery;
 377                 case -NFS4ERR_EXPIRED:
 378                         if (state != NULL) {
 379                                 ret = nfs4_schedule_stateid_recovery(server, state);
 380                                 if (ret < 0)
 381                                         break;
 382                         }
 383                 case -NFS4ERR_STALE_STATEID:
 384                 case -NFS4ERR_STALE_CLIENTID:
 385                         nfs4_schedule_lease_recovery(clp);
 386                         goto wait_on_recovery;
 387                 case -NFS4ERR_MOVED:
 388                         ret = nfs4_schedule_migration_recovery(server);
 389                         if (ret < 0)
 390                                 break;
 391                         goto wait_on_recovery;
 392                 case -NFS4ERR_LEASE_MOVED:
 393                         nfs4_schedule_lease_moved_recovery(clp);
 394                         goto wait_on_recovery;
 395 #if defined(CONFIG_NFS_V4_1)
 396                 case -NFS4ERR_BADSESSION:
 397                 case -NFS4ERR_BADSLOT:
 398                 case -NFS4ERR_BAD_HIGH_SLOT:
 399                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 400                 case -NFS4ERR_DEADSESSION:
 401                 case -NFS4ERR_SEQ_FALSE_RETRY:
 402                 case -NFS4ERR_SEQ_MISORDERED:
 403                         dprintk("%s ERROR: %d Reset session\n", __func__,
 404                                 errorcode);
 405                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 406                         goto wait_on_recovery;
 407 #endif /* defined(CONFIG_NFS_V4_1) */
 408                 case -NFS4ERR_FILE_OPEN:
 409                         if (exception->timeout > HZ) {
 410                                 /* We have retried a decent amount, time to
 411                                  * fail
 412                                  */
 413                                 ret = -EBUSY;
 414                                 break;
 415                         }
 416                 case -NFS4ERR_GRACE:
 417                 case -NFS4ERR_DELAY:
 418                         ret = nfs4_delay(server->client, &exception->timeout);
 419                         if (ret != 0)
 420                                 break;
 421                 case -NFS4ERR_RETRY_UNCACHED_REP:
 422                 case -NFS4ERR_OLD_STATEID:
 423                         exception->retry = 1;
 424                         break;
 425                 case -NFS4ERR_BADOWNER:
 426                         /* The following works around a Linux server bug! */
 427                 case -NFS4ERR_BADNAME:
 428                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 429                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 430                                 exception->retry = 1;
 431                                 printk(KERN_WARNING "NFS: v4 server %s "
 432                                                 "does not accept raw "
 433                                                 "uid/gids. "
 434                                                 "Reenabling the idmapper.\n",
 435                                                 server->nfs_client->cl_hostname);
 436                         }
 437         }
 438         /* We failed to handle the error */
 439         return nfs4_map_errors(ret);
 440 wait_on_recovery:
 441         ret = nfs4_wait_clnt_recover(clp);
 442         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 443                 return -EIO;
 444         if (ret == 0)
 445                 exception->retry = 1;
 446         return ret;
 447 }
 448
 449 /*
 450  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 451  * or 'false' otherwise.
 452  */
 453 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 454 {
 455         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 456
 457         if (flavor == RPC_AUTH_GSS_KRB5I ||
 458             flavor == RPC_AUTH_GSS_KRB5P)
 459                 return true;
 460
 461         return false;
 462 }
 463
 464 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 465 {
 466         spin_lock(&clp->cl_lock);
 467         if (time_before(clp->cl_last_renewal,timestamp))
 468                 clp->cl_last_renewal = timestamp;
 469         spin_unlock(&clp->cl_lock);
 470 }
 471
 472 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 473 {
 474         do_renew_lease(server->nfs_client, timestamp);
 475 }
 476
 477 struct nfs4_call_sync_data {
 478         const struct nfs_server *seq_server;
 479         struct nfs4_sequence_args *seq_args;
 480         struct nfs4_sequence_res *seq_res;
 481 };
 482
 483 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
 484                                struct nfs4_sequence_res *res, int cache_reply)
 485 {
 486         args->sa_slot = NULL;
 487         args->sa_cache_this = cache_reply;
 488         args->sa_privileged = 0;
 489
 490         res->sr_slot = NULL;
 491 }
 492
 493 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 494 {
 495         args->sa_privileged = 1;
 496 }
 497
 498 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 499                          struct nfs4_sequence_args *args,
 500                          struct nfs4_sequence_res *res,
 501                          struct rpc_task *task)
 502 {
 503         struct nfs4_slot *slot;
 504
 505         /* slot already allocated? */
 506         if (res->sr_slot != NULL)
 507                 goto out_start;
 508
 509         spin_lock(&tbl->slot_tbl_lock);
 510         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 511                 goto out_sleep;
 512
 513         slot = nfs4_alloc_slot(tbl);
 514         if (IS_ERR(slot)) {
 515                 if (slot == ERR_PTR(-ENOMEM))
 516                         task->tk_timeout = HZ >> 2;
 517                 goto out_sleep;
 518         }
 519         spin_unlock(&tbl->slot_tbl_lock);
 520
 521         args->sa_slot = slot;
 522         res->sr_slot = slot;
 523
 524 out_start:
 525         rpc_call_start(task);
 526         return 0;
 527
 528 out_sleep:
 529         if (args->sa_privileged)
 530                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 531                                 NULL, RPC_PRIORITY_PRIVILEGED);
 532         else
 533                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 534         spin_unlock(&tbl->slot_tbl_lock);
 535         return -EAGAIN;
 536 }
 537 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 538
 539 static int nfs40_sequence_done(struct rpc_task *task,
 540                                struct nfs4_sequence_res *res)
 541 {
 542         struct nfs4_slot *slot = res->sr_slot;
 543         struct nfs4_slot_table *tbl;
 544
 545         if (slot == NULL)
 546                 goto out;
 547
 548         tbl = slot->table;
 549         spin_lock(&tbl->slot_tbl_lock);
 550         if (!nfs41_wake_and_assign_slot(tbl, slot))
 551                 nfs4_free_slot(tbl, slot);
 552         spin_unlock(&tbl->slot_tbl_lock);
 553
 554         res->sr_slot = NULL;
 555 out:
 556         return 1;
 557 }
 558
 559 #if defined(CONFIG_NFS_V4_1)
 560
 561 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 562 {
 563         struct nfs4_session *session;
 564         struct nfs4_slot_table *tbl;
 565         struct nfs4_slot *slot = res->sr_slot;
 566         bool send_new_highest_used_slotid = false;
 567
 568         tbl = slot->table;
 569         session = tbl->session;
 570
 571         spin_lock(&tbl->slot_tbl_lock);
 572         /* Be nice to the server: try to ensure that the last transmitted
 573          * value for highest_user_slotid <= target_highest_slotid
 574          */
 575         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 576                 send_new_highest_used_slotid = true;
 577
 578         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 579                 send_new_highest_used_slotid = false;
 580                 goto out_unlock;
 581         }
 582         nfs4_free_slot(tbl, slot);
 583
 584         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 585                 send_new_highest_used_slotid = false;
 586 out_unlock:
 587         spin_unlock(&tbl->slot_tbl_lock);
 588         res->sr_slot = NULL;
 589         if (send_new_highest_used_slotid)
 590                 nfs41_server_notify_highest_slotid_update(session->clp);
 591 }
 592
 593 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 594 {
 595         struct nfs4_session *session;
 596         struct nfs4_slot *slot = res->sr_slot;
 597         struct nfs_client *clp;
 598         bool interrupted = false;
 599         int ret = 1;
 600
 601         if (slot == NULL)
 602                 goto out_noaction;
 603         /* don't increment the sequence number if the task wasn't sent */
 604         if (!RPC_WAS_SENT(task))
 605                 goto out;
 606
 607         session = slot->table->session;
 608
 609         if (slot->interrupted) {
 610                 slot->interrupted = 0;
 611                 interrupted = true;
 612         }
 613
 614         trace_nfs4_sequence_done(session, res);
 615         /* Check the SEQUENCE operation status */
 616         switch (res->sr_status) {
 617         case 0:
 618                 /* Update the slot's sequence and clientid lease timer */
 619                 ++slot->seq_nr;
 620                 clp = session->clp;
 621                 do_renew_lease(clp, res->sr_timestamp);
 622                 /* Check sequence flags */
 623                 if (res->sr_status_flags != 0)
 624                         nfs4_schedule_lease_recovery(clp);
 625                 nfs41_update_target_slotid(slot->table, slot, res);
 626                 break;
 627         case 1:
 628                 /*
 629                  * sr_status remains 1 if an RPC level error occurred.
 630                  * The server may or may not have processed the sequence
 631                  * operation..
 632                  * Mark the slot as having hosted an interrupted RPC call.
 633                  */
 634                 slot->interrupted = 1;
 635                 goto out;
 636         case -NFS4ERR_DELAY:
 637                 /* The server detected a resend of the RPC call and
 638                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 639                  * of RFC5661.
 640                  */
 641                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 642                         __func__,
 643                         slot->slot_nr,
 644                         slot->seq_nr);
 645                 goto out_retry;
 646         case -NFS4ERR_BADSLOT:
 647                 /*
 648                  * The slot id we used was probably retired. Try again
 649                  * using a different slot id.
 650                  */
 651                 goto retry_nowait;
 652         case -NFS4ERR_SEQ_MISORDERED:
 653                 /*
 654                  * Was the last operation on this sequence interrupted?
 655                  * If so, retry after bumping the sequence number.
 656                  */
 657                 if (interrupted) {
 658                         ++slot->seq_nr;
 659                         goto retry_nowait;
 660                 }
 661                 /*
 662                  * Could this slot have been previously retired?
 663                  * If so, then the server may be expecting seq_nr = 1!
 664                  */
 665                 if (slot->seq_nr != 1) {
 666                         slot->seq_nr = 1;
 667                         goto retry_nowait;
 668                 }
 669                 break;
 670         case -NFS4ERR_SEQ_FALSE_RETRY:
 671                 ++slot->seq_nr;
 672                 goto retry_nowait;
 673         default:
 674                 /* Just update the slot sequence no. */
 675                 ++slot->seq_nr;
 676         }
 677 out:
 678         /* The session may be reset by one of the error handlers. */
 679         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 680         nfs41_sequence_free_slot(res);
 681 out_noaction:
 682         return ret;
 683 retry_nowait:
 684         if (rpc_restart_call_prepare(task)) {
 685                 task->tk_status = 0;
 686                 ret = 0;
 687         }
 688         goto out;
 689 out_retry:
 690         if (!rpc_restart_call(task))
 691                 goto out;
 692         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 693         return 0;
 694 }
 695 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 696
 697 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 698 {
 699         if (res->sr_slot == NULL)
 700                 return 1;
 701         if (!res->sr_slot->table->session)
 702                 return nfs40_sequence_done(task, res);
 703         return nfs41_sequence_done(task, res);
 704 }
 705 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 706
 707 int nfs41_setup_sequence(struct nfs4_session *session,
 708                                 struct nfs4_sequence_args *args,
 709                                 struct nfs4_sequence_res *res,
 710                                 struct rpc_task *task)
 711 {
 712         struct nfs4_slot *slot;
 713         struct nfs4_slot_table *tbl;
 714
 715         dprintk("--> %s\n", __func__);
 716         /* slot already allocated? */
 717         if (res->sr_slot != NULL)
 718                 goto out_success;
 719
 720         tbl = &session->fc_slot_table;
 721
 722         task->tk_timeout = 0;
 723
 724         spin_lock(&tbl->slot_tbl_lock);
 725         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 726             !args->sa_privileged) {
 727                 /* The state manager will wait until the slot table is empty */
 728                 dprintk("%s session is draining\n", __func__);
 729                 goto out_sleep;
 730         }
 731
 732         slot = nfs4_alloc_slot(tbl);
 733         if (IS_ERR(slot)) {
 734                 /* If out of memory, try again in 1/4 second */
 735                 if (slot == ERR_PTR(-ENOMEM))
 736                         task->tk_timeout = HZ >> 2;
 737                 dprintk("<-- %s: no free slots\n", __func__);
 738                 goto out_sleep;
 739         }
 740         spin_unlock(&tbl->slot_tbl_lock);
 741
 742         args->sa_slot = slot;
 743
 744         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 745                         slot->slot_nr, slot->seq_nr);
 746
 747         res->sr_slot = slot;
 748         res->sr_timestamp = jiffies;
 749         res->sr_status_flags = 0;
 750         /*
 751          * sr_status is only set in decode_sequence, and so will remain
 752          * set to 1 if an rpc level failure occurs.
 753          */
 754         res->sr_status = 1;
 755         trace_nfs4_setup_sequence(session, args);
 756 out_success:
 757         rpc_call_start(task);
 758         return 0;
 759 out_sleep:
 760         /* Privileged tasks are queued with top priority */
 761         if (args->sa_privileged)
 762                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 763                                 NULL, RPC_PRIORITY_PRIVILEGED);
 764         else
 765                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 766         spin_unlock(&tbl->slot_tbl_lock);
 767         return -EAGAIN;
 768 }
 769 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 770
 771 static int nfs4_setup_sequence(const struct nfs_server *server,
 772                                struct nfs4_sequence_args *args,
 773                                struct nfs4_sequence_res *res,
 774                                struct rpc_task *task)
 775 {
 776         struct nfs4_session *session = nfs4_get_session(server);
 777         int ret = 0;
 778
 779         if (!session)
 780                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 781                                             args, res, task);
 782
 783         dprintk("--> %s clp %p session %p sr_slot %u\n",
 784                 __func__, session->clp, session, res->sr_slot ?
 785                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 786
 787         ret = nfs41_setup_sequence(session, args, res, task);
 788
 789         dprintk("<-- %s status=%d\n", __func__, ret);
 790         return ret;
 791 }
 792
 793 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 794 {
 795         struct nfs4_call_sync_data *data = calldata;
 796         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 797
 798         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 799
 800         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 801 }
 802
 803 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 804 {
 805         struct nfs4_call_sync_data *data = calldata;
 806
 807         nfs41_sequence_done(task, data->seq_res);
 808 }
 809
 810 static const struct rpc_call_ops nfs41_call_sync_ops = {
 811         .rpc_call_prepare = nfs41_call_sync_prepare,
 812         .rpc_call_done = nfs41_call_sync_done,
 813 };
 814
 815 #else   /* !CONFIG_NFS_V4_1 */
 816
 817 static int nfs4_setup_sequence(const struct nfs_server *server,
 818                                struct nfs4_sequence_args *args,
 819                                struct nfs4_sequence_res *res,
 820                                struct rpc_task *task)
 821 {
 822         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 823                                     args, res, task);
 824 }
 825
 826 int nfs4_sequence_done(struct rpc_task *task,
 827                        struct nfs4_sequence_res *res)
 828 {
 829         return nfs40_sequence_done(task, res);
 830 }
 831 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 832
 833 #endif  /* !CONFIG_NFS_V4_1 */
 834
 835 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 836 {
 837         struct nfs4_call_sync_data *data = calldata;
 838         nfs4_setup_sequence(data->seq_server,
 839                                 data->seq_args, data->seq_res, task);
 840 }
 841
 842 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 843 {
 844         struct nfs4_call_sync_data *data = calldata;
 845         nfs4_sequence_done(task, data->seq_res);
 846 }
 847
 848 static const struct rpc_call_ops nfs40_call_sync_ops = {
 849         .rpc_call_prepare = nfs40_call_sync_prepare,
 850         .rpc_call_done = nfs40_call_sync_done,
 851 };
 852
 853 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 854                                    struct nfs_server *server,
 855                                    struct rpc_message *msg,
 856                                    struct nfs4_sequence_args *args,
 857                                    struct nfs4_sequence_res *res)
 858 {
 859         int ret;
 860         struct rpc_task *task;
 861         struct nfs_client *clp = server->nfs_client;
 862         struct nfs4_call_sync_data data = {
 863                 .seq_server = server,
 864                 .seq_args = args,
 865                 .seq_res = res,
 866         };
 867         struct rpc_task_setup task_setup = {
 868                 .rpc_client = clnt,
 869                 .rpc_message = msg,
 870                 .callback_ops = clp->cl_mvops->call_sync_ops,
 871                 .callback_data = &data
 872         };
 873
 874         task = rpc_run_task(&task_setup);
 875         if (IS_ERR(task))
 876                 ret = PTR_ERR(task);
 877         else {
 878                 ret = task->tk_status;
 879                 rpc_put_task(task);
 880         }
 881         return ret;
 882 }
 883
 884 int nfs4_call_sync(struct rpc_clnt *clnt,
 885                    struct nfs_server *server,
 886                    struct rpc_message *msg,
 887                    struct nfs4_sequence_args *args,
 888                    struct nfs4_sequence_res *res,
 889                    int cache_reply)
 890 {
 891         nfs4_init_sequence(args, res, cache_reply);
 892         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 893 }
 894
 895 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 896 {
 897         struct nfs_inode *nfsi = NFS_I(dir);
 898
 899         spin_lock(&dir->i_lock);
 900         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 901         if (!cinfo->atomic || cinfo->before != dir->i_version)
 902                 nfs_force_lookup_revalidate(dir);
 903         dir->i_version = cinfo->after;
 904         nfs_fscache_invalidate(dir);
 905         spin_unlock(&dir->i_lock);
 906 }
 907
 908 struct nfs4_opendata {
 909         struct kref kref;
 910         struct nfs_openargs o_arg;
 911         struct nfs_openres o_res;
 912         struct nfs_open_confirmargs c_arg;
 913         struct nfs_open_confirmres c_res;
 914         struct nfs4_string owner_name;
 915         struct nfs4_string group_name;
 916         struct nfs_fattr f_attr;
 917         struct nfs4_label *f_label;
 918         struct dentry *dir;
 919         struct dentry *dentry;
 920         struct nfs4_state_owner *owner;
 921         struct nfs4_state *state;
 922         struct iattr attrs;
 923         unsigned long timestamp;
 924         unsigned int rpc_done : 1;
 925         unsigned int file_created : 1;
 926         unsigned int is_recover : 1;
 927         int rpc_status;
 928         int cancelled;
 929 };
 930
 931 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 932                 int err, struct nfs4_exception *exception)
 933 {
 934         if (err != -EINVAL)
 935                 return false;
 936         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 937                 return false;
 938         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 939         exception->retry = 1;
 940         return true;
 941 }
 942
 943 static enum open_claim_type4
 944 nfs4_map_atomic_open_claim(struct nfs_server *server,
 945                 enum open_claim_type4 claim)
 946 {
 947         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 948                 return claim;
 949         switch (claim) {
 950         default:
 951                 return claim;
 952         case NFS4_OPEN_CLAIM_FH:
 953                 return NFS4_OPEN_CLAIM_NULL;
 954         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 955                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 956         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 957                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 958         }
 959 }
 960
 961 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 962 {
 963         p->o_res.f_attr = &p->f_attr;
 964         p->o_res.f_label = p->f_label;
 965         p->o_res.seqid = p->o_arg.seqid;
 966         p->c_res.seqid = p->c_arg.seqid;
 967         p->o_res.server = p->o_arg.server;
 968         p->o_res.access_request = p->o_arg.access;
 969         nfs_fattr_init(&p->f_attr);
 970         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 971 }
 972
 973 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 974                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 975                 const struct iattr *attrs,
 976                 struct nfs4_label *label,
 977                 enum open_claim_type4 claim,
 978                 gfp_t gfp_mask)
 979 {
 980         struct dentry *parent = dget_parent(dentry);
 981         struct inode *dir = parent->d_inode;
 982         struct nfs_server *server = NFS_SERVER(dir);
 983         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
 984         struct nfs4_opendata *p;
 985
 986         p = kzalloc(sizeof(*p), gfp_mask);
 987         if (p == NULL)
 988                 goto err;
 989
 990         p->f_label = nfs4_label_alloc(server, gfp_mask);
 991         if (IS_ERR(p->f_label))
 992                 goto err_free_p;
 993
 994         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
 995         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
 996         if (IS_ERR(p->o_arg.seqid))
 997                 goto err_free_label;
 998         nfs_sb_active(dentry->d_sb);
 999         p->dentry = dget(dentry);
1000         p->dir = parent;
1001         p->owner = sp;
1002         atomic_inc(&sp->so_count);
1003         p->o_arg.open_flags = flags;
1004         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1005         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1006          * will return permission denied for all bits until close */
1007         if (!(flags & O_EXCL)) {
1008                 /* ask server to check for all possible rights as results
1009                  * are cached */
1010                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1011                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1012         }
1013         p->o_arg.clientid = server->nfs_client->cl_clientid;
1014         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1015         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1016         p->o_arg.name = &dentry->d_name;
1017         p->o_arg.server = server;
1018         p->o_arg.bitmask = nfs4_bitmask(server, label);
1019         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1020         p->o_arg.label = label;
1021         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1022         switch (p->o_arg.claim) {
1023         case NFS4_OPEN_CLAIM_NULL:
1024         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1025         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1026                 p->o_arg.fh = NFS_FH(dir);
1027                 break;
1028         case NFS4_OPEN_CLAIM_PREVIOUS:
1029         case NFS4_OPEN_CLAIM_FH:
1030         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1031         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1032                 p->o_arg.fh = NFS_FH(dentry->d_inode);
1033         }
1034         if (attrs != NULL && attrs->ia_valid != 0) {
1035                 __u32 verf[2];
1036
1037                 p->o_arg.u.attrs = &p->attrs;
1038                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1039
1040                 verf[0] = jiffies;
1041                 verf[1] = current->pid;
1042                 memcpy(p->o_arg.u.verifier.data, verf,
1043                                 sizeof(p->o_arg.u.verifier.data));
1044         }
1045         p->c_arg.fh = &p->o_res.fh;
1046         p->c_arg.stateid = &p->o_res.stateid;
1047         p->c_arg.seqid = p->o_arg.seqid;
1048         nfs4_init_opendata_res(p);
1049         kref_init(&p->kref);
1050         return p;
1051
1052 err_free_label:
1053         nfs4_label_free(p->f_label);
1054 err_free_p:
1055         kfree(p);
1056 err:
1057         dput(parent);
1058         return NULL;
1059 }
1060
1061 static void nfs4_opendata_free(struct kref *kref)
1062 {
1063         struct nfs4_opendata *p = container_of(kref,
1064                         struct nfs4_opendata, kref);
1065         struct super_block *sb = p->dentry->d_sb;
1066
1067         nfs_free_seqid(p->o_arg.seqid);
1068         if (p->state != NULL)
1069                 nfs4_put_open_state(p->state);
1070         nfs4_put_state_owner(p->owner);
1071
1072         nfs4_label_free(p->f_label);
1073
1074         dput(p->dir);
1075         dput(p->dentry);
1076         nfs_sb_deactive(sb);
1077         nfs_fattr_free_names(&p->f_attr);
1078         kfree(p->f_attr.mdsthreshold);
1079         kfree(p);
1080 }
1081
1082 static void nfs4_opendata_put(struct nfs4_opendata *p)
1083 {
1084         if (p != NULL)
1085                 kref_put(&p->kref, nfs4_opendata_free);
1086 }
1087
1088 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1089 {
1090         int ret;
1091
1092         ret = rpc_wait_for_completion_task(task);
1093         return ret;
1094 }
1095
1096 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1097 {
1098         int ret = 0;
1099
1100         if (open_mode & (O_EXCL|O_TRUNC))
1101                 goto out;
1102         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1103                 case FMODE_READ:
1104                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1105                                 && state->n_rdonly != 0;
1106                         break;
1107                 case FMODE_WRITE:
1108                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1109                                 && state->n_wronly != 0;
1110                         break;
1111                 case FMODE_READ|FMODE_WRITE:
1112                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1113                                 && state->n_rdwr != 0;
1114         }
1115 out:
1116         return ret;
1117 }
1118
1119 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1120 {
1121         if (delegation == NULL)
1122                 return 0;
1123         if ((delegation->type & fmode) != fmode)
1124                 return 0;
1125         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1126                 return 0;
1127         nfs_mark_delegation_referenced(delegation);
1128         return 1;
1129 }
1130
1131 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1132 {
1133         switch (fmode) {
1134                 case FMODE_WRITE:
1135                         state->n_wronly++;
1136                         break;
1137                 case FMODE_READ:
1138                         state->n_rdonly++;
1139                         break;
1140                 case FMODE_READ|FMODE_WRITE:
1141                         state->n_rdwr++;
1142         }
1143         nfs4_state_set_mode_locked(state, state->state | fmode);
1144 }
1145
1146 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1147 {
1148         struct nfs_client *clp = state->owner->so_server->nfs_client;
1149         bool need_recover = false;
1150
1151         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1152                 need_recover = true;
1153         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1154                 need_recover = true;
1155         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1156                 need_recover = true;
1157         if (need_recover)
1158                 nfs4_state_mark_reclaim_nograce(clp, state);
1159 }
1160
1161 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1162                 nfs4_stateid *stateid)
1163 {
1164         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1165                 return true;
1166         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1167                 nfs_test_and_clear_all_open_stateid(state);
1168                 return true;
1169         }
1170         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1171                 return true;
1172         return false;
1173 }
1174
1175 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1176 {
1177         if (state->n_wronly)
1178                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1179         if (state->n_rdonly)
1180                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1181         if (state->n_rdwr)
1182                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1183 }
1184
1185 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1186                 nfs4_stateid *stateid, fmode_t fmode)
1187 {
1188         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1189         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1190         case FMODE_WRITE:
1191                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1192                 break;
1193         case FMODE_READ:
1194                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1195                 break;
1196         case 0:
1197                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1198                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1199                 clear_bit(NFS_OPEN_STATE, &state->flags);
1200         }
1201         if (stateid == NULL)
1202                 return;
1203         /* Handle races with OPEN */
1204         if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1205             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1206                 nfs_resync_open_stateid_locked(state);
1207                 return;
1208         }
1209         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1210                 nfs4_stateid_copy(&state->stateid, stateid);
1211         nfs4_stateid_copy(&state->open_stateid, stateid);
1212 }
1213
1214 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1215 {
1216         write_seqlock(&state->seqlock);
1217         nfs_clear_open_stateid_locked(state, stateid, fmode);
1218         write_sequnlock(&state->seqlock);
1219         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1220                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1221 }
1222
1223 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1224 {
1225         switch (fmode) {
1226                 case FMODE_READ:
1227                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1228                         break;
1229                 case FMODE_WRITE:
1230                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1231                         break;
1232                 case FMODE_READ|FMODE_WRITE:
1233                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1234         }
1235         if (!nfs_need_update_open_stateid(state, stateid))
1236                 return;
1237         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1238                 nfs4_stateid_copy(&state->stateid, stateid);
1239         nfs4_stateid_copy(&state->open_stateid, stateid);
1240 }
1241
1242 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1243 {
1244         /*
1245          * Protect the call to nfs4_state_set_mode_locked and
1246          * serialise the stateid update
1247          */
1248         write_seqlock(&state->seqlock);
1249         if (deleg_stateid != NULL) {
1250                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1251                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1252         }
1253         if (open_stateid != NULL)
1254                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1255         write_sequnlock(&state->seqlock);
1256         spin_lock(&state->owner->so_lock);
1257         update_open_stateflags(state, fmode);
1258         spin_unlock(&state->owner->so_lock);
1259 }
1260
1261 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1262 {
1263         struct nfs_inode *nfsi = NFS_I(state->inode);
1264         struct nfs_delegation *deleg_cur;
1265         int ret = 0;
1266
1267         fmode &= (FMODE_READ|FMODE_WRITE);
1268
1269         rcu_read_lock();
1270         deleg_cur = rcu_dereference(nfsi->delegation);
1271         if (deleg_cur == NULL)
1272                 goto no_delegation;
1273
1274         spin_lock(&deleg_cur->lock);
1275         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1276            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1277             (deleg_cur->type & fmode) != fmode)
1278                 goto no_delegation_unlock;
1279
1280         if (delegation == NULL)
1281                 delegation = &deleg_cur->stateid;
1282         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1283                 goto no_delegation_unlock;
1284
1285         nfs_mark_delegation_referenced(deleg_cur);
1286         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1287         ret = 1;
1288 no_delegation_unlock:
1289         spin_unlock(&deleg_cur->lock);
1290 no_delegation:
1291         rcu_read_unlock();
1292
1293         if (!ret && open_stateid != NULL) {
1294                 __update_open_stateid(state, open_stateid, NULL, fmode);
1295                 ret = 1;
1296         }
1297         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1298                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1299
1300         return ret;
1301 }
1302
1303 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1304                 const nfs4_stateid *stateid)
1305 {
1306         struct nfs4_state *state = lsp->ls_state;
1307         bool ret = false;
1308
1309         spin_lock(&state->state_lock);
1310         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1311                 goto out_noupdate;
1312         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1313                 goto out_noupdate;
1314         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1315         ret = true;
1316 out_noupdate:
1317         spin_unlock(&state->state_lock);
1318         return ret;
1319 }
1320
1321 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1322 {
1323         struct nfs_delegation *delegation;
1324
1325         rcu_read_lock();
1326         delegation = rcu_dereference(NFS_I(inode)->delegation);
1327         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1328                 rcu_read_unlock();
1329                 return;
1330         }
1331         rcu_read_unlock();
1332         nfs4_inode_return_delegation(inode);
1333 }
1334
1335 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1336 {
1337         struct nfs4_state *state = opendata->state;
1338         struct nfs_inode *nfsi = NFS_I(state->inode);
1339         struct nfs_delegation *delegation;
1340         int open_mode = opendata->o_arg.open_flags;
1341         fmode_t fmode = opendata->o_arg.fmode;
1342         nfs4_stateid stateid;
1343         int ret = -EAGAIN;
1344
1345         for (;;) {
1346                 spin_lock(&state->owner->so_lock);
1347                 if (can_open_cached(state, fmode, open_mode)) {
1348                         update_open_stateflags(state, fmode);
1349                         spin_unlock(&state->owner->so_lock);
1350                         goto out_return_state;
1351                 }
1352                 spin_unlock(&state->owner->so_lock);
1353                 rcu_read_lock();
1354                 delegation = rcu_dereference(nfsi->delegation);
1355                 if (!can_open_delegated(delegation, fmode)) {
1356                         rcu_read_unlock();
1357                         break;
1358                 }
1359                 /* Save the delegation */
1360                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1361                 rcu_read_unlock();
1362                 nfs_release_seqid(opendata->o_arg.seqid);
1363                 if (!opendata->is_recover) {
1364                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1365                         if (ret != 0)
1366                                 goto out;
1367                 }
1368                 ret = -EAGAIN;
1369
1370                 /* Try to update the stateid using the delegation */
1371                 if (update_open_stateid(state, NULL, &stateid, fmode))
1372                         goto out_return_state;
1373         }
1374 out:
1375         return ERR_PTR(ret);
1376 out_return_state:
1377         atomic_inc(&state->count);
1378         return state;
1379 }
1380
1381 static void
1382 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1383 {
1384         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1385         struct nfs_delegation *delegation;
1386         int delegation_flags = 0;
1387
1388         rcu_read_lock();
1389         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1390         if (delegation)
1391                 delegation_flags = delegation->flags;
1392         rcu_read_unlock();
1393         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1394                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1395                                    "returning a delegation for "
1396                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1397                                    clp->cl_hostname);
1398         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1399                 nfs_inode_set_delegation(state->inode,
1400                                          data->owner->so_cred,
1401                                          &data->o_res);
1402         else
1403                 nfs_inode_reclaim_delegation(state->inode,
1404                                              data->owner->so_cred,
1405                                              &data->o_res);
1406 }
1407
1408 /*
1409  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1410  * and update the nfs4_state.
1411  */
1412 static struct nfs4_state *
1413 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1414 {
1415         struct inode *inode = data->state->inode;
1416         struct nfs4_state *state = data->state;
1417         int ret;
1418
1419         if (!data->rpc_done) {
1420                 if (data->rpc_status) {
1421                         ret = data->rpc_status;
1422                         goto err;
1423                 }
1424                 /* cached opens have already been processed */
1425                 goto update;
1426         }
1427
1428         ret = nfs_refresh_inode(inode, &data->f_attr);
1429         if (ret)
1430                 goto err;
1431
1432         if (data->o_res.delegation_type != 0)
1433                 nfs4_opendata_check_deleg(data, state);
1434 update:
1435         update_open_stateid(state, &data->o_res.stateid, NULL,
1436                             data->o_arg.fmode);
1437         atomic_inc(&state->count);
1438
1439         return state;
1440 err:
1441         return ERR_PTR(ret);
1442
1443 }
1444
1445 static struct nfs4_state *
1446 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1447 {
1448         struct inode *inode;
1449         struct nfs4_state *state = NULL;
1450         int ret;
1451
1452         if (!data->rpc_done) {
1453                 state = nfs4_try_open_cached(data);
1454                 goto out;
1455         }
1456
1457         ret = -EAGAIN;
1458         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1459                 goto err;
1460         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1461         ret = PTR_ERR(inode);
1462         if (IS_ERR(inode))
1463                 goto err;
1464         ret = -ENOMEM;
1465         state = nfs4_get_open_state(inode, data->owner);
1466         if (state == NULL)
1467                 goto err_put_inode;
1468         if (data->o_res.delegation_type != 0)
1469                 nfs4_opendata_check_deleg(data, state);
1470         update_open_stateid(state, &data->o_res.stateid, NULL,
1471                         data->o_arg.fmode);
1472         iput(inode);
1473 out:
1474         nfs_release_seqid(data->o_arg.seqid);
1475         return state;
1476 err_put_inode:
1477         iput(inode);
1478 err:
1479         return ERR_PTR(ret);
1480 }
1481
1482 static struct nfs4_state *
1483 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1484 {
1485         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1486                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1487         return _nfs4_opendata_to_nfs4_state(data);
1488 }
1489
1490 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1491 {
1492         struct nfs_inode *nfsi = NFS_I(state->inode);
1493         struct nfs_open_context *ctx;
1494
1495         spin_lock(&state->inode->i_lock);
1496         list_for_each_entry(ctx, &nfsi->open_files, list) {
1497                 if (ctx->state != state)
1498                         continue;
1499                 get_nfs_open_context(ctx);
1500                 spin_unlock(&state->inode->i_lock);
1501                 return ctx;
1502         }
1503         spin_unlock(&state->inode->i_lock);
1504         return ERR_PTR(-ENOENT);
1505 }
1506
1507 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1508                 struct nfs4_state *state, enum open_claim_type4 claim)
1509 {
1510         struct nfs4_opendata *opendata;
1511
1512         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1513                         NULL, NULL, claim, GFP_NOFS);
1514         if (opendata == NULL)
1515                 return ERR_PTR(-ENOMEM);
1516         opendata->state = state;
1517         atomic_inc(&state->count);
1518         return opendata;
1519 }
1520
1521 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1522 {
1523         struct nfs4_state *newstate;
1524         int ret;
1525
1526         opendata->o_arg.open_flags = 0;
1527         opendata->o_arg.fmode = fmode;
1528         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1529         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1530         nfs4_init_opendata_res(opendata);
1531         ret = _nfs4_recover_proc_open(opendata);
1532         if (ret != 0)
1533                 return ret;
1534         newstate = nfs4_opendata_to_nfs4_state(opendata);
1535         if (IS_ERR(newstate))
1536                 return PTR_ERR(newstate);
1537         nfs4_close_state(newstate, fmode);
1538         *res = newstate;
1539         return 0;
1540 }
1541
1542 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1543 {
1544         struct nfs4_state *newstate;
1545         int ret;
1546
1547         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1548         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1549         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1550         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1551         /* memory barrier prior to reading state->n_* */
1552         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1553         clear_bit(NFS_OPEN_STATE, &state->flags);
1554         smp_rmb();
1555         if (state->n_rdwr != 0) {
1556                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1557                 if (ret != 0)
1558                         return ret;
1559                 if (newstate != state)
1560                         return -ESTALE;
1561         }
1562         if (state->n_wronly != 0) {
1563                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1564                 if (ret != 0)
1565                         return ret;
1566                 if (newstate != state)
1567                         return -ESTALE;
1568         }
1569         if (state->n_rdonly != 0) {
1570                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1571                 if (ret != 0)
1572                         return ret;
1573                 if (newstate != state)
1574                         return -ESTALE;
1575         }
1576         /*
1577          * We may have performed cached opens for all three recoveries.
1578          * Check if we need to update the current stateid.
1579          */
1580         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1581             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1582                 write_seqlock(&state->seqlock);
1583                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1584                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1585                 write_sequnlock(&state->seqlock);
1586         }
1587         return 0;
1588 }
1589
1590 /*
1591  * OPEN_RECLAIM:
1592  *      reclaim state on the server after a reboot.
1593  */
1594 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1595 {
1596         struct nfs_delegation *delegation;
1597         struct nfs4_opendata *opendata;
1598         fmode_t delegation_type = 0;
1599         int status;
1600
1601         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1602                         NFS4_OPEN_CLAIM_PREVIOUS);
1603         if (IS_ERR(opendata))
1604                 return PTR_ERR(opendata);
1605         rcu_read_lock();
1606         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1607         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1608                 delegation_type = delegation->type;
1609         rcu_read_unlock();
1610         opendata->o_arg.u.delegation_type = delegation_type;
1611         status = nfs4_open_recover(opendata, state);
1612         nfs4_opendata_put(opendata);
1613         return status;
1614 }
1615
1616 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1617 {
1618         struct nfs_server *server = NFS_SERVER(state->inode);
1619         struct nfs4_exception exception = { };
1620         int err;
1621         do {
1622                 err = _nfs4_do_open_reclaim(ctx, state);
1623                 trace_nfs4_open_reclaim(ctx, 0, err);
1624                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1625                         continue;
1626                 if (err != -NFS4ERR_DELAY)
1627                         break;
1628                 nfs4_handle_exception(server, err, &exception);
1629         } while (exception.retry);
1630         return err;
1631 }
1632
1633 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1634 {
1635         struct nfs_open_context *ctx;
1636         int ret;
1637
1638         ctx = nfs4_state_find_open_context(state);
1639         if (IS_ERR(ctx))
1640                 return -EAGAIN;
1641         ret = nfs4_do_open_reclaim(ctx, state);
1642         put_nfs_open_context(ctx);
1643         return ret;
1644 }
1645
1646 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1647 {
1648         switch (err) {
1649                 default:
1650                         printk(KERN_ERR "NFS: %s: unhandled error "
1651                                         "%d.\n", __func__, err);
1652                 case 0:
1653                 case -ENOENT:
1654                 case -ESTALE:
1655                         break;
1656                 case -NFS4ERR_BADSESSION:
1657                 case -NFS4ERR_BADSLOT:
1658                 case -NFS4ERR_BAD_HIGH_SLOT:
1659                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1660                 case -NFS4ERR_DEADSESSION:
1661                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1662                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1663                         return -EAGAIN;
1664                 case -NFS4ERR_STALE_CLIENTID:
1665                 case -NFS4ERR_STALE_STATEID:
1666                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1667                 case -NFS4ERR_EXPIRED:
1668                         /* Don't recall a delegation if it was lost */
1669                         nfs4_schedule_lease_recovery(server->nfs_client);
1670                         return -EAGAIN;
1671                 case -NFS4ERR_MOVED:
1672                         nfs4_schedule_migration_recovery(server);
1673                         return -EAGAIN;
1674                 case -NFS4ERR_LEASE_MOVED:
1675                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1676                         return -EAGAIN;
1677                 case -NFS4ERR_DELEG_REVOKED:
1678                 case -NFS4ERR_ADMIN_REVOKED:
1679                 case -NFS4ERR_BAD_STATEID:
1680                 case -NFS4ERR_OPENMODE:
1681                         nfs_inode_find_state_and_recover(state->inode,
1682                                         stateid);
1683                         nfs4_schedule_stateid_recovery(server, state);
1684                         return -EAGAIN;
1685                 case -NFS4ERR_DELAY:
1686                 case -NFS4ERR_GRACE:
1687                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1688                         ssleep(1);
1689                         return -EAGAIN;
1690                 case -ENOMEM:
1691                 case -NFS4ERR_DENIED:
1692                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1693                         return 0;
1694         }
1695         return err;
1696 }
1697
1698 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1699 {
1700         struct nfs_server *server = NFS_SERVER(state->inode);
1701         struct nfs4_opendata *opendata;
1702         int err;
1703
1704         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1705                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1706         if (IS_ERR(opendata))
1707                 return PTR_ERR(opendata);
1708         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1709         err = nfs4_open_recover(opendata, state);
1710         nfs4_opendata_put(opendata);
1711         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1712 }
1713
1714 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1715 {
1716         struct nfs4_opendata *data = calldata;
1717
1718         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1719                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1720 }
1721
1722 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1723 {
1724         struct nfs4_opendata *data = calldata;
1725
1726         nfs40_sequence_done(task, &data->c_res.seq_res);
1727
1728         data->rpc_status = task->tk_status;
1729         if (data->rpc_status == 0) {
1730                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1731                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1732                 renew_lease(data->o_res.server, data->timestamp);
1733                 data->rpc_done = 1;
1734         }
1735 }
1736
1737 static void nfs4_open_confirm_release(void *calldata)
1738 {
1739         struct nfs4_opendata *data = calldata;
1740         struct nfs4_state *state = NULL;
1741
1742         /* If this request hasn't been cancelled, do nothing */
1743         if (data->cancelled == 0)
1744                 goto out_free;
1745         /* In case of error, no cleanup! */
1746         if (!data->rpc_done)
1747                 goto out_free;
1748         state = nfs4_opendata_to_nfs4_state(data);
1749         if (!IS_ERR(state))
1750                 nfs4_close_state(state, data->o_arg.fmode);
1751 out_free:
1752         nfs4_opendata_put(data);
1753 }
1754
1755 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1756         .rpc_call_prepare = nfs4_open_confirm_prepare,
1757         .rpc_call_done = nfs4_open_confirm_done,
1758         .rpc_release = nfs4_open_confirm_release,
1759 };
1760
1761 /*
1762  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1763  */
1764 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1765 {
1766         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1767         struct rpc_task *task;
1768         struct  rpc_message msg = {
1769                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1770                 .rpc_argp = &data->c_arg,
1771                 .rpc_resp = &data->c_res,
1772                 .rpc_cred = data->owner->so_cred,
1773         };
1774         struct rpc_task_setup task_setup_data = {
1775                 .rpc_client = server->client,
1776                 .rpc_message = &msg,
1777                 .callback_ops = &nfs4_open_confirm_ops,
1778                 .callback_data = data,
1779                 .workqueue = nfsiod_workqueue,
1780                 .flags = RPC_TASK_ASYNC,
1781         };
1782         int status;
1783
1784         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1785         kref_get(&data->kref);
1786         data->rpc_done = 0;
1787         data->rpc_status = 0;
1788         data->timestamp = jiffies;
1789         task = rpc_run_task(&task_setup_data);
1790         if (IS_ERR(task))
1791                 return PTR_ERR(task);
1792         status = nfs4_wait_for_completion_rpc_task(task);
1793         if (status != 0) {
1794                 data->cancelled = 1;
1795                 smp_wmb();
1796         } else
1797                 status = data->rpc_status;
1798         rpc_put_task(task);
1799         return status;
1800 }
1801
1802 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1803 {
1804         struct nfs4_opendata *data = calldata;
1805         struct nfs4_state_owner *sp = data->owner;
1806         struct nfs_client *clp = sp->so_server->nfs_client;
1807
1808         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1809                 goto out_wait;
1810         /*
1811          * Check if we still need to send an OPEN call, or if we can use
1812          * a delegation instead.
1813          */
1814         if (data->state != NULL) {
1815                 struct nfs_delegation *delegation;
1816
1817                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1818                         goto out_no_action;
1819                 rcu_read_lock();
1820                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1821                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1822                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1823                     can_open_delegated(delegation, data->o_arg.fmode))
1824                         goto unlock_no_action;
1825                 rcu_read_unlock();
1826         }
1827         /* Update client id. */
1828         data->o_arg.clientid = clp->cl_clientid;
1829         switch (data->o_arg.claim) {
1830         case NFS4_OPEN_CLAIM_PREVIOUS:
1831         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1832         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1833                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1834         case NFS4_OPEN_CLAIM_FH:
1835                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1836                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1837         }
1838         data->timestamp = jiffies;
1839         if (nfs4_setup_sequence(data->o_arg.server,
1840                                 &data->o_arg.seq_args,
1841                                 &data->o_res.seq_res,
1842                                 task) != 0)
1843                 nfs_release_seqid(data->o_arg.seqid);
1844
1845         /* Set the create mode (note dependency on the session type) */
1846         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1847         if (data->o_arg.open_flags & O_EXCL) {
1848                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1849                 if (nfs4_has_persistent_session(clp))
1850                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1851                 else if (clp->cl_mvops->minor_version > 0)
1852                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1853         }
1854         return;
1855 unlock_no_action:
1856         rcu_read_unlock();
1857 out_no_action:
1858         task->tk_action = NULL;
1859 out_wait:
1860         nfs4_sequence_done(task, &data->o_res.seq_res);
1861 }
1862
1863 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1864 {
1865         struct nfs4_opendata *data = calldata;
1866
1867         data->rpc_status = task->tk_status;
1868
1869         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1870                 return;
1871
1872         if (task->tk_status == 0) {
1873                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1874                         switch (data->o_res.f_attr->mode & S_IFMT) {
1875                         case S_IFREG:
1876                                 break;
1877                         case S_IFLNK:
1878                                 data->rpc_status = -ELOOP;
1879                                 break;
1880                         case S_IFDIR:
1881                                 data->rpc_status = -EISDIR;
1882                                 break;
1883                         default:
1884                                 data->rpc_status = -ENOTDIR;
1885                         }
1886                 }
1887                 renew_lease(data->o_res.server, data->timestamp);
1888                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1889                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1890         }
1891         data->rpc_done = 1;
1892 }
1893
1894 static void nfs4_open_release(void *calldata)
1895 {
1896         struct nfs4_opendata *data = calldata;
1897         struct nfs4_state *state = NULL;
1898
1899         /* If this request hasn't been cancelled, do nothing */
1900         if (data->cancelled == 0)
1901                 goto out_free;
1902         /* In case of error, no cleanup! */
1903         if (data->rpc_status != 0 || !data->rpc_done)
1904                 goto out_free;
1905         /* In case we need an open_confirm, no cleanup! */
1906         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1907                 goto out_free;
1908         state = nfs4_opendata_to_nfs4_state(data);
1909         if (!IS_ERR(state))
1910                 nfs4_close_state(state, data->o_arg.fmode);
1911 out_free:
1912         nfs4_opendata_put(data);
1913 }
1914
1915 static const struct rpc_call_ops nfs4_open_ops = {
1916         .rpc_call_prepare = nfs4_open_prepare,
1917         .rpc_call_done = nfs4_open_done,
1918         .rpc_release = nfs4_open_release,
1919 };
1920
1921 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1922 {
1923         struct inode *dir = data->dir->d_inode;
1924         struct nfs_server *server = NFS_SERVER(dir);
1925         struct nfs_openargs *o_arg = &data->o_arg;
1926         struct nfs_openres *o_res = &data->o_res;
1927         struct rpc_task *task;
1928         struct rpc_message msg = {
1929                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1930                 .rpc_argp = o_arg,
1931                 .rpc_resp = o_res,
1932                 .rpc_cred = data->owner->so_cred,
1933         };
1934         struct rpc_task_setup task_setup_data = {
1935                 .rpc_client = server->client,
1936                 .rpc_message = &msg,
1937                 .callback_ops = &nfs4_open_ops,
1938                 .callback_data = data,
1939                 .workqueue = nfsiod_workqueue,
1940                 .flags = RPC_TASK_ASYNC,
1941         };
1942         int status;
1943
1944         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1945         kref_get(&data->kref);
1946         data->rpc_done = 0;
1947         data->rpc_status = 0;
1948         data->cancelled = 0;
1949         data->is_recover = 0;
1950         if (isrecover) {
1951                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1952                 data->is_recover = 1;
1953         }
1954         task = rpc_run_task(&task_setup_data);
1955         if (IS_ERR(task))
1956                 return PTR_ERR(task);
1957         status = nfs4_wait_for_completion_rpc_task(task);
1958         if (status != 0) {
1959                 data->cancelled = 1;
1960                 smp_wmb();
1961         } else
1962                 status = data->rpc_status;
1963         rpc_put_task(task);
1964
1965         return status;
1966 }
1967
1968 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1969 {
1970         struct inode *dir = data->dir->d_inode;
1971         struct nfs_openres *o_res = &data->o_res;
1972         int status;
1973
1974         status = nfs4_run_open_task(data, 1);
1975         if (status != 0 || !data->rpc_done)
1976                 return status;
1977
1978         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1979
1980         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1981                 status = _nfs4_proc_open_confirm(data);
1982                 if (status != 0)
1983                         return status;
1984         }
1985
1986         return status;
1987 }
1988
1989 /*
1990  * Additional permission checks in order to distinguish between an
1991  * open for read, and an open for execute. This works around the
1992  * fact that NFSv4 OPEN treats read and execute permissions as being
1993  * the same.
1994  * Note that in the non-execute case, we want to turn off permission
1995  * checking if we just created a new file (POSIX open() semantics).
1996  */
1997 static int nfs4_opendata_access(struct rpc_cred *cred,
1998                                 struct nfs4_opendata *opendata,
1999                                 struct nfs4_state *state, fmode_t fmode,
2000                                 int openflags)
2001 {
2002         struct nfs_access_entry cache;
2003         u32 mask;
2004
2005         /* access call failed or for some reason the server doesn't
2006          * support any access modes -- defer access call until later */
2007         if (opendata->o_res.access_supported == 0)
2008                 return 0;
2009
2010         mask = 0;
2011         /*
2012          * Use openflags to check for exec, because fmode won't
2013          * always have FMODE_EXEC set when file open for exec.
2014          */
2015         if (openflags & __FMODE_EXEC) {
2016                 /* ONLY check for exec rights */
2017                 mask = MAY_EXEC;
2018         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2019                 mask = MAY_READ;
2020
2021         cache.cred = cred;
2022         cache.jiffies = jiffies;
2023         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2024         nfs_access_add_cache(state->inode, &cache);
2025
2026         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2027                 return 0;
2028
2029         /* even though OPEN succeeded, access is denied. Close the file */
2030         nfs4_close_state(state, fmode);
2031         return -EACCES;
2032 }
2033
2034 /*
2035  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2036  */
2037 static int _nfs4_proc_open(struct nfs4_opendata *data)
2038 {
2039         struct inode *dir = data->dir->d_inode;
2040         struct nfs_server *server = NFS_SERVER(dir);
2041         struct nfs_openargs *o_arg = &data->o_arg;
2042         struct nfs_openres *o_res = &data->o_res;
2043         int status;
2044
2045         status = nfs4_run_open_task(data, 0);
2046         if (!data->rpc_done)
2047                 return status;
2048         if (status != 0) {
2049                 if (status == -NFS4ERR_BADNAME &&
2050                                 !(o_arg->open_flags & O_CREAT))
2051                         return -ENOENT;
2052                 return status;
2053         }
2054
2055         nfs_fattr_map_and_free_names(server, &data->f_attr);
2056
2057         if (o_arg->open_flags & O_CREAT) {
2058                 update_changeattr(dir, &o_res->cinfo);
2059                 if (o_arg->open_flags & O_EXCL)
2060                         data->file_created = 1;
2061                 else if (o_res->cinfo.before != o_res->cinfo.after)
2062                         data->file_created = 1;
2063         }
2064         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2065                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2066         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2067                 status = _nfs4_proc_open_confirm(data);
2068                 if (status != 0)
2069                         return status;
2070         }
2071         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2072                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2073         return 0;
2074 }
2075
2076 static int nfs4_recover_expired_lease(struct nfs_server *server)
2077 {
2078         return nfs4_client_recover_expired_lease(server->nfs_client);
2079 }
2080
2081 /*
2082  * OPEN_EXPIRED:
2083  *      reclaim state on the server after a network partition.
2084  *      Assumes caller holds the appropriate lock
2085  */
2086 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2087 {
2088         struct nfs4_opendata *opendata;
2089         int ret;
2090
2091         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2092                         NFS4_OPEN_CLAIM_FH);
2093         if (IS_ERR(opendata))
2094                 return PTR_ERR(opendata);
2095         ret = nfs4_open_recover(opendata, state);
2096         if (ret == -ESTALE)
2097                 d_drop(ctx->dentry);
2098         nfs4_opendata_put(opendata);
2099         return ret;
2100 }
2101
2102 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2103 {
2104         struct nfs_server *server = NFS_SERVER(state->inode);
2105         struct nfs4_exception exception = { };
2106         int err;
2107
2108         do {
2109                 err = _nfs4_open_expired(ctx, state);
2110                 trace_nfs4_open_expired(ctx, 0, err);
2111                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2112                         continue;
2113                 switch (err) {
2114                 default:
2115                         goto out;
2116                 case -NFS4ERR_GRACE:
2117                 case -NFS4ERR_DELAY:
2118                         nfs4_handle_exception(server, err, &exception);
2119                         err = 0;
2120                 }
2121         } while (exception.retry);
2122 out:
2123         return err;
2124 }
2125
2126 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2127 {
2128         struct nfs_open_context *ctx;
2129         int ret;
2130
2131         ctx = nfs4_state_find_open_context(state);
2132         if (IS_ERR(ctx))
2133                 return -EAGAIN;
2134         ret = nfs4_do_open_expired(ctx, state);
2135         put_nfs_open_context(ctx);
2136         return ret;
2137 }
2138
2139 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2140 {
2141         nfs_remove_bad_delegation(state->inode);
2142         write_seqlock(&state->seqlock);
2143         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2144         write_sequnlock(&state->seqlock);
2145         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2146 }
2147
2148 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2149 {
2150         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2151                 nfs_finish_clear_delegation_stateid(state);
2152 }
2153
2154 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2155 {
2156         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2157         nfs40_clear_delegation_stateid(state);
2158         return nfs4_open_expired(sp, state);
2159 }
2160
2161 #if defined(CONFIG_NFS_V4_1)
2162 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2163 {
2164         struct nfs_server *server = NFS_SERVER(state->inode);
2165         nfs4_stateid stateid;
2166         struct nfs_delegation *delegation;
2167         struct rpc_cred *cred;
2168         int status;
2169
2170         /* Get the delegation credential for use by test/free_stateid */
2171         rcu_read_lock();
2172         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2173         if (delegation == NULL) {
2174                 rcu_read_unlock();
2175                 return;
2176         }
2177
2178         nfs4_stateid_copy(&stateid, &delegation->stateid);
2179         cred = get_rpccred(delegation->cred);
2180         rcu_read_unlock();
2181         status = nfs41_test_stateid(server, &stateid, cred);
2182         trace_nfs4_test_delegation_stateid(state, NULL, status);
2183
2184         if (status != NFS_OK) {
2185                 /* Free the stateid unless the server explicitly
2186                  * informs us the stateid is unrecognized. */
2187                 if (status != -NFS4ERR_BAD_STATEID)
2188                         nfs41_free_stateid(server, &stateid, cred);
2189                 nfs_finish_clear_delegation_stateid(state);
2190         }
2191
2192         put_rpccred(cred);
2193 }
2194
2195 /**
2196  * nfs41_check_open_stateid - possibly free an open stateid
2197  *
2198  * @state: NFSv4 state for an inode
2199  *
2200  * Returns NFS_OK if recovery for this stateid is now finished.
2201  * Otherwise a negative NFS4ERR value is returned.
2202  */
2203 static int nfs41_check_open_stateid(struct nfs4_state *state)
2204 {
2205         struct nfs_server *server = NFS_SERVER(state->inode);
2206         nfs4_stateid *stateid = &state->open_stateid;
2207         struct rpc_cred *cred = state->owner->so_cred;
2208         int status;
2209
2210         /* If a state reset has been done, test_stateid is unneeded */
2211         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2212             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2213             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2214                 return -NFS4ERR_BAD_STATEID;
2215
2216         status = nfs41_test_stateid(server, stateid, cred);
2217         trace_nfs4_test_open_stateid(state, NULL, status);
2218         if (status != NFS_OK) {
2219                 /* Free the stateid unless the server explicitly
2220                  * informs us the stateid is unrecognized. */
2221                 if (status != -NFS4ERR_BAD_STATEID)
2222                         nfs41_free_stateid(server, stateid, cred);
2223
2224                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2225                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2226                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2227                 clear_bit(NFS_OPEN_STATE, &state->flags);
2228         }
2229         return status;
2230 }
2231
2232 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2233 {
2234         int status;
2235
2236         nfs41_check_delegation_stateid(state);
2237         status = nfs41_check_open_stateid(state);
2238         if (status != NFS_OK)
2239                 status = nfs4_open_expired(sp, state);
2240         return status;
2241 }
2242 #endif
2243
2244 /*
2245  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2246  * fields corresponding to attributes that were used to store the verifier.
2247  * Make sure we clobber those fields in the later setattr call
2248  */
2249 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2250 {
2251         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2252             !(sattr->ia_valid & ATTR_ATIME_SET))
2253                 sattr->ia_valid |= ATTR_ATIME;
2254
2255         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2256             !(sattr->ia_valid & ATTR_MTIME_SET))
2257                 sattr->ia_valid |= ATTR_MTIME;
2258 }
2259
2260 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2261                 fmode_t fmode,
2262                 int flags,
2263                 struct nfs_open_context *ctx)
2264 {
2265         struct nfs4_state_owner *sp = opendata->owner;
2266         struct nfs_server *server = sp->so_server;
2267         struct dentry *dentry;
2268         struct nfs4_state *state;
2269         unsigned int seq;
2270         int ret;
2271
2272         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2273
2274         ret = _nfs4_proc_open(opendata);
2275         if (ret != 0)
2276                 goto out;
2277
2278         state = nfs4_opendata_to_nfs4_state(opendata);
2279         ret = PTR_ERR(state);
2280         if (IS_ERR(state))
2281                 goto out;
2282         if (server->caps & NFS_CAP_POSIX_LOCK)
2283                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2284
2285         dentry = opendata->dentry;
2286         if (dentry->d_inode == NULL) {
2287                 /* FIXME: Is this d_drop() ever needed? */
2288                 d_drop(dentry);
2289                 dentry = d_add_unique(dentry, igrab(state->inode));
2290                 if (dentry == NULL) {
2291                         dentry = opendata->dentry;
2292                 } else if (dentry != ctx->dentry) {
2293                         dput(ctx->dentry);
2294                         ctx->dentry = dget(dentry);
2295                 }
2296                 nfs_set_verifier(dentry,
2297                                 nfs_save_change_attribute(opendata->dir->d_inode));
2298         }
2299
2300         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2301         if (ret != 0)
2302                 goto out;
2303
2304         ctx->state = state;
2305         if (dentry->d_inode == state->inode) {
2306                 nfs_inode_attach_open_context(ctx);
2307                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2308                         nfs4_schedule_stateid_recovery(server, state);
2309         }
2310 out:
2311         return ret;
2312 }
2313
2314 /*
2315  * Returns a referenced nfs4_state
2316  */
2317 static int _nfs4_do_open(struct inode *dir,
2318                         struct nfs_open_context *ctx,
2319                         int flags,
2320                         struct iattr *sattr,
2321                         struct nfs4_label *label,
2322                         int *opened)
2323 {
2324         struct nfs4_state_owner  *sp;
2325         struct nfs4_state     *state = NULL;
2326         struct nfs_server       *server = NFS_SERVER(dir);
2327         struct nfs4_opendata *opendata;
2328         struct dentry *dentry = ctx->dentry;
2329         struct rpc_cred *cred = ctx->cred;
2330         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2331         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2332         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2333         struct nfs4_label *olabel = NULL;
2334         int status;
2335
2336         /* Protect against reboot recovery conflicts */
2337         status = -ENOMEM;
2338         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2339         if (sp == NULL) {
2340                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2341                 goto out_err;
2342         }
2343         status = nfs4_recover_expired_lease(server);
2344         if (status != 0)
2345                 goto err_put_state_owner;
2346         if (dentry->d_inode != NULL)
2347                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2348         status = -ENOMEM;
2349         if (dentry->d_inode)
2350                 claim = NFS4_OPEN_CLAIM_FH;
2351         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2352                         label, claim, GFP_KERNEL);
2353         if (opendata == NULL)
2354                 goto err_put_state_owner;
2355
2356         if (label) {
2357                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2358                 if (IS_ERR(olabel)) {
2359                         status = PTR_ERR(olabel);
2360                         goto err_opendata_put;
2361                 }
2362         }
2363
2364         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2365                 if (!opendata->f_attr.mdsthreshold) {
2366                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2367                         if (!opendata->f_attr.mdsthreshold)
2368                                 goto err_free_label;
2369                 }
2370                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2371         }
2372         if (dentry->d_inode != NULL)
2373                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2374
2375         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2376         if (status != 0)
2377                 goto err_free_label;
2378         state = ctx->state;
2379
2380         if ((opendata->o_arg.open_flags & O_EXCL) &&
2381             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2382                 nfs4_exclusive_attrset(opendata, sattr);
2383
2384                 nfs_fattr_init(opendata->o_res.f_attr);
2385                 status = nfs4_do_setattr(state->inode, cred,
2386                                 opendata->o_res.f_attr, sattr,
2387                                 state, label, olabel);
2388                 if (status == 0) {
2389                         nfs_setattr_update_inode(state->inode, sattr);
2390                         nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2391                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2392                 }
2393         }
2394         if (opendata->file_created)
2395                 *opened |= FILE_CREATED;
2396
2397         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2398                 *ctx_th = opendata->f_attr.mdsthreshold;
2399                 opendata->f_attr.mdsthreshold = NULL;
2400         }
2401
2402         nfs4_label_free(olabel);
2403
2404         nfs4_opendata_put(opendata);
2405         nfs4_put_state_owner(sp);
2406         return 0;
2407 err_free_label:
2408         nfs4_label_free(olabel);
2409 err_opendata_put:
2410         nfs4_opendata_put(opendata);
2411 err_put_state_owner:
2412         nfs4_put_state_owner(sp);
2413 out_err:
2414         return status;
2415 }
2416
2417
2418 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2419                                         struct nfs_open_context *ctx,
2420                                         int flags,
2421                                         struct iattr *sattr,
2422                                         struct nfs4_label *label,
2423                                         int *opened)
2424 {
2425         struct nfs_server *server = NFS_SERVER(dir);
2426         struct nfs4_exception exception = { };
2427         struct nfs4_state *res;
2428         int status;
2429
2430         do {
2431                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2432                 res = ctx->state;
2433                 trace_nfs4_open_file(ctx, flags, status);
2434                 if (status == 0)
2435                         break;
2436                 /* NOTE: BAD_SEQID means the server and client disagree about the
2437                  * book-keeping w.r.t. state-changing operations
2438                  * (OPEN/CLOSE/LOCK/LOCKU...)
2439                  * It is actually a sign of a bug on the client or on the server.
2440                  *
2441                  * If we receive a BAD_SEQID error in the particular case of
2442                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2443                  * have unhashed the old state_owner for us, and that we can
2444                  * therefore safely retry using a new one. We should still warn
2445                  * the user though...
2446                  */
2447                 if (status == -NFS4ERR_BAD_SEQID) {
2448                         pr_warn_ratelimited("NFS: v4 server %s "
2449                                         " returned a bad sequence-id error!\n",
2450                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2451                         exception.retry = 1;
2452                         continue;
2453                 }
2454                 /*
2455                  * BAD_STATEID on OPEN means that the server cancelled our
2456                  * state before it received the OPEN_CONFIRM.
2457                  * Recover by retrying the request as per the discussion
2458                  * on Page 181 of RFC3530.
2459                  */
2460                 if (status == -NFS4ERR_BAD_STATEID) {
2461                         exception.retry = 1;
2462                         continue;
2463                 }
2464                 if (status == -EAGAIN) {
2465                         /* We must have found a delegation */
2466                         exception.retry = 1;
2467                         continue;
2468                 }
2469                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2470                         continue;
2471                 res = ERR_PTR(nfs4_handle_exception(server,
2472                                         status, &exception));
2473         } while (exception.retry);
2474         return res;
2475 }
2476
2477 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2478                             struct nfs_fattr *fattr, struct iattr *sattr,
2479                             struct nfs4_state *state, struct nfs4_label *ilabel,
2480                             struct nfs4_label *olabel)
2481 {
2482         struct nfs_server *server = NFS_SERVER(inode);
2483         struct nfs_setattrargs  arg = {
2484                 .fh             = NFS_FH(inode),
2485                 .iap            = sattr,
2486                 .server         = server,
2487                 .bitmask = server->attr_bitmask,
2488                 .label          = ilabel,
2489         };
2490         struct nfs_setattrres  res = {
2491                 .fattr          = fattr,
2492                 .label          = olabel,
2493                 .server         = server,
2494         };
2495         struct rpc_message msg = {
2496                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2497                 .rpc_argp       = &arg,
2498                 .rpc_resp       = &res,
2499                 .rpc_cred       = cred,
2500         };
2501         unsigned long timestamp = jiffies;
2502         fmode_t fmode;
2503         bool truncate;
2504         int status;
2505
2506         arg.bitmask = nfs4_bitmask(server, ilabel);
2507         if (ilabel)
2508                 arg.bitmask = nfs4_bitmask(server, olabel);
2509
2510         nfs_fattr_init(fattr);
2511
2512         /* Servers should only apply open mode checks for file size changes */
2513         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2514         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2515
2516         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2517                 /* Use that stateid */
2518         } else if (truncate && state != NULL) {
2519                 struct nfs_lockowner lockowner = {
2520                         .l_owner = current->files,
2521                         .l_pid = current->tgid,
2522                 };
2523                 if (!nfs4_valid_open_stateid(state))
2524                         return -EBADF;
2525                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2526                                 &lockowner) == -EIO)
2527                         return -EBADF;
2528         } else
2529                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2530
2531         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2532         if (status == 0 && state != NULL)
2533                 renew_lease(server, timestamp);
2534         return status;
2535 }
2536
2537 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2538                            struct nfs_fattr *fattr, struct iattr *sattr,
2539                            struct nfs4_state *state, struct nfs4_label *ilabel,
2540                            struct nfs4_label *olabel)
2541 {
2542         struct nfs_server *server = NFS_SERVER(inode);
2543         struct nfs4_exception exception = {
2544                 .state = state,
2545                 .inode = inode,
2546         };
2547         int err;
2548         do {
2549                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2550                 trace_nfs4_setattr(inode, err);
2551                 switch (err) {
2552                 case -NFS4ERR_OPENMODE:
2553                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2554                                 pr_warn_once("NFSv4: server %s is incorrectly "
2555                                                 "applying open mode checks to "
2556                                                 "a SETATTR that is not "
2557                                                 "changing file size.\n",
2558                                                 server->nfs_client->cl_hostname);
2559                         }
2560                         if (state && !(state->state & FMODE_WRITE)) {
2561                                 err = -EBADF;
2562                                 if (sattr->ia_valid & ATTR_OPEN)
2563                                         err = -EACCES;
2564                                 goto out;
2565                         }
2566                 }
2567                 err = nfs4_handle_exception(server, err, &exception);
2568         } while (exception.retry);
2569 out:
2570         return err;
2571 }
2572
2573 struct nfs4_closedata {
2574         struct inode *inode;
2575         struct nfs4_state *state;
2576         struct nfs_closeargs arg;
2577         struct nfs_closeres res;
2578         struct nfs_fattr fattr;
2579         unsigned long timestamp;
2580         bool roc;
2581         u32 roc_barrier;
2582 };
2583
2584 static void nfs4_free_closedata(void *data)
2585 {
2586         struct nfs4_closedata *calldata = data;
2587         struct nfs4_state_owner *sp = calldata->state->owner;
2588         struct super_block *sb = calldata->state->inode->i_sb;
2589
2590         if (calldata->roc)
2591                 pnfs_roc_release(calldata->state->inode);
2592         nfs4_put_open_state(calldata->state);
2593         nfs_free_seqid(calldata->arg.seqid);
2594         nfs4_put_state_owner(sp);
2595         nfs_sb_deactive(sb);
2596         kfree(calldata);
2597 }
2598
2599 static void nfs4_close_done(struct rpc_task *task, void *data)
2600 {
2601         struct nfs4_closedata *calldata = data;
2602         struct nfs4_state *state = calldata->state;
2603         struct nfs_server *server = NFS_SERVER(calldata->inode);
2604         nfs4_stateid *res_stateid = NULL;
2605
2606         dprintk("%s: begin!\n", __func__);
2607         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2608                 return;
2609         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2610         /* hmm. we are done with the inode, and in the process of freeing
2611          * the state_owner. we keep this around to process errors
2612          */
2613         switch (task->tk_status) {
2614                 case 0:
2615                         res_stateid = &calldata->res.stateid;
2616                         if (calldata->arg.fmode == 0 && calldata->roc)
2617                                 pnfs_roc_set_barrier(state->inode,
2618                                                      calldata->roc_barrier);
2619                         renew_lease(server, calldata->timestamp);
2620                         break;
2621                 case -NFS4ERR_ADMIN_REVOKED:
2622                 case -NFS4ERR_STALE_STATEID:
2623                 case -NFS4ERR_OLD_STATEID:
2624                 case -NFS4ERR_BAD_STATEID:
2625                 case -NFS4ERR_EXPIRED:
2626                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2627                                                 &state->stateid)) {
2628                                 rpc_restart_call_prepare(task);
2629                                 goto out_release;
2630                         }
2631                         if (calldata->arg.fmode == 0)
2632                                 break;
2633                 default:
2634                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2635                                 rpc_restart_call_prepare(task);
2636                                 goto out_release;
2637                         }
2638         }
2639         nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2640 out_release:
2641         nfs_release_seqid(calldata->arg.seqid);
2642         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2643         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2644 }
2645
2646 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2647 {
2648         struct nfs4_closedata *calldata = data;
2649         struct nfs4_state *state = calldata->state;
2650         struct inode *inode = calldata->inode;
2651         bool is_rdonly, is_wronly, is_rdwr;
2652         int call_close = 0;
2653
2654         dprintk("%s: begin!\n", __func__);
2655         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2656                 goto out_wait;
2657
2658         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2659         spin_lock(&state->owner->so_lock);
2660         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2661         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2662         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2663         nfs4_stateid_copy(&calldata->arg.stateid, &state->stateid);
2664         /* Calculate the change in open mode */
2665         calldata->arg.fmode = 0;
2666         if (state->n_rdwr == 0) {
2667                 if (state->n_rdonly == 0)
2668                         call_close |= is_rdonly;
2669                 else if (is_rdonly)
2670                         calldata->arg.fmode |= FMODE_READ;
2671                 if (state->n_wronly == 0)
2672                         call_close |= is_wronly;
2673                 else if (is_wronly)
2674                         calldata->arg.fmode |= FMODE_WRITE;
2675         } else if (is_rdwr)
2676                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2677
2678         if (calldata->arg.fmode == 0)
2679                 call_close |= is_rdwr;
2680
2681         if (!nfs4_valid_open_stateid(state))
2682                 call_close = 0;
2683         spin_unlock(&state->owner->so_lock);
2684
2685         if (!call_close) {
2686                 /* Note: exit _without_ calling nfs4_close_done */
2687                 goto out_no_action;
2688         }
2689
2690         if (calldata->arg.fmode == 0) {
2691                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2692                 if (calldata->roc &&
2693                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2694                         nfs_release_seqid(calldata->arg.seqid);
2695                         goto out_wait;
2696                     }
2697         }
2698
2699         nfs_fattr_init(calldata->res.fattr);
2700         calldata->timestamp = jiffies;
2701         if (nfs4_setup_sequence(NFS_SERVER(inode),
2702                                 &calldata->arg.seq_args,
2703                                 &calldata->res.seq_res,
2704                                 task) != 0)
2705                 nfs_release_seqid(calldata->arg.seqid);
2706         dprintk("%s: done!\n", __func__);
2707         return;
2708 out_no_action:
2709         task->tk_action = NULL;
2710 out_wait:
2711         nfs4_sequence_done(task, &calldata->res.seq_res);
2712 }
2713
2714 static const struct rpc_call_ops nfs4_close_ops = {
2715         .rpc_call_prepare = nfs4_close_prepare,
2716         .rpc_call_done = nfs4_close_done,
2717         .rpc_release = nfs4_free_closedata,
2718 };
2719
2720 static bool nfs4_roc(struct inode *inode)
2721 {
2722         if (!nfs_have_layout(inode))
2723                 return false;
2724         return pnfs_roc(inode);
2725 }
2726
2727 /*
2728  * It is possible for data to be read/written from a mem-mapped file
2729  * after the sys_close call (which hits the vfs layer as a flush).
2730  * This means that we can't safely call nfsv4 close on a file until
2731  * the inode is cleared. This in turn means that we are not good
2732  * NFSv4 citizens - we do not indicate to the server to update the file's
2733  * share state even when we are done with one of the three share
2734  * stateid's in the inode.
2735  *
2736  * NOTE: Caller must be holding the sp->so_owner semaphore!
2737  */
2738 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2739 {
2740         struct nfs_server *server = NFS_SERVER(state->inode);
2741         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2742         struct nfs4_closedata *calldata;
2743         struct nfs4_state_owner *sp = state->owner;
2744         struct rpc_task *task;
2745         struct rpc_message msg = {
2746                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2747                 .rpc_cred = state->owner->so_cred,
2748         };
2749         struct rpc_task_setup task_setup_data = {
2750                 .rpc_client = server->client,
2751                 .rpc_message = &msg,
2752                 .callback_ops = &nfs4_close_ops,
2753                 .workqueue = nfsiod_workqueue,
2754                 .flags = RPC_TASK_ASYNC,
2755         };
2756         int status = -ENOMEM;
2757
2758         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2759                 &task_setup_data.rpc_client, &msg);
2760
2761         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2762         if (calldata == NULL)
2763                 goto out;
2764         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2765         calldata->inode = state->inode;
2766         calldata->state = state;
2767         calldata->arg.fh = NFS_FH(state->inode);
2768         /* Serialization for the sequence id */
2769         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2770         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2771         if (IS_ERR(calldata->arg.seqid))
2772                 goto out_free_calldata;
2773         calldata->arg.fmode = 0;
2774         calldata->arg.bitmask = server->cache_consistency_bitmask;
2775         calldata->res.fattr = &calldata->fattr;
2776         calldata->res.seqid = calldata->arg.seqid;
2777         calldata->res.server = server;
2778         calldata->roc = nfs4_roc(state->inode);
2779         nfs_sb_active(calldata->inode->i_sb);
2780
2781         msg.rpc_argp = &calldata->arg;
2782         msg.rpc_resp = &calldata->res;
2783         task_setup_data.callback_data = calldata;
2784         task = rpc_run_task(&task_setup_data);
2785         if (IS_ERR(task))
2786                 return PTR_ERR(task);
2787         status = 0;
2788         if (wait)
2789                 status = rpc_wait_for_completion_task(task);
2790         rpc_put_task(task);
2791         return status;
2792 out_free_calldata:
2793         kfree(calldata);
2794 out:
2795         nfs4_put_open_state(state);
2796         nfs4_put_state_owner(sp);
2797         return status;
2798 }
2799
2800 static struct inode *
2801 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2802                 int open_flags, struct iattr *attr, int *opened)
2803 {
2804         struct nfs4_state *state;
2805         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2806
2807         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2808
2809         /* Protect against concurrent sillydeletes */
2810         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2811
2812         nfs4_label_release_security(label);
2813
2814         if (IS_ERR(state))
2815                 return ERR_CAST(state);
2816         return state->inode;
2817 }
2818
2819 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2820 {
2821         if (ctx->state == NULL)
2822                 return;
2823         if (is_sync)
2824                 nfs4_close_sync(ctx->state, ctx->mode);
2825         else
2826                 nfs4_close_state(ctx->state, ctx->mode);
2827 }
2828
2829 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2830 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2831 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2832
2833 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2834 {
2835         struct nfs4_server_caps_arg args = {
2836                 .fhandle = fhandle,
2837         };
2838         struct nfs4_server_caps_res res = {};
2839         struct rpc_message msg = {
2840                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2841                 .rpc_argp = &args,
2842                 .rpc_resp = &res,
2843         };
2844         int status;
2845
2846         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2847         if (status == 0) {
2848                 /* Sanity check the server answers */
2849                 switch (server->nfs_client->cl_minorversion) {
2850                 case 0:
2851                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2852                         res.attr_bitmask[2] = 0;
2853                         break;
2854                 case 1:
2855                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2856                         break;
2857                 case 2:
2858                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2859                 }
2860                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2861                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2862                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2863                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2864                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2865                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2866                                 NFS_CAP_SECURITY_LABEL);
2867                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2868                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2869                         server->caps |= NFS_CAP_ACLS;
2870                 if (res.has_links != 0)
2871                         server->caps |= NFS_CAP_HARDLINKS;
2872                 if (res.has_symlinks != 0)
2873                         server->caps |= NFS_CAP_SYMLINKS;
2874                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2875                         server->caps |= NFS_CAP_FILEID;
2876                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2877                         server->caps |= NFS_CAP_MODE;
2878                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2879                         server->caps |= NFS_CAP_NLINK;
2880                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2881                         server->caps |= NFS_CAP_OWNER;
2882                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2883                         server->caps |= NFS_CAP_OWNER_GROUP;
2884                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2885                         server->caps |= NFS_CAP_ATIME;
2886                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2887                         server->caps |= NFS_CAP_CTIME;
2888                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2889                         server->caps |= NFS_CAP_MTIME;
2890 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2891                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2892                         server->caps |= NFS_CAP_SECURITY_LABEL;
2893 #endif
2894                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2895                                 sizeof(server->attr_bitmask));
2896                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2897
2898                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2899                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2900                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2901                 server->cache_consistency_bitmask[2] = 0;
2902                 server->acl_bitmask = res.acl_bitmask;
2903                 server->fh_expire_type = res.fh_expire_type;
2904         }
2905
2906         return status;
2907 }
2908
2909 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2910 {
2911         struct nfs4_exception exception = { };
2912         int err;
2913         do {
2914                 err = nfs4_handle_exception(server,
2915                                 _nfs4_server_capabilities(server, fhandle),
2916                                 &exception);
2917         } while (exception.retry);
2918         return err;
2919 }
2920
2921 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2922                 struct nfs_fsinfo *info)
2923 {
2924         u32 bitmask[3];
2925         struct nfs4_lookup_root_arg args = {
2926                 .bitmask = bitmask,
2927         };
2928         struct nfs4_lookup_res res = {
2929                 .server = server,
2930                 .fattr = info->fattr,
2931                 .fh = fhandle,
2932         };
2933         struct rpc_message msg = {
2934                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2935                 .rpc_argp = &args,
2936                 .rpc_resp = &res,
2937         };
2938
2939         bitmask[0] = nfs4_fattr_bitmap[0];
2940         bitmask[1] = nfs4_fattr_bitmap[1];
2941         /*
2942          * Process the label in the upcoming getfattr
2943          */
2944         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2945
2946         nfs_fattr_init(info->fattr);
2947         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2948 }
2949
2950 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2951                 struct nfs_fsinfo *info)
2952 {
2953         struct nfs4_exception exception = { };
2954         int err;
2955         do {
2956                 err = _nfs4_lookup_root(server, fhandle, info);
2957                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2958                 switch (err) {
2959                 case 0:
2960                 case -NFS4ERR_WRONGSEC:
2961                         goto out;
2962                 default:
2963                         err = nfs4_handle_exception(server, err, &exception);
2964                 }
2965         } while (exception.retry);
2966 out:
2967         return err;
2968 }
2969
2970 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2971                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2972 {
2973         struct rpc_auth_create_args auth_args = {
2974                 .pseudoflavor = flavor,
2975         };
2976         struct rpc_auth *auth;
2977         int ret;
2978
2979         auth = rpcauth_create(&auth_args, server->client);
2980         if (IS_ERR(auth)) {
2981                 ret = -EACCES;
2982                 goto out;
2983         }
2984         ret = nfs4_lookup_root(server, fhandle, info);
2985 out:
2986         return ret;
2987 }
2988
2989 /*
2990  * Retry pseudoroot lookup with various security flavors.  We do this when:
2991  *
2992  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2993  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2994  *
2995  * Returns zero on success, or a negative NFS4ERR value, or a
2996  * negative errno value.
2997  */
2998 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2999                               struct nfs_fsinfo *info)
3000 {
3001         /* Per 3530bis 15.33.5 */
3002         static const rpc_authflavor_t flav_array[] = {
3003                 RPC_AUTH_GSS_KRB5P,
3004                 RPC_AUTH_GSS_KRB5I,
3005                 RPC_AUTH_GSS_KRB5,
3006                 RPC_AUTH_UNIX,                  /* courtesy */
3007                 RPC_AUTH_NULL,
3008         };
3009         int status = -EPERM;
3010         size_t i;
3011
3012         if (server->auth_info.flavor_len > 0) {
3013                 /* try each flavor specified by user */
3014                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3015                         status = nfs4_lookup_root_sec(server, fhandle, info,
3016                                                 server->auth_info.flavors[i]);
3017                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3018                                 continue;
3019                         break;
3020                 }
3021         } else {
3022                 /* no flavors specified by user, try default list */
3023                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3024                         status = nfs4_lookup_root_sec(server, fhandle, info,
3025                                                       flav_array[i]);
3026                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3027                                 continue;
3028                         break;
3029                 }
3030         }
3031
3032         /*
3033          * -EACCESS could mean that the user doesn't have correct permissions
3034          * to access the mount.  It could also mean that we tried to mount
3035          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3036          * existing mount programs don't handle -EACCES very well so it should
3037          * be mapped to -EPERM instead.
3038          */
3039         if (status == -EACCES)
3040                 status = -EPERM;
3041         return status;
3042 }
3043
3044 static int nfs4_do_find_root_sec(struct nfs_server *server,
3045                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3046 {
3047         int mv = server->nfs_client->cl_minorversion;
3048         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3049 }
3050
3051 /**
3052  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3053  * @server: initialized nfs_server handle
3054  * @fhandle: we fill in the pseudo-fs root file handle
3055  * @info: we fill in an FSINFO struct
3056  * @auth_probe: probe the auth flavours
3057  *
3058  * Returns zero on success, or a negative errno.
3059  */
3060 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3061                          struct nfs_fsinfo *info,
3062                          bool auth_probe)
3063 {
3064         int status;
3065
3066         switch (auth_probe) {
3067         case false:
3068                 status = nfs4_lookup_root(server, fhandle, info);
3069                 if (status != -NFS4ERR_WRONGSEC)
3070                         break;
3071         default:
3072                 status = nfs4_do_find_root_sec(server, fhandle, info);
3073         }
3074
3075         if (status == 0)
3076                 status = nfs4_server_capabilities(server, fhandle);
3077         if (status == 0)
3078                 status = nfs4_do_fsinfo(server, fhandle, info);
3079
3080         return nfs4_map_errors(status);
3081 }
3082
3083 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3084                               struct nfs_fsinfo *info)
3085 {
3086         int error;
3087         struct nfs_fattr *fattr = info->fattr;
3088         struct nfs4_label *label = NULL;
3089
3090         error = nfs4_server_capabilities(server, mntfh);
3091         if (error < 0) {
3092                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3093                 return error;
3094         }
3095
3096         label = nfs4_label_alloc(server, GFP_KERNEL);
3097         if (IS_ERR(label))
3098                 return PTR_ERR(label);
3099
3100         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3101         if (error < 0) {
3102                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3103                 goto err_free_label;
3104         }
3105
3106         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3107             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3108                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3109
3110 err_free_label:
3111         nfs4_label_free(label);
3112
3113         return error;
3114 }
3115
3116 /*
3117  * Get locations and (maybe) other attributes of a referral.
3118  * Note that we'll actually follow the referral later when
3119  * we detect fsid mismatch in inode revalidation
3120  */
3121 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3122                              const struct qstr *name, struct nfs_fattr *fattr,
3123                              struct nfs_fh *fhandle)
3124 {
3125         int status = -ENOMEM;
3126         struct page *page = NULL;
3127         struct nfs4_fs_locations *locations = NULL;
3128
3129         page = alloc_page(GFP_KERNEL);
3130         if (page == NULL)
3131                 goto out;
3132         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3133         if (locations == NULL)
3134                 goto out;
3135
3136         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3137         if (status != 0)
3138                 goto out;
3139
3140         /*
3141          * If the fsid didn't change, this is a migration event, not a
3142          * referral.  Cause us to drop into the exception handler, which
3143          * will kick off migration recovery.
3144          */
3145         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3146                 dprintk("%s: server did not return a different fsid for"
3147                         " a referral at %s\n", __func__, name->name);
3148                 status = -NFS4ERR_MOVED;
3149                 goto out;
3150         }
3151         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3152         nfs_fixup_referral_attributes(&locations->fattr);
3153
3154         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3155         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3156         memset(fhandle, 0, sizeof(struct nfs_fh));
3157 out:
3158         if (page)
3159                 __free_page(page);
3160         kfree(locations);
3161         return status;
3162 }
3163
3164 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3165                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3166 {
3167         struct nfs4_getattr_arg args = {
3168                 .fh = fhandle,
3169                 .bitmask = server->attr_bitmask,
3170         };
3171         struct nfs4_getattr_res res = {
3172                 .fattr = fattr,
3173                 .label = label,
3174                 .server = server,
3175         };
3176         struct rpc_message msg = {
3177                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3178                 .rpc_argp = &args,
3179                 .rpc_resp = &res,
3180         };
3181
3182         args.bitmask = nfs4_bitmask(server, label);
3183
3184         nfs_fattr_init(fattr);
3185         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3186 }
3187
3188 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3189                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3190 {
3191         struct nfs4_exception exception = { };
3192         int err;
3193         do {
3194                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3195                 trace_nfs4_getattr(server, fhandle, fattr, err);
3196                 err = nfs4_handle_exception(server, err,
3197                                 &exception);
3198         } while (exception.retry);
3199         return err;
3200 }
3201
3202 /*
3203  * The file is not closed if it is opened due to the a request to change
3204  * the size of the file. The open call will not be needed once the
3205  * VFS layer lookup-intents are implemented.
3206  *
3207  * Close is called when the inode is destroyed.
3208  * If we haven't opened the file for O_WRONLY, we
3209  * need to in the size_change case to obtain a stateid.
3210  *
3211  * Got race?
3212  * Because OPEN is always done by name in nfsv4, it is
3213  * possible that we opened a different file by the same
3214  * name.  We can recognize this race condition, but we
3215  * can't do anything about it besides returning an error.
3216  *
3217  * This will be fixed with VFS changes (lookup-intent).
3218  */
3219 static int
3220 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3221                   struct iattr *sattr)
3222 {
3223         struct inode *inode = dentry->d_inode;
3224         struct rpc_cred *cred = NULL;
3225         struct nfs4_state *state = NULL;
3226         struct nfs4_label *label = NULL;
3227         int status;
3228
3229         if (pnfs_ld_layoutret_on_setattr(inode) &&
3230             sattr->ia_valid & ATTR_SIZE &&
3231             sattr->ia_size < i_size_read(inode))
3232                 pnfs_commit_and_return_layout(inode);
3233
3234         nfs_fattr_init(fattr);
3235
3236         /* Deal with open(O_TRUNC) */
3237         if (sattr->ia_valid & ATTR_OPEN)
3238                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3239
3240         /* Optimization: if the end result is no change, don't RPC */
3241         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3242                 return 0;
3243
3244         /* Search for an existing open(O_WRITE) file */
3245         if (sattr->ia_valid & ATTR_FILE) {
3246                 struct nfs_open_context *ctx;
3247
3248                 ctx = nfs_file_open_context(sattr->ia_file);
3249                 if (ctx) {
3250                         cred = ctx->cred;
3251                         state = ctx->state;
3252                 }
3253         }
3254
3255         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3256         if (IS_ERR(label))
3257                 return PTR_ERR(label);
3258
3259         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3260         if (status == 0) {
3261                 nfs_setattr_update_inode(inode, sattr);
3262                 nfs_setsecurity(inode, fattr, label);
3263         }
3264         nfs4_label_free(label);
3265         return status;
3266 }
3267
3268 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3269                 const struct qstr *name, struct nfs_fh *fhandle,
3270                 struct nfs_fattr *fattr, struct nfs4_label *label)
3271 {
3272         struct nfs_server *server = NFS_SERVER(dir);
3273         int                    status;
3274         struct nfs4_lookup_arg args = {
3275                 .bitmask = server->attr_bitmask,
3276                 .dir_fh = NFS_FH(dir),
3277                 .name = name,
3278         };
3279         struct nfs4_lookup_res res = {
3280                 .server = server,
3281                 .fattr = fattr,
3282                 .label = label,
3283                 .fh = fhandle,
3284         };
3285         struct rpc_message msg = {
3286                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3287                 .rpc_argp = &args,
3288                 .rpc_resp = &res,
3289         };
3290
3291         args.bitmask = nfs4_bitmask(server, label);
3292
3293         nfs_fattr_init(fattr);
3294
3295         dprintk("NFS call  lookup %s\n", name->name);
3296         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3297         dprintk("NFS reply lookup: %d\n", status);
3298         return status;
3299 }
3300
3301 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3302 {
3303         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3304                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3305         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3306         fattr->nlink = 2;
3307 }
3308
3309 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3310                                    struct qstr *name, struct nfs_fh *fhandle,
3311                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3312 {
3313         struct nfs4_exception exception = { };
3314         struct rpc_clnt *client = *clnt;
3315         int err;
3316         do {
3317                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3318                 trace_nfs4_lookup(dir, name, err);
3319                 switch (err) {
3320                 case -NFS4ERR_BADNAME:
3321                         err = -ENOENT;
3322                         goto out;
3323                 case -NFS4ERR_MOVED:
3324                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3325                         goto out;
3326                 case -NFS4ERR_WRONGSEC:
3327                         err = -EPERM;
3328                         if (client != *clnt)
3329                                 goto out;
3330                         client = nfs4_negotiate_security(client, dir, name);
3331                         if (IS_ERR(client))
3332                                 return PTR_ERR(client);
3333
3334                         exception.retry = 1;
3335                         break;
3336                 default:
3337                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3338                 }
3339         } while (exception.retry);
3340
3341 out:
3342         if (err == 0)
3343                 *clnt = client;
3344         else if (client != *clnt)
3345                 rpc_shutdown_client(client);
3346
3347         return err;
3348 }
3349
3350 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3351                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3352                             struct nfs4_label *label)
3353 {
3354         int status;
3355         struct rpc_clnt *client = NFS_CLIENT(dir);
3356
3357         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3358         if (client != NFS_CLIENT(dir)) {
3359                 rpc_shutdown_client(client);
3360                 nfs_fixup_secinfo_attributes(fattr);
3361         }
3362         return status;
3363 }
3364
3365 struct rpc_clnt *
3366 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3367                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3368 {
3369         struct rpc_clnt *client = NFS_CLIENT(dir);
3370         int status;
3371
3372         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3373         if (status < 0)
3374                 return ERR_PTR(status);
3375         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3376 }
3377
3378 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3379 {
3380         struct nfs_server *server = NFS_SERVER(inode);
3381         struct nfs4_accessargs args = {
3382                 .fh = NFS_FH(inode),
3383                 .bitmask = server->cache_consistency_bitmask,
3384         };
3385         struct nfs4_accessres res = {
3386                 .server = server,
3387         };
3388         struct rpc_message msg = {
3389                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3390                 .rpc_argp = &args,
3391                 .rpc_resp = &res,
3392                 .rpc_cred = entry->cred,
3393         };
3394         int mode = entry->mask;
3395         int status = 0;
3396
3397         /*
3398          * Determine which access bits we want to ask for...
3399          */
3400         if (mode & MAY_READ)
3401                 args.access |= NFS4_ACCESS_READ;
3402         if (S_ISDIR(inode->i_mode)) {
3403                 if (mode & MAY_WRITE)
3404                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3405                 if (mode & MAY_EXEC)
3406                         args.access |= NFS4_ACCESS_LOOKUP;
3407         } else {
3408                 if (mode & MAY_WRITE)
3409                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3410                 if (mode & MAY_EXEC)
3411                         args.access |= NFS4_ACCESS_EXECUTE;
3412         }
3413
3414         res.fattr = nfs_alloc_fattr();
3415         if (res.fattr == NULL)
3416                 return -ENOMEM;
3417
3418         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3419         if (!status) {
3420                 nfs_access_set_mask(entry, res.access);
3421                 nfs_refresh_inode(inode, res.fattr);
3422         }
3423         nfs_free_fattr(res.fattr);
3424         return status;
3425 }
3426
3427 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3428 {
3429         struct nfs4_exception exception = { };
3430         int err;
3431         do {
3432                 err = _nfs4_proc_access(inode, entry);
3433                 trace_nfs4_access(inode, err);
3434                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3435                                 &exception);
3436         } while (exception.retry);
3437         return err;
3438 }
3439
3440 /*
3441  * TODO: For the time being, we don't try to get any attributes
3442  * along with any of the zero-copy operations READ, READDIR,
3443  * READLINK, WRITE.
3444  *
3445  * In the case of the first three, we want to put the GETATTR
3446  * after the read-type operation -- this is because it is hard
3447  * to predict the length of a GETATTR response in v4, and thus
3448  * align the READ data correctly.  This means that the GETATTR
3449  * may end up partially falling into the page cache, and we should
3450  * shift it into the 'tail' of the xdr_buf before processing.
3451  * To do this efficiently, we need to know the total length
3452  * of data received, which doesn't seem to be available outside
3453  * of the RPC layer.
3454  *
3455  * In the case of WRITE, we also want to put the GETATTR after
3456  * the operation -- in this case because we want to make sure
3457  * we get the post-operation mtime and size.
3458  *
3459  * Both of these changes to the XDR layer would in fact be quite
3460  * minor, but I decided to leave them for a subsequent patch.
3461  */
3462 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3463                 unsigned int pgbase, unsigned int pglen)
3464 {
3465         struct nfs4_readlink args = {
3466                 .fh       = NFS_FH(inode),
3467                 .pgbase   = pgbase,
3468                 .pglen    = pglen,
3469                 .pages    = &page,
3470         };
3471         struct nfs4_readlink_res res;
3472         struct rpc_message msg = {
3473                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3474                 .rpc_argp = &args,
3475                 .rpc_resp = &res,
3476         };
3477
3478         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3479 }
3480
3481 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3482                 unsigned int pgbase, unsigned int pglen)
3483 {
3484         struct nfs4_exception exception = { };
3485         int err;
3486         do {
3487                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3488                 trace_nfs4_readlink(inode, err);
3489                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3490                                 &exception);
3491         } while (exception.retry);
3492         return err;
3493 }
3494
3495 /*
3496  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3497  */
3498 static int
3499 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3500                  int flags)
3501 {
3502         struct nfs4_label l, *ilabel = NULL;
3503         struct nfs_open_context *ctx;
3504         struct nfs4_state *state;
3505         int opened = 0;
3506         int status = 0;
3507
3508         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3509         if (IS_ERR(ctx))
3510                 return PTR_ERR(ctx);
3511
3512         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3513
3514         sattr->ia_mode &= ~current_umask();
3515         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3516         if (IS_ERR(state)) {
3517                 status = PTR_ERR(state);
3518                 goto out;
3519         }
3520 out:
3521         nfs4_label_release_security(ilabel);
3522         put_nfs_open_context(ctx);
3523         return status;
3524 }
3525
3526 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3527 {
3528         struct nfs_server *server = NFS_SERVER(dir);
3529         struct nfs_removeargs args = {
3530                 .fh = NFS_FH(dir),
3531                 .name = *name,
3532         };
3533         struct nfs_removeres res = {
3534                 .server = server,
3535         };
3536         struct rpc_message msg = {
3537                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3538                 .rpc_argp = &args,
3539                 .rpc_resp = &res,
3540         };
3541         int status;
3542
3543         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3544         if (status == 0)
3545                 update_changeattr(dir, &res.cinfo);
3546         return status;
3547 }
3548
3549 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3550 {
3551         struct nfs4_exception exception = { };
3552         int err;
3553         do {
3554                 err = _nfs4_proc_remove(dir, name);
3555                 trace_nfs4_remove(dir, name, err);
3556                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3557                                 &exception);
3558         } while (exception.retry);
3559         return err;
3560 }
3561
3562 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3563 {
3564         struct nfs_server *server = NFS_SERVER(dir);
3565         struct nfs_removeargs *args = msg->rpc_argp;
3566         struct nfs_removeres *res = msg->rpc_resp;
3567
3568         res->server = server;
3569         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3570         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3571
3572         nfs_fattr_init(res->dir_attr);
3573 }
3574
3575 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3576 {
3577         nfs4_setup_sequence(NFS_SERVER(data->dir),
3578                         &data->args.seq_args,
3579                         &data->res.seq_res,
3580                         task);
3581 }
3582
3583 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3584 {
3585         struct nfs_unlinkdata *data = task->tk_calldata;
3586         struct nfs_removeres *res = &data->res;
3587
3588         if (!nfs4_sequence_done(task, &res->seq_res))
3589                 return 0;
3590         if (nfs4_async_handle_error(task, res->server, NULL,
3591                                     &data->timeout) == -EAGAIN)
3592                 return 0;
3593         update_changeattr(dir, &res->cinfo);
3594         return 1;
3595 }
3596
3597 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3598 {
3599         struct nfs_server *server = NFS_SERVER(dir);
3600         struct nfs_renameargs *arg = msg->rpc_argp;
3601         struct nfs_renameres *res = msg->rpc_resp;
3602
3603         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3604         res->server = server;
3605         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3606 }
3607
3608 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3609 {
3610         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3611                         &data->args.seq_args,
3612                         &data->res.seq_res,
3613                         task);
3614 }
3615
3616 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3617                                  struct inode *new_dir)
3618 {
3619         struct nfs_renamedata *data = task->tk_calldata;
3620         struct nfs_renameres *res = &data->res;
3621
3622         if (!nfs4_sequence_done(task, &res->seq_res))
3623                 return 0;
3624         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3625                 return 0;
3626
3627         update_changeattr(old_dir, &res->old_cinfo);
3628         update_changeattr(new_dir, &res->new_cinfo);
3629         return 1;
3630 }
3631
3632 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3633 {
3634         struct nfs_server *server = NFS_SERVER(inode);
3635         struct nfs4_link_arg arg = {
3636                 .fh     = NFS_FH(inode),
3637                 .dir_fh = NFS_FH(dir),
3638                 .name   = name,
3639                 .bitmask = server->attr_bitmask,
3640         };
3641         struct nfs4_link_res res = {
3642                 .server = server,
3643                 .label = NULL,
3644         };
3645         struct rpc_message msg = {
3646                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3647                 .rpc_argp = &arg,
3648                 .rpc_resp = &res,
3649         };
3650         int status = -ENOMEM;
3651
3652         res.fattr = nfs_alloc_fattr();
3653         if (res.fattr == NULL)
3654                 goto out;
3655
3656         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3657         if (IS_ERR(res.label)) {
3658                 status = PTR_ERR(res.label);
3659                 goto out;
3660         }
3661         arg.bitmask = nfs4_bitmask(server, res.label);
3662
3663         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3664         if (!status) {
3665                 update_changeattr(dir, &res.cinfo);
3666                 status = nfs_post_op_update_inode(inode, res.fattr);
3667                 if (!status)
3668                         nfs_setsecurity(inode, res.fattr, res.label);
3669         }
3670
3671
3672         nfs4_label_free(res.label);
3673
3674 out:
3675         nfs_free_fattr(res.fattr);
3676         return status;
3677 }
3678
3679 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3680 {
3681         struct nfs4_exception exception = { };
3682         int err;
3683         do {
3684                 err = nfs4_handle_exception(NFS_SERVER(inode),
3685                                 _nfs4_proc_link(inode, dir, name),
3686                                 &exception);
3687         } while (exception.retry);
3688         return err;
3689 }
3690
3691 struct nfs4_createdata {
3692         struct rpc_message msg;
3693         struct nfs4_create_arg arg;
3694         struct nfs4_create_res res;
3695         struct nfs_fh fh;
3696         struct nfs_fattr fattr;
3697         struct nfs4_label *label;
3698 };
3699
3700 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3701                 struct qstr *name, struct iattr *sattr, u32 ftype)
3702 {
3703         struct nfs4_createdata *data;
3704
3705         data = kzalloc(sizeof(*data), GFP_KERNEL);
3706         if (data != NULL) {
3707                 struct nfs_server *server = NFS_SERVER(dir);
3708
3709                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3710                 if (IS_ERR(data->label))
3711                         goto out_free;
3712
3713                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3714                 data->msg.rpc_argp = &data->arg;
3715                 data->msg.rpc_resp = &data->res;
3716                 data->arg.dir_fh = NFS_FH(dir);
3717                 data->arg.server = server;
3718                 data->arg.name = name;
3719                 data->arg.attrs = sattr;
3720                 data->arg.ftype = ftype;
3721                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3722                 data->res.server = server;
3723                 data->res.fh = &data->fh;
3724                 data->res.fattr = &data->fattr;
3725                 data->res.label = data->label;
3726                 nfs_fattr_init(data->res.fattr);
3727         }
3728         return data;
3729 out_free:
3730         kfree(data);
3731         return NULL;
3732 }
3733
3734 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3735 {
3736         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3737                                     &data->arg.seq_args, &data->res.seq_res, 1);
3738         if (status == 0) {
3739                 update_changeattr(dir, &data->res.dir_cinfo);
3740                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3741         }
3742         return status;
3743 }
3744
3745 static void nfs4_free_createdata(struct nfs4_createdata *data)
3746 {
3747         nfs4_label_free(data->label);
3748         kfree(data);
3749 }
3750
3751 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3752                 struct page *page, unsigned int len, struct iattr *sattr,
3753                 struct nfs4_label *label)
3754 {
3755         struct nfs4_createdata *data;
3756         int status = -ENAMETOOLONG;
3757
3758         if (len > NFS4_MAXPATHLEN)
3759                 goto out;
3760
3761         status = -ENOMEM;
3762         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3763         if (data == NULL)
3764                 goto out;
3765
3766         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3767         data->arg.u.symlink.pages = &page;
3768         data->arg.u.symlink.len = len;
3769         data->arg.label = label;
3770
3771         status = nfs4_do_create(dir, dentry, data);
3772
3773         nfs4_free_createdata(data);
3774 out:
3775         return status;
3776 }
3777
3778 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3779                 struct page *page, unsigned int len, struct iattr *sattr)
3780 {
3781         struct nfs4_exception exception = { };
3782         struct nfs4_label l, *label = NULL;
3783         int err;
3784
3785         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3786
3787         do {
3788                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3789                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3790                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3791                                 &exception);
3792         } while (exception.retry);
3793
3794         nfs4_label_release_security(label);
3795         return err;
3796 }
3797
3798 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3799                 struct iattr *sattr, struct nfs4_label *label)
3800 {
3801         struct nfs4_createdata *data;
3802         int status = -ENOMEM;
3803
3804         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3805         if (data == NULL)
3806                 goto out;
3807
3808         data->arg.label = label;
3809         status = nfs4_do_create(dir, dentry, data);
3810
3811         nfs4_free_createdata(data);
3812 out:
3813         return status;
3814 }
3815
3816 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3817                 struct iattr *sattr)
3818 {
3819         struct nfs4_exception exception = { };
3820         struct nfs4_label l, *label = NULL;
3821         int err;
3822
3823         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3824
3825         sattr->ia_mode &= ~current_umask();
3826         do {
3827                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3828                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3829                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3830                                 &exception);
3831         } while (exception.retry);
3832         nfs4_label_release_security(label);
3833
3834         return err;
3835 }
3836
3837 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3838                 u64 cookie, struct page **pages, unsigned int count, int plus)
3839 {
3840         struct inode            *dir = dentry->d_inode;
3841         struct nfs4_readdir_arg args = {
3842                 .fh = NFS_FH(dir),
3843                 .pages = pages,
3844                 .pgbase = 0,
3845                 .count = count,
3846                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3847                 .plus = plus,
3848         };
3849         struct nfs4_readdir_res res;
3850         struct rpc_message msg = {
3851                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3852                 .rpc_argp = &args,
3853                 .rpc_resp = &res,
3854                 .rpc_cred = cred,
3855         };
3856         int                     status;
3857
3858         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3859                         dentry,
3860                         (unsigned long long)cookie);
3861         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3862         res.pgbase = args.pgbase;
3863         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3864         if (status >= 0) {
3865                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3866                 status += args.pgbase;
3867         }
3868
3869         nfs_invalidate_atime(dir);
3870
3871         dprintk("%s: returns %d\n", __func__, status);
3872         return status;
3873 }
3874
3875 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3876                 u64 cookie, struct page **pages, unsigned int count, int plus)
3877 {
3878         struct nfs4_exception exception = { };
3879         int err;
3880         do {
3881                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3882                                 pages, count, plus);
3883                 trace_nfs4_readdir(dentry->d_inode, err);
3884                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode), err,
3885                                 &exception);
3886         } while (exception.retry);
3887         return err;
3888 }
3889
3890 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3891                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3892 {
3893         struct nfs4_createdata *data;
3894         int mode = sattr->ia_mode;
3895         int status = -ENOMEM;
3896
3897         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3898         if (data == NULL)
3899                 goto out;
3900
3901         if (S_ISFIFO(mode))
3902                 data->arg.ftype = NF4FIFO;
3903         else if (S_ISBLK(mode)) {
3904                 data->arg.ftype = NF4BLK;
3905                 data->arg.u.device.specdata1 = MAJOR(rdev);
3906                 data->arg.u.device.specdata2 = MINOR(rdev);
3907         }
3908         else if (S_ISCHR(mode)) {
3909                 data->arg.ftype = NF4CHR;
3910                 data->arg.u.device.specdata1 = MAJOR(rdev);
3911                 data->arg.u.device.specdata2 = MINOR(rdev);
3912         } else if (!S_ISSOCK(mode)) {
3913                 status = -EINVAL;
3914                 goto out_free;
3915         }
3916
3917         data->arg.label = label;
3918         status = nfs4_do_create(dir, dentry, data);
3919 out_free:
3920         nfs4_free_createdata(data);
3921 out:
3922         return status;
3923 }
3924
3925 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3926                 struct iattr *sattr, dev_t rdev)
3927 {
3928         struct nfs4_exception exception = { };
3929         struct nfs4_label l, *label = NULL;
3930         int err;
3931
3932         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3933
3934         sattr->ia_mode &= ~current_umask();
3935         do {
3936                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3937                 trace_nfs4_mknod(dir, &dentry->d_name, err);
3938                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3939                                 &exception);
3940         } while (exception.retry);
3941
3942         nfs4_label_release_security(label);
3943
3944         return err;
3945 }
3946
3947 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3948                  struct nfs_fsstat *fsstat)
3949 {
3950         struct nfs4_statfs_arg args = {
3951                 .fh = fhandle,
3952                 .bitmask = server->attr_bitmask,
3953         };
3954         struct nfs4_statfs_res res = {
3955                 .fsstat = fsstat,
3956         };
3957         struct rpc_message msg = {
3958                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3959                 .rpc_argp = &args,
3960                 .rpc_resp = &res,
3961         };
3962
3963         nfs_fattr_init(fsstat->fattr);
3964         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3965 }
3966
3967 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3968 {
3969         struct nfs4_exception exception = { };
3970         int err;
3971         do {
3972                 err = nfs4_handle_exception(server,
3973                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3974                                 &exception);
3975         } while (exception.retry);
3976         return err;
3977 }
3978
3979 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3980                 struct nfs_fsinfo *fsinfo)
3981 {
3982         struct nfs4_fsinfo_arg args = {
3983                 .fh = fhandle,
3984                 .bitmask = server->attr_bitmask,
3985         };
3986         struct nfs4_fsinfo_res res = {
3987                 .fsinfo = fsinfo,
3988         };
3989         struct rpc_message msg = {
3990                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3991                 .rpc_argp = &args,
3992                 .rpc_resp = &res,
3993         };
3994
3995         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3996 }
3997
3998 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3999 {
4000         struct nfs4_exception exception = { };
4001         unsigned long now = jiffies;
4002         int err;
4003
4004         do {
4005                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4006                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4007                 if (err == 0) {
4008                         struct nfs_client *clp = server->nfs_client;
4009
4010                         spin_lock(&clp->cl_lock);
4011                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4012                         clp->cl_last_renewal = now;
4013                         spin_unlock(&clp->cl_lock);
4014                         break;
4015                 }
4016                 err = nfs4_handle_exception(server, err, &exception);
4017         } while (exception.retry);
4018         return err;
4019 }
4020
4021 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4022 {
4023         int error;
4024
4025         nfs_fattr_init(fsinfo->fattr);
4026         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4027         if (error == 0) {
4028                 /* block layout checks this! */
4029                 server->pnfs_blksize = fsinfo->blksize;
4030                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4031         }
4032
4033         return error;
4034 }
4035
4036 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4037                 struct nfs_pathconf *pathconf)
4038 {
4039         struct nfs4_pathconf_arg args = {
4040                 .fh = fhandle,
4041                 .bitmask = server->attr_bitmask,
4042         };
4043         struct nfs4_pathconf_res res = {
4044                 .pathconf = pathconf,
4045         };
4046         struct rpc_message msg = {
4047                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4048                 .rpc_argp = &args,
4049                 .rpc_resp = &res,
4050         };
4051
4052         /* None of the pathconf attributes are mandatory to implement */
4053         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4054                 memset(pathconf, 0, sizeof(*pathconf));
4055                 return 0;
4056         }
4057
4058         nfs_fattr_init(pathconf->fattr);
4059         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4060 }
4061
4062 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4063                 struct nfs_pathconf *pathconf)
4064 {
4065         struct nfs4_exception exception = { };
4066         int err;
4067
4068         do {
4069                 err = nfs4_handle_exception(server,
4070                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4071                                 &exception);
4072         } while (exception.retry);
4073         return err;
4074 }
4075
4076 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4077                 const struct nfs_open_context *ctx,
4078                 const struct nfs_lock_context *l_ctx,
4079                 fmode_t fmode)
4080 {
4081         const struct nfs_lockowner *lockowner = NULL;
4082
4083         if (l_ctx != NULL)
4084                 lockowner = &l_ctx->lockowner;
4085         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4086 }
4087 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4088
4089 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4090                 const struct nfs_open_context *ctx,
4091                 const struct nfs_lock_context *l_ctx,
4092                 fmode_t fmode)
4093 {
4094         nfs4_stateid current_stateid;
4095
4096         /* If the current stateid represents a lost lock, then exit */
4097         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4098                 return true;
4099         return nfs4_stateid_match(stateid, &current_stateid);
4100 }
4101
4102 static bool nfs4_error_stateid_expired(int err)
4103 {
4104         switch (err) {
4105         case -NFS4ERR_DELEG_REVOKED:
4106         case -NFS4ERR_ADMIN_REVOKED:
4107         case -NFS4ERR_BAD_STATEID:
4108         case -NFS4ERR_STALE_STATEID:
4109         case -NFS4ERR_OLD_STATEID:
4110         case -NFS4ERR_OPENMODE:
4111         case -NFS4ERR_EXPIRED:
4112                 return true;
4113         }
4114         return false;
4115 }
4116
4117 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4118 {
4119         nfs_invalidate_atime(hdr->inode);
4120 }
4121
4122 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4123 {
4124         struct nfs_server *server = NFS_SERVER(hdr->inode);
4125
4126         trace_nfs4_read(hdr, task->tk_status);
4127         if (nfs4_async_handle_error(task, server,
4128                                     hdr->args.context->state,
4129                                     NULL) == -EAGAIN) {
4130                 rpc_restart_call_prepare(task);
4131                 return -EAGAIN;
4132         }
4133
4134         __nfs4_read_done_cb(hdr);
4135         if (task->tk_status > 0)
4136                 renew_lease(server, hdr->timestamp);
4137         return 0;
4138 }
4139
4140 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4141                 struct nfs_pgio_args *args)
4142 {
4143
4144         if (!nfs4_error_stateid_expired(task->tk_status) ||
4145                 nfs4_stateid_is_current(&args->stateid,
4146                                 args->context,
4147                                 args->lock_context,
4148                                 FMODE_READ))
4149                 return false;
4150         rpc_restart_call_prepare(task);
4151         return true;
4152 }
4153
4154 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4155 {
4156
4157         dprintk("--> %s\n", __func__);
4158
4159         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4160                 return -EAGAIN;
4161         if (nfs4_read_stateid_changed(task, &hdr->args))
4162                 return -EAGAIN;
4163         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4164                                     nfs4_read_done_cb(task, hdr);
4165 }
4166
4167 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4168                                  struct rpc_message *msg)
4169 {
4170         hdr->timestamp   = jiffies;
4171         hdr->pgio_done_cb = nfs4_read_done_cb;
4172         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4173         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4174 }
4175
4176 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4177                                       struct nfs_pgio_header *hdr)
4178 {
4179         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4180                         &hdr->args.seq_args,
4181                         &hdr->res.seq_res,
4182                         task))
4183                 return 0;
4184         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4185                                 hdr->args.lock_context,
4186                                 hdr->rw_ops->rw_mode) == -EIO)
4187                 return -EIO;
4188         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4189                 return -EIO;
4190         return 0;
4191 }
4192
4193 static int nfs4_write_done_cb(struct rpc_task *task,
4194                               struct nfs_pgio_header *hdr)
4195 {
4196         struct inode *inode = hdr->inode;
4197
4198         trace_nfs4_write(hdr, task->tk_status);
4199         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4200                                     hdr->args.context->state,
4201                                     NULL) == -EAGAIN) {
4202                 rpc_restart_call_prepare(task);
4203                 return -EAGAIN;
4204         }
4205         if (task->tk_status >= 0) {
4206                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4207                 nfs_post_op_update_inode_force_wcc(inode, &hdr->fattr);
4208         }
4209         return 0;
4210 }
4211
4212 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4213                 struct nfs_pgio_args *args)
4214 {
4215
4216         if (!nfs4_error_stateid_expired(task->tk_status) ||
4217                 nfs4_stateid_is_current(&args->stateid,
4218                                 args->context,
4219                                 args->lock_context,
4220                                 FMODE_WRITE))
4221                 return false;
4222         rpc_restart_call_prepare(task);
4223         return true;
4224 }
4225
4226 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4227 {
4228         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4229                 return -EAGAIN;
4230         if (nfs4_write_stateid_changed(task, &hdr->args))
4231                 return -EAGAIN;
4232         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4233                 nfs4_write_done_cb(task, hdr);
4234 }
4235
4236 static
4237 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4238 {
4239         /* Don't request attributes for pNFS or O_DIRECT writes */
4240         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4241                 return false;
4242         /* Otherwise, request attributes if and only if we don't hold
4243          * a delegation
4244          */
4245         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4246 }
4247
4248 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4249                                   struct rpc_message *msg)
4250 {
4251         struct nfs_server *server = NFS_SERVER(hdr->inode);
4252
4253         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4254                 hdr->args.bitmask = NULL;
4255                 hdr->res.fattr = NULL;
4256         } else
4257                 hdr->args.bitmask = server->cache_consistency_bitmask;
4258
4259         if (!hdr->pgio_done_cb)
4260                 hdr->pgio_done_cb = nfs4_write_done_cb;
4261         hdr->res.server = server;
4262         hdr->timestamp   = jiffies;
4263
4264         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4265         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4266 }
4267
4268 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4269 {
4270         nfs4_setup_sequence(NFS_SERVER(data->inode),
4271                         &data->args.seq_args,
4272                         &data->res.seq_res,
4273                         task);
4274 }
4275
4276 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4277 {
4278         struct inode *inode = data->inode;
4279
4280         trace_nfs4_commit(data, task->tk_status);
4281         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4282                                     NULL, NULL) == -EAGAIN) {
4283                 rpc_restart_call_prepare(task);
4284                 return -EAGAIN;
4285         }
4286         return 0;
4287 }
4288
4289 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4290 {
4291         if (!nfs4_sequence_done(task, &data->res.seq_res))
4292                 return -EAGAIN;
4293         return data->commit_done_cb(task, data);
4294 }
4295
4296 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4297 {
4298         struct nfs_server *server = NFS_SERVER(data->inode);
4299
4300         if (data->commit_done_cb == NULL)
4301                 data->commit_done_cb = nfs4_commit_done_cb;
4302         data->res.server = server;
4303         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4304         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4305 }
4306
4307 struct nfs4_renewdata {
4308         struct nfs_client       *client;
4309         unsigned long           timestamp;
4310 };
4311
4312 /*
4313  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4314  * standalone procedure for queueing an asynchronous RENEW.
4315  */
4316 static void nfs4_renew_release(void *calldata)
4317 {
4318         struct nfs4_renewdata *data = calldata;
4319         struct nfs_client *clp = data->client;
4320
4321         if (atomic_read(&clp->cl_count) > 1)
4322                 nfs4_schedule_state_renewal(clp);
4323         nfs_put_client(clp);
4324         kfree(data);
4325 }
4326
4327 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4328 {
4329         struct nfs4_renewdata *data = calldata;
4330         struct nfs_client *clp = data->client;
4331         unsigned long timestamp = data->timestamp;
4332
4333         trace_nfs4_renew_async(clp, task->tk_status);
4334         switch (task->tk_status) {
4335         case 0:
4336                 break;
4337         case -NFS4ERR_LEASE_MOVED:
4338                 nfs4_schedule_lease_moved_recovery(clp);
4339                 break;
4340         default:
4341                 /* Unless we're shutting down, schedule state recovery! */
4342                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4343                         return;
4344                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4345                         nfs4_schedule_lease_recovery(clp);
4346                         return;
4347                 }
4348                 nfs4_schedule_path_down_recovery(clp);
4349         }
4350         do_renew_lease(clp, timestamp);
4351 }
4352
4353 static const struct rpc_call_ops nfs4_renew_ops = {
4354         .rpc_call_done = nfs4_renew_done,
4355         .rpc_release = nfs4_renew_release,
4356 };
4357
4358 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4359 {
4360         struct rpc_message msg = {
4361                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4362                 .rpc_argp       = clp,
4363                 .rpc_cred       = cred,
4364         };
4365         struct nfs4_renewdata *data;
4366
4367         if (renew_flags == 0)
4368                 return 0;
4369         if (!atomic_inc_not_zero(&clp->cl_count))
4370                 return -EIO;
4371         data = kmalloc(sizeof(*data), GFP_NOFS);
4372         if (data == NULL)
4373                 return -ENOMEM;
4374         data->client = clp;
4375         data->timestamp = jiffies;
4376         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4377                         &nfs4_renew_ops, data);
4378 }
4379
4380 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4381 {
4382         struct rpc_message msg = {
4383                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4384                 .rpc_argp       = clp,
4385                 .rpc_cred       = cred,
4386         };
4387         unsigned long now = jiffies;
4388         int status;
4389
4390         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4391         if (status < 0)
4392                 return status;
4393         do_renew_lease(clp, now);
4394         return 0;
4395 }
4396
4397 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4398 {
4399         return server->caps & NFS_CAP_ACLS;
4400 }
4401
4402 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4403  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4404  * the stack.
4405  */
4406 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4407
4408 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4409                 struct page **pages, unsigned int *pgbase)
4410 {
4411         struct page *newpage, **spages;
4412         int rc = 0;
4413         size_t len;
4414         spages = pages;
4415
4416         do {
4417                 len = min_t(size_t, PAGE_SIZE, buflen);
4418                 newpage = alloc_page(GFP_KERNEL);
4419
4420                 if (newpage == NULL)
4421                         goto unwind;
4422                 memcpy(page_address(newpage), buf, len);
4423                 buf += len;
4424                 buflen -= len;
4425                 *pages++ = newpage;
4426                 rc++;
4427         } while (buflen != 0);
4428
4429         return rc;
4430
4431 unwind:
4432         for(; rc > 0; rc--)
4433                 __free_page(spages[rc-1]);
4434         return -ENOMEM;
4435 }
4436
4437 struct nfs4_cached_acl {
4438         int cached;
4439         size_t len;
4440         char data[0];
4441 };
4442
4443 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4444 {
4445         struct nfs_inode *nfsi = NFS_I(inode);
4446
4447         spin_lock(&inode->i_lock);
4448         kfree(nfsi->nfs4_acl);
4449         nfsi->nfs4_acl = acl;
4450         spin_unlock(&inode->i_lock);
4451 }
4452
4453 static void nfs4_zap_acl_attr(struct inode *inode)
4454 {
4455         nfs4_set_cached_acl(inode, NULL);
4456 }
4457
4458 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4459 {
4460         struct nfs_inode *nfsi = NFS_I(inode);
4461         struct nfs4_cached_acl *acl;
4462         int ret = -ENOENT;
4463
4464         spin_lock(&inode->i_lock);
4465         acl = nfsi->nfs4_acl;
4466         if (acl == NULL)
4467                 goto out;
4468         if (buf == NULL) /* user is just asking for length */
4469                 goto out_len;
4470         if (acl->cached == 0)
4471                 goto out;
4472         ret = -ERANGE; /* see getxattr(2) man page */
4473         if (acl->len > buflen)
4474                 goto out;
4475         memcpy(buf, acl->data, acl->len);
4476 out_len:
4477         ret = acl->len;
4478 out:
4479         spin_unlock(&inode->i_lock);
4480         return ret;
4481 }
4482
4483 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4484 {
4485         struct nfs4_cached_acl *acl;
4486         size_t buflen = sizeof(*acl) + acl_len;
4487
4488         if (buflen <= PAGE_SIZE) {
4489                 acl = kmalloc(buflen, GFP_KERNEL);
4490                 if (acl == NULL)
4491                         goto out;
4492                 acl->cached = 1;
4493                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4494         } else {
4495                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4496                 if (acl == NULL)
4497                         goto out;
4498                 acl->cached = 0;
4499         }
4500         acl->len = acl_len;
4501 out:
4502         nfs4_set_cached_acl(inode, acl);
4503 }
4504
4505 /*
4506  * The getxattr API returns the required buffer length when called with a
4507  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4508  * the required buf.  On a NULL buf, we send a page of data to the server
4509  * guessing that the ACL request can be serviced by a page. If so, we cache
4510  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4511  * the cache. If not so, we throw away the page, and cache the required
4512  * length. The next getxattr call will then produce another round trip to
4513  * the server, this time with the input buf of the required size.
4514  */
4515 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4516 {
4517         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4518         struct nfs_getaclargs args = {
4519                 .fh = NFS_FH(inode),
4520                 .acl_pages = pages,
4521                 .acl_len = buflen,
4522         };
4523         struct nfs_getaclres res = {
4524                 .acl_len = buflen,
4525         };
4526         struct rpc_message msg = {
4527                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4528                 .rpc_argp = &args,
4529                 .rpc_resp = &res,
4530         };
4531         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4532         int ret = -ENOMEM, i;
4533
4534         /* As long as we're doing a round trip to the server anyway,
4535          * let's be prepared for a page of acl data. */
4536         if (npages == 0)
4537                 npages = 1;
4538         if (npages > ARRAY_SIZE(pages))
4539                 return -ERANGE;
4540
4541         for (i = 0; i < npages; i++) {
4542                 pages[i] = alloc_page(GFP_KERNEL);
4543                 if (!pages[i])
4544                         goto out_free;
4545         }
4546
4547         /* for decoding across pages */
4548         res.acl_scratch = alloc_page(GFP_KERNEL);
4549         if (!res.acl_scratch)
4550                 goto out_free;
4551
4552         args.acl_len = npages * PAGE_SIZE;
4553         args.acl_pgbase = 0;
4554
4555         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4556                 __func__, buf, buflen, npages, args.acl_len);
4557         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4558                              &msg, &args.seq_args, &res.seq_res, 0);
4559         if (ret)
4560                 goto out_free;
4561
4562         /* Handle the case where the passed-in buffer is too short */
4563         if (res.acl_flags & NFS4_ACL_TRUNC) {
4564                 /* Did the user only issue a request for the acl length? */
4565                 if (buf == NULL)
4566                         goto out_ok;
4567                 ret = -ERANGE;
4568                 goto out_free;
4569         }
4570         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4571         if (buf) {
4572                 if (res.acl_len > buflen) {
4573                         ret = -ERANGE;
4574                         goto out_free;
4575                 }
4576                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4577         }
4578 out_ok:
4579         ret = res.acl_len;
4580 out_free:
4581         for (i = 0; i < npages; i++)
4582                 if (pages[i])
4583                         __free_page(pages[i]);
4584         if (res.acl_scratch)
4585                 __free_page(res.acl_scratch);
4586         return ret;
4587 }
4588
4589 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4590 {
4591         struct nfs4_exception exception = { };
4592         ssize_t ret;
4593         do {
4594                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4595                 trace_nfs4_get_acl(inode, ret);
4596                 if (ret >= 0)
4597                         break;
4598                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4599         } while (exception.retry);
4600         return ret;
4601 }
4602
4603 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4604 {
4605         struct nfs_server *server = NFS_SERVER(inode);
4606         int ret;
4607
4608         if (!nfs4_server_supports_acls(server))
4609                 return -EOPNOTSUPP;
4610         ret = nfs_revalidate_inode(server, inode);
4611         if (ret < 0)
4612                 return ret;
4613         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4614                 nfs_zap_acl_cache(inode);
4615         ret = nfs4_read_cached_acl(inode, buf, buflen);
4616         if (ret != -ENOENT)
4617                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4618                  * but no cached acl data, just the acl length */
4619                 return ret;
4620         return nfs4_get_acl_uncached(inode, buf, buflen);
4621 }
4622
4623 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4624 {
4625         struct nfs_server *server = NFS_SERVER(inode);
4626         struct page *pages[NFS4ACL_MAXPAGES];
4627         struct nfs_setaclargs arg = {
4628                 .fh             = NFS_FH(inode),
4629                 .acl_pages      = pages,
4630                 .acl_len        = buflen,
4631         };
4632         struct nfs_setaclres res;
4633         struct rpc_message msg = {
4634                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4635                 .rpc_argp       = &arg,
4636                 .rpc_resp       = &res,
4637         };
4638         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4639         int ret, i;
4640
4641         if (!nfs4_server_supports_acls(server))
4642                 return -EOPNOTSUPP;
4643         if (npages > ARRAY_SIZE(pages))
4644                 return -ERANGE;
4645         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4646         if (i < 0)
4647                 return i;
4648         nfs4_inode_return_delegation(inode);
4649         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4650
4651         /*
4652          * Free each page after tx, so the only ref left is
4653          * held by the network stack
4654          */
4655         for (; i > 0; i--)
4656                 put_page(pages[i-1]);
4657
4658         /*
4659          * Acl update can result in inode attribute update.
4660          * so mark the attribute cache invalid.
4661          */
4662         spin_lock(&inode->i_lock);
4663         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4664         spin_unlock(&inode->i_lock);
4665         nfs_access_zap_cache(inode);
4666         nfs_zap_acl_cache(inode);
4667         return ret;
4668 }
4669
4670 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4671 {
4672         struct nfs4_exception exception = { };
4673         int err;
4674         do {
4675                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4676                 trace_nfs4_set_acl(inode, err);
4677                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4678                                 &exception);
4679         } while (exception.retry);
4680         return err;
4681 }
4682
4683 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4684 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4685                                         size_t buflen)
4686 {
4687         struct nfs_server *server = NFS_SERVER(inode);
4688         struct nfs_fattr fattr;
4689         struct nfs4_label label = {0, 0, buflen, buf};
4690
4691         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4692         struct nfs4_getattr_arg arg = {
4693                 .fh             = NFS_FH(inode),
4694                 .bitmask        = bitmask,
4695         };
4696         struct nfs4_getattr_res res = {
4697                 .fattr          = &fattr,
4698                 .label          = &label,
4699                 .server         = server,
4700         };
4701         struct rpc_message msg = {
4702                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4703                 .rpc_argp       = &arg,
4704                 .rpc_resp       = &res,
4705         };
4706         int ret;
4707
4708         nfs_fattr_init(&fattr);
4709
4710         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4711         if (ret)
4712                 return ret;
4713         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4714                 return -ENOENT;
4715         if (buflen < label.len)
4716                 return -ERANGE;
4717         return 0;
4718 }
4719
4720 static int nfs4_get_security_label(struct inode *inode, void *buf,
4721                                         size_t buflen)
4722 {
4723         struct nfs4_exception exception = { };
4724         int err;
4725
4726         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4727                 return -EOPNOTSUPP;
4728
4729         do {
4730                 err = _nfs4_get_security_label(inode, buf, buflen);
4731                 trace_nfs4_get_security_label(inode, err);
4732                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4733                                 &exception);
4734         } while (exception.retry);
4735         return err;
4736 }
4737
4738 static int _nfs4_do_set_security_label(struct inode *inode,
4739                 struct nfs4_label *ilabel,
4740                 struct nfs_fattr *fattr,
4741                 struct nfs4_label *olabel)
4742 {
4743
4744         struct iattr sattr = {0};
4745         struct nfs_server *server = NFS_SERVER(inode);
4746         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4747         struct nfs_setattrargs arg = {
4748                 .fh             = NFS_FH(inode),
4749                 .iap            = &sattr,
4750                 .server         = server,
4751                 .bitmask        = bitmask,
4752                 .label          = ilabel,
4753         };
4754         struct nfs_setattrres res = {
4755                 .fattr          = fattr,
4756                 .label          = olabel,
4757                 .server         = server,
4758         };
4759         struct rpc_message msg = {
4760                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4761                 .rpc_argp       = &arg,
4762                 .rpc_resp       = &res,
4763         };
4764         int status;
4765
4766         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4767
4768         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4769         if (status)
4770                 dprintk("%s failed: %d\n", __func__, status);
4771
4772         return status;
4773 }
4774
4775 static int nfs4_do_set_security_label(struct inode *inode,
4776                 struct nfs4_label *ilabel,
4777                 struct nfs_fattr *fattr,
4778                 struct nfs4_label *olabel)
4779 {
4780         struct nfs4_exception exception = { };
4781         int err;
4782
4783         do {
4784                 err = _nfs4_do_set_security_label(inode, ilabel,
4785                                 fattr, olabel);
4786                 trace_nfs4_set_security_label(inode, err);
4787                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4788                                 &exception);
4789         } while (exception.retry);
4790         return err;
4791 }
4792
4793 static int
4794 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4795 {
4796         struct nfs4_label ilabel, *olabel = NULL;
4797         struct nfs_fattr fattr;
4798         struct rpc_cred *cred;
4799         struct inode *inode = dentry->d_inode;
4800         int status;
4801
4802         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4803                 return -EOPNOTSUPP;
4804
4805         nfs_fattr_init(&fattr);
4806
4807         ilabel.pi = 0;
4808         ilabel.lfs = 0;
4809         ilabel.label = (char *)buf;
4810         ilabel.len = buflen;
4811
4812         cred = rpc_lookup_cred();
4813         if (IS_ERR(cred))
4814                 return PTR_ERR(cred);
4815
4816         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4817         if (IS_ERR(olabel)) {
4818                 status = -PTR_ERR(olabel);
4819                 goto out;
4820         }
4821
4822         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4823         if (status == 0)
4824                 nfs_setsecurity(inode, &fattr, olabel);
4825
4826         nfs4_label_free(olabel);
4827 out:
4828         put_rpccred(cred);
4829         return status;
4830 }
4831 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4832
4833
4834 static int
4835 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4836                         struct nfs4_state *state, long *timeout)
4837 {
4838         struct nfs_client *clp = server->nfs_client;
4839
4840         if (task->tk_status >= 0)
4841                 return 0;
4842         switch(task->tk_status) {
4843                 case -NFS4ERR_DELEG_REVOKED:
4844                 case -NFS4ERR_ADMIN_REVOKED:
4845                 case -NFS4ERR_BAD_STATEID:
4846                 case -NFS4ERR_OPENMODE:
4847                         if (state == NULL)
4848                                 break;
4849                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4850                                 goto recovery_failed;
4851                         goto wait_on_recovery;
4852                 case -NFS4ERR_EXPIRED:
4853                         if (state != NULL) {
4854                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4855                                         goto recovery_failed;
4856                         }
4857                 case -NFS4ERR_STALE_STATEID:
4858                 case -NFS4ERR_STALE_CLIENTID:
4859                         nfs4_schedule_lease_recovery(clp);
4860                         goto wait_on_recovery;
4861                 case -NFS4ERR_MOVED:
4862                         if (nfs4_schedule_migration_recovery(server) < 0)
4863                                 goto recovery_failed;
4864                         goto wait_on_recovery;
4865                 case -NFS4ERR_LEASE_MOVED:
4866                         nfs4_schedule_lease_moved_recovery(clp);
4867                         goto wait_on_recovery;
4868 #if defined(CONFIG_NFS_V4_1)
4869                 case -NFS4ERR_BADSESSION:
4870                 case -NFS4ERR_BADSLOT:
4871                 case -NFS4ERR_BAD_HIGH_SLOT:
4872                 case -NFS4ERR_DEADSESSION:
4873                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4874                 case -NFS4ERR_SEQ_FALSE_RETRY:
4875                 case -NFS4ERR_SEQ_MISORDERED:
4876                         dprintk("%s ERROR %d, Reset session\n", __func__,
4877                                 task->tk_status);
4878                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4879                         goto wait_on_recovery;
4880 #endif /* CONFIG_NFS_V4_1 */
4881                 case -NFS4ERR_DELAY:
4882                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4883                         rpc_delay(task, nfs4_update_delay(timeout));
4884                         goto restart_call;
4885                 case -NFS4ERR_GRACE:
4886                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4887                 case -NFS4ERR_RETRY_UNCACHED_REP:
4888                 case -NFS4ERR_OLD_STATEID:
4889                         goto restart_call;
4890         }
4891         task->tk_status = nfs4_map_errors(task->tk_status);
4892         return 0;
4893 recovery_failed:
4894         task->tk_status = -EIO;
4895         return 0;
4896 wait_on_recovery:
4897         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4898         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4899                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4900         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4901                 goto recovery_failed;
4902 restart_call:
4903         task->tk_status = 0;
4904         return -EAGAIN;
4905 }
4906
4907 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4908                                     nfs4_verifier *bootverf)
4909 {
4910         __be32 verf[2];
4911
4912         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4913                 /* An impossible timestamp guarantees this value
4914                  * will never match a generated boot time. */
4915                 verf[0] = 0;
4916                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4917         } else {
4918                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4919                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4920                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4921         }
4922         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4923 }
4924
4925 static unsigned int
4926 nfs4_init_nonuniform_client_string(struct nfs_client *clp,
4927                                    char *buf, size_t len)
4928 {
4929         unsigned int result;
4930
4931         if (clp->cl_owner_id != NULL)
4932                 return strlcpy(buf, clp->cl_owner_id, len);
4933
4934         rcu_read_lock();
4935         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4936                                 clp->cl_ipaddr,
4937                                 rpc_peeraddr2str(clp->cl_rpcclient,
4938                                                         RPC_DISPLAY_ADDR),
4939                                 rpc_peeraddr2str(clp->cl_rpcclient,
4940                                                         RPC_DISPLAY_PROTO));
4941         rcu_read_unlock();
4942         clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4943         return result;
4944 }
4945
4946 static unsigned int
4947 nfs4_init_uniform_client_string(struct nfs_client *clp,
4948                                 char *buf, size_t len)
4949 {
4950         const char *nodename = clp->cl_rpcclient->cl_nodename;
4951         unsigned int result;
4952
4953         if (clp->cl_owner_id != NULL)
4954                 return strlcpy(buf, clp->cl_owner_id, len);
4955
4956         if (nfs4_client_id_uniquifier[0] != '\0')
4957                 result = scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4958                                 clp->rpc_ops->version,
4959                                 clp->cl_minorversion,
4960                                 nfs4_client_id_uniquifier,
4961                                 nodename);
4962         else
4963                 result = scnprintf(buf, len, "Linux NFSv%u.%u %s",
4964                                 clp->rpc_ops->version, clp->cl_minorversion,
4965                                 nodename);
4966         clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4967         return result;
4968 }
4969
4970 /*
4971  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
4972  * services.  Advertise one based on the address family of the
4973  * clientaddr.
4974  */
4975 static unsigned int
4976 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
4977 {
4978         if (strchr(clp->cl_ipaddr, ':') != NULL)
4979                 return scnprintf(buf, len, "tcp6");
4980         else
4981                 return scnprintf(buf, len, "tcp");
4982 }
4983
4984 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
4985 {
4986         struct nfs4_setclientid *sc = calldata;
4987
4988         if (task->tk_status == 0)
4989                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
4990 }
4991
4992 static const struct rpc_call_ops nfs4_setclientid_ops = {
4993         .rpc_call_done = nfs4_setclientid_done,
4994 };
4995
4996 /**
4997  * nfs4_proc_setclientid - Negotiate client ID
4998  * @clp: state data structure
4999  * @program: RPC program for NFSv4 callback service
5000  * @port: IP port number for NFS4 callback service
5001  * @cred: RPC credential to use for this call
5002  * @res: where to place the result
5003  *
5004  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5005  */
5006 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5007                 unsigned short port, struct rpc_cred *cred,
5008                 struct nfs4_setclientid_res *res)
5009 {
5010         nfs4_verifier sc_verifier;
5011         struct nfs4_setclientid setclientid = {
5012                 .sc_verifier = &sc_verifier,
5013                 .sc_prog = program,
5014                 .sc_cb_ident = clp->cl_cb_ident,
5015         };
5016         struct rpc_message msg = {
5017                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5018                 .rpc_argp = &setclientid,
5019                 .rpc_resp = res,
5020                 .rpc_cred = cred,
5021         };
5022         struct rpc_task *task;
5023         struct rpc_task_setup task_setup_data = {
5024                 .rpc_client = clp->cl_rpcclient,
5025                 .rpc_message = &msg,
5026                 .callback_ops = &nfs4_setclientid_ops,
5027                 .callback_data = &setclientid,
5028                 .flags = RPC_TASK_TIMEOUT,
5029         };
5030         int status;
5031
5032         /* nfs_client_id4 */
5033         nfs4_init_boot_verifier(clp, &sc_verifier);
5034         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5035                 setclientid.sc_name_len =
5036                                 nfs4_init_uniform_client_string(clp,
5037                                                 setclientid.sc_name,
5038                                                 sizeof(setclientid.sc_name));
5039         else
5040                 setclientid.sc_name_len =
5041                                 nfs4_init_nonuniform_client_string(clp,
5042                                                 setclientid.sc_name,
5043                                                 sizeof(setclientid.sc_name));
5044         /* cb_client4 */
5045         setclientid.sc_netid_len =
5046                                 nfs4_init_callback_netid(clp,
5047                                                 setclientid.sc_netid,
5048                                                 sizeof(setclientid.sc_netid));
5049         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5050                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5051                                 clp->cl_ipaddr, port >> 8, port & 255);
5052
5053         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5054                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5055                 setclientid.sc_name_len, setclientid.sc_name);
5056         task = rpc_run_task(&task_setup_data);
5057         if (IS_ERR(task)) {
5058                 status = PTR_ERR(task);
5059                 goto out;
5060         }
5061         status = task->tk_status;
5062         if (setclientid.sc_cred) {
5063                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5064                 put_rpccred(setclientid.sc_cred);
5065         }
5066         rpc_put_task(task);
5067 out:
5068         trace_nfs4_setclientid(clp, status);
5069         dprintk("NFS reply setclientid: %d\n", status);
5070         return status;
5071 }
5072
5073 /**
5074  * nfs4_proc_setclientid_confirm - Confirm client ID
5075  * @clp: state data structure
5076  * @res: result of a previous SETCLIENTID
5077  * @cred: RPC credential to use for this call
5078  *
5079  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5080  */
5081 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5082                 struct nfs4_setclientid_res *arg,
5083                 struct rpc_cred *cred)
5084 {
5085         struct rpc_message msg = {
5086                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5087                 .rpc_argp = arg,
5088                 .rpc_cred = cred,
5089         };
5090         int status;
5091
5092         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5093                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5094                 clp->cl_clientid);
5095         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5096         trace_nfs4_setclientid_confirm(clp, status);
5097         dprintk("NFS reply setclientid_confirm: %d\n", status);
5098         return status;
5099 }
5100
5101 struct nfs4_delegreturndata {
5102         struct nfs4_delegreturnargs args;
5103         struct nfs4_delegreturnres res;
5104         struct nfs_fh fh;
5105         nfs4_stateid stateid;
5106         unsigned long timestamp;
5107         struct nfs_fattr fattr;
5108         int rpc_status;
5109         struct inode *inode;
5110         bool roc;
5111         u32 roc_barrier;
5112 };
5113
5114 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5115 {
5116         struct nfs4_delegreturndata *data = calldata;
5117
5118         if (!nfs4_sequence_done(task, &data->res.seq_res))
5119                 return;
5120
5121         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5122         switch (task->tk_status) {
5123         case 0:
5124                 renew_lease(data->res.server, data->timestamp);
5125         case -NFS4ERR_ADMIN_REVOKED:
5126         case -NFS4ERR_DELEG_REVOKED:
5127         case -NFS4ERR_BAD_STATEID:
5128         case -NFS4ERR_OLD_STATEID:
5129         case -NFS4ERR_STALE_STATEID:
5130         case -NFS4ERR_EXPIRED:
5131                 task->tk_status = 0;
5132                 if (data->roc)
5133                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5134                 break;
5135         default:
5136                 if (nfs4_async_handle_error(task, data->res.server,
5137                                             NULL, NULL) == -EAGAIN) {
5138                         rpc_restart_call_prepare(task);
5139                         return;
5140                 }
5141         }
5142         data->rpc_status = task->tk_status;
5143 }
5144
5145 static void nfs4_delegreturn_release(void *calldata)
5146 {
5147         struct nfs4_delegreturndata *data = calldata;
5148
5149         if (data->roc)
5150                 pnfs_roc_release(data->inode);
5151         kfree(calldata);
5152 }
5153
5154 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5155 {
5156         struct nfs4_delegreturndata *d_data;
5157
5158         d_data = (struct nfs4_delegreturndata *)data;
5159
5160         if (d_data->roc &&
5161             pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5162                 return;
5163
5164         nfs4_setup_sequence(d_data->res.server,
5165                         &d_data->args.seq_args,
5166                         &d_data->res.seq_res,
5167                         task);
5168 }
5169
5170 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5171         .rpc_call_prepare = nfs4_delegreturn_prepare,
5172         .rpc_call_done = nfs4_delegreturn_done,
5173         .rpc_release = nfs4_delegreturn_release,
5174 };
5175
5176 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5177 {
5178         struct nfs4_delegreturndata *data;
5179         struct nfs_server *server = NFS_SERVER(inode);
5180         struct rpc_task *task;
5181         struct rpc_message msg = {
5182                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5183                 .rpc_cred = cred,
5184         };
5185         struct rpc_task_setup task_setup_data = {
5186                 .rpc_client = server->client,
5187                 .rpc_message = &msg,
5188                 .callback_ops = &nfs4_delegreturn_ops,
5189                 .flags = RPC_TASK_ASYNC,
5190         };
5191         int status = 0;
5192
5193         data = kzalloc(sizeof(*data), GFP_NOFS);
5194         if (data == NULL)
5195                 return -ENOMEM;
5196         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5197         data->args.fhandle = &data->fh;
5198         data->args.stateid = &data->stateid;
5199         data->args.bitmask = server->cache_consistency_bitmask;
5200         nfs_copy_fh(&data->fh, NFS_FH(inode));
5201         nfs4_stateid_copy(&data->stateid, stateid);
5202         data->res.fattr = &data->fattr;
5203         data->res.server = server;
5204         nfs_fattr_init(data->res.fattr);
5205         data->timestamp = jiffies;
5206         data->rpc_status = 0;
5207         data->inode = inode;
5208         data->roc = list_empty(&NFS_I(inode)->open_files) ?
5209                     pnfs_roc(inode) : false;
5210
5211         task_setup_data.callback_data = data;
5212         msg.rpc_argp = &data->args;
5213         msg.rpc_resp = &data->res;
5214         task = rpc_run_task(&task_setup_data);
5215         if (IS_ERR(task))
5216                 return PTR_ERR(task);
5217         if (!issync)
5218                 goto out;
5219         status = nfs4_wait_for_completion_rpc_task(task);
5220         if (status != 0)
5221                 goto out;
5222         status = data->rpc_status;
5223         if (status == 0)
5224                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5225         else
5226                 nfs_refresh_inode(inode, &data->fattr);
5227 out:
5228         rpc_put_task(task);
5229         return status;
5230 }
5231
5232 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5233 {
5234         struct nfs_server *server = NFS_SERVER(inode);
5235         struct nfs4_exception exception = { };
5236         int err;
5237         do {
5238                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5239                 trace_nfs4_delegreturn(inode, err);
5240                 switch (err) {
5241                         case -NFS4ERR_STALE_STATEID:
5242                         case -NFS4ERR_EXPIRED:
5243                         case 0:
5244                                 return 0;
5245                 }
5246                 err = nfs4_handle_exception(server, err, &exception);
5247         } while (exception.retry);
5248         return err;
5249 }
5250
5251 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5252 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5253
5254 /*
5255  * sleep, with exponential backoff, and retry the LOCK operation.
5256  */
5257 static unsigned long
5258 nfs4_set_lock_task_retry(unsigned long timeout)
5259 {
5260         freezable_schedule_timeout_killable_unsafe(timeout);
5261         timeout <<= 1;
5262         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5263                 return NFS4_LOCK_MAXTIMEOUT;
5264         return timeout;
5265 }
5266
5267 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5268 {
5269         struct inode *inode = state->inode;
5270         struct nfs_server *server = NFS_SERVER(inode);
5271         struct nfs_client *clp = server->nfs_client;
5272         struct nfs_lockt_args arg = {
5273                 .fh = NFS_FH(inode),
5274                 .fl = request,
5275         };
5276         struct nfs_lockt_res res = {
5277                 .denied = request,
5278         };
5279         struct rpc_message msg = {
5280                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5281                 .rpc_argp       = &arg,
5282                 .rpc_resp       = &res,
5283                 .rpc_cred       = state->owner->so_cred,
5284         };
5285         struct nfs4_lock_state *lsp;
5286         int status;
5287
5288         arg.lock_owner.clientid = clp->cl_clientid;
5289         status = nfs4_set_lock_state(state, request);
5290         if (status != 0)
5291                 goto out;
5292         lsp = request->fl_u.nfs4_fl.owner;
5293         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5294         arg.lock_owner.s_dev = server->s_dev;
5295         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5296         switch (status) {
5297                 case 0:
5298                         request->fl_type = F_UNLCK;
5299                         break;
5300                 case -NFS4ERR_DENIED:
5301                         status = 0;
5302         }
5303         request->fl_ops->fl_release_private(request);
5304         request->fl_ops = NULL;
5305 out:
5306         return status;
5307 }
5308
5309 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5310 {
5311         struct nfs4_exception exception = { };
5312         int err;
5313
5314         do {
5315                 err = _nfs4_proc_getlk(state, cmd, request);
5316                 trace_nfs4_get_lock(request, state, cmd, err);
5317                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5318                                 &exception);
5319         } while (exception.retry);
5320         return err;
5321 }
5322
5323 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5324 {
5325         int res = 0;
5326         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5327                 case FL_POSIX:
5328                         res = posix_lock_file_wait(file, fl);
5329                         break;
5330                 case FL_FLOCK:
5331                         res = flock_lock_file_wait(file, fl);
5332                         break;
5333                 default:
5334                         BUG();
5335         }
5336         return res;
5337 }
5338
5339 struct nfs4_unlockdata {
5340         struct nfs_locku_args arg;
5341         struct nfs_locku_res res;
5342         struct nfs4_lock_state *lsp;
5343         struct nfs_open_context *ctx;
5344         struct file_lock fl;
5345         const struct nfs_server *server;
5346         unsigned long timestamp;
5347 };
5348
5349 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5350                 struct nfs_open_context *ctx,
5351                 struct nfs4_lock_state *lsp,
5352                 struct nfs_seqid *seqid)
5353 {
5354         struct nfs4_unlockdata *p;
5355         struct inode *inode = lsp->ls_state->inode;
5356
5357         p = kzalloc(sizeof(*p), GFP_NOFS);
5358         if (p == NULL)
5359                 return NULL;
5360         p->arg.fh = NFS_FH(inode);
5361         p->arg.fl = &p->fl;
5362         p->arg.seqid = seqid;
5363         p->res.seqid = seqid;
5364         p->lsp = lsp;
5365         atomic_inc(&lsp->ls_count);
5366         /* Ensure we don't close file until we're done freeing locks! */
5367         p->ctx = get_nfs_open_context(ctx);
5368         memcpy(&p->fl, fl, sizeof(p->fl));
5369         p->server = NFS_SERVER(inode);
5370         return p;
5371 }
5372
5373 static void nfs4_locku_release_calldata(void *data)
5374 {
5375         struct nfs4_unlockdata *calldata = data;
5376         nfs_free_seqid(calldata->arg.seqid);
5377         nfs4_put_lock_state(calldata->lsp);
5378         put_nfs_open_context(calldata->ctx);
5379         kfree(calldata);
5380 }
5381
5382 static void nfs4_locku_done(struct rpc_task *task, void *data)
5383 {
5384         struct nfs4_unlockdata *calldata = data;
5385
5386         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5387                 return;
5388         switch (task->tk_status) {
5389                 case 0:
5390                         renew_lease(calldata->server, calldata->timestamp);
5391                         do_vfs_lock(calldata->fl.fl_file, &calldata->fl);
5392                         if (nfs4_update_lock_stateid(calldata->lsp,
5393                                         &calldata->res.stateid))
5394                                 break;
5395                 case -NFS4ERR_BAD_STATEID:
5396                 case -NFS4ERR_OLD_STATEID:
5397                 case -NFS4ERR_STALE_STATEID:
5398                 case -NFS4ERR_EXPIRED:
5399                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5400                                                 &calldata->lsp->ls_stateid))
5401                                 rpc_restart_call_prepare(task);
5402                         break;
5403                 default:
5404                         if (nfs4_async_handle_error(task, calldata->server,
5405                                                     NULL, NULL) == -EAGAIN)
5406                                 rpc_restart_call_prepare(task);
5407         }
5408         nfs_release_seqid(calldata->arg.seqid);
5409 }
5410
5411 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5412 {
5413         struct nfs4_unlockdata *calldata = data;
5414
5415         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5416                 goto out_wait;
5417         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5418         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5419                 /* Note: exit _without_ running nfs4_locku_done */
5420                 goto out_no_action;
5421         }
5422         calldata->timestamp = jiffies;
5423         if (nfs4_setup_sequence(calldata->server,
5424                                 &calldata->arg.seq_args,
5425                                 &calldata->res.seq_res,
5426                                 task) != 0)
5427                 nfs_release_seqid(calldata->arg.seqid);
5428         return;
5429 out_no_action:
5430         task->tk_action = NULL;
5431 out_wait:
5432         nfs4_sequence_done(task, &calldata->res.seq_res);
5433 }
5434
5435 static const struct rpc_call_ops nfs4_locku_ops = {
5436         .rpc_call_prepare = nfs4_locku_prepare,
5437         .rpc_call_done = nfs4_locku_done,
5438         .rpc_release = nfs4_locku_release_calldata,
5439 };
5440
5441 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5442                 struct nfs_open_context *ctx,
5443                 struct nfs4_lock_state *lsp,
5444                 struct nfs_seqid *seqid)
5445 {
5446         struct nfs4_unlockdata *data;
5447         struct rpc_message msg = {
5448                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5449                 .rpc_cred = ctx->cred,
5450         };
5451         struct rpc_task_setup task_setup_data = {
5452                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5453                 .rpc_message = &msg,
5454                 .callback_ops = &nfs4_locku_ops,
5455                 .workqueue = nfsiod_workqueue,
5456                 .flags = RPC_TASK_ASYNC,
5457         };
5458
5459         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5460                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5461
5462         /* Ensure this is an unlock - when canceling a lock, the
5463          * canceled lock is passed in, and it won't be an unlock.
5464          */
5465         fl->fl_type = F_UNLCK;
5466
5467         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5468         if (data == NULL) {
5469                 nfs_free_seqid(seqid);
5470                 return ERR_PTR(-ENOMEM);
5471         }
5472
5473         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5474         msg.rpc_argp = &data->arg;
5475         msg.rpc_resp = &data->res;
5476         task_setup_data.callback_data = data;
5477         return rpc_run_task(&task_setup_data);
5478 }
5479
5480 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5481 {
5482         struct inode *inode = state->inode;
5483         struct nfs4_state_owner *sp = state->owner;
5484         struct nfs_inode *nfsi = NFS_I(inode);
5485         struct nfs_seqid *seqid;
5486         struct nfs4_lock_state *lsp;
5487         struct rpc_task *task;
5488         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5489         int status = 0;
5490         unsigned char fl_flags = request->fl_flags;
5491
5492         status = nfs4_set_lock_state(state, request);
5493         /* Unlock _before_ we do the RPC call */
5494         request->fl_flags |= FL_EXISTS;
5495         /* Exclude nfs_delegation_claim_locks() */
5496         mutex_lock(&sp->so_delegreturn_mutex);
5497         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5498         down_read(&nfsi->rwsem);
5499         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5500                 up_read(&nfsi->rwsem);
5501                 mutex_unlock(&sp->so_delegreturn_mutex);
5502                 goto out;
5503         }
5504         up_read(&nfsi->rwsem);
5505         mutex_unlock(&sp->so_delegreturn_mutex);
5506         if (status != 0)
5507                 goto out;
5508         /* Is this a delegated lock? */
5509         lsp = request->fl_u.nfs4_fl.owner;
5510         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5511                 goto out;
5512         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5513         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5514         status = -ENOMEM;
5515         if (IS_ERR(seqid))
5516                 goto out;
5517         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5518         status = PTR_ERR(task);
5519         if (IS_ERR(task))
5520                 goto out;
5521         status = nfs4_wait_for_completion_rpc_task(task);
5522         rpc_put_task(task);
5523 out:
5524         request->fl_flags = fl_flags;
5525         trace_nfs4_unlock(request, state, F_SETLK, status);
5526         return status;
5527 }
5528
5529 struct nfs4_lockdata {
5530         struct nfs_lock_args arg;
5531         struct nfs_lock_res res;
5532         struct nfs4_lock_state *lsp;
5533         struct nfs_open_context *ctx;
5534         struct file_lock fl;
5535         unsigned long timestamp;
5536         int rpc_status;
5537         int cancelled;
5538         struct nfs_server *server;
5539 };
5540
5541 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5542                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5543                 gfp_t gfp_mask)
5544 {
5545         struct nfs4_lockdata *p;
5546         struct inode *inode = lsp->ls_state->inode;
5547         struct nfs_server *server = NFS_SERVER(inode);
5548         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5549
5550         p = kzalloc(sizeof(*p), gfp_mask);
5551         if (p == NULL)
5552                 return NULL;
5553
5554         p->arg.fh = NFS_FH(inode);
5555         p->arg.fl = &p->fl;
5556         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5557         if (IS_ERR(p->arg.open_seqid))
5558                 goto out_free;
5559         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5560         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5561         if (IS_ERR(p->arg.lock_seqid))
5562                 goto out_free_seqid;
5563         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5564         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5565         p->arg.lock_owner.s_dev = server->s_dev;
5566         p->res.lock_seqid = p->arg.lock_seqid;
5567         p->lsp = lsp;
5568         p->server = server;
5569         atomic_inc(&lsp->ls_count);
5570         p->ctx = get_nfs_open_context(ctx);
5571         memcpy(&p->fl, fl, sizeof(p->fl));
5572         return p;
5573 out_free_seqid:
5574         nfs_free_seqid(p->arg.open_seqid);
5575 out_free:
5576         kfree(p);
5577         return NULL;
5578 }
5579
5580 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5581 {
5582         struct nfs4_lockdata *data = calldata;
5583         struct nfs4_state *state = data->lsp->ls_state;
5584
5585         dprintk("%s: begin!\n", __func__);
5586         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5587                 goto out_wait;
5588         /* Do we need to do an open_to_lock_owner? */
5589         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5590                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5591                         goto out_release_lock_seqid;
5592                 }
5593                 nfs4_stateid_copy(&data->arg.open_stateid,
5594                                 &state->open_stateid);
5595                 data->arg.new_lock_owner = 1;
5596                 data->res.open_seqid = data->arg.open_seqid;
5597         } else {
5598                 data->arg.new_lock_owner = 0;
5599                 nfs4_stateid_copy(&data->arg.lock_stateid,
5600                                 &data->lsp->ls_stateid);
5601         }
5602         if (!nfs4_valid_open_stateid(state)) {
5603                 data->rpc_status = -EBADF;
5604                 task->tk_action = NULL;
5605                 goto out_release_open_seqid;
5606         }
5607         data->timestamp = jiffies;
5608         if (nfs4_setup_sequence(data->server,
5609                                 &data->arg.seq_args,
5610                                 &data->res.seq_res,
5611                                 task) == 0)
5612                 return;
5613 out_release_open_seqid:
5614         nfs_release_seqid(data->arg.open_seqid);
5615 out_release_lock_seqid:
5616         nfs_release_seqid(data->arg.lock_seqid);
5617 out_wait:
5618         nfs4_sequence_done(task, &data->res.seq_res);
5619         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5620 }
5621
5622 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5623 {
5624         struct nfs4_lockdata *data = calldata;
5625         struct nfs4_lock_state *lsp = data->lsp;
5626
5627         dprintk("%s: begin!\n", __func__);
5628
5629         if (!nfs4_sequence_done(task, &data->res.seq_res))
5630                 return;
5631
5632         data->rpc_status = task->tk_status;
5633         switch (task->tk_status) {
5634         case 0:
5635                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode),
5636                                 data->timestamp);
5637                 if (data->arg.new_lock) {
5638                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5639                         if (do_vfs_lock(data->fl.fl_file, &data->fl) < 0) {
5640                                 rpc_restart_call_prepare(task);
5641                                 break;
5642                         }
5643                 }
5644                 if (data->arg.new_lock_owner != 0) {
5645                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5646                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5647                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5648                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5649                         rpc_restart_call_prepare(task);
5650                 break;
5651         case -NFS4ERR_BAD_STATEID:
5652         case -NFS4ERR_OLD_STATEID:
5653         case -NFS4ERR_STALE_STATEID:
5654         case -NFS4ERR_EXPIRED:
5655                 if (data->arg.new_lock_owner != 0) {
5656                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5657                                                 &lsp->ls_state->open_stateid))
5658                                 rpc_restart_call_prepare(task);
5659                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5660                                                 &lsp->ls_stateid))
5661                                 rpc_restart_call_prepare(task);
5662         }
5663         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5664 }
5665
5666 static void nfs4_lock_release(void *calldata)
5667 {
5668         struct nfs4_lockdata *data = calldata;
5669
5670         dprintk("%s: begin!\n", __func__);
5671         nfs_free_seqid(data->arg.open_seqid);
5672         if (data->cancelled != 0) {
5673                 struct rpc_task *task;
5674                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5675                                 data->arg.lock_seqid);
5676                 if (!IS_ERR(task))
5677                         rpc_put_task_async(task);
5678                 dprintk("%s: cancelling lock!\n", __func__);
5679         } else
5680                 nfs_free_seqid(data->arg.lock_seqid);
5681         nfs4_put_lock_state(data->lsp);
5682         put_nfs_open_context(data->ctx);
5683         kfree(data);
5684         dprintk("%s: done!\n", __func__);
5685 }
5686
5687 static const struct rpc_call_ops nfs4_lock_ops = {
5688         .rpc_call_prepare = nfs4_lock_prepare,
5689         .rpc_call_done = nfs4_lock_done,
5690         .rpc_release = nfs4_lock_release,
5691 };
5692
5693 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5694 {
5695         switch (error) {
5696         case -NFS4ERR_ADMIN_REVOKED:
5697         case -NFS4ERR_BAD_STATEID:
5698                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5699                 if (new_lock_owner != 0 ||
5700                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5701                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5702                 break;
5703         case -NFS4ERR_STALE_STATEID:
5704                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5705         case -NFS4ERR_EXPIRED:
5706                 nfs4_schedule_lease_recovery(server->nfs_client);
5707         };
5708 }
5709
5710 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5711 {
5712         struct nfs4_lockdata *data;
5713         struct rpc_task *task;
5714         struct rpc_message msg = {
5715                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5716                 .rpc_cred = state->owner->so_cred,
5717         };
5718         struct rpc_task_setup task_setup_data = {
5719                 .rpc_client = NFS_CLIENT(state->inode),
5720                 .rpc_message = &msg,
5721                 .callback_ops = &nfs4_lock_ops,
5722                 .workqueue = nfsiod_workqueue,
5723                 .flags = RPC_TASK_ASYNC,
5724         };
5725         int ret;
5726
5727         dprintk("%s: begin!\n", __func__);
5728         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5729                         fl->fl_u.nfs4_fl.owner,
5730                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5731         if (data == NULL)
5732                 return -ENOMEM;
5733         if (IS_SETLKW(cmd))
5734                 data->arg.block = 1;
5735         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5736         msg.rpc_argp = &data->arg;
5737         msg.rpc_resp = &data->res;
5738         task_setup_data.callback_data = data;
5739         if (recovery_type > NFS_LOCK_NEW) {
5740                 if (recovery_type == NFS_LOCK_RECLAIM)
5741                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5742                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5743         } else
5744                 data->arg.new_lock = 1;
5745         task = rpc_run_task(&task_setup_data);
5746         if (IS_ERR(task))
5747                 return PTR_ERR(task);
5748         ret = nfs4_wait_for_completion_rpc_task(task);
5749         if (ret == 0) {
5750                 ret = data->rpc_status;
5751                 if (ret)
5752                         nfs4_handle_setlk_error(data->server, data->lsp,
5753                                         data->arg.new_lock_owner, ret);
5754         } else
5755                 data->cancelled = 1;
5756         rpc_put_task(task);
5757         dprintk("%s: done, ret = %d!\n", __func__, ret);
5758         return ret;
5759 }
5760
5761 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5762 {
5763         struct nfs_server *server = NFS_SERVER(state->inode);
5764         struct nfs4_exception exception = {
5765                 .inode = state->inode,
5766         };
5767         int err;
5768
5769         do {
5770                 /* Cache the lock if possible... */
5771                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5772                         return 0;
5773                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5774                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5775                 if (err != -NFS4ERR_DELAY)
5776                         break;
5777                 nfs4_handle_exception(server, err, &exception);
5778         } while (exception.retry);
5779         return err;
5780 }
5781
5782 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5783 {
5784         struct nfs_server *server = NFS_SERVER(state->inode);
5785         struct nfs4_exception exception = {
5786                 .inode = state->inode,
5787         };
5788         int err;
5789
5790         err = nfs4_set_lock_state(state, request);
5791         if (err != 0)
5792                 return err;
5793         if (!recover_lost_locks) {
5794                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5795                 return 0;
5796         }
5797         do {
5798                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5799                         return 0;
5800                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5801                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5802                 switch (err) {
5803                 default:
5804                         goto out;
5805                 case -NFS4ERR_GRACE:
5806                 case -NFS4ERR_DELAY:
5807                         nfs4_handle_exception(server, err, &exception);
5808                         err = 0;
5809                 }
5810         } while (exception.retry);
5811 out:
5812         return err;
5813 }
5814
5815 #if defined(CONFIG_NFS_V4_1)
5816 /**
5817  * nfs41_check_expired_locks - possibly free a lock stateid
5818  *
5819  * @state: NFSv4 state for an inode
5820  *
5821  * Returns NFS_OK if recovery for this stateid is now finished.
5822  * Otherwise a negative NFS4ERR value is returned.
5823  */
5824 static int nfs41_check_expired_locks(struct nfs4_state *state)
5825 {
5826         int status, ret = -NFS4ERR_BAD_STATEID;
5827         struct nfs4_lock_state *lsp;
5828         struct nfs_server *server = NFS_SERVER(state->inode);
5829
5830         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5831                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5832                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5833
5834                         status = nfs41_test_stateid(server,
5835                                         &lsp->ls_stateid,
5836                                         cred);
5837                         trace_nfs4_test_lock_stateid(state, lsp, status);
5838                         if (status != NFS_OK) {
5839                                 /* Free the stateid unless the server
5840                                  * informs us the stateid is unrecognized. */
5841                                 if (status != -NFS4ERR_BAD_STATEID)
5842                                         nfs41_free_stateid(server,
5843                                                         &lsp->ls_stateid,
5844                                                         cred);
5845                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5846                                 ret = status;
5847                         }
5848                 }
5849         };
5850
5851         return ret;
5852 }
5853
5854 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5855 {
5856         int status = NFS_OK;
5857
5858         if (test_bit(LK_STATE_IN_USE, &state->flags))
5859                 status = nfs41_check_expired_locks(state);
5860         if (status != NFS_OK)
5861                 status = nfs4_lock_expired(state, request);
5862         return status;
5863 }
5864 #endif
5865
5866 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5867 {
5868         struct nfs_inode *nfsi = NFS_I(state->inode);
5869         unsigned char fl_flags = request->fl_flags;
5870         int status = -ENOLCK;
5871
5872         if ((fl_flags & FL_POSIX) &&
5873                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5874                 goto out;
5875         /* Is this a delegated open? */
5876         status = nfs4_set_lock_state(state, request);
5877         if (status != 0)
5878                 goto out;
5879         request->fl_flags |= FL_ACCESS;
5880         status = do_vfs_lock(request->fl_file, request);
5881         if (status < 0)
5882                 goto out;
5883         down_read(&nfsi->rwsem);
5884         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5885                 /* Yes: cache locks! */
5886                 /* ...but avoid races with delegation recall... */
5887                 request->fl_flags = fl_flags & ~FL_SLEEP;
5888                 status = do_vfs_lock(request->fl_file, request);
5889                 up_read(&nfsi->rwsem);
5890                 goto out;
5891         }
5892         up_read(&nfsi->rwsem);
5893         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5894 out:
5895         request->fl_flags = fl_flags;
5896         return status;
5897 }
5898
5899 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5900 {
5901         struct nfs4_exception exception = {
5902                 .state = state,
5903                 .inode = state->inode,
5904         };
5905         int err;
5906
5907         do {
5908                 err = _nfs4_proc_setlk(state, cmd, request);
5909                 trace_nfs4_set_lock(request, state, cmd, err);
5910                 if (err == -NFS4ERR_DENIED)
5911                         err = -EAGAIN;
5912                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5913                                 err, &exception);
5914         } while (exception.retry);
5915         return err;
5916 }
5917
5918 static int
5919 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5920 {
5921         struct nfs_open_context *ctx;
5922         struct nfs4_state *state;
5923         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5924         int status;
5925
5926         /* verify open state */
5927         ctx = nfs_file_open_context(filp);
5928         state = ctx->state;
5929
5930         if (request->fl_start < 0 || request->fl_end < 0)
5931                 return -EINVAL;
5932
5933         if (IS_GETLK(cmd)) {
5934                 if (state != NULL)
5935                         return nfs4_proc_getlk(state, F_GETLK, request);
5936                 return 0;
5937         }
5938
5939         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5940                 return -EINVAL;
5941
5942         if (request->fl_type == F_UNLCK) {
5943                 if (state != NULL)
5944                         return nfs4_proc_unlck(state, cmd, request);
5945                 return 0;
5946         }
5947
5948         if (state == NULL)
5949                 return -ENOLCK;
5950         /*
5951          * Don't rely on the VFS having checked the file open mode,
5952          * since it won't do this for flock() locks.
5953          */
5954         switch (request->fl_type) {
5955         case F_RDLCK:
5956                 if (!(filp->f_mode & FMODE_READ))
5957                         return -EBADF;
5958                 break;
5959         case F_WRLCK:
5960                 if (!(filp->f_mode & FMODE_WRITE))
5961                         return -EBADF;
5962         }
5963
5964         do {
5965                 status = nfs4_proc_setlk(state, cmd, request);
5966                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5967                         break;
5968                 timeout = nfs4_set_lock_task_retry(timeout);
5969                 status = -ERESTARTSYS;
5970                 if (signalled())
5971                         break;
5972         } while(status < 0);
5973         return status;
5974 }
5975
5976 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5977 {
5978         struct nfs_server *server = NFS_SERVER(state->inode);
5979         int err;
5980
5981         err = nfs4_set_lock_state(state, fl);
5982         if (err != 0)
5983                 return err;
5984         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5985         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5986 }
5987
5988 struct nfs_release_lockowner_data {
5989         struct nfs4_lock_state *lsp;
5990         struct nfs_server *server;
5991         struct nfs_release_lockowner_args args;
5992         struct nfs_release_lockowner_res res;
5993         unsigned long timestamp;
5994 };
5995
5996 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
5997 {
5998         struct nfs_release_lockowner_data *data = calldata;
5999         struct nfs_server *server = data->server;
6000         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6001                              &data->args.seq_args, &data->res.seq_res, task);
6002         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6003         data->timestamp = jiffies;
6004 }
6005
6006 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6007 {
6008         struct nfs_release_lockowner_data *data = calldata;
6009         struct nfs_server *server = data->server;
6010
6011         nfs40_sequence_done(task, &data->res.seq_res);
6012
6013         switch (task->tk_status) {
6014         case 0:
6015                 renew_lease(server, data->timestamp);
6016                 break;
6017         case -NFS4ERR_STALE_CLIENTID:
6018         case -NFS4ERR_EXPIRED:
6019                 nfs4_schedule_lease_recovery(server->nfs_client);
6020                 break;
6021         case -NFS4ERR_LEASE_MOVED:
6022         case -NFS4ERR_DELAY:
6023                 if (nfs4_async_handle_error(task, server,
6024                                             NULL, NULL) == -EAGAIN)
6025                         rpc_restart_call_prepare(task);
6026         }
6027 }
6028
6029 static void nfs4_release_lockowner_release(void *calldata)
6030 {
6031         struct nfs_release_lockowner_data *data = calldata;
6032         nfs4_free_lock_state(data->server, data->lsp);
6033         kfree(calldata);
6034 }
6035
6036 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6037         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6038         .rpc_call_done = nfs4_release_lockowner_done,
6039         .rpc_release = nfs4_release_lockowner_release,
6040 };
6041
6042 static void
6043 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6044 {
6045         struct nfs_release_lockowner_data *data;
6046         struct rpc_message msg = {
6047                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6048         };
6049
6050         if (server->nfs_client->cl_mvops->minor_version != 0)
6051                 return;
6052
6053         data = kmalloc(sizeof(*data), GFP_NOFS);
6054         if (!data)
6055                 return;
6056         data->lsp = lsp;
6057         data->server = server;
6058         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6059         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6060         data->args.lock_owner.s_dev = server->s_dev;
6061
6062         msg.rpc_argp = &data->args;
6063         msg.rpc_resp = &data->res;
6064         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6065         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6066 }
6067
6068 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6069
6070 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6071                                    const void *buf, size_t buflen,
6072                                    int flags, int type)
6073 {
6074         if (strcmp(key, "") != 0)
6075                 return -EINVAL;
6076
6077         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
6078 }
6079
6080 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6081                                    void *buf, size_t buflen, int type)
6082 {
6083         if (strcmp(key, "") != 0)
6084                 return -EINVAL;
6085
6086         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
6087 }
6088
6089 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6090                                        size_t list_len, const char *name,
6091                                        size_t name_len, int type)
6092 {
6093         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6094
6095         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
6096                 return 0;
6097
6098         if (list && len <= list_len)
6099                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6100         return len;
6101 }
6102
6103 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6104 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6105 {
6106         return server->caps & NFS_CAP_SECURITY_LABEL;
6107 }
6108
6109 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6110                                    const void *buf, size_t buflen,
6111                                    int flags, int type)
6112 {
6113         if (security_ismaclabel(key))
6114                 return nfs4_set_security_label(dentry, buf, buflen);
6115
6116         return -EOPNOTSUPP;
6117 }
6118
6119 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6120                                    void *buf, size_t buflen, int type)
6121 {
6122         if (security_ismaclabel(key))
6123                 return nfs4_get_security_label(dentry->d_inode, buf, buflen);
6124         return -EOPNOTSUPP;
6125 }
6126
6127 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6128                                        size_t list_len, const char *name,
6129                                        size_t name_len, int type)
6130 {
6131         size_t len = 0;
6132
6133         if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
6134                 len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
6135                 if (list && len <= list_len)
6136                         security_inode_listsecurity(dentry->d_inode, list, len);
6137         }
6138         return len;
6139 }
6140
6141 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6142         .prefix = XATTR_SECURITY_PREFIX,
6143         .list   = nfs4_xattr_list_nfs4_label,
6144         .get    = nfs4_xattr_get_nfs4_label,
6145         .set    = nfs4_xattr_set_nfs4_label,
6146 };
6147 #endif
6148
6149
6150 /*
6151  * nfs_fhget will use either the mounted_on_fileid or the fileid
6152  */
6153 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6154 {
6155         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6156                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6157               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6158               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6159                 return;
6160
6161         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6162                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6163         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6164         fattr->nlink = 2;
6165 }
6166
6167 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6168                                    const struct qstr *name,
6169                                    struct nfs4_fs_locations *fs_locations,
6170                                    struct page *page)
6171 {
6172         struct nfs_server *server = NFS_SERVER(dir);
6173         u32 bitmask[3] = {
6174                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6175         };
6176         struct nfs4_fs_locations_arg args = {
6177                 .dir_fh = NFS_FH(dir),
6178                 .name = name,
6179                 .page = page,
6180                 .bitmask = bitmask,
6181         };
6182         struct nfs4_fs_locations_res res = {
6183                 .fs_locations = fs_locations,
6184         };
6185         struct rpc_message msg = {
6186                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6187                 .rpc_argp = &args,
6188                 .rpc_resp = &res,
6189         };
6190         int status;
6191
6192         dprintk("%s: start\n", __func__);
6193
6194         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6195          * is not supported */
6196         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6197                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6198         else
6199                 bitmask[0] |= FATTR4_WORD0_FILEID;
6200
6201         nfs_fattr_init(&fs_locations->fattr);
6202         fs_locations->server = server;
6203         fs_locations->nlocations = 0;
6204         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6205         dprintk("%s: returned status = %d\n", __func__, status);
6206         return status;
6207 }
6208
6209 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6210                            const struct qstr *name,
6211                            struct nfs4_fs_locations *fs_locations,
6212                            struct page *page)
6213 {
6214         struct nfs4_exception exception = { };
6215         int err;
6216         do {
6217                 err = _nfs4_proc_fs_locations(client, dir, name,
6218                                 fs_locations, page);
6219                 trace_nfs4_get_fs_locations(dir, name, err);
6220                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6221                                 &exception);
6222         } while (exception.retry);
6223         return err;
6224 }
6225
6226 /*
6227  * This operation also signals the server that this client is
6228  * performing migration recovery.  The server can stop returning
6229  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6230  * appended to this compound to identify the client ID which is
6231  * performing recovery.
6232  */
6233 static int _nfs40_proc_get_locations(struct inode *inode,
6234                                      struct nfs4_fs_locations *locations,
6235                                      struct page *page, struct rpc_cred *cred)
6236 {
6237         struct nfs_server *server = NFS_SERVER(inode);
6238         struct rpc_clnt *clnt = server->client;
6239         u32 bitmask[2] = {
6240                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6241         };
6242         struct nfs4_fs_locations_arg args = {
6243                 .clientid       = server->nfs_client->cl_clientid,
6244                 .fh             = NFS_FH(inode),
6245                 .page           = page,
6246                 .bitmask        = bitmask,
6247                 .migration      = 1,            /* skip LOOKUP */
6248                 .renew          = 1,            /* append RENEW */
6249         };
6250         struct nfs4_fs_locations_res res = {
6251                 .fs_locations   = locations,
6252                 .migration      = 1,
6253                 .renew          = 1,
6254         };
6255         struct rpc_message msg = {
6256                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6257                 .rpc_argp       = &args,
6258                 .rpc_resp       = &res,
6259                 .rpc_cred       = cred,
6260         };
6261         unsigned long now = jiffies;
6262         int status;
6263
6264         nfs_fattr_init(&locations->fattr);
6265         locations->server = server;
6266         locations->nlocations = 0;
6267
6268         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6269         nfs4_set_sequence_privileged(&args.seq_args);
6270         status = nfs4_call_sync_sequence(clnt, server, &msg,
6271                                         &args.seq_args, &res.seq_res);
6272         if (status)
6273                 return status;
6274
6275         renew_lease(server, now);
6276         return 0;
6277 }
6278
6279 #ifdef CONFIG_NFS_V4_1
6280
6281 /*
6282  * This operation also signals the server that this client is
6283  * performing migration recovery.  The server can stop asserting
6284  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6285  * performing this operation is identified in the SEQUENCE
6286  * operation in this compound.
6287  *
6288  * When the client supports GETATTR(fs_locations_info), it can
6289  * be plumbed in here.
6290  */
6291 static int _nfs41_proc_get_locations(struct inode *inode,
6292                                      struct nfs4_fs_locations *locations,
6293                                      struct page *page, struct rpc_cred *cred)
6294 {
6295         struct nfs_server *server = NFS_SERVER(inode);
6296         struct rpc_clnt *clnt = server->client;
6297         u32 bitmask[2] = {
6298                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6299         };
6300         struct nfs4_fs_locations_arg args = {
6301                 .fh             = NFS_FH(inode),
6302                 .page           = page,
6303                 .bitmask        = bitmask,
6304                 .migration      = 1,            /* skip LOOKUP */
6305         };
6306         struct nfs4_fs_locations_res res = {
6307                 .fs_locations   = locations,
6308                 .migration      = 1,
6309         };
6310         struct rpc_message msg = {
6311                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6312                 .rpc_argp       = &args,
6313                 .rpc_resp       = &res,
6314                 .rpc_cred       = cred,
6315         };
6316         int status;
6317
6318         nfs_fattr_init(&locations->fattr);
6319         locations->server = server;
6320         locations->nlocations = 0;
6321
6322         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6323         nfs4_set_sequence_privileged(&args.seq_args);
6324         status = nfs4_call_sync_sequence(clnt, server, &msg,
6325                                         &args.seq_args, &res.seq_res);
6326         if (status == NFS4_OK &&
6327             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6328                 status = -NFS4ERR_LEASE_MOVED;
6329         return status;
6330 }
6331
6332 #endif  /* CONFIG_NFS_V4_1 */
6333
6334 /**
6335  * nfs4_proc_get_locations - discover locations for a migrated FSID
6336  * @inode: inode on FSID that is migrating
6337  * @locations: result of query
6338  * @page: buffer
6339  * @cred: credential to use for this operation
6340  *
6341  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6342  * operation failed, or a negative errno if a local error occurred.
6343  *
6344  * On success, "locations" is filled in, but if the server has
6345  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6346  * asserted.
6347  *
6348  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6349  * from this client that require migration recovery.
6350  */
6351 int nfs4_proc_get_locations(struct inode *inode,
6352                             struct nfs4_fs_locations *locations,
6353                             struct page *page, struct rpc_cred *cred)
6354 {
6355         struct nfs_server *server = NFS_SERVER(inode);
6356         struct nfs_client *clp = server->nfs_client;
6357         const struct nfs4_mig_recovery_ops *ops =
6358                                         clp->cl_mvops->mig_recovery_ops;
6359         struct nfs4_exception exception = { };
6360         int status;
6361
6362         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6363                 (unsigned long long)server->fsid.major,
6364                 (unsigned long long)server->fsid.minor,
6365                 clp->cl_hostname);
6366         nfs_display_fhandle(NFS_FH(inode), __func__);
6367
6368         do {
6369                 status = ops->get_locations(inode, locations, page, cred);
6370                 if (status != -NFS4ERR_DELAY)
6371                         break;
6372                 nfs4_handle_exception(server, status, &exception);
6373         } while (exception.retry);
6374         return status;
6375 }
6376
6377 /*
6378  * This operation also signals the server that this client is
6379  * performing "lease moved" recovery.  The server can stop
6380  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6381  * is appended to this compound to identify the client ID which is
6382  * performing recovery.
6383  */
6384 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6385 {
6386         struct nfs_server *server = NFS_SERVER(inode);
6387         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6388         struct rpc_clnt *clnt = server->client;
6389         struct nfs4_fsid_present_arg args = {
6390                 .fh             = NFS_FH(inode),
6391                 .clientid       = clp->cl_clientid,
6392                 .renew          = 1,            /* append RENEW */
6393         };
6394         struct nfs4_fsid_present_res res = {
6395                 .renew          = 1,
6396         };
6397         struct rpc_message msg = {
6398                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6399                 .rpc_argp       = &args,
6400                 .rpc_resp       = &res,
6401                 .rpc_cred       = cred,
6402         };
6403         unsigned long now = jiffies;
6404         int status;
6405
6406         res.fh = nfs_alloc_fhandle();
6407         if (res.fh == NULL)
6408                 return -ENOMEM;
6409
6410         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6411         nfs4_set_sequence_privileged(&args.seq_args);
6412         status = nfs4_call_sync_sequence(clnt, server, &msg,
6413                                                 &args.seq_args, &res.seq_res);
6414         nfs_free_fhandle(res.fh);
6415         if (status)
6416                 return status;
6417
6418         do_renew_lease(clp, now);
6419         return 0;
6420 }
6421
6422 #ifdef CONFIG_NFS_V4_1
6423
6424 /*
6425  * This operation also signals the server that this client is
6426  * performing "lease moved" recovery.  The server can stop asserting
6427  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6428  * this operation is identified in the SEQUENCE operation in this
6429  * compound.
6430  */
6431 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6432 {
6433         struct nfs_server *server = NFS_SERVER(inode);
6434         struct rpc_clnt *clnt = server->client;
6435         struct nfs4_fsid_present_arg args = {
6436                 .fh             = NFS_FH(inode),
6437         };
6438         struct nfs4_fsid_present_res res = {
6439         };
6440         struct rpc_message msg = {
6441                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6442                 .rpc_argp       = &args,
6443                 .rpc_resp       = &res,
6444                 .rpc_cred       = cred,
6445         };
6446         int status;
6447
6448         res.fh = nfs_alloc_fhandle();
6449         if (res.fh == NULL)
6450                 return -ENOMEM;
6451
6452         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6453         nfs4_set_sequence_privileged(&args.seq_args);
6454         status = nfs4_call_sync_sequence(clnt, server, &msg,
6455                                                 &args.seq_args, &res.seq_res);
6456         nfs_free_fhandle(res.fh);
6457         if (status == NFS4_OK &&
6458             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6459                 status = -NFS4ERR_LEASE_MOVED;
6460         return status;
6461 }
6462
6463 #endif  /* CONFIG_NFS_V4_1 */
6464
6465 /**
6466  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6467  * @inode: inode on FSID to check
6468  * @cred: credential to use for this operation
6469  *
6470  * Server indicates whether the FSID is present, moved, or not
6471  * recognized.  This operation is necessary to clear a LEASE_MOVED
6472  * condition for this client ID.
6473  *
6474  * Returns NFS4_OK if the FSID is present on this server,
6475  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6476  *  NFS4ERR code if some error occurred on the server, or a
6477  *  negative errno if a local failure occurred.
6478  */
6479 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6480 {
6481         struct nfs_server *server = NFS_SERVER(inode);
6482         struct nfs_client *clp = server->nfs_client;
6483         const struct nfs4_mig_recovery_ops *ops =
6484                                         clp->cl_mvops->mig_recovery_ops;
6485         struct nfs4_exception exception = { };
6486         int status;
6487
6488         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6489                 (unsigned long long)server->fsid.major,
6490                 (unsigned long long)server->fsid.minor,
6491                 clp->cl_hostname);
6492         nfs_display_fhandle(NFS_FH(inode), __func__);
6493
6494         do {
6495                 status = ops->fsid_present(inode, cred);
6496                 if (status != -NFS4ERR_DELAY)
6497                         break;
6498                 nfs4_handle_exception(server, status, &exception);
6499         } while (exception.retry);
6500         return status;
6501 }
6502
6503 /**
6504  * If 'use_integrity' is true and the state managment nfs_client
6505  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6506  * and the machine credential as per RFC3530bis and RFC5661 Security
6507  * Considerations sections. Otherwise, just use the user cred with the
6508  * filesystem's rpc_client.
6509  */
6510 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6511 {
6512         int status;
6513         struct nfs4_secinfo_arg args = {
6514                 .dir_fh = NFS_FH(dir),
6515                 .name   = name,
6516         };
6517         struct nfs4_secinfo_res res = {
6518                 .flavors     = flavors,
6519         };
6520         struct rpc_message msg = {
6521                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6522                 .rpc_argp = &args,
6523                 .rpc_resp = &res,
6524         };
6525         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6526         struct rpc_cred *cred = NULL;
6527
6528         if (use_integrity) {
6529                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6530                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6531                 msg.rpc_cred = cred;
6532         }
6533
6534         dprintk("NFS call  secinfo %s\n", name->name);
6535
6536         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6537                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6538
6539         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6540                                 &res.seq_res, 0);
6541         dprintk("NFS reply  secinfo: %d\n", status);
6542
6543         if (cred)
6544                 put_rpccred(cred);
6545
6546         return status;
6547 }
6548
6549 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6550                       struct nfs4_secinfo_flavors *flavors)
6551 {
6552         struct nfs4_exception exception = { };
6553         int err;
6554         do {
6555                 err = -NFS4ERR_WRONGSEC;
6556
6557                 /* try to use integrity protection with machine cred */
6558                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6559                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6560
6561                 /*
6562                  * if unable to use integrity protection, or SECINFO with
6563                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6564                  * disallowed by spec, but exists in deployed servers) use
6565                  * the current filesystem's rpc_client and the user cred.
6566                  */
6567                 if (err == -NFS4ERR_WRONGSEC)
6568                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6569
6570                 trace_nfs4_secinfo(dir, name, err);
6571                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6572                                 &exception);
6573         } while (exception.retry);
6574         return err;
6575 }
6576
6577 #ifdef CONFIG_NFS_V4_1
6578 /*
6579  * Check the exchange flags returned by the server for invalid flags, having
6580  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6581  * DS flags set.
6582  */
6583 static int nfs4_check_cl_exchange_flags(u32 flags)
6584 {
6585         if (flags & ~EXCHGID4_FLAG_MASK_R)
6586                 goto out_inval;
6587         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6588             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6589                 goto out_inval;
6590         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6591                 goto out_inval;
6592         return NFS_OK;
6593 out_inval:
6594         return -NFS4ERR_INVAL;
6595 }
6596
6597 static bool
6598 nfs41_same_server_scope(struct nfs41_server_scope *a,
6599                         struct nfs41_server_scope *b)
6600 {
6601         if (a->server_scope_sz == b->server_scope_sz &&
6602             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6603                 return true;
6604
6605         return false;
6606 }
6607
6608 /*
6609  * nfs4_proc_bind_conn_to_session()
6610  *
6611  * The 4.1 client currently uses the same TCP connection for the
6612  * fore and backchannel.
6613  */
6614 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6615 {
6616         int status;
6617         struct nfs41_bind_conn_to_session_res res;
6618         struct rpc_message msg = {
6619                 .rpc_proc =
6620                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6621                 .rpc_argp = clp,
6622                 .rpc_resp = &res,
6623                 .rpc_cred = cred,
6624         };
6625
6626         dprintk("--> %s\n", __func__);
6627
6628         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
6629         if (unlikely(res.session == NULL)) {
6630                 status = -ENOMEM;
6631                 goto out;
6632         }
6633
6634         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6635         trace_nfs4_bind_conn_to_session(clp, status);
6636         if (status == 0) {
6637                 if (memcmp(res.session->sess_id.data,
6638                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6639                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6640                         status = -EIO;
6641                         goto out_session;
6642                 }
6643                 if (res.dir != NFS4_CDFS4_BOTH) {
6644                         dprintk("NFS: %s: Unexpected direction from server\n",
6645                                 __func__);
6646                         status = -EIO;
6647                         goto out_session;
6648                 }
6649                 if (res.use_conn_in_rdma_mode) {
6650                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6651                                 __func__);
6652                         status = -EIO;
6653                         goto out_session;
6654                 }
6655         }
6656 out_session:
6657         kfree(res.session);
6658 out:
6659         dprintk("<-- %s status= %d\n", __func__, status);
6660         return status;
6661 }
6662
6663 /*
6664  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6665  * and operations we'd like to see to enable certain features in the allow map
6666  */
6667 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6668         .how = SP4_MACH_CRED,
6669         .enforce.u.words = {
6670                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6671                       1 << (OP_EXCHANGE_ID - 32) |
6672                       1 << (OP_CREATE_SESSION - 32) |
6673                       1 << (OP_DESTROY_SESSION - 32) |
6674                       1 << (OP_DESTROY_CLIENTID - 32)
6675         },
6676         .allow.u.words = {
6677                 [0] = 1 << (OP_CLOSE) |
6678                       1 << (OP_LOCKU) |
6679                       1 << (OP_COMMIT),
6680                 [1] = 1 << (OP_SECINFO - 32) |
6681                       1 << (OP_SECINFO_NO_NAME - 32) |
6682                       1 << (OP_TEST_STATEID - 32) |
6683                       1 << (OP_FREE_STATEID - 32) |
6684                       1 << (OP_WRITE - 32)
6685         }
6686 };
6687
6688 /*
6689  * Select the state protection mode for client `clp' given the server results
6690  * from exchange_id in `sp'.
6691  *
6692  * Returns 0 on success, negative errno otherwise.
6693  */
6694 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6695                                  struct nfs41_state_protection *sp)
6696 {
6697         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6698                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6699                       1 << (OP_EXCHANGE_ID - 32) |
6700                       1 << (OP_CREATE_SESSION - 32) |
6701                       1 << (OP_DESTROY_SESSION - 32) |
6702                       1 << (OP_DESTROY_CLIENTID - 32)
6703         };
6704         unsigned int i;
6705
6706         if (sp->how == SP4_MACH_CRED) {
6707                 /* Print state protect result */
6708                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6709                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6710                         if (test_bit(i, sp->enforce.u.longs))
6711                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6712                         if (test_bit(i, sp->allow.u.longs))
6713                                 dfprintk(MOUNT, "  allow op %d\n", i);
6714                 }
6715
6716                 /* make sure nothing is on enforce list that isn't supported */
6717                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6718                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6719                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6720                                 return -EINVAL;
6721                         }
6722                 }
6723
6724                 /*
6725                  * Minimal mode - state operations are allowed to use machine
6726                  * credential.  Note this already happens by default, so the
6727                  * client doesn't have to do anything more than the negotiation.
6728                  *
6729                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6730                  *       we're already using the machine cred for exchange_id
6731                  *       and will never use a different cred.
6732                  */
6733                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6734                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6735                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6736                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6737                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6738                         dfprintk(MOUNT, "  minimal mode enabled\n");
6739                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6740                 } else {
6741                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6742                         return -EINVAL;
6743                 }
6744
6745                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6746                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6747                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6748                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6749                 }
6750
6751                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6752                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6753                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6754                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6755                 }
6756
6757                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6758                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6759                         dfprintk(MOUNT, "  stateid mode enabled\n");
6760                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6761                 }
6762
6763                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6764                         dfprintk(MOUNT, "  write mode enabled\n");
6765                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6766                 }
6767
6768                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6769                         dfprintk(MOUNT, "  commit mode enabled\n");
6770                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6771                 }
6772         }
6773
6774         return 0;
6775 }
6776
6777 /*
6778  * _nfs4_proc_exchange_id()
6779  *
6780  * Wrapper for EXCHANGE_ID operation.
6781  */
6782 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6783         u32 sp4_how)
6784 {
6785         nfs4_verifier verifier;
6786         struct nfs41_exchange_id_args args = {
6787                 .verifier = &verifier,
6788                 .client = clp,
6789 #ifdef CONFIG_NFS_V4_1_MIGRATION
6790                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6791                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6792                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6793 #else
6794                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6795                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6796 #endif
6797         };
6798         struct nfs41_exchange_id_res res = {
6799                 0
6800         };
6801         int status;
6802         struct rpc_message msg = {
6803                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6804                 .rpc_argp = &args,
6805                 .rpc_resp = &res,
6806                 .rpc_cred = cred,
6807         };
6808
6809         nfs4_init_boot_verifier(clp, &verifier);
6810         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6811                                                         sizeof(args.id));
6812         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6813                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6814                 args.id_len, args.id);
6815
6816         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6817                                         GFP_NOFS);
6818         if (unlikely(res.server_owner == NULL)) {
6819                 status = -ENOMEM;
6820                 goto out;
6821         }
6822
6823         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6824                                         GFP_NOFS);
6825         if (unlikely(res.server_scope == NULL)) {
6826                 status = -ENOMEM;
6827                 goto out_server_owner;
6828         }
6829
6830         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6831         if (unlikely(res.impl_id == NULL)) {
6832                 status = -ENOMEM;
6833                 goto out_server_scope;
6834         }
6835
6836         switch (sp4_how) {
6837         case SP4_NONE:
6838                 args.state_protect.how = SP4_NONE;
6839                 break;
6840
6841         case SP4_MACH_CRED:
6842                 args.state_protect = nfs4_sp4_mach_cred_request;
6843                 break;
6844
6845         default:
6846                 /* unsupported! */
6847                 WARN_ON_ONCE(1);
6848                 status = -EINVAL;
6849                 goto out_server_scope;
6850         }
6851
6852         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6853         trace_nfs4_exchange_id(clp, status);
6854         if (status == 0)
6855                 status = nfs4_check_cl_exchange_flags(res.flags);
6856
6857         if (status == 0)
6858                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
6859
6860         if (status == 0) {
6861                 clp->cl_clientid = res.clientid;
6862                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
6863                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
6864                         clp->cl_seqid = res.seqid;
6865
6866                 kfree(clp->cl_serverowner);
6867                 clp->cl_serverowner = res.server_owner;
6868                 res.server_owner = NULL;
6869
6870                 /* use the most recent implementation id */
6871                 kfree(clp->cl_implid);
6872                 clp->cl_implid = res.impl_id;
6873
6874                 if (clp->cl_serverscope != NULL &&
6875                     !nfs41_same_server_scope(clp->cl_serverscope,
6876                                              res.server_scope)) {
6877                         dprintk("%s: server_scope mismatch detected\n",
6878                                 __func__);
6879                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6880                         kfree(clp->cl_serverscope);
6881                         clp->cl_serverscope = NULL;
6882                 }
6883
6884                 if (clp->cl_serverscope == NULL) {
6885                         clp->cl_serverscope = res.server_scope;
6886                         goto out;
6887                 }
6888         } else
6889                 kfree(res.impl_id);
6890
6891 out_server_owner:
6892         kfree(res.server_owner);
6893 out_server_scope:
6894         kfree(res.server_scope);
6895 out:
6896         if (clp->cl_implid != NULL)
6897                 dprintk("NFS reply exchange_id: Server Implementation ID: "
6898                         "domain: %s, name: %s, date: %llu,%u\n",
6899                         clp->cl_implid->domain, clp->cl_implid->name,
6900                         clp->cl_implid->date.seconds,
6901                         clp->cl_implid->date.nseconds);
6902         dprintk("NFS reply exchange_id: %d\n", status);
6903         return status;
6904 }
6905
6906 /*
6907  * nfs4_proc_exchange_id()
6908  *
6909  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6910  *
6911  * Since the clientid has expired, all compounds using sessions
6912  * associated with the stale clientid will be returning
6913  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6914  * be in some phase of session reset.
6915  *
6916  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6917  */
6918 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6919 {
6920         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6921         int status;
6922
6923         /* try SP4_MACH_CRED if krb5i/p */
6924         if (authflavor == RPC_AUTH_GSS_KRB5I ||
6925             authflavor == RPC_AUTH_GSS_KRB5P) {
6926                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6927                 if (!status)
6928                         return 0;
6929         }
6930
6931         /* try SP4_NONE */
6932         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6933 }
6934
6935 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6936                 struct rpc_cred *cred)
6937 {
6938         struct rpc_message msg = {
6939                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6940                 .rpc_argp = clp,
6941                 .rpc_cred = cred,
6942         };
6943         int status;
6944
6945         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6946         trace_nfs4_destroy_clientid(clp, status);
6947         if (status)
6948                 dprintk("NFS: Got error %d from the server %s on "
6949                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
6950         return status;
6951 }
6952
6953 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6954                 struct rpc_cred *cred)
6955 {
6956         unsigned int loop;
6957         int ret;
6958
6959         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6960                 ret = _nfs4_proc_destroy_clientid(clp, cred);
6961                 switch (ret) {
6962                 case -NFS4ERR_DELAY:
6963                 case -NFS4ERR_CLIENTID_BUSY:
6964                         ssleep(1);
6965                         break;
6966                 default:
6967                         return ret;
6968                 }
6969         }
6970         return 0;
6971 }
6972
6973 int nfs4_destroy_clientid(struct nfs_client *clp)
6974 {
6975         struct rpc_cred *cred;
6976         int ret = 0;
6977
6978         if (clp->cl_mvops->minor_version < 1)
6979                 goto out;
6980         if (clp->cl_exchange_flags == 0)
6981                 goto out;
6982         if (clp->cl_preserve_clid)
6983                 goto out;
6984         cred = nfs4_get_clid_cred(clp);
6985         ret = nfs4_proc_destroy_clientid(clp, cred);
6986         if (cred)
6987                 put_rpccred(cred);
6988         switch (ret) {
6989         case 0:
6990         case -NFS4ERR_STALE_CLIENTID:
6991                 clp->cl_exchange_flags = 0;
6992         }
6993 out:
6994         return ret;
6995 }
6996
6997 struct nfs4_get_lease_time_data {
6998         struct nfs4_get_lease_time_args *args;
6999         struct nfs4_get_lease_time_res *res;
7000         struct nfs_client *clp;
7001 };
7002
7003 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7004                                         void *calldata)
7005 {
7006         struct nfs4_get_lease_time_data *data =
7007                         (struct nfs4_get_lease_time_data *)calldata;
7008
7009         dprintk("--> %s\n", __func__);
7010         /* just setup sequence, do not trigger session recovery
7011            since we're invoked within one */
7012         nfs41_setup_sequence(data->clp->cl_session,
7013                         &data->args->la_seq_args,
7014                         &data->res->lr_seq_res,
7015                         task);
7016         dprintk("<-- %s\n", __func__);
7017 }
7018
7019 /*
7020  * Called from nfs4_state_manager thread for session setup, so don't recover
7021  * from sequence operation or clientid errors.
7022  */
7023 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7024 {
7025         struct nfs4_get_lease_time_data *data =
7026                         (struct nfs4_get_lease_time_data *)calldata;
7027
7028         dprintk("--> %s\n", __func__);
7029         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7030                 return;
7031         switch (task->tk_status) {
7032         case -NFS4ERR_DELAY:
7033         case -NFS4ERR_GRACE:
7034                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7035                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7036                 task->tk_status = 0;
7037                 /* fall through */
7038         case -NFS4ERR_RETRY_UNCACHED_REP:
7039                 rpc_restart_call_prepare(task);
7040                 return;
7041         }
7042         dprintk("<-- %s\n", __func__);
7043 }
7044
7045 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7046         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7047         .rpc_call_done = nfs4_get_lease_time_done,
7048 };
7049
7050 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7051 {
7052         struct rpc_task *task;
7053         struct nfs4_get_lease_time_args args;
7054         struct nfs4_get_lease_time_res res = {
7055                 .lr_fsinfo = fsinfo,
7056         };
7057         struct nfs4_get_lease_time_data data = {
7058                 .args = &args,
7059                 .res = &res,
7060                 .clp = clp,
7061         };
7062         struct rpc_message msg = {
7063                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7064                 .rpc_argp = &args,
7065                 .rpc_resp = &res,
7066         };
7067         struct rpc_task_setup task_setup = {
7068                 .rpc_client = clp->cl_rpcclient,
7069                 .rpc_message = &msg,
7070                 .callback_ops = &nfs4_get_lease_time_ops,
7071                 .callback_data = &data,
7072                 .flags = RPC_TASK_TIMEOUT,
7073         };
7074         int status;
7075
7076         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7077         nfs4_set_sequence_privileged(&args.la_seq_args);
7078         dprintk("--> %s\n", __func__);
7079         task = rpc_run_task(&task_setup);
7080
7081         if (IS_ERR(task))
7082                 status = PTR_ERR(task);
7083         else {
7084                 status = task->tk_status;
7085                 rpc_put_task(task);
7086         }
7087         dprintk("<-- %s return %d\n", __func__, status);
7088
7089         return status;
7090 }
7091
7092 /*
7093  * Initialize the values to be used by the client in CREATE_SESSION
7094  * If nfs4_init_session set the fore channel request and response sizes,
7095  * use them.
7096  *
7097  * Set the back channel max_resp_sz_cached to zero to force the client to
7098  * always set csa_cachethis to FALSE because the current implementation
7099  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7100  */
7101 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7102 {
7103         unsigned int max_rqst_sz, max_resp_sz;
7104
7105         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7106         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7107
7108         /* Fore channel attributes */
7109         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7110         args->fc_attrs.max_resp_sz = max_resp_sz;
7111         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7112         args->fc_attrs.max_reqs = max_session_slots;
7113
7114         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7115                 "max_ops=%u max_reqs=%u\n",
7116                 __func__,
7117                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7118                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7119
7120         /* Back channel attributes */
7121         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7122         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7123         args->bc_attrs.max_resp_sz_cached = 0;
7124         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7125         args->bc_attrs.max_reqs = 1;
7126
7127         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7128                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7129                 __func__,
7130                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7131                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7132                 args->bc_attrs.max_reqs);
7133 }
7134
7135 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7136 {
7137         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7138         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
7139
7140         if (rcvd->max_resp_sz > sent->max_resp_sz)
7141                 return -EINVAL;
7142         /*
7143          * Our requested max_ops is the minimum we need; we're not
7144          * prepared to break up compounds into smaller pieces than that.
7145          * So, no point even trying to continue if the server won't
7146          * cooperate:
7147          */
7148         if (rcvd->max_ops < sent->max_ops)
7149                 return -EINVAL;
7150         if (rcvd->max_reqs == 0)
7151                 return -EINVAL;
7152         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7153                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7154         return 0;
7155 }
7156
7157 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7158 {
7159         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7160         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
7161
7162         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7163                 return -EINVAL;
7164         if (rcvd->max_resp_sz < sent->max_resp_sz)
7165                 return -EINVAL;
7166         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7167                 return -EINVAL;
7168         /* These would render the backchannel useless: */
7169         if (rcvd->max_ops != sent->max_ops)
7170                 return -EINVAL;
7171         if (rcvd->max_reqs != sent->max_reqs)
7172                 return -EINVAL;
7173         return 0;
7174 }
7175
7176 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7177                                      struct nfs4_session *session)
7178 {
7179         int ret;
7180
7181         ret = nfs4_verify_fore_channel_attrs(args, session);
7182         if (ret)
7183                 return ret;
7184         return nfs4_verify_back_channel_attrs(args, session);
7185 }
7186
7187 static int _nfs4_proc_create_session(struct nfs_client *clp,
7188                 struct rpc_cred *cred)
7189 {
7190         struct nfs4_session *session = clp->cl_session;
7191         struct nfs41_create_session_args args = {
7192                 .client = clp,
7193                 .cb_program = NFS4_CALLBACK,
7194         };
7195         struct nfs41_create_session_res res = {
7196                 .client = clp,
7197         };
7198         struct rpc_message msg = {
7199                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7200                 .rpc_argp = &args,
7201                 .rpc_resp = &res,
7202                 .rpc_cred = cred,
7203         };
7204         int status;
7205
7206         nfs4_init_channel_attrs(&args);
7207         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7208
7209         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7210         trace_nfs4_create_session(clp, status);
7211
7212         if (!status) {
7213                 /* Verify the session's negotiated channel_attrs values */
7214                 status = nfs4_verify_channel_attrs(&args, session);
7215                 /* Increment the clientid slot sequence id */
7216                 clp->cl_seqid++;
7217         }
7218
7219         return status;
7220 }
7221
7222 /*
7223  * Issues a CREATE_SESSION operation to the server.
7224  * It is the responsibility of the caller to verify the session is
7225  * expired before calling this routine.
7226  */
7227 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7228 {
7229         int status;
7230         unsigned *ptr;
7231         struct nfs4_session *session = clp->cl_session;
7232
7233         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7234
7235         status = _nfs4_proc_create_session(clp, cred);
7236         if (status)
7237                 goto out;
7238
7239         /* Init or reset the session slot tables */
7240         status = nfs4_setup_session_slot_tables(session);
7241         dprintk("slot table setup returned %d\n", status);
7242         if (status)
7243                 goto out;
7244
7245         ptr = (unsigned *)&session->sess_id.data[0];
7246         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7247                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7248 out:
7249         dprintk("<-- %s\n", __func__);
7250         return status;
7251 }
7252
7253 /*
7254  * Issue the over-the-wire RPC DESTROY_SESSION.
7255  * The caller must serialize access to this routine.
7256  */
7257 int nfs4_proc_destroy_session(struct nfs4_session *session,
7258                 struct rpc_cred *cred)
7259 {
7260         struct rpc_message msg = {
7261                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7262                 .rpc_argp = session,
7263                 .rpc_cred = cred,
7264         };
7265         int status = 0;
7266
7267         dprintk("--> nfs4_proc_destroy_session\n");
7268
7269         /* session is still being setup */
7270         if (session->clp->cl_cons_state != NFS_CS_READY)
7271                 return status;
7272
7273         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7274         trace_nfs4_destroy_session(session->clp, status);
7275
7276         if (status)
7277                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7278                         "Session has been destroyed regardless...\n", status);
7279
7280         dprintk("<-- nfs4_proc_destroy_session\n");
7281         return status;
7282 }
7283
7284 /*
7285  * Renew the cl_session lease.
7286  */
7287 struct nfs4_sequence_data {
7288         struct nfs_client *clp;
7289         struct nfs4_sequence_args args;
7290         struct nfs4_sequence_res res;
7291 };
7292
7293 static void nfs41_sequence_release(void *data)
7294 {
7295         struct nfs4_sequence_data *calldata = data;
7296         struct nfs_client *clp = calldata->clp;
7297
7298         if (atomic_read(&clp->cl_count) > 1)
7299                 nfs4_schedule_state_renewal(clp);
7300         nfs_put_client(clp);
7301         kfree(calldata);
7302 }
7303
7304 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7305 {
7306         switch(task->tk_status) {
7307         case -NFS4ERR_DELAY:
7308                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7309                 return -EAGAIN;
7310         default:
7311                 nfs4_schedule_lease_recovery(clp);
7312         }
7313         return 0;
7314 }
7315
7316 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7317 {
7318         struct nfs4_sequence_data *calldata = data;
7319         struct nfs_client *clp = calldata->clp;
7320
7321         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7322                 return;
7323
7324         trace_nfs4_sequence(clp, task->tk_status);
7325         if (task->tk_status < 0) {
7326                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7327                 if (atomic_read(&clp->cl_count) == 1)
7328                         goto out;
7329
7330                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7331                         rpc_restart_call_prepare(task);
7332                         return;
7333                 }
7334         }
7335         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7336 out:
7337         dprintk("<-- %s\n", __func__);
7338 }
7339
7340 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7341 {
7342         struct nfs4_sequence_data *calldata = data;
7343         struct nfs_client *clp = calldata->clp;
7344         struct nfs4_sequence_args *args;
7345         struct nfs4_sequence_res *res;
7346
7347         args = task->tk_msg.rpc_argp;
7348         res = task->tk_msg.rpc_resp;
7349
7350         nfs41_setup_sequence(clp->cl_session, args, res, task);
7351 }
7352
7353 static const struct rpc_call_ops nfs41_sequence_ops = {
7354         .rpc_call_done = nfs41_sequence_call_done,
7355         .rpc_call_prepare = nfs41_sequence_prepare,
7356         .rpc_release = nfs41_sequence_release,
7357 };
7358
7359 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7360                 struct rpc_cred *cred,
7361                 bool is_privileged)
7362 {
7363         struct nfs4_sequence_data *calldata;
7364         struct rpc_message msg = {
7365                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7366                 .rpc_cred = cred,
7367         };
7368         struct rpc_task_setup task_setup_data = {
7369                 .rpc_client = clp->cl_rpcclient,
7370                 .rpc_message = &msg,
7371                 .callback_ops = &nfs41_sequence_ops,
7372                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7373         };
7374
7375         if (!atomic_inc_not_zero(&clp->cl_count))
7376                 return ERR_PTR(-EIO);
7377         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7378         if (calldata == NULL) {
7379                 nfs_put_client(clp);
7380                 return ERR_PTR(-ENOMEM);
7381         }
7382         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7383         if (is_privileged)
7384                 nfs4_set_sequence_privileged(&calldata->args);
7385         msg.rpc_argp = &calldata->args;
7386         msg.rpc_resp = &calldata->res;
7387         calldata->clp = clp;
7388         task_setup_data.callback_data = calldata;
7389
7390         return rpc_run_task(&task_setup_data);
7391 }
7392
7393 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7394 {
7395         struct rpc_task *task;
7396         int ret = 0;
7397
7398         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7399                 return -EAGAIN;
7400         task = _nfs41_proc_sequence(clp, cred, false);
7401         if (IS_ERR(task))
7402                 ret = PTR_ERR(task);
7403         else
7404                 rpc_put_task_async(task);
7405         dprintk("<-- %s status=%d\n", __func__, ret);
7406         return ret;
7407 }
7408
7409 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7410 {
7411         struct rpc_task *task;
7412         int ret;
7413
7414         task = _nfs41_proc_sequence(clp, cred, true);
7415         if (IS_ERR(task)) {
7416                 ret = PTR_ERR(task);
7417                 goto out;
7418         }
7419         ret = rpc_wait_for_completion_task(task);
7420         if (!ret) {
7421                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7422
7423                 if (task->tk_status == 0)
7424                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7425                 ret = task->tk_status;
7426         }
7427         rpc_put_task(task);
7428 out:
7429         dprintk("<-- %s status=%d\n", __func__, ret);
7430         return ret;
7431 }
7432
7433 struct nfs4_reclaim_complete_data {
7434         struct nfs_client *clp;
7435         struct nfs41_reclaim_complete_args arg;
7436         struct nfs41_reclaim_complete_res res;
7437 };
7438
7439 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7440 {
7441         struct nfs4_reclaim_complete_data *calldata = data;
7442
7443         nfs41_setup_sequence(calldata->clp->cl_session,
7444                         &calldata->arg.seq_args,
7445                         &calldata->res.seq_res,
7446                         task);
7447 }
7448
7449 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7450 {
7451         switch(task->tk_status) {
7452         case 0:
7453         case -NFS4ERR_COMPLETE_ALREADY:
7454         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7455                 break;
7456         case -NFS4ERR_DELAY:
7457                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7458                 /* fall through */
7459         case -NFS4ERR_RETRY_UNCACHED_REP:
7460                 return -EAGAIN;
7461         default:
7462                 nfs4_schedule_lease_recovery(clp);
7463         }
7464         return 0;
7465 }
7466
7467 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7468 {
7469         struct nfs4_reclaim_complete_data *calldata = data;
7470         struct nfs_client *clp = calldata->clp;
7471         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7472
7473         dprintk("--> %s\n", __func__);
7474         if (!nfs41_sequence_done(task, res))
7475                 return;
7476
7477         trace_nfs4_reclaim_complete(clp, task->tk_status);
7478         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7479                 rpc_restart_call_prepare(task);
7480                 return;
7481         }
7482         dprintk("<-- %s\n", __func__);
7483 }
7484
7485 static void nfs4_free_reclaim_complete_data(void *data)
7486 {
7487         struct nfs4_reclaim_complete_data *calldata = data;
7488
7489         kfree(calldata);
7490 }
7491
7492 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7493         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7494         .rpc_call_done = nfs4_reclaim_complete_done,
7495         .rpc_release = nfs4_free_reclaim_complete_data,
7496 };
7497
7498 /*
7499  * Issue a global reclaim complete.
7500  */
7501 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7502                 struct rpc_cred *cred)
7503 {
7504         struct nfs4_reclaim_complete_data *calldata;
7505         struct rpc_task *task;
7506         struct rpc_message msg = {
7507                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7508                 .rpc_cred = cred,
7509         };
7510         struct rpc_task_setup task_setup_data = {
7511                 .rpc_client = clp->cl_rpcclient,
7512                 .rpc_message = &msg,
7513                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7514                 .flags = RPC_TASK_ASYNC,
7515         };
7516         int status = -ENOMEM;
7517
7518         dprintk("--> %s\n", __func__);
7519         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7520         if (calldata == NULL)
7521                 goto out;
7522         calldata->clp = clp;
7523         calldata->arg.one_fs = 0;
7524
7525         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7526         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7527         msg.rpc_argp = &calldata->arg;
7528         msg.rpc_resp = &calldata->res;
7529         task_setup_data.callback_data = calldata;
7530         task = rpc_run_task(&task_setup_data);
7531         if (IS_ERR(task)) {
7532                 status = PTR_ERR(task);
7533                 goto out;
7534         }
7535         status = nfs4_wait_for_completion_rpc_task(task);
7536         if (status == 0)
7537                 status = task->tk_status;
7538         rpc_put_task(task);
7539         return 0;
7540 out:
7541         dprintk("<-- %s status=%d\n", __func__, status);
7542         return status;
7543 }
7544
7545 static void
7546 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7547 {
7548         struct nfs4_layoutget *lgp = calldata;
7549         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7550         struct nfs4_session *session = nfs4_get_session(server);
7551
7552         dprintk("--> %s\n", __func__);
7553         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7554          * right now covering the LAYOUTGET we are about to send.
7555          * However, that is not so catastrophic, and there seems
7556          * to be no way to prevent it completely.
7557          */
7558         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7559                                 &lgp->res.seq_res, task))
7560                 return;
7561         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7562                                           NFS_I(lgp->args.inode)->layout,
7563                                           &lgp->args.range,
7564                                           lgp->args.ctx->state)) {
7565                 rpc_exit(task, NFS4_OK);
7566         }
7567 }
7568
7569 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7570 {
7571         struct nfs4_layoutget *lgp = calldata;
7572         struct inode *inode = lgp->args.inode;
7573         struct nfs_server *server = NFS_SERVER(inode);
7574         struct pnfs_layout_hdr *lo;
7575         struct nfs4_state *state = NULL;
7576         unsigned long timeo, now, giveup;
7577
7578         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7579
7580         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7581                 goto out;
7582
7583         switch (task->tk_status) {
7584         case 0:
7585                 goto out;
7586         /*
7587          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7588          * (or clients) writing to the same RAID stripe
7589          */
7590         case -NFS4ERR_LAYOUTTRYLATER:
7591         /*
7592          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7593          * existing layout before getting a new one).
7594          */
7595         case -NFS4ERR_RECALLCONFLICT:
7596                 timeo = rpc_get_timeout(task->tk_client);
7597                 giveup = lgp->args.timestamp + timeo;
7598                 now = jiffies;
7599                 if (time_after(giveup, now)) {
7600                         unsigned long delay;
7601
7602                         /* Delay for:
7603                          * - Not less then NFS4_POLL_RETRY_MIN.
7604                          * - One last time a jiffie before we give up
7605                          * - exponential backoff (time_now minus start_attempt)
7606                          */
7607                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7608                                     min((giveup - now - 1),
7609                                         now - lgp->args.timestamp));
7610
7611                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7612                                 __func__, delay);
7613                         rpc_delay(task, delay);
7614                         task->tk_status = 0;
7615                         rpc_restart_call_prepare(task);
7616                         goto out; /* Do not call nfs4_async_handle_error() */
7617                 }
7618                 break;
7619         case -NFS4ERR_EXPIRED:
7620         case -NFS4ERR_BAD_STATEID:
7621                 spin_lock(&inode->i_lock);
7622                 lo = NFS_I(inode)->layout;
7623                 if (!lo || list_empty(&lo->plh_segs)) {
7624                         spin_unlock(&inode->i_lock);
7625                         /* If the open stateid was bad, then recover it. */
7626                         state = lgp->args.ctx->state;
7627                 } else {
7628                         LIST_HEAD(head);
7629
7630                         /*
7631                          * Mark the bad layout state as invalid, then retry
7632                          * with the current stateid.
7633                          */
7634                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7635                         spin_unlock(&inode->i_lock);
7636                         pnfs_free_lseg_list(&head);
7637
7638                         task->tk_status = 0;
7639                         rpc_restart_call_prepare(task);
7640                 }
7641         }
7642         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7643                 rpc_restart_call_prepare(task);
7644 out:
7645         dprintk("<-- %s\n", __func__);
7646 }
7647
7648 static size_t max_response_pages(struct nfs_server *server)
7649 {
7650         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7651         return nfs_page_array_len(0, max_resp_sz);
7652 }
7653
7654 static void nfs4_free_pages(struct page **pages, size_t size)
7655 {
7656         int i;
7657
7658         if (!pages)
7659                 return;
7660
7661         for (i = 0; i < size; i++) {
7662                 if (!pages[i])
7663                         break;
7664                 __free_page(pages[i]);
7665         }
7666         kfree(pages);
7667 }
7668
7669 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7670 {
7671         struct page **pages;
7672         int i;
7673
7674         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7675         if (!pages) {
7676                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7677                 return NULL;
7678         }
7679
7680         for (i = 0; i < size; i++) {
7681                 pages[i] = alloc_page(gfp_flags);
7682                 if (!pages[i]) {
7683                         dprintk("%s: failed to allocate page\n", __func__);
7684                         nfs4_free_pages(pages, size);
7685                         return NULL;
7686                 }
7687         }
7688
7689         return pages;
7690 }
7691
7692 static void nfs4_layoutget_release(void *calldata)
7693 {
7694         struct nfs4_layoutget *lgp = calldata;
7695         struct inode *inode = lgp->args.inode;
7696         struct nfs_server *server = NFS_SERVER(inode);
7697         size_t max_pages = max_response_pages(server);
7698
7699         dprintk("--> %s\n", __func__);
7700         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7701         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7702         put_nfs_open_context(lgp->args.ctx);
7703         kfree(calldata);
7704         dprintk("<-- %s\n", __func__);
7705 }
7706
7707 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7708         .rpc_call_prepare = nfs4_layoutget_prepare,
7709         .rpc_call_done = nfs4_layoutget_done,
7710         .rpc_release = nfs4_layoutget_release,
7711 };
7712
7713 struct pnfs_layout_segment *
7714 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7715 {
7716         struct inode *inode = lgp->args.inode;
7717         struct nfs_server *server = NFS_SERVER(inode);
7718         size_t max_pages = max_response_pages(server);
7719         struct rpc_task *task;
7720         struct rpc_message msg = {
7721                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7722                 .rpc_argp = &lgp->args,
7723                 .rpc_resp = &lgp->res,
7724                 .rpc_cred = lgp->cred,
7725         };
7726         struct rpc_task_setup task_setup_data = {
7727                 .rpc_client = server->client,
7728                 .rpc_message = &msg,
7729                 .callback_ops = &nfs4_layoutget_call_ops,
7730                 .callback_data = lgp,
7731                 .flags = RPC_TASK_ASYNC,
7732         };
7733         struct pnfs_layout_segment *lseg = NULL;
7734         int status = 0;
7735
7736         dprintk("--> %s\n", __func__);
7737
7738         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7739         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7740
7741         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7742         if (!lgp->args.layout.pages) {
7743                 nfs4_layoutget_release(lgp);
7744                 return ERR_PTR(-ENOMEM);
7745         }
7746         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7747         lgp->args.timestamp = jiffies;
7748
7749         lgp->res.layoutp = &lgp->args.layout;
7750         lgp->res.seq_res.sr_slot = NULL;
7751         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7752
7753         task = rpc_run_task(&task_setup_data);
7754         if (IS_ERR(task))
7755                 return ERR_CAST(task);
7756         status = nfs4_wait_for_completion_rpc_task(task);
7757         if (status == 0)
7758                 status = task->tk_status;
7759         trace_nfs4_layoutget(lgp->args.ctx,
7760                         &lgp->args.range,
7761                         &lgp->res.range,
7762                         status);
7763         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7764         if (status == 0 && lgp->res.layoutp->len)
7765                 lseg = pnfs_layout_process(lgp);
7766         rpc_put_task(task);
7767         dprintk("<-- %s status=%d\n", __func__, status);
7768         if (status)
7769                 return ERR_PTR(status);
7770         return lseg;
7771 }
7772
7773 static void
7774 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7775 {
7776         struct nfs4_layoutreturn *lrp = calldata;
7777
7778         dprintk("--> %s\n", __func__);
7779         nfs41_setup_sequence(lrp->clp->cl_session,
7780                         &lrp->args.seq_args,
7781                         &lrp->res.seq_res,
7782                         task);
7783 }
7784
7785 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7786 {
7787         struct nfs4_layoutreturn *lrp = calldata;
7788         struct nfs_server *server;
7789
7790         dprintk("--> %s\n", __func__);
7791
7792         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7793                 return;
7794
7795         server = NFS_SERVER(lrp->args.inode);
7796         switch (task->tk_status) {
7797         default:
7798                 task->tk_status = 0;
7799         case 0:
7800                 break;
7801         case -NFS4ERR_DELAY:
7802                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7803                         break;
7804                 rpc_restart_call_prepare(task);
7805                 return;
7806         }
7807         dprintk("<-- %s\n", __func__);
7808 }
7809
7810 static void nfs4_layoutreturn_release(void *calldata)
7811 {
7812         struct nfs4_layoutreturn *lrp = calldata;
7813         struct pnfs_layout_hdr *lo = lrp->args.layout;
7814
7815         dprintk("--> %s\n", __func__);
7816         spin_lock(&lo->plh_inode->i_lock);
7817         if (lrp->res.lrs_present)
7818                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7819         pnfs_clear_layoutreturn_waitbit(lo);
7820         clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, &lo->plh_flags);
7821         rpc_wake_up(&NFS_SERVER(lo->plh_inode)->roc_rpcwaitq);
7822         lo->plh_block_lgets--;
7823         spin_unlock(&lo->plh_inode->i_lock);
7824         pnfs_put_layout_hdr(lrp->args.layout);
7825         kfree(calldata);
7826         dprintk("<-- %s\n", __func__);
7827 }
7828
7829 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7830         .rpc_call_prepare = nfs4_layoutreturn_prepare,
7831         .rpc_call_done = nfs4_layoutreturn_done,
7832         .rpc_release = nfs4_layoutreturn_release,
7833 };
7834
7835 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
7836 {
7837         struct rpc_task *task;
7838         struct rpc_message msg = {
7839                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7840                 .rpc_argp = &lrp->args,
7841                 .rpc_resp = &lrp->res,
7842                 .rpc_cred = lrp->cred,
7843         };
7844         struct rpc_task_setup task_setup_data = {
7845                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
7846                 .rpc_message = &msg,
7847                 .callback_ops = &nfs4_layoutreturn_call_ops,
7848                 .callback_data = lrp,
7849                 .flags = RPC_TASK_ASYNC,
7850         };
7851         int status = 0;
7852
7853         dprintk("--> %s\n", __func__);
7854         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7855         task = rpc_run_task(&task_setup_data);
7856         if (IS_ERR(task))
7857                 return PTR_ERR(task);
7858         if (sync == false)
7859                 goto out;
7860         status = nfs4_wait_for_completion_rpc_task(task);
7861         if (status != 0)
7862                 goto out;
7863         status = task->tk_status;
7864         trace_nfs4_layoutreturn(lrp->args.inode, status);
7865 out:
7866         dprintk("<-- %s status=%d\n", __func__, status);
7867         rpc_put_task(task);
7868         return status;
7869 }
7870
7871 static int
7872 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7873                 struct pnfs_device *pdev,
7874                 struct rpc_cred *cred)
7875 {
7876         struct nfs4_getdeviceinfo_args args = {
7877                 .pdev = pdev,
7878         };
7879         struct nfs4_getdeviceinfo_res res = {
7880                 .pdev = pdev,
7881         };
7882         struct rpc_message msg = {
7883                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7884                 .rpc_argp = &args,
7885                 .rpc_resp = &res,
7886                 .rpc_cred = cred,
7887         };
7888         int status;
7889
7890         dprintk("--> %s\n", __func__);
7891         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7892         dprintk("<-- %s status=%d\n", __func__, status);
7893
7894         return status;
7895 }
7896
7897 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7898                 struct pnfs_device *pdev,
7899                 struct rpc_cred *cred)
7900 {
7901         struct nfs4_exception exception = { };
7902         int err;
7903
7904         do {
7905                 err = nfs4_handle_exception(server,
7906                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
7907                                         &exception);
7908         } while (exception.retry);
7909         return err;
7910 }
7911 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7912
7913 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7914 {
7915         struct nfs4_layoutcommit_data *data = calldata;
7916         struct nfs_server *server = NFS_SERVER(data->args.inode);
7917         struct nfs4_session *session = nfs4_get_session(server);
7918
7919         nfs41_setup_sequence(session,
7920                         &data->args.seq_args,
7921                         &data->res.seq_res,
7922                         task);
7923 }
7924
7925 static void
7926 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7927 {
7928         struct nfs4_layoutcommit_data *data = calldata;
7929         struct nfs_server *server = NFS_SERVER(data->args.inode);
7930
7931         if (!nfs41_sequence_done(task, &data->res.seq_res))
7932                 return;
7933
7934         switch (task->tk_status) { /* Just ignore these failures */
7935         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7936         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7937         case -NFS4ERR_BADLAYOUT:     /* no layout */
7938         case -NFS4ERR_GRACE:        /* loca_recalim always false */
7939                 task->tk_status = 0;
7940         case 0:
7941                 break;
7942         default:
7943                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
7944                         rpc_restart_call_prepare(task);
7945                         return;
7946                 }
7947         }
7948 }
7949
7950 static void nfs4_layoutcommit_release(void *calldata)
7951 {
7952         struct nfs4_layoutcommit_data *data = calldata;
7953
7954         pnfs_cleanup_layoutcommit(data);
7955         nfs_post_op_update_inode_force_wcc(data->args.inode,
7956                                            data->res.fattr);
7957         put_rpccred(data->cred);
7958         kfree(data);
7959 }
7960
7961 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
7962         .rpc_call_prepare = nfs4_layoutcommit_prepare,
7963         .rpc_call_done = nfs4_layoutcommit_done,
7964         .rpc_release = nfs4_layoutcommit_release,
7965 };
7966
7967 int
7968 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
7969 {
7970         struct rpc_message msg = {
7971                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
7972                 .rpc_argp = &data->args,
7973                 .rpc_resp = &data->res,
7974                 .rpc_cred = data->cred,
7975         };
7976         struct rpc_task_setup task_setup_data = {
7977                 .task = &data->task,
7978                 .rpc_client = NFS_CLIENT(data->args.inode),
7979                 .rpc_message = &msg,
7980                 .callback_ops = &nfs4_layoutcommit_ops,
7981                 .callback_data = data,
7982                 .flags = RPC_TASK_ASYNC,
7983         };
7984         struct rpc_task *task;
7985         int status = 0;
7986
7987         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
7988                 "lbw: %llu inode %lu\n",
7989                 data->task.tk_pid, sync,
7990                 data->args.lastbytewritten,
7991                 data->args.inode->i_ino);
7992
7993         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
7994         task = rpc_run_task(&task_setup_data);
7995         if (IS_ERR(task))
7996                 return PTR_ERR(task);
7997         if (sync == false)
7998                 goto out;
7999         status = nfs4_wait_for_completion_rpc_task(task);
8000         if (status != 0)
8001                 goto out;
8002         status = task->tk_status;
8003         trace_nfs4_layoutcommit(data->args.inode, status);
8004 out:
8005         dprintk("%s: status %d\n", __func__, status);
8006         rpc_put_task(task);
8007         return status;
8008 }
8009
8010 /**
8011  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8012  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8013  */
8014 static int
8015 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8016                     struct nfs_fsinfo *info,
8017                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8018 {
8019         struct nfs41_secinfo_no_name_args args = {
8020                 .style = SECINFO_STYLE_CURRENT_FH,
8021         };
8022         struct nfs4_secinfo_res res = {
8023                 .flavors = flavors,
8024         };
8025         struct rpc_message msg = {
8026                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8027                 .rpc_argp = &args,
8028                 .rpc_resp = &res,
8029         };
8030         struct rpc_clnt *clnt = server->client;
8031         struct rpc_cred *cred = NULL;
8032         int status;
8033
8034         if (use_integrity) {
8035                 clnt = server->nfs_client->cl_rpcclient;
8036                 cred = nfs4_get_clid_cred(server->nfs_client);
8037                 msg.rpc_cred = cred;
8038         }
8039
8040         dprintk("--> %s\n", __func__);
8041         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8042                                 &res.seq_res, 0);
8043         dprintk("<-- %s status=%d\n", __func__, status);
8044
8045         if (cred)
8046                 put_rpccred(cred);
8047
8048         return status;
8049 }
8050
8051 static int
8052 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8053                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8054 {
8055         struct nfs4_exception exception = { };
8056         int err;
8057         do {
8058                 /* first try using integrity protection */
8059                 err = -NFS4ERR_WRONGSEC;
8060
8061                 /* try to use integrity protection with machine cred */
8062                 if (_nfs4_is_integrity_protected(server->nfs_client))
8063                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8064                                                           flavors, true);
8065
8066                 /*
8067                  * if unable to use integrity protection, or SECINFO with
8068                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8069                  * disallowed by spec, but exists in deployed servers) use
8070                  * the current filesystem's rpc_client and the user cred.
8071                  */
8072                 if (err == -NFS4ERR_WRONGSEC)
8073                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8074                                                           flavors, false);
8075
8076                 switch (err) {
8077                 case 0:
8078                 case -NFS4ERR_WRONGSEC:
8079                 case -ENOTSUPP:
8080                         goto out;
8081                 default:
8082                         err = nfs4_handle_exception(server, err, &exception);
8083                 }
8084         } while (exception.retry);
8085 out:
8086         return err;
8087 }
8088
8089 static int
8090 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8091                     struct nfs_fsinfo *info)
8092 {
8093         int err;
8094         struct page *page;
8095         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8096         struct nfs4_secinfo_flavors *flavors;
8097         struct nfs4_secinfo4 *secinfo;
8098         int i;
8099
8100         page = alloc_page(GFP_KERNEL);
8101         if (!page) {
8102                 err = -ENOMEM;
8103                 goto out;
8104         }
8105
8106         flavors = page_address(page);
8107         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8108
8109         /*
8110          * Fall back on "guess and check" method if
8111          * the server doesn't support SECINFO_NO_NAME
8112          */
8113         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8114                 err = nfs4_find_root_sec(server, fhandle, info);
8115                 goto out_freepage;
8116         }
8117         if (err)
8118                 goto out_freepage;
8119
8120         for (i = 0; i < flavors->num_flavors; i++) {
8121                 secinfo = &flavors->flavors[i];
8122
8123                 switch (secinfo->flavor) {
8124                 case RPC_AUTH_NULL:
8125                 case RPC_AUTH_UNIX:
8126                 case RPC_AUTH_GSS:
8127                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8128                                         &secinfo->flavor_info);
8129                         break;
8130                 default:
8131                         flavor = RPC_AUTH_MAXFLAVOR;
8132                         break;
8133                 }
8134
8135                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8136                         flavor = RPC_AUTH_MAXFLAVOR;
8137
8138                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8139                         err = nfs4_lookup_root_sec(server, fhandle,
8140                                                    info, flavor);
8141                         if (!err)
8142                                 break;
8143                 }
8144         }
8145
8146         if (flavor == RPC_AUTH_MAXFLAVOR)
8147                 err = -EPERM;
8148
8149 out_freepage:
8150         put_page(page);
8151         if (err == -EACCES)
8152                 return -EPERM;
8153 out:
8154         return err;
8155 }
8156
8157 static int _nfs41_test_stateid(struct nfs_server *server,
8158                 nfs4_stateid *stateid,
8159                 struct rpc_cred *cred)
8160 {
8161         int status;
8162         struct nfs41_test_stateid_args args = {
8163                 .stateid = stateid,
8164         };
8165         struct nfs41_test_stateid_res res;
8166         struct rpc_message msg = {
8167                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8168                 .rpc_argp = &args,
8169                 .rpc_resp = &res,
8170                 .rpc_cred = cred,
8171         };
8172         struct rpc_clnt *rpc_client = server->client;
8173
8174         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8175                 &rpc_client, &msg);
8176
8177         dprintk("NFS call  test_stateid %p\n", stateid);
8178         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8179         nfs4_set_sequence_privileged(&args.seq_args);
8180         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8181                         &args.seq_args, &res.seq_res);
8182         if (status != NFS_OK) {
8183                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8184                 return status;
8185         }
8186         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8187         return -res.status;
8188 }
8189
8190 /**
8191  * nfs41_test_stateid - perform a TEST_STATEID operation
8192  *
8193  * @server: server / transport on which to perform the operation
8194  * @stateid: state ID to test
8195  * @cred: credential
8196  *
8197  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8198  * Otherwise a negative NFS4ERR value is returned if the operation
8199  * failed or the state ID is not currently valid.
8200  */
8201 static int nfs41_test_stateid(struct nfs_server *server,
8202                 nfs4_stateid *stateid,
8203                 struct rpc_cred *cred)
8204 {
8205         struct nfs4_exception exception = { };
8206         int err;
8207         do {
8208                 err = _nfs41_test_stateid(server, stateid, cred);
8209                 if (err != -NFS4ERR_DELAY)
8210                         break;
8211                 nfs4_handle_exception(server, err, &exception);
8212         } while (exception.retry);
8213         return err;
8214 }
8215
8216 struct nfs_free_stateid_data {
8217         struct nfs_server *server;
8218         struct nfs41_free_stateid_args args;
8219         struct nfs41_free_stateid_res res;
8220 };
8221
8222 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8223 {
8224         struct nfs_free_stateid_data *data = calldata;
8225         nfs41_setup_sequence(nfs4_get_session(data->server),
8226                         &data->args.seq_args,
8227                         &data->res.seq_res,
8228                         task);
8229 }
8230
8231 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8232 {
8233         struct nfs_free_stateid_data *data = calldata;
8234
8235         nfs41_sequence_done(task, &data->res.seq_res);
8236
8237         switch (task->tk_status) {
8238         case -NFS4ERR_DELAY:
8239                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8240                         rpc_restart_call_prepare(task);
8241         }
8242 }
8243
8244 static void nfs41_free_stateid_release(void *calldata)
8245 {
8246         kfree(calldata);
8247 }
8248
8249 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8250         .rpc_call_prepare = nfs41_free_stateid_prepare,
8251         .rpc_call_done = nfs41_free_stateid_done,
8252         .rpc_release = nfs41_free_stateid_release,
8253 };
8254
8255 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8256                 nfs4_stateid *stateid,
8257                 struct rpc_cred *cred,
8258                 bool privileged)
8259 {
8260         struct rpc_message msg = {
8261                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8262                 .rpc_cred = cred,
8263         };
8264         struct rpc_task_setup task_setup = {
8265                 .rpc_client = server->client,
8266                 .rpc_message = &msg,
8267                 .callback_ops = &nfs41_free_stateid_ops,
8268                 .flags = RPC_TASK_ASYNC,
8269         };
8270         struct nfs_free_stateid_data *data;
8271
8272         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8273                 &task_setup.rpc_client, &msg);
8274
8275         dprintk("NFS call  free_stateid %p\n", stateid);
8276         data = kmalloc(sizeof(*data), GFP_NOFS);
8277         if (!data)
8278                 return ERR_PTR(-ENOMEM);
8279         data->server = server;
8280         nfs4_stateid_copy(&data->args.stateid, stateid);
8281
8282         task_setup.callback_data = data;
8283
8284         msg.rpc_argp = &data->args;
8285         msg.rpc_resp = &data->res;
8286         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8287         if (privileged)
8288                 nfs4_set_sequence_privileged(&data->args.seq_args);
8289
8290         return rpc_run_task(&task_setup);
8291 }
8292
8293 /**
8294  * nfs41_free_stateid - perform a FREE_STATEID operation
8295  *
8296  * @server: server / transport on which to perform the operation
8297  * @stateid: state ID to release
8298  * @cred: credential
8299  *
8300  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8301  * negative NFS4ERR value is returned.
8302  */
8303 static int nfs41_free_stateid(struct nfs_server *server,
8304                 nfs4_stateid *stateid,
8305                 struct rpc_cred *cred)
8306 {
8307         struct rpc_task *task;
8308         int ret;
8309
8310         task = _nfs41_free_stateid(server, stateid, cred, true);
8311         if (IS_ERR(task))
8312                 return PTR_ERR(task);
8313         ret = rpc_wait_for_completion_task(task);
8314         if (!ret)
8315                 ret = task->tk_status;
8316         rpc_put_task(task);
8317         return ret;
8318 }
8319
8320 static void
8321 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8322 {
8323         struct rpc_task *task;
8324         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8325
8326         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8327         nfs4_free_lock_state(server, lsp);
8328         if (IS_ERR(task))
8329                 return;
8330         rpc_put_task(task);
8331 }
8332
8333 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8334                 const nfs4_stateid *s2)
8335 {
8336         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8337                 return false;
8338
8339         if (s1->seqid == s2->seqid)
8340                 return true;
8341         if (s1->seqid == 0 || s2->seqid == 0)
8342                 return true;
8343
8344         return false;
8345 }
8346
8347 #endif /* CONFIG_NFS_V4_1 */
8348
8349 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8350                 const nfs4_stateid *s2)
8351 {
8352         return nfs4_stateid_match(s1, s2);
8353 }
8354
8355
8356 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8357         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8358         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8359         .recover_open   = nfs4_open_reclaim,
8360         .recover_lock   = nfs4_lock_reclaim,
8361         .establish_clid = nfs4_init_clientid,
8362         .detect_trunking = nfs40_discover_server_trunking,
8363 };
8364
8365 #if defined(CONFIG_NFS_V4_1)
8366 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8367         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8368         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8369         .recover_open   = nfs4_open_reclaim,
8370         .recover_lock   = nfs4_lock_reclaim,
8371         .establish_clid = nfs41_init_clientid,
8372         .reclaim_complete = nfs41_proc_reclaim_complete,
8373         .detect_trunking = nfs41_discover_server_trunking,
8374 };
8375 #endif /* CONFIG_NFS_V4_1 */
8376
8377 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8378         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8379         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8380         .recover_open   = nfs40_open_expired,
8381         .recover_lock   = nfs4_lock_expired,
8382         .establish_clid = nfs4_init_clientid,
8383 };
8384
8385 #if defined(CONFIG_NFS_V4_1)
8386 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8387         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8388         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8389         .recover_open   = nfs41_open_expired,
8390         .recover_lock   = nfs41_lock_expired,
8391         .establish_clid = nfs41_init_clientid,
8392 };
8393 #endif /* CONFIG_NFS_V4_1 */
8394
8395 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8396         .sched_state_renewal = nfs4_proc_async_renew,
8397         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8398         .renew_lease = nfs4_proc_renew,
8399 };
8400
8401 #if defined(CONFIG_NFS_V4_1)
8402 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8403         .sched_state_renewal = nfs41_proc_async_sequence,
8404         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8405         .renew_lease = nfs4_proc_sequence,
8406 };
8407 #endif
8408
8409 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8410         .get_locations = _nfs40_proc_get_locations,
8411         .fsid_present = _nfs40_proc_fsid_present,
8412 };
8413
8414 #if defined(CONFIG_NFS_V4_1)
8415 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8416         .get_locations = _nfs41_proc_get_locations,
8417         .fsid_present = _nfs41_proc_fsid_present,
8418 };
8419 #endif  /* CONFIG_NFS_V4_1 */
8420
8421 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8422         .minor_version = 0,
8423         .init_caps = NFS_CAP_READDIRPLUS
8424                 | NFS_CAP_ATOMIC_OPEN
8425                 | NFS_CAP_CHANGE_ATTR
8426                 | NFS_CAP_POSIX_LOCK,
8427         .init_client = nfs40_init_client,
8428         .shutdown_client = nfs40_shutdown_client,
8429         .match_stateid = nfs4_match_stateid,
8430         .find_root_sec = nfs4_find_root_sec,
8431         .free_lock_state = nfs4_release_lockowner,
8432         .alloc_seqid = nfs_alloc_seqid,
8433         .call_sync_ops = &nfs40_call_sync_ops,
8434         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8435         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8436         .state_renewal_ops = &nfs40_state_renewal_ops,
8437         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8438 };
8439
8440 #if defined(CONFIG_NFS_V4_1)
8441 static struct nfs_seqid *
8442 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8443 {
8444         return NULL;
8445 }
8446
8447 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8448         .minor_version = 1,
8449         .init_caps = NFS_CAP_READDIRPLUS
8450                 | NFS_CAP_ATOMIC_OPEN
8451                 | NFS_CAP_CHANGE_ATTR
8452                 | NFS_CAP_POSIX_LOCK
8453                 | NFS_CAP_STATEID_NFSV41
8454                 | NFS_CAP_ATOMIC_OPEN_V1,
8455         .init_client = nfs41_init_client,
8456         .shutdown_client = nfs41_shutdown_client,
8457         .match_stateid = nfs41_match_stateid,
8458         .find_root_sec = nfs41_find_root_sec,
8459         .free_lock_state = nfs41_free_lock_state,
8460         .alloc_seqid = nfs_alloc_no_seqid,
8461         .call_sync_ops = &nfs41_call_sync_ops,
8462         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8463         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8464         .state_renewal_ops = &nfs41_state_renewal_ops,
8465         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8466 };
8467 #endif
8468
8469 #if defined(CONFIG_NFS_V4_2)
8470 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8471         .minor_version = 2,
8472         .init_caps = NFS_CAP_READDIRPLUS
8473                 | NFS_CAP_ATOMIC_OPEN
8474                 | NFS_CAP_CHANGE_ATTR
8475                 | NFS_CAP_POSIX_LOCK
8476                 | NFS_CAP_STATEID_NFSV41
8477                 | NFS_CAP_ATOMIC_OPEN_V1
8478                 | NFS_CAP_ALLOCATE
8479                 | NFS_CAP_DEALLOCATE
8480                 | NFS_CAP_SEEK,
8481         .init_client = nfs41_init_client,
8482         .shutdown_client = nfs41_shutdown_client,
8483         .match_stateid = nfs41_match_stateid,
8484         .find_root_sec = nfs41_find_root_sec,
8485         .free_lock_state = nfs41_free_lock_state,
8486         .call_sync_ops = &nfs41_call_sync_ops,
8487         .alloc_seqid = nfs_alloc_no_seqid,
8488         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8489         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8490         .state_renewal_ops = &nfs41_state_renewal_ops,
8491 };
8492 #endif
8493
8494 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8495         [0] = &nfs_v4_0_minor_ops,
8496 #if defined(CONFIG_NFS_V4_1)
8497         [1] = &nfs_v4_1_minor_ops,
8498 #endif
8499 #if defined(CONFIG_NFS_V4_2)
8500         [2] = &nfs_v4_2_minor_ops,
8501 #endif
8502 };
8503
8504 static const struct inode_operations nfs4_dir_inode_operations = {
8505         .create         = nfs_create,
8506         .lookup         = nfs_lookup,
8507         .atomic_open    = nfs_atomic_open,
8508         .link           = nfs_link,
8509         .unlink         = nfs_unlink,
8510         .symlink        = nfs_symlink,
8511         .mkdir          = nfs_mkdir,
8512         .rmdir          = nfs_rmdir,
8513         .mknod          = nfs_mknod,
8514         .rename         = nfs_rename,
8515         .permission     = nfs_permission,
8516         .getattr        = nfs_getattr,
8517         .setattr        = nfs_setattr,
8518         .getxattr       = generic_getxattr,
8519         .setxattr       = generic_setxattr,
8520         .listxattr      = generic_listxattr,
8521         .removexattr    = generic_removexattr,
8522 };
8523
8524 static const struct inode_operations nfs4_file_inode_operations = {
8525         .permission     = nfs_permission,
8526         .getattr        = nfs_getattr,
8527         .setattr        = nfs_setattr,
8528         .getxattr       = generic_getxattr,
8529         .setxattr       = generic_setxattr,
8530         .listxattr      = generic_listxattr,
8531         .removexattr    = generic_removexattr,
8532 };
8533
8534 const struct nfs_rpc_ops nfs_v4_clientops = {
8535         .version        = 4,                    /* protocol version */
8536         .dentry_ops     = &nfs4_dentry_operations,
8537         .dir_inode_ops  = &nfs4_dir_inode_operations,
8538         .file_inode_ops = &nfs4_file_inode_operations,
8539         .file_ops       = &nfs4_file_operations,
8540         .getroot        = nfs4_proc_get_root,
8541         .submount       = nfs4_submount,
8542         .try_mount      = nfs4_try_mount,
8543         .getattr        = nfs4_proc_getattr,
8544         .setattr        = nfs4_proc_setattr,
8545         .lookup         = nfs4_proc_lookup,
8546         .access         = nfs4_proc_access,
8547         .readlink       = nfs4_proc_readlink,
8548         .create         = nfs4_proc_create,
8549         .remove         = nfs4_proc_remove,
8550         .unlink_setup   = nfs4_proc_unlink_setup,
8551         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8552         .unlink_done    = nfs4_proc_unlink_done,
8553         .rename_setup   = nfs4_proc_rename_setup,
8554         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8555         .rename_done    = nfs4_proc_rename_done,
8556         .link           = nfs4_proc_link,
8557         .symlink        = nfs4_proc_symlink,
8558         .mkdir          = nfs4_proc_mkdir,
8559         .rmdir          = nfs4_proc_remove,
8560         .readdir        = nfs4_proc_readdir,
8561         .mknod          = nfs4_proc_mknod,
8562         .statfs         = nfs4_proc_statfs,
8563         .fsinfo         = nfs4_proc_fsinfo,
8564         .pathconf       = nfs4_proc_pathconf,
8565         .set_capabilities = nfs4_server_capabilities,
8566         .decode_dirent  = nfs4_decode_dirent,
8567         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8568         .read_setup     = nfs4_proc_read_setup,
8569         .read_done      = nfs4_read_done,
8570         .write_setup    = nfs4_proc_write_setup,
8571         .write_done     = nfs4_write_done,
8572         .commit_setup   = nfs4_proc_commit_setup,
8573         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8574         .commit_done    = nfs4_commit_done,
8575         .lock           = nfs4_proc_lock,
8576         .clear_acl_cache = nfs4_zap_acl_attr,
8577         .close_context  = nfs4_close_context,
8578         .open_context   = nfs4_atomic_open,
8579         .have_delegation = nfs4_have_delegation,
8580         .return_delegation = nfs4_inode_return_delegation,
8581         .alloc_client   = nfs4_alloc_client,
8582         .init_client    = nfs4_init_client,
8583         .free_client    = nfs4_free_client,
8584         .create_server  = nfs4_create_server,
8585         .clone_server   = nfs_clone_server,
8586 };
8587
8588 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8589         .prefix = XATTR_NAME_NFSV4_ACL,
8590         .list   = nfs4_xattr_list_nfs4_acl,
8591         .get    = nfs4_xattr_get_nfs4_acl,
8592         .set    = nfs4_xattr_set_nfs4_acl,
8593 };
8594
8595 const struct xattr_handler *nfs4_xattr_handlers[] = {
8596         &nfs4_xattr_nfs4_acl_handler,
8597 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8598         &nfs4_xattr_nfs4_label_handler,
8599 #endif
8600         NULL
8601 };
8602
8603 /*
8604  * Local variables:
8605  *  c-basic-offset: 8
8606  * End:
8607  */