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 #define NFSDBG_FACILITY         NFSDBG_PROC
  70
  71 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  72 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  73
  74 struct nfs4_opendata;
  75 static int _nfs4_proc_open(struct nfs4_opendata *data);
  76 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  77 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  78 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  79 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  80 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *);
  81 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  82 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  83                             struct nfs_fattr *fattr, struct iattr *sattr,
  84                             struct nfs4_state *state);
  85 #ifdef CONFIG_NFS_V4_1
  86 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
  87 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
  88 #endif
  89 /* Prevent leaks of NFSv4 errors into userland */
  90 static int nfs4_map_errors(int err)
  91 {
  92         if (err >= -1000)
  93                 return err;
  94         switch (err) {
  95         case -NFS4ERR_RESOURCE:
  96         case -NFS4ERR_LAYOUTTRYLATER:
  97         case -NFS4ERR_RECALLCONFLICT:
  98                 return -EREMOTEIO;
  99         case -NFS4ERR_WRONGSEC:
 100                 return -EPERM;
 101         case -NFS4ERR_BADOWNER:
 102         case -NFS4ERR_BADNAME:
 103                 return -EINVAL;
 104         case -NFS4ERR_SHARE_DENIED:
 105                 return -EACCES;
 106         case -NFS4ERR_MINOR_VERS_MISMATCH:
 107                 return -EPROTONOSUPPORT;
 108         case -NFS4ERR_ACCESS:
 109                 return -EACCES;
 110         case -NFS4ERR_FILE_OPEN:
 111                 return -EBUSY;
 112         default:
 113                 dprintk("%s could not handle NFSv4 error %d\n",
 114                                 __func__, -err);
 115                 break;
 116         }
 117         return -EIO;
 118 }
 119
 120 /*
 121  * This is our standard bitmap for GETATTR requests.
 122  */
 123 const u32 nfs4_fattr_bitmap[3] = {
 124         FATTR4_WORD0_TYPE
 125         | FATTR4_WORD0_CHANGE
 126         | FATTR4_WORD0_SIZE
 127         | FATTR4_WORD0_FSID
 128         | FATTR4_WORD0_FILEID,
 129         FATTR4_WORD1_MODE
 130         | FATTR4_WORD1_NUMLINKS
 131         | FATTR4_WORD1_OWNER
 132         | FATTR4_WORD1_OWNER_GROUP
 133         | FATTR4_WORD1_RAWDEV
 134         | FATTR4_WORD1_SPACE_USED
 135         | FATTR4_WORD1_TIME_ACCESS
 136         | FATTR4_WORD1_TIME_METADATA
 137         | FATTR4_WORD1_TIME_MODIFY
 138 };
 139
 140 static const u32 nfs4_pnfs_open_bitmap[3] = {
 141         FATTR4_WORD0_TYPE
 142         | FATTR4_WORD0_CHANGE
 143         | FATTR4_WORD0_SIZE
 144         | FATTR4_WORD0_FSID
 145         | FATTR4_WORD0_FILEID,
 146         FATTR4_WORD1_MODE
 147         | FATTR4_WORD1_NUMLINKS
 148         | FATTR4_WORD1_OWNER
 149         | FATTR4_WORD1_OWNER_GROUP
 150         | FATTR4_WORD1_RAWDEV
 151         | FATTR4_WORD1_SPACE_USED
 152         | FATTR4_WORD1_TIME_ACCESS
 153         | FATTR4_WORD1_TIME_METADATA
 154         | FATTR4_WORD1_TIME_MODIFY,
 155         FATTR4_WORD2_MDSTHRESHOLD
 156 };
 157
 158 static const u32 nfs4_open_noattr_bitmap[3] = {
 159         FATTR4_WORD0_TYPE
 160         | FATTR4_WORD0_CHANGE
 161         | FATTR4_WORD0_FILEID,
 162 };
 163
 164 const u32 nfs4_statfs_bitmap[2] = {
 165         FATTR4_WORD0_FILES_AVAIL
 166         | FATTR4_WORD0_FILES_FREE
 167         | FATTR4_WORD0_FILES_TOTAL,
 168         FATTR4_WORD1_SPACE_AVAIL
 169         | FATTR4_WORD1_SPACE_FREE
 170         | FATTR4_WORD1_SPACE_TOTAL
 171 };
 172
 173 const u32 nfs4_pathconf_bitmap[2] = {
 174         FATTR4_WORD0_MAXLINK
 175         | FATTR4_WORD0_MAXNAME,
 176         0
 177 };
 178
 179 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 180                         | FATTR4_WORD0_MAXREAD
 181                         | FATTR4_WORD0_MAXWRITE
 182                         | FATTR4_WORD0_LEASE_TIME,
 183                         FATTR4_WORD1_TIME_DELTA
 184                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 185                         FATTR4_WORD2_LAYOUT_BLKSIZE
 186 };
 187
 188 const u32 nfs4_fs_locations_bitmap[2] = {
 189         FATTR4_WORD0_TYPE
 190         | FATTR4_WORD0_CHANGE
 191         | FATTR4_WORD0_SIZE
 192         | FATTR4_WORD0_FSID
 193         | FATTR4_WORD0_FILEID
 194         | FATTR4_WORD0_FS_LOCATIONS,
 195         FATTR4_WORD1_MODE
 196         | FATTR4_WORD1_NUMLINKS
 197         | FATTR4_WORD1_OWNER
 198         | FATTR4_WORD1_OWNER_GROUP
 199         | FATTR4_WORD1_RAWDEV
 200         | FATTR4_WORD1_SPACE_USED
 201         | FATTR4_WORD1_TIME_ACCESS
 202         | FATTR4_WORD1_TIME_METADATA
 203         | FATTR4_WORD1_TIME_MODIFY
 204         | FATTR4_WORD1_MOUNTED_ON_FILEID
 205 };
 206
 207 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 208                 struct nfs4_readdir_arg *readdir)
 209 {
 210         __be32 *start, *p;
 211
 212         if (cookie > 2) {
 213                 readdir->cookie = cookie;
 214                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 215                 return;
 216         }
 217
 218         readdir->cookie = 0;
 219         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 220         if (cookie == 2)
 221                 return;
 222
 223         /*
 224          * NFSv4 servers do not return entries for '.' and '..'
 225          * Therefore, we fake these entries here.  We let '.'
 226          * have cookie 0 and '..' have cookie 1.  Note that
 227          * when talking to the server, we always send cookie 0
 228          * instead of 1 or 2.
 229          */
 230         start = p = kmap_atomic(*readdir->pages);
 231
 232         if (cookie == 0) {
 233                 *p++ = xdr_one;                                  /* next */
 234                 *p++ = xdr_zero;                   /* cookie, first word */
 235                 *p++ = xdr_one;                   /* cookie, second word */
 236                 *p++ = xdr_one;                             /* entry len */
 237                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 238                 p++;
 239                 *p++ = xdr_one;                         /* bitmap length */
 240                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 241                 *p++ = htonl(8);              /* attribute buffer length */
 242                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 243         }
 244
 245         *p++ = xdr_one;                                  /* next */
 246         *p++ = xdr_zero;                   /* cookie, first word */
 247         *p++ = xdr_two;                   /* cookie, second word */
 248         *p++ = xdr_two;                             /* entry len */
 249         memcpy(p, "..\0\0", 4);                         /* entry */
 250         p++;
 251         *p++ = xdr_one;                         /* bitmap length */
 252         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 253         *p++ = htonl(8);              /* attribute buffer length */
 254         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 255
 256         readdir->pgbase = (char *)p - (char *)start;
 257         readdir->count -= readdir->pgbase;
 258         kunmap_atomic(start);
 259 }
 260
 261 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 262 {
 263         int res = 0;
 264
 265         might_sleep();
 266
 267         if (*timeout <= 0)
 268                 *timeout = NFS4_POLL_RETRY_MIN;
 269         if (*timeout > NFS4_POLL_RETRY_MAX)
 270                 *timeout = NFS4_POLL_RETRY_MAX;
 271         freezable_schedule_timeout_killable(*timeout);
 272         if (fatal_signal_pending(current))
 273                 res = -ERESTARTSYS;
 274         *timeout <<= 1;
 275         return res;
 276 }
 277
 278 /* This is the error handling routine for processes that are allowed
 279  * to sleep.
 280  */
 281 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 282 {
 283         struct nfs_client *clp = server->nfs_client;
 284         struct nfs4_state *state = exception->state;
 285         struct inode *inode = exception->inode;
 286         int ret = errorcode;
 287
 288         exception->retry = 0;
 289         switch(errorcode) {
 290                 case 0:
 291                         return 0;
 292                 case -NFS4ERR_OPENMODE:
 293                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 294                                 nfs4_inode_return_delegation(inode);
 295                                 exception->retry = 1;
 296                                 return 0;
 297                         }
 298                         if (state == NULL)
 299                                 break;
 300                         ret = nfs4_schedule_stateid_recovery(server, state);
 301                         if (ret < 0)
 302                                 break;
 303                         goto wait_on_recovery;
 304                 case -NFS4ERR_DELEG_REVOKED:
 305                 case -NFS4ERR_ADMIN_REVOKED:
 306                 case -NFS4ERR_BAD_STATEID:
 307                         if (state == NULL)
 308                                 break;
 309                         nfs_remove_bad_delegation(state->inode);
 310                         ret = nfs4_schedule_stateid_recovery(server, state);
 311                         if (ret < 0)
 312                                 break;
 313                         goto wait_on_recovery;
 314                 case -NFS4ERR_EXPIRED:
 315                         if (state != NULL) {
 316                                 ret = nfs4_schedule_stateid_recovery(server, state);
 317                                 if (ret < 0)
 318                                         break;
 319                         }
 320                 case -NFS4ERR_STALE_STATEID:
 321                 case -NFS4ERR_STALE_CLIENTID:
 322                         nfs4_schedule_lease_recovery(clp);
 323                         goto wait_on_recovery;
 324 #if defined(CONFIG_NFS_V4_1)
 325                 case -NFS4ERR_BADSESSION:
 326                 case -NFS4ERR_BADSLOT:
 327                 case -NFS4ERR_BAD_HIGH_SLOT:
 328                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 329                 case -NFS4ERR_DEADSESSION:
 330                 case -NFS4ERR_SEQ_FALSE_RETRY:
 331                 case -NFS4ERR_SEQ_MISORDERED:
 332                         dprintk("%s ERROR: %d Reset session\n", __func__,
 333                                 errorcode);
 334                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 335                         goto wait_on_recovery;
 336 #endif /* defined(CONFIG_NFS_V4_1) */
 337                 case -NFS4ERR_FILE_OPEN:
 338                         if (exception->timeout > HZ) {
 339                                 /* We have retried a decent amount, time to
 340                                  * fail
 341                                  */
 342                                 ret = -EBUSY;
 343                                 break;
 344                         }
 345                 case -NFS4ERR_GRACE:
 346                 case -NFS4ERR_DELAY:
 347                         ret = nfs4_delay(server->client, &exception->timeout);
 348                         if (ret != 0)
 349                                 break;
 350                 case -NFS4ERR_RETRY_UNCACHED_REP:
 351                 case -NFS4ERR_OLD_STATEID:
 352                         exception->retry = 1;
 353                         break;
 354                 case -NFS4ERR_BADOWNER:
 355                         /* The following works around a Linux server bug! */
 356                 case -NFS4ERR_BADNAME:
 357                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 358                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 359                                 exception->retry = 1;
 360                                 printk(KERN_WARNING "NFS: v4 server %s "
 361                                                 "does not accept raw "
 362                                                 "uid/gids. "
 363                                                 "Reenabling the idmapper.\n",
 364                                                 server->nfs_client->cl_hostname);
 365                         }
 366         }
 367         /* We failed to handle the error */
 368         return nfs4_map_errors(ret);
 369 wait_on_recovery:
 370         ret = nfs4_wait_clnt_recover(clp);
 371         if (ret == 0)
 372                 exception->retry = 1;
 373         return ret;
 374 }
 375
 376
 377 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 378 {
 379         spin_lock(&clp->cl_lock);
 380         if (time_before(clp->cl_last_renewal,timestamp))
 381                 clp->cl_last_renewal = timestamp;
 382         spin_unlock(&clp->cl_lock);
 383 }
 384
 385 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 386 {
 387         do_renew_lease(server->nfs_client, timestamp);
 388 }
 389
 390 #if defined(CONFIG_NFS_V4_1)
 391
 392 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 393 {
 394         struct nfs4_session *session;
 395         struct nfs4_slot_table *tbl;
 396         bool send_new_highest_used_slotid = false;
 397
 398         if (!res->sr_slot) {
 399                 /* just wake up the next guy waiting since
 400                  * we may have not consumed a slot after all */
 401                 dprintk("%s: No slot\n", __func__);
 402                 return;
 403         }
 404         tbl = res->sr_slot->table;
 405         session = tbl->session;
 406
 407         spin_lock(&tbl->slot_tbl_lock);
 408         /* Be nice to the server: try to ensure that the last transmitted
 409          * value for highest_user_slotid <= target_highest_slotid
 410          */
 411         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 412                 send_new_highest_used_slotid = true;
 413
 414         if (nfs41_wake_and_assign_slot(tbl, res->sr_slot)) {
 415                 send_new_highest_used_slotid = false;
 416                 goto out_unlock;
 417         }
 418         nfs4_free_slot(tbl, res->sr_slot);
 419
 420         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 421                 send_new_highest_used_slotid = false;
 422 out_unlock:
 423         spin_unlock(&tbl->slot_tbl_lock);
 424         res->sr_slot = NULL;
 425         if (send_new_highest_used_slotid)
 426                 nfs41_server_notify_highest_slotid_update(session->clp);
 427 }
 428
 429 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 430 {
 431         struct nfs4_session *session;
 432         struct nfs4_slot *slot;
 433         struct nfs_client *clp;
 434         bool interrupted = false;
 435         int ret = 1;
 436
 437         /* don't increment the sequence number if the task wasn't sent */
 438         if (!RPC_WAS_SENT(task))
 439                 goto out;
 440
 441         slot = res->sr_slot;
 442         session = slot->table->session;
 443
 444         if (slot->interrupted) {
 445                 slot->interrupted = 0;
 446                 interrupted = true;
 447         }
 448
 449         /* Check the SEQUENCE operation status */
 450         switch (res->sr_status) {
 451         case 0:
 452                 /* Update the slot's sequence and clientid lease timer */
 453                 ++slot->seq_nr;
 454                 clp = session->clp;
 455                 do_renew_lease(clp, res->sr_timestamp);
 456                 /* Check sequence flags */
 457                 if (res->sr_status_flags != 0)
 458                         nfs4_schedule_lease_recovery(clp);
 459                 nfs41_update_target_slotid(slot->table, slot, res);
 460                 break;
 461         case 1:
 462                 /*
 463                  * sr_status remains 1 if an RPC level error occurred.
 464                  * The server may or may not have processed the sequence
 465                  * operation..
 466                  * Mark the slot as having hosted an interrupted RPC call.
 467                  */
 468                 slot->interrupted = 1;
 469                 goto out;
 470         case -NFS4ERR_DELAY:
 471                 /* The server detected a resend of the RPC call and
 472                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 473                  * of RFC5661.
 474                  */
 475                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 476                         __func__,
 477                         slot->slot_nr,
 478                         slot->seq_nr);
 479                 goto out_retry;
 480         case -NFS4ERR_BADSLOT:
 481                 /*
 482                  * The slot id we used was probably retired. Try again
 483                  * using a different slot id.
 484                  */
 485                 goto retry_nowait;
 486         case -NFS4ERR_SEQ_MISORDERED:
 487                 /*
 488                  * Was the last operation on this sequence interrupted?
 489                  * If so, retry after bumping the sequence number.
 490                  */
 491                 if (interrupted) {
 492                         ++slot->seq_nr;
 493                         goto retry_nowait;
 494                 }
 495                 /*
 496                  * Could this slot have been previously retired?
 497                  * If so, then the server may be expecting seq_nr = 1!
 498                  */
 499                 if (slot->seq_nr != 1) {
 500                         slot->seq_nr = 1;
 501                         goto retry_nowait;
 502                 }
 503                 break;
 504         case -NFS4ERR_SEQ_FALSE_RETRY:
 505                 ++slot->seq_nr;
 506                 goto retry_nowait;
 507         default:
 508                 /* Just update the slot sequence no. */
 509                 ++slot->seq_nr;
 510         }
 511 out:
 512         /* The session may be reset by one of the error handlers. */
 513         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 514         nfs41_sequence_free_slot(res);
 515         return ret;
 516 retry_nowait:
 517         if (rpc_restart_call_prepare(task)) {
 518                 task->tk_status = 0;
 519                 ret = 0;
 520         }
 521         goto out;
 522 out_retry:
 523         if (!rpc_restart_call(task))
 524                 goto out;
 525         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 526         return 0;
 527 }
 528
 529 static int nfs4_sequence_done(struct rpc_task *task,
 530                                struct nfs4_sequence_res *res)
 531 {
 532         if (res->sr_slot == NULL)
 533                 return 1;
 534         return nfs41_sequence_done(task, res);
 535 }
 536
 537 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 538                 struct nfs4_sequence_res *res, int cache_reply)
 539 {
 540         args->sa_slot = NULL;
 541         args->sa_cache_this = 0;
 542         args->sa_privileged = 0;
 543         if (cache_reply)
 544                 args->sa_cache_this = 1;
 545         res->sr_slot = NULL;
 546 }
 547
 548 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 549 {
 550         args->sa_privileged = 1;
 551 }
 552
 553 int nfs41_setup_sequence(struct nfs4_session *session,
 554                                 struct nfs4_sequence_args *args,
 555                                 struct nfs4_sequence_res *res,
 556                                 struct rpc_task *task)
 557 {
 558         struct nfs4_slot *slot;
 559         struct nfs4_slot_table *tbl;
 560
 561         dprintk("--> %s\n", __func__);
 562         /* slot already allocated? */
 563         if (res->sr_slot != NULL)
 564                 goto out_success;
 565
 566         tbl = &session->fc_slot_table;
 567
 568         task->tk_timeout = 0;
 569
 570         spin_lock(&tbl->slot_tbl_lock);
 571         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 572             !args->sa_privileged) {
 573                 /* The state manager will wait until the slot table is empty */
 574                 dprintk("%s session is draining\n", __func__);
 575                 goto out_sleep;
 576         }
 577
 578         slot = nfs4_alloc_slot(tbl);
 579         if (IS_ERR(slot)) {
 580                 /* If out of memory, try again in 1/4 second */
 581                 if (slot == ERR_PTR(-ENOMEM))
 582                         task->tk_timeout = HZ >> 2;
 583                 dprintk("<-- %s: no free slots\n", __func__);
 584                 goto out_sleep;
 585         }
 586         spin_unlock(&tbl->slot_tbl_lock);
 587
 588         args->sa_slot = slot;
 589
 590         dprintk("<-- %s slotid=%d seqid=%d\n", __func__,
 591                         slot->slot_nr, slot->seq_nr);
 592
 593         res->sr_slot = slot;
 594         res->sr_timestamp = jiffies;
 595         res->sr_status_flags = 0;
 596         /*
 597          * sr_status is only set in decode_sequence, and so will remain
 598          * set to 1 if an rpc level failure occurs.
 599          */
 600         res->sr_status = 1;
 601 out_success:
 602         rpc_call_start(task);
 603         return 0;
 604 out_sleep:
 605         /* Privileged tasks are queued with top priority */
 606         if (args->sa_privileged)
 607                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 608                                 NULL, RPC_PRIORITY_PRIVILEGED);
 609         else
 610                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 611         spin_unlock(&tbl->slot_tbl_lock);
 612         return -EAGAIN;
 613 }
 614 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 615
 616 int nfs4_setup_sequence(const struct nfs_server *server,
 617                         struct nfs4_sequence_args *args,
 618                         struct nfs4_sequence_res *res,
 619                         struct rpc_task *task)
 620 {
 621         struct nfs4_session *session = nfs4_get_session(server);
 622         int ret = 0;
 623
 624         if (session == NULL) {
 625                 rpc_call_start(task);
 626                 goto out;
 627         }
 628
 629         dprintk("--> %s clp %p session %p sr_slot %d\n",
 630                 __func__, session->clp, session, res->sr_slot ?
 631                         res->sr_slot->slot_nr : -1);
 632
 633         ret = nfs41_setup_sequence(session, args, res, task);
 634 out:
 635         dprintk("<-- %s status=%d\n", __func__, ret);
 636         return ret;
 637 }
 638
 639 struct nfs41_call_sync_data {
 640         const struct nfs_server *seq_server;
 641         struct nfs4_sequence_args *seq_args;
 642         struct nfs4_sequence_res *seq_res;
 643 };
 644
 645 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 646 {
 647         struct nfs41_call_sync_data *data = calldata;
 648         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 649
 650         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 651
 652         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 653 }
 654
 655 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 656 {
 657         struct nfs41_call_sync_data *data = calldata;
 658
 659         nfs41_sequence_done(task, data->seq_res);
 660 }
 661
 662 static const struct rpc_call_ops nfs41_call_sync_ops = {
 663         .rpc_call_prepare = nfs41_call_sync_prepare,
 664         .rpc_call_done = nfs41_call_sync_done,
 665 };
 666
 667 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 668                                    struct nfs_server *server,
 669                                    struct rpc_message *msg,
 670                                    struct nfs4_sequence_args *args,
 671                                    struct nfs4_sequence_res *res)
 672 {
 673         int ret;
 674         struct rpc_task *task;
 675         struct nfs41_call_sync_data data = {
 676                 .seq_server = server,
 677                 .seq_args = args,
 678                 .seq_res = res,
 679         };
 680         struct rpc_task_setup task_setup = {
 681                 .rpc_client = clnt,
 682                 .rpc_message = msg,
 683                 .callback_ops = &nfs41_call_sync_ops,
 684                 .callback_data = &data
 685         };
 686
 687         task = rpc_run_task(&task_setup);
 688         if (IS_ERR(task))
 689                 ret = PTR_ERR(task);
 690         else {
 691                 ret = task->tk_status;
 692                 rpc_put_task(task);
 693         }
 694         return ret;
 695 }
 696
 697 #else
 698 static
 699 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 700                 struct nfs4_sequence_res *res, int cache_reply)
 701 {
 702 }
 703
 704 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 705 {
 706 }
 707
 708
 709 static int nfs4_sequence_done(struct rpc_task *task,
 710                                struct nfs4_sequence_res *res)
 711 {
 712         return 1;
 713 }
 714 #endif /* CONFIG_NFS_V4_1 */
 715
 716 static
 717 int _nfs4_call_sync(struct rpc_clnt *clnt,
 718                     struct nfs_server *server,
 719                     struct rpc_message *msg,
 720                     struct nfs4_sequence_args *args,
 721                     struct nfs4_sequence_res *res)
 722 {
 723         return rpc_call_sync(clnt, msg, 0);
 724 }
 725
 726 static
 727 int nfs4_call_sync(struct rpc_clnt *clnt,
 728                    struct nfs_server *server,
 729                    struct rpc_message *msg,
 730                    struct nfs4_sequence_args *args,
 731                    struct nfs4_sequence_res *res,
 732                    int cache_reply)
 733 {
 734         nfs41_init_sequence(args, res, cache_reply);
 735         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 736                                                 args, res);
 737 }
 738
 739 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 740 {
 741         struct nfs_inode *nfsi = NFS_I(dir);
 742
 743         spin_lock(&dir->i_lock);
 744         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 745         if (!cinfo->atomic || cinfo->before != dir->i_version)
 746                 nfs_force_lookup_revalidate(dir);
 747         dir->i_version = cinfo->after;
 748         nfs_fscache_invalidate(dir);
 749         spin_unlock(&dir->i_lock);
 750 }
 751
 752 struct nfs4_opendata {
 753         struct kref kref;
 754         struct nfs_openargs o_arg;
 755         struct nfs_openres o_res;
 756         struct nfs_open_confirmargs c_arg;
 757         struct nfs_open_confirmres c_res;
 758         struct nfs4_string owner_name;
 759         struct nfs4_string group_name;
 760         struct nfs_fattr f_attr;
 761         struct dentry *dir;
 762         struct dentry *dentry;
 763         struct nfs4_state_owner *owner;
 764         struct nfs4_state *state;
 765         struct iattr attrs;
 766         unsigned long timestamp;
 767         unsigned int rpc_done : 1;
 768         int rpc_status;
 769         int cancelled;
 770 };
 771
 772 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 773                 int err, struct nfs4_exception *exception)
 774 {
 775         if (err != -EINVAL)
 776                 return false;
 777         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 778                 return false;
 779         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 780         exception->retry = 1;
 781         return true;
 782 }
 783
 784 static enum open_claim_type4
 785 nfs4_map_atomic_open_claim(struct nfs_server *server,
 786                 enum open_claim_type4 claim)
 787 {
 788         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 789                 return claim;
 790         switch (claim) {
 791         default:
 792                 return claim;
 793         case NFS4_OPEN_CLAIM_FH:
 794                 return NFS4_OPEN_CLAIM_NULL;
 795         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 796                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 797         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 798                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 799         }
 800 }
 801
 802 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 803 {
 804         p->o_res.f_attr = &p->f_attr;
 805         p->o_res.seqid = p->o_arg.seqid;
 806         p->c_res.seqid = p->c_arg.seqid;
 807         p->o_res.server = p->o_arg.server;
 808         p->o_res.access_request = p->o_arg.access;
 809         nfs_fattr_init(&p->f_attr);
 810         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 811 }
 812
 813 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 814                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 815                 const struct iattr *attrs,
 816                 enum open_claim_type4 claim,
 817                 gfp_t gfp_mask)
 818 {
 819         struct dentry *parent = dget_parent(dentry);
 820         struct inode *dir = parent->d_inode;
 821         struct nfs_server *server = NFS_SERVER(dir);
 822         struct nfs4_opendata *p;
 823
 824         p = kzalloc(sizeof(*p), gfp_mask);
 825         if (p == NULL)
 826                 goto err;
 827         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 828         if (p->o_arg.seqid == NULL)
 829                 goto err_free;
 830         nfs_sb_active(dentry->d_sb);
 831         p->dentry = dget(dentry);
 832         p->dir = parent;
 833         p->owner = sp;
 834         atomic_inc(&sp->so_count);
 835         p->o_arg.open_flags = flags;
 836         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 837         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
 838          * will return permission denied for all bits until close */
 839         if (!(flags & O_EXCL)) {
 840                 /* ask server to check for all possible rights as results
 841                  * are cached */
 842                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
 843                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
 844         }
 845         p->o_arg.clientid = server->nfs_client->cl_clientid;
 846         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
 847         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 848         p->o_arg.name = &dentry->d_name;
 849         p->o_arg.server = server;
 850         p->o_arg.bitmask = server->attr_bitmask;
 851         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
 852         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
 853         switch (p->o_arg.claim) {
 854         case NFS4_OPEN_CLAIM_NULL:
 855         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
 856         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
 857                 p->o_arg.fh = NFS_FH(dir);
 858                 break;
 859         case NFS4_OPEN_CLAIM_PREVIOUS:
 860         case NFS4_OPEN_CLAIM_FH:
 861         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 862         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 863                 p->o_arg.fh = NFS_FH(dentry->d_inode);
 864         }
 865         if (attrs != NULL && attrs->ia_valid != 0) {
 866                 __be32 verf[2];
 867
 868                 p->o_arg.u.attrs = &p->attrs;
 869                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 870
 871                 verf[0] = jiffies;
 872                 verf[1] = current->pid;
 873                 memcpy(p->o_arg.u.verifier.data, verf,
 874                                 sizeof(p->o_arg.u.verifier.data));
 875         }
 876         p->c_arg.fh = &p->o_res.fh;
 877         p->c_arg.stateid = &p->o_res.stateid;
 878         p->c_arg.seqid = p->o_arg.seqid;
 879         nfs4_init_opendata_res(p);
 880         kref_init(&p->kref);
 881         return p;
 882 err_free:
 883         kfree(p);
 884 err:
 885         dput(parent);
 886         return NULL;
 887 }
 888
 889 static void nfs4_opendata_free(struct kref *kref)
 890 {
 891         struct nfs4_opendata *p = container_of(kref,
 892                         struct nfs4_opendata, kref);
 893         struct super_block *sb = p->dentry->d_sb;
 894
 895         nfs_free_seqid(p->o_arg.seqid);
 896         if (p->state != NULL)
 897                 nfs4_put_open_state(p->state);
 898         nfs4_put_state_owner(p->owner);
 899         dput(p->dir);
 900         dput(p->dentry);
 901         nfs_sb_deactive(sb);
 902         nfs_fattr_free_names(&p->f_attr);
 903         kfree(p);
 904 }
 905
 906 static void nfs4_opendata_put(struct nfs4_opendata *p)
 907 {
 908         if (p != NULL)
 909                 kref_put(&p->kref, nfs4_opendata_free);
 910 }
 911
 912 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 913 {
 914         int ret;
 915
 916         ret = rpc_wait_for_completion_task(task);
 917         return ret;
 918 }
 919
 920 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 921 {
 922         int ret = 0;
 923
 924         if (open_mode & (O_EXCL|O_TRUNC))
 925                 goto out;
 926         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 927                 case FMODE_READ:
 928                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 929                                 && state->n_rdonly != 0;
 930                         break;
 931                 case FMODE_WRITE:
 932                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 933                                 && state->n_wronly != 0;
 934                         break;
 935                 case FMODE_READ|FMODE_WRITE:
 936                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 937                                 && state->n_rdwr != 0;
 938         }
 939 out:
 940         return ret;
 941 }
 942
 943 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 944 {
 945         if (delegation == NULL)
 946                 return 0;
 947         if ((delegation->type & fmode) != fmode)
 948                 return 0;
 949         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 950                 return 0;
 951         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
 952                 return 0;
 953         nfs_mark_delegation_referenced(delegation);
 954         return 1;
 955 }
 956
 957 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 958 {
 959         switch (fmode) {
 960                 case FMODE_WRITE:
 961                         state->n_wronly++;
 962                         break;
 963                 case FMODE_READ:
 964                         state->n_rdonly++;
 965                         break;
 966                 case FMODE_READ|FMODE_WRITE:
 967                         state->n_rdwr++;
 968         }
 969         nfs4_state_set_mode_locked(state, state->state | fmode);
 970 }
 971
 972 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 973 {
 974         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 975                 nfs4_stateid_copy(&state->stateid, stateid);
 976         nfs4_stateid_copy(&state->open_stateid, stateid);
 977         switch (fmode) {
 978                 case FMODE_READ:
 979                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 980                         break;
 981                 case FMODE_WRITE:
 982                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 983                         break;
 984                 case FMODE_READ|FMODE_WRITE:
 985                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 986         }
 987 }
 988
 989 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 990 {
 991         write_seqlock(&state->seqlock);
 992         nfs_set_open_stateid_locked(state, stateid, fmode);
 993         write_sequnlock(&state->seqlock);
 994 }
 995
 996 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 997 {
 998         /*
 999          * Protect the call to nfs4_state_set_mode_locked and
1000          * serialise the stateid update
1001          */
1002         write_seqlock(&state->seqlock);
1003         if (deleg_stateid != NULL) {
1004                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1005                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1006         }
1007         if (open_stateid != NULL)
1008                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1009         write_sequnlock(&state->seqlock);
1010         spin_lock(&state->owner->so_lock);
1011         update_open_stateflags(state, fmode);
1012         spin_unlock(&state->owner->so_lock);
1013 }
1014
1015 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1016 {
1017         struct nfs_inode *nfsi = NFS_I(state->inode);
1018         struct nfs_delegation *deleg_cur;
1019         int ret = 0;
1020
1021         fmode &= (FMODE_READ|FMODE_WRITE);
1022
1023         rcu_read_lock();
1024         deleg_cur = rcu_dereference(nfsi->delegation);
1025         if (deleg_cur == NULL)
1026                 goto no_delegation;
1027
1028         spin_lock(&deleg_cur->lock);
1029         if (nfsi->delegation != deleg_cur ||
1030            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1031             (deleg_cur->type & fmode) != fmode)
1032                 goto no_delegation_unlock;
1033
1034         if (delegation == NULL)
1035                 delegation = &deleg_cur->stateid;
1036         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1037                 goto no_delegation_unlock;
1038
1039         nfs_mark_delegation_referenced(deleg_cur);
1040         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1041         ret = 1;
1042 no_delegation_unlock:
1043         spin_unlock(&deleg_cur->lock);
1044 no_delegation:
1045         rcu_read_unlock();
1046
1047         if (!ret && open_stateid != NULL) {
1048                 __update_open_stateid(state, open_stateid, NULL, fmode);
1049                 ret = 1;
1050         }
1051
1052         return ret;
1053 }
1054
1055
1056 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1057 {
1058         struct nfs_delegation *delegation;
1059
1060         rcu_read_lock();
1061         delegation = rcu_dereference(NFS_I(inode)->delegation);
1062         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1063                 rcu_read_unlock();
1064                 return;
1065         }
1066         rcu_read_unlock();
1067         nfs4_inode_return_delegation(inode);
1068 }
1069
1070 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1071 {
1072         struct nfs4_state *state = opendata->state;
1073         struct nfs_inode *nfsi = NFS_I(state->inode);
1074         struct nfs_delegation *delegation;
1075         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1076         fmode_t fmode = opendata->o_arg.fmode;
1077         nfs4_stateid stateid;
1078         int ret = -EAGAIN;
1079
1080         for (;;) {
1081                 if (can_open_cached(state, fmode, open_mode)) {
1082                         spin_lock(&state->owner->so_lock);
1083                         if (can_open_cached(state, fmode, open_mode)) {
1084                                 update_open_stateflags(state, fmode);
1085                                 spin_unlock(&state->owner->so_lock);
1086                                 goto out_return_state;
1087                         }
1088                         spin_unlock(&state->owner->so_lock);
1089                 }
1090                 rcu_read_lock();
1091                 delegation = rcu_dereference(nfsi->delegation);
1092                 if (!can_open_delegated(delegation, fmode)) {
1093                         rcu_read_unlock();
1094                         break;
1095                 }
1096                 /* Save the delegation */
1097                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1098                 rcu_read_unlock();
1099                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1100                 if (ret != 0)
1101                         goto out;
1102                 ret = -EAGAIN;
1103
1104                 /* Try to update the stateid using the delegation */
1105                 if (update_open_stateid(state, NULL, &stateid, fmode))
1106                         goto out_return_state;
1107         }
1108 out:
1109         return ERR_PTR(ret);
1110 out_return_state:
1111         atomic_inc(&state->count);
1112         return state;
1113 }
1114
1115 static void
1116 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1117 {
1118         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1119         struct nfs_delegation *delegation;
1120         int delegation_flags = 0;
1121
1122         rcu_read_lock();
1123         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1124         if (delegation)
1125                 delegation_flags = delegation->flags;
1126         rcu_read_unlock();
1127         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1128                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1129                                    "returning a delegation for "
1130                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1131                                    clp->cl_hostname);
1132         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1133                 nfs_inode_set_delegation(state->inode,
1134                                          data->owner->so_cred,
1135                                          &data->o_res);
1136         else
1137                 nfs_inode_reclaim_delegation(state->inode,
1138                                              data->owner->so_cred,
1139                                              &data->o_res);
1140 }
1141
1142 /*
1143  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1144  * and update the nfs4_state.
1145  */
1146 static struct nfs4_state *
1147 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1148 {
1149         struct inode *inode = data->state->inode;
1150         struct nfs4_state *state = data->state;
1151         int ret;
1152
1153         if (!data->rpc_done) {
1154                 ret = data->rpc_status;
1155                 goto err;
1156         }
1157
1158         ret = -ESTALE;
1159         if (!(data->f_attr.valid & NFS_ATTR_FATTR_TYPE) ||
1160             !(data->f_attr.valid & NFS_ATTR_FATTR_FILEID) ||
1161             !(data->f_attr.valid & NFS_ATTR_FATTR_CHANGE))
1162                 goto err;
1163
1164         ret = -ENOMEM;
1165         state = nfs4_get_open_state(inode, data->owner);
1166         if (state == NULL)
1167                 goto err;
1168
1169         ret = nfs_refresh_inode(inode, &data->f_attr);
1170         if (ret)
1171                 goto err;
1172
1173         if (data->o_res.delegation_type != 0)
1174                 nfs4_opendata_check_deleg(data, state);
1175         update_open_stateid(state, &data->o_res.stateid, NULL,
1176                             data->o_arg.fmode);
1177
1178         return state;
1179 err:
1180         return ERR_PTR(ret);
1181
1182 }
1183
1184 static struct nfs4_state *
1185 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1186 {
1187         struct inode *inode;
1188         struct nfs4_state *state = NULL;
1189         int ret;
1190
1191         if (!data->rpc_done) {
1192                 state = nfs4_try_open_cached(data);
1193                 goto out;
1194         }
1195
1196         ret = -EAGAIN;
1197         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1198                 goto err;
1199         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1200         ret = PTR_ERR(inode);
1201         if (IS_ERR(inode))
1202                 goto err;
1203         ret = -ENOMEM;
1204         state = nfs4_get_open_state(inode, data->owner);
1205         if (state == NULL)
1206                 goto err_put_inode;
1207         if (data->o_res.delegation_type != 0)
1208                 nfs4_opendata_check_deleg(data, state);
1209         update_open_stateid(state, &data->o_res.stateid, NULL,
1210                         data->o_arg.fmode);
1211         iput(inode);
1212 out:
1213         nfs_release_seqid(data->o_arg.seqid);
1214         return state;
1215 err_put_inode:
1216         iput(inode);
1217 err:
1218         return ERR_PTR(ret);
1219 }
1220
1221 static struct nfs4_state *
1222 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1223 {
1224         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1225                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1226         return _nfs4_opendata_to_nfs4_state(data);
1227 }
1228
1229 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1230 {
1231         struct nfs_inode *nfsi = NFS_I(state->inode);
1232         struct nfs_open_context *ctx;
1233
1234         spin_lock(&state->inode->i_lock);
1235         list_for_each_entry(ctx, &nfsi->open_files, list) {
1236                 if (ctx->state != state)
1237                         continue;
1238                 get_nfs_open_context(ctx);
1239                 spin_unlock(&state->inode->i_lock);
1240                 return ctx;
1241         }
1242         spin_unlock(&state->inode->i_lock);
1243         return ERR_PTR(-ENOENT);
1244 }
1245
1246 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1247                 struct nfs4_state *state, enum open_claim_type4 claim)
1248 {
1249         struct nfs4_opendata *opendata;
1250
1251         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1252                         NULL, claim, GFP_NOFS);
1253         if (opendata == NULL)
1254                 return ERR_PTR(-ENOMEM);
1255         opendata->state = state;
1256         atomic_inc(&state->count);
1257         return opendata;
1258 }
1259
1260 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1261 {
1262         struct nfs4_state *newstate;
1263         int ret;
1264
1265         opendata->o_arg.open_flags = 0;
1266         opendata->o_arg.fmode = fmode;
1267         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1268         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1269         nfs4_init_opendata_res(opendata);
1270         ret = _nfs4_recover_proc_open(opendata);
1271         if (ret != 0)
1272                 return ret;
1273         newstate = nfs4_opendata_to_nfs4_state(opendata);
1274         if (IS_ERR(newstate))
1275                 return PTR_ERR(newstate);
1276         nfs4_close_state(newstate, fmode);
1277         *res = newstate;
1278         return 0;
1279 }
1280
1281 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1282 {
1283         struct nfs4_state *newstate;
1284         int ret;
1285
1286         /* memory barrier prior to reading state->n_* */
1287         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1288         smp_rmb();
1289         if (state->n_rdwr != 0) {
1290                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1291                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1292                 if (ret != 0)
1293                         return ret;
1294                 if (newstate != state)
1295                         return -ESTALE;
1296         }
1297         if (state->n_wronly != 0) {
1298                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1299                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1300                 if (ret != 0)
1301                         return ret;
1302                 if (newstate != state)
1303                         return -ESTALE;
1304         }
1305         if (state->n_rdonly != 0) {
1306                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1307                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1308                 if (ret != 0)
1309                         return ret;
1310                 if (newstate != state)
1311                         return -ESTALE;
1312         }
1313         /*
1314          * We may have performed cached opens for all three recoveries.
1315          * Check if we need to update the current stateid.
1316          */
1317         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1318             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1319                 write_seqlock(&state->seqlock);
1320                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1321                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1322                 write_sequnlock(&state->seqlock);
1323         }
1324         return 0;
1325 }
1326
1327 /*
1328  * OPEN_RECLAIM:
1329  *      reclaim state on the server after a reboot.
1330  */
1331 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1332 {
1333         struct nfs_delegation *delegation;
1334         struct nfs4_opendata *opendata;
1335         fmode_t delegation_type = 0;
1336         int status;
1337
1338         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1339                         NFS4_OPEN_CLAIM_PREVIOUS);
1340         if (IS_ERR(opendata))
1341                 return PTR_ERR(opendata);
1342         rcu_read_lock();
1343         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1344         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1345                 delegation_type = delegation->type;
1346         rcu_read_unlock();
1347         opendata->o_arg.u.delegation_type = delegation_type;
1348         status = nfs4_open_recover(opendata, state);
1349         nfs4_opendata_put(opendata);
1350         return status;
1351 }
1352
1353 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1354 {
1355         struct nfs_server *server = NFS_SERVER(state->inode);
1356         struct nfs4_exception exception = { };
1357         int err;
1358         do {
1359                 err = _nfs4_do_open_reclaim(ctx, state);
1360                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1361                         continue;
1362                 if (err != -NFS4ERR_DELAY)
1363                         break;
1364                 nfs4_handle_exception(server, err, &exception);
1365         } while (exception.retry);
1366         return err;
1367 }
1368
1369 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1370 {
1371         struct nfs_open_context *ctx;
1372         int ret;
1373
1374         ctx = nfs4_state_find_open_context(state);
1375         if (IS_ERR(ctx))
1376                 return -EAGAIN;
1377         ret = nfs4_do_open_reclaim(ctx, state);
1378         put_nfs_open_context(ctx);
1379         return ret;
1380 }
1381
1382 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1383 {
1384         struct nfs4_opendata *opendata;
1385         int ret;
1386
1387         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1388                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1389         if (IS_ERR(opendata))
1390                 return PTR_ERR(opendata);
1391         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1392         ret = nfs4_open_recover(opendata, state);
1393         nfs4_opendata_put(opendata);
1394         return ret;
1395 }
1396
1397 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1398 {
1399         struct nfs4_exception exception = { };
1400         struct nfs_server *server = NFS_SERVER(state->inode);
1401         int err;
1402         do {
1403                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1404                 switch (err) {
1405                         case 0:
1406                         case -ENOENT:
1407                         case -ESTALE:
1408                                 goto out;
1409                         case -NFS4ERR_BADSESSION:
1410                         case -NFS4ERR_BADSLOT:
1411                         case -NFS4ERR_BAD_HIGH_SLOT:
1412                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1413                         case -NFS4ERR_DEADSESSION:
1414                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1415                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1416                                 err = -EAGAIN;
1417                                 goto out;
1418                         case -NFS4ERR_STALE_CLIENTID:
1419                         case -NFS4ERR_STALE_STATEID:
1420                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1421                         case -NFS4ERR_EXPIRED:
1422                                 /* Don't recall a delegation if it was lost */
1423                                 nfs4_schedule_lease_recovery(server->nfs_client);
1424                                 err = -EAGAIN;
1425                                 goto out;
1426                         case -NFS4ERR_DELEG_REVOKED:
1427                         case -NFS4ERR_ADMIN_REVOKED:
1428                         case -NFS4ERR_BAD_STATEID:
1429                                 nfs_inode_find_state_and_recover(state->inode,
1430                                                 stateid);
1431                                 nfs4_schedule_stateid_recovery(server, state);
1432                         case -ENOMEM:
1433                                 err = 0;
1434                                 goto out;
1435                         case -NFS4ERR_DELAY:
1436                         case -NFS4ERR_GRACE:
1437                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1438                                 ssleep(1);
1439                                 err = -EAGAIN;
1440                                 goto out;
1441                 }
1442                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1443                 err = nfs4_handle_exception(server, err, &exception);
1444         } while (exception.retry);
1445 out:
1446         return err;
1447 }
1448
1449 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1450 {
1451         struct nfs4_opendata *data = calldata;
1452
1453         data->rpc_status = task->tk_status;
1454         if (data->rpc_status == 0) {
1455                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1456                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1457                 renew_lease(data->o_res.server, data->timestamp);
1458                 data->rpc_done = 1;
1459         }
1460 }
1461
1462 static void nfs4_open_confirm_release(void *calldata)
1463 {
1464         struct nfs4_opendata *data = calldata;
1465         struct nfs4_state *state = NULL;
1466
1467         /* If this request hasn't been cancelled, do nothing */
1468         if (data->cancelled == 0)
1469                 goto out_free;
1470         /* In case of error, no cleanup! */
1471         if (!data->rpc_done)
1472                 goto out_free;
1473         state = nfs4_opendata_to_nfs4_state(data);
1474         if (!IS_ERR(state))
1475                 nfs4_close_state(state, data->o_arg.fmode);
1476 out_free:
1477         nfs4_opendata_put(data);
1478 }
1479
1480 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1481         .rpc_call_done = nfs4_open_confirm_done,
1482         .rpc_release = nfs4_open_confirm_release,
1483 };
1484
1485 /*
1486  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1487  */
1488 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1489 {
1490         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1491         struct rpc_task *task;
1492         struct  rpc_message msg = {
1493                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1494                 .rpc_argp = &data->c_arg,
1495                 .rpc_resp = &data->c_res,
1496                 .rpc_cred = data->owner->so_cred,
1497         };
1498         struct rpc_task_setup task_setup_data = {
1499                 .rpc_client = server->client,
1500                 .rpc_message = &msg,
1501                 .callback_ops = &nfs4_open_confirm_ops,
1502                 .callback_data = data,
1503                 .workqueue = nfsiod_workqueue,
1504                 .flags = RPC_TASK_ASYNC,
1505         };
1506         int status;
1507
1508         kref_get(&data->kref);
1509         data->rpc_done = 0;
1510         data->rpc_status = 0;
1511         data->timestamp = jiffies;
1512         task = rpc_run_task(&task_setup_data);
1513         if (IS_ERR(task))
1514                 return PTR_ERR(task);
1515         status = nfs4_wait_for_completion_rpc_task(task);
1516         if (status != 0) {
1517                 data->cancelled = 1;
1518                 smp_wmb();
1519         } else
1520                 status = data->rpc_status;
1521         rpc_put_task(task);
1522         return status;
1523 }
1524
1525 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1526 {
1527         struct nfs4_opendata *data = calldata;
1528         struct nfs4_state_owner *sp = data->owner;
1529
1530         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1531                 goto out_wait;
1532         /*
1533          * Check if we still need to send an OPEN call, or if we can use
1534          * a delegation instead.
1535          */
1536         if (data->state != NULL) {
1537                 struct nfs_delegation *delegation;
1538
1539                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1540                         goto out_no_action;
1541                 rcu_read_lock();
1542                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1543                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1544                     can_open_delegated(delegation, data->o_arg.fmode))
1545                         goto unlock_no_action;
1546                 rcu_read_unlock();
1547         }
1548         /* Update client id. */
1549         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1550         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1551                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1552                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1553                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1554         }
1555         data->timestamp = jiffies;
1556         if (nfs4_setup_sequence(data->o_arg.server,
1557                                 &data->o_arg.seq_args,
1558                                 &data->o_res.seq_res,
1559                                 task) != 0)
1560                 nfs_release_seqid(data->o_arg.seqid);
1561         return;
1562 unlock_no_action:
1563         rcu_read_unlock();
1564 out_no_action:
1565         task->tk_action = NULL;
1566 out_wait:
1567         nfs4_sequence_done(task, &data->o_res.seq_res);
1568 }
1569
1570 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1571 {
1572         struct nfs4_opendata *data = calldata;
1573
1574         data->rpc_status = task->tk_status;
1575
1576         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1577                 return;
1578
1579         if (task->tk_status == 0) {
1580                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1581                         switch (data->o_res.f_attr->mode & S_IFMT) {
1582                         case S_IFREG:
1583                                 break;
1584                         case S_IFLNK:
1585                                 data->rpc_status = -ELOOP;
1586                                 break;
1587                         case S_IFDIR:
1588                                 data->rpc_status = -EISDIR;
1589                                 break;
1590                         default:
1591                                 data->rpc_status = -ENOTDIR;
1592                         }
1593                 }
1594                 renew_lease(data->o_res.server, data->timestamp);
1595                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1596                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1597         }
1598         data->rpc_done = 1;
1599 }
1600
1601 static void nfs4_open_release(void *calldata)
1602 {
1603         struct nfs4_opendata *data = calldata;
1604         struct nfs4_state *state = NULL;
1605
1606         /* If this request hasn't been cancelled, do nothing */
1607         if (data->cancelled == 0)
1608                 goto out_free;
1609         /* In case of error, no cleanup! */
1610         if (data->rpc_status != 0 || !data->rpc_done)
1611                 goto out_free;
1612         /* In case we need an open_confirm, no cleanup! */
1613         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1614                 goto out_free;
1615         state = nfs4_opendata_to_nfs4_state(data);
1616         if (!IS_ERR(state))
1617                 nfs4_close_state(state, data->o_arg.fmode);
1618 out_free:
1619         nfs4_opendata_put(data);
1620 }
1621
1622 static const struct rpc_call_ops nfs4_open_ops = {
1623         .rpc_call_prepare = nfs4_open_prepare,
1624         .rpc_call_done = nfs4_open_done,
1625         .rpc_release = nfs4_open_release,
1626 };
1627
1628 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1629 {
1630         struct inode *dir = data->dir->d_inode;
1631         struct nfs_server *server = NFS_SERVER(dir);
1632         struct nfs_openargs *o_arg = &data->o_arg;
1633         struct nfs_openres *o_res = &data->o_res;
1634         struct rpc_task *task;
1635         struct rpc_message msg = {
1636                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1637                 .rpc_argp = o_arg,
1638                 .rpc_resp = o_res,
1639                 .rpc_cred = data->owner->so_cred,
1640         };
1641         struct rpc_task_setup task_setup_data = {
1642                 .rpc_client = server->client,
1643                 .rpc_message = &msg,
1644                 .callback_ops = &nfs4_open_ops,
1645                 .callback_data = data,
1646                 .workqueue = nfsiod_workqueue,
1647                 .flags = RPC_TASK_ASYNC,
1648         };
1649         int status;
1650
1651         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1652         kref_get(&data->kref);
1653         data->rpc_done = 0;
1654         data->rpc_status = 0;
1655         data->cancelled = 0;
1656         if (isrecover)
1657                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1658         task = rpc_run_task(&task_setup_data);
1659         if (IS_ERR(task))
1660                 return PTR_ERR(task);
1661         status = nfs4_wait_for_completion_rpc_task(task);
1662         if (status != 0) {
1663                 data->cancelled = 1;
1664                 smp_wmb();
1665         } else
1666                 status = data->rpc_status;
1667         rpc_put_task(task);
1668
1669         return status;
1670 }
1671
1672 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1673 {
1674         struct inode *dir = data->dir->d_inode;
1675         struct nfs_openres *o_res = &data->o_res;
1676         int status;
1677
1678         status = nfs4_run_open_task(data, 1);
1679         if (status != 0 || !data->rpc_done)
1680                 return status;
1681
1682         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1683
1684         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1685                 status = _nfs4_proc_open_confirm(data);
1686                 if (status != 0)
1687                         return status;
1688         }
1689
1690         return status;
1691 }
1692
1693 static int nfs4_opendata_access(struct rpc_cred *cred,
1694                                 struct nfs4_opendata *opendata,
1695                                 struct nfs4_state *state, fmode_t fmode,
1696                                 int openflags)
1697 {
1698         struct nfs_access_entry cache;
1699         u32 mask;
1700
1701         /* access call failed or for some reason the server doesn't
1702          * support any access modes -- defer access call until later */
1703         if (opendata->o_res.access_supported == 0)
1704                 return 0;
1705
1706         mask = 0;
1707         /* don't check MAY_WRITE - a newly created file may not have
1708          * write mode bits, but POSIX allows the creating process to write.
1709          * use openflags to check for exec, because fmode won't
1710          * always have FMODE_EXEC set when file open for exec. */
1711         if (openflags & __FMODE_EXEC) {
1712                 /* ONLY check for exec rights */
1713                 mask = MAY_EXEC;
1714         } else if (fmode & FMODE_READ)
1715                 mask = MAY_READ;
1716
1717         cache.cred = cred;
1718         cache.jiffies = jiffies;
1719         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1720         nfs_access_add_cache(state->inode, &cache);
1721
1722         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1723                 return 0;
1724
1725         /* even though OPEN succeeded, access is denied. Close the file */
1726         nfs4_close_state(state, fmode);
1727         return -EACCES;
1728 }
1729
1730 /*
1731  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1732  */
1733 static int _nfs4_proc_open(struct nfs4_opendata *data)
1734 {
1735         struct inode *dir = data->dir->d_inode;
1736         struct nfs_server *server = NFS_SERVER(dir);
1737         struct nfs_openargs *o_arg = &data->o_arg;
1738         struct nfs_openres *o_res = &data->o_res;
1739         int status;
1740
1741         status = nfs4_run_open_task(data, 0);
1742         if (!data->rpc_done)
1743                 return status;
1744         if (status != 0) {
1745                 if (status == -NFS4ERR_BADNAME &&
1746                                 !(o_arg->open_flags & O_CREAT))
1747                         return -ENOENT;
1748                 return status;
1749         }
1750
1751         nfs_fattr_map_and_free_names(server, &data->f_attr);
1752
1753         if (o_arg->open_flags & O_CREAT)
1754                 update_changeattr(dir, &o_res->cinfo);
1755         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1756                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1757         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1758                 status = _nfs4_proc_open_confirm(data);
1759                 if (status != 0)
1760                         return status;
1761         }
1762         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1763                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1764         return 0;
1765 }
1766
1767 static int nfs4_recover_expired_lease(struct nfs_server *server)
1768 {
1769         return nfs4_client_recover_expired_lease(server->nfs_client);
1770 }
1771
1772 /*
1773  * OPEN_EXPIRED:
1774  *      reclaim state on the server after a network partition.
1775  *      Assumes caller holds the appropriate lock
1776  */
1777 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1778 {
1779         struct nfs4_opendata *opendata;
1780         int ret;
1781
1782         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1783                         NFS4_OPEN_CLAIM_FH);
1784         if (IS_ERR(opendata))
1785                 return PTR_ERR(opendata);
1786         ret = nfs4_open_recover(opendata, state);
1787         if (ret == -ESTALE)
1788                 d_drop(ctx->dentry);
1789         nfs4_opendata_put(opendata);
1790         return ret;
1791 }
1792
1793 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1794 {
1795         struct nfs_server *server = NFS_SERVER(state->inode);
1796         struct nfs4_exception exception = { };
1797         int err;
1798
1799         do {
1800                 err = _nfs4_open_expired(ctx, state);
1801                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1802                         continue;
1803                 switch (err) {
1804                 default:
1805                         goto out;
1806                 case -NFS4ERR_GRACE:
1807                 case -NFS4ERR_DELAY:
1808                         nfs4_handle_exception(server, err, &exception);
1809                         err = 0;
1810                 }
1811         } while (exception.retry);
1812 out:
1813         return err;
1814 }
1815
1816 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1817 {
1818         struct nfs_open_context *ctx;
1819         int ret;
1820
1821         ctx = nfs4_state_find_open_context(state);
1822         if (IS_ERR(ctx))
1823                 return -EAGAIN;
1824         ret = nfs4_do_open_expired(ctx, state);
1825         put_nfs_open_context(ctx);
1826         return ret;
1827 }
1828
1829 #if defined(CONFIG_NFS_V4_1)
1830 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1831 {
1832         struct nfs_server *server = NFS_SERVER(state->inode);
1833         nfs4_stateid *stateid = &state->stateid;
1834         int status;
1835
1836         /* If a state reset has been done, test_stateid is unneeded */
1837         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1838                 return;
1839
1840         status = nfs41_test_stateid(server, stateid);
1841         if (status != NFS_OK) {
1842                 /* Free the stateid unless the server explicitly
1843                  * informs us the stateid is unrecognized. */
1844                 if (status != -NFS4ERR_BAD_STATEID)
1845                         nfs41_free_stateid(server, stateid);
1846                 nfs_remove_bad_delegation(state->inode);
1847
1848                 write_seqlock(&state->seqlock);
1849                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1850                 write_sequnlock(&state->seqlock);
1851                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1852         }
1853 }
1854
1855 /**
1856  * nfs41_check_open_stateid - possibly free an open stateid
1857  *
1858  * @state: NFSv4 state for an inode
1859  *
1860  * Returns NFS_OK if recovery for this stateid is now finished.
1861  * Otherwise a negative NFS4ERR value is returned.
1862  */
1863 static int nfs41_check_open_stateid(struct nfs4_state *state)
1864 {
1865         struct nfs_server *server = NFS_SERVER(state->inode);
1866         nfs4_stateid *stateid = &state->open_stateid;
1867         int status;
1868
1869         /* If a state reset has been done, test_stateid is unneeded */
1870         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1871             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1872             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1873                 return -NFS4ERR_BAD_STATEID;
1874
1875         status = nfs41_test_stateid(server, stateid);
1876         if (status != NFS_OK) {
1877                 /* Free the stateid unless the server explicitly
1878                  * informs us the stateid is unrecognized. */
1879                 if (status != -NFS4ERR_BAD_STATEID)
1880                         nfs41_free_stateid(server, stateid);
1881
1882                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1883                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1884                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1885         }
1886         return status;
1887 }
1888
1889 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1890 {
1891         int status;
1892
1893         nfs41_clear_delegation_stateid(state);
1894         status = nfs41_check_open_stateid(state);
1895         if (status != NFS_OK)
1896                 status = nfs4_open_expired(sp, state);
1897         return status;
1898 }
1899 #endif
1900
1901 /*
1902  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1903  * fields corresponding to attributes that were used to store the verifier.
1904  * Make sure we clobber those fields in the later setattr call
1905  */
1906 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1907 {
1908         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1909             !(sattr->ia_valid & ATTR_ATIME_SET))
1910                 sattr->ia_valid |= ATTR_ATIME;
1911
1912         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1913             !(sattr->ia_valid & ATTR_MTIME_SET))
1914                 sattr->ia_valid |= ATTR_MTIME;
1915 }
1916
1917 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
1918                 fmode_t fmode,
1919                 int flags,
1920                 struct nfs4_state **res)
1921 {
1922         struct nfs4_state_owner *sp = opendata->owner;
1923         struct nfs_server *server = sp->so_server;
1924         struct nfs4_state *state;
1925         unsigned int seq;
1926         int ret;
1927
1928         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
1929
1930         ret = _nfs4_proc_open(opendata);
1931         if (ret != 0)
1932                 goto out;
1933
1934         state = nfs4_opendata_to_nfs4_state(opendata);
1935         ret = PTR_ERR(state);
1936         if (IS_ERR(state))
1937                 goto out;
1938         if (server->caps & NFS_CAP_POSIX_LOCK)
1939                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1940
1941         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
1942         if (ret != 0)
1943                 goto out;
1944
1945         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
1946                 nfs4_schedule_stateid_recovery(server, state);
1947         *res = state;
1948 out:
1949         return ret;
1950 }
1951
1952 /*
1953  * Returns a referenced nfs4_state
1954  */
1955 static int _nfs4_do_open(struct inode *dir,
1956                         struct dentry *dentry,
1957                         fmode_t fmode,
1958                         int flags,
1959                         struct iattr *sattr,
1960                         struct rpc_cred *cred,
1961                         struct nfs4_state **res,
1962                         struct nfs4_threshold **ctx_th)
1963 {
1964         struct nfs4_state_owner  *sp;
1965         struct nfs4_state     *state = NULL;
1966         struct nfs_server       *server = NFS_SERVER(dir);
1967         struct nfs4_opendata *opendata;
1968         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
1969         int status;
1970
1971         /* Protect against reboot recovery conflicts */
1972         status = -ENOMEM;
1973         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1974         if (sp == NULL) {
1975                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1976                 goto out_err;
1977         }
1978         status = nfs4_recover_expired_lease(server);
1979         if (status != 0)
1980                 goto err_put_state_owner;
1981         if (dentry->d_inode != NULL)
1982                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1983         status = -ENOMEM;
1984         if (dentry->d_inode)
1985                 claim = NFS4_OPEN_CLAIM_FH;
1986         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
1987                         claim, GFP_KERNEL);
1988         if (opendata == NULL)
1989                 goto err_put_state_owner;
1990
1991         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1992                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1993                 if (!opendata->f_attr.mdsthreshold)
1994                         goto err_opendata_put;
1995                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
1996         }
1997         if (dentry->d_inode != NULL)
1998                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1999
2000         status = _nfs4_open_and_get_state(opendata, fmode, flags, &state);
2001         if (status != 0)
2002                 goto err_opendata_put;
2003
2004         if (opendata->o_arg.open_flags & O_EXCL) {
2005                 nfs4_exclusive_attrset(opendata, sattr);
2006
2007                 nfs_fattr_init(opendata->o_res.f_attr);
2008                 status = nfs4_do_setattr(state->inode, cred,
2009                                 opendata->o_res.f_attr, sattr,
2010                                 state);
2011                 if (status == 0)
2012                         nfs_setattr_update_inode(state->inode, sattr);
2013                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2014         }
2015
2016         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2017                 *ctx_th = opendata->f_attr.mdsthreshold;
2018         else
2019                 kfree(opendata->f_attr.mdsthreshold);
2020         opendata->f_attr.mdsthreshold = NULL;
2021
2022         nfs4_opendata_put(opendata);
2023         nfs4_put_state_owner(sp);
2024         *res = state;
2025         return 0;
2026 err_opendata_put:
2027         kfree(opendata->f_attr.mdsthreshold);
2028         nfs4_opendata_put(opendata);
2029 err_put_state_owner:
2030         nfs4_put_state_owner(sp);
2031 out_err:
2032         *res = NULL;
2033         return status;
2034 }
2035
2036
2037 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2038                                         struct dentry *dentry,
2039                                         fmode_t fmode,
2040                                         int flags,
2041                                         struct iattr *sattr,
2042                                         struct rpc_cred *cred,
2043                                         struct nfs4_threshold **ctx_th)
2044 {
2045         struct nfs_server *server = NFS_SERVER(dir);
2046         struct nfs4_exception exception = { };
2047         struct nfs4_state *res;
2048         int status;
2049
2050         fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
2051         do {
2052                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
2053                                        &res, ctx_th);
2054                 if (status == 0)
2055                         break;
2056                 /* NOTE: BAD_SEQID means the server and client disagree about the
2057                  * book-keeping w.r.t. state-changing operations
2058                  * (OPEN/CLOSE/LOCK/LOCKU...)
2059                  * It is actually a sign of a bug on the client or on the server.
2060                  *
2061                  * If we receive a BAD_SEQID error in the particular case of
2062                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2063                  * have unhashed the old state_owner for us, and that we can
2064                  * therefore safely retry using a new one. We should still warn
2065                  * the user though...
2066                  */
2067                 if (status == -NFS4ERR_BAD_SEQID) {
2068                         pr_warn_ratelimited("NFS: v4 server %s "
2069                                         " returned a bad sequence-id error!\n",
2070                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2071                         exception.retry = 1;
2072                         continue;
2073                 }
2074                 /*
2075                  * BAD_STATEID on OPEN means that the server cancelled our
2076                  * state before it received the OPEN_CONFIRM.
2077                  * Recover by retrying the request as per the discussion
2078                  * on Page 181 of RFC3530.
2079                  */
2080                 if (status == -NFS4ERR_BAD_STATEID) {
2081                         exception.retry = 1;
2082                         continue;
2083                 }
2084                 if (status == -EAGAIN) {
2085                         /* We must have found a delegation */
2086                         exception.retry = 1;
2087                         continue;
2088                 }
2089                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2090                         continue;
2091                 res = ERR_PTR(nfs4_handle_exception(server,
2092                                         status, &exception));
2093         } while (exception.retry);
2094         return res;
2095 }
2096
2097 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2098                             struct nfs_fattr *fattr, struct iattr *sattr,
2099                             struct nfs4_state *state)
2100 {
2101         struct nfs_server *server = NFS_SERVER(inode);
2102         struct nfs_setattrargs  arg = {
2103                 .fh             = NFS_FH(inode),
2104                 .iap            = sattr,
2105                 .server         = server,
2106                 .bitmask = server->attr_bitmask,
2107         };
2108         struct nfs_setattrres  res = {
2109                 .fattr          = fattr,
2110                 .server         = server,
2111         };
2112         struct rpc_message msg = {
2113                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2114                 .rpc_argp       = &arg,
2115                 .rpc_resp       = &res,
2116                 .rpc_cred       = cred,
2117         };
2118         unsigned long timestamp = jiffies;
2119         int status;
2120
2121         nfs_fattr_init(fattr);
2122
2123         if (state != NULL && nfs4_valid_open_stateid(state)) {
2124                 struct nfs_lockowner lockowner = {
2125                         .l_owner = current->files,
2126                         .l_pid = current->tgid,
2127                 };
2128                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2129                                 &lockowner);
2130         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
2131                                 FMODE_WRITE)) {
2132                 /* Use that stateid */
2133         } else
2134                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2135
2136         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2137         if (status == 0 && state != NULL)
2138                 renew_lease(server, timestamp);
2139         return status;
2140 }
2141
2142 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2143                            struct nfs_fattr *fattr, struct iattr *sattr,
2144                            struct nfs4_state *state)
2145 {
2146         struct nfs_server *server = NFS_SERVER(inode);
2147         struct nfs4_exception exception = {
2148                 .state = state,
2149                 .inode = inode,
2150         };
2151         int err;
2152         do {
2153                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2154                 switch (err) {
2155                 case -NFS4ERR_OPENMODE:
2156                         if (state && !(state->state & FMODE_WRITE)) {
2157                                 err = -EBADF;
2158                                 if (sattr->ia_valid & ATTR_OPEN)
2159                                         err = -EACCES;
2160                                 goto out;
2161                         }
2162                 }
2163                 err = nfs4_handle_exception(server, err, &exception);
2164         } while (exception.retry);
2165 out:
2166         return err;
2167 }
2168
2169 struct nfs4_closedata {
2170         struct inode *inode;
2171         struct nfs4_state *state;
2172         struct nfs_closeargs arg;
2173         struct nfs_closeres res;
2174         struct nfs_fattr fattr;
2175         unsigned long timestamp;
2176         bool roc;
2177         u32 roc_barrier;
2178 };
2179
2180 static void nfs4_free_closedata(void *data)
2181 {
2182         struct nfs4_closedata *calldata = data;
2183         struct nfs4_state_owner *sp = calldata->state->owner;
2184         struct super_block *sb = calldata->state->inode->i_sb;
2185
2186         if (calldata->roc)
2187                 pnfs_roc_release(calldata->state->inode);
2188         nfs4_put_open_state(calldata->state);
2189         nfs_free_seqid(calldata->arg.seqid);
2190         nfs4_put_state_owner(sp);
2191         nfs_sb_deactive(sb);
2192         kfree(calldata);
2193 }
2194
2195 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2196                 fmode_t fmode)
2197 {
2198         spin_lock(&state->owner->so_lock);
2199         if (!(fmode & FMODE_READ))
2200                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2201         if (!(fmode & FMODE_WRITE))
2202                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2203         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2204         spin_unlock(&state->owner->so_lock);
2205 }
2206
2207 static void nfs4_close_done(struct rpc_task *task, void *data)
2208 {
2209         struct nfs4_closedata *calldata = data;
2210         struct nfs4_state *state = calldata->state;
2211         struct nfs_server *server = NFS_SERVER(calldata->inode);
2212
2213         dprintk("%s: begin!\n", __func__);
2214         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2215                 return;
2216         /* hmm. we are done with the inode, and in the process of freeing
2217          * the state_owner. we keep this around to process errors
2218          */
2219         switch (task->tk_status) {
2220                 case 0:
2221                         if (calldata->roc)
2222                                 pnfs_roc_set_barrier(state->inode,
2223                                                      calldata->roc_barrier);
2224                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2225                         renew_lease(server, calldata->timestamp);
2226                         nfs4_close_clear_stateid_flags(state,
2227                                         calldata->arg.fmode);
2228                         break;
2229                 case -NFS4ERR_STALE_STATEID:
2230                 case -NFS4ERR_OLD_STATEID:
2231                 case -NFS4ERR_BAD_STATEID:
2232                 case -NFS4ERR_EXPIRED:
2233                         if (calldata->arg.fmode == 0)
2234                                 break;
2235                 default:
2236                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2237                                 rpc_restart_call_prepare(task);
2238         }
2239         nfs_release_seqid(calldata->arg.seqid);
2240         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2241         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2242 }
2243
2244 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2245 {
2246         struct nfs4_closedata *calldata = data;
2247         struct nfs4_state *state = calldata->state;
2248         struct inode *inode = calldata->inode;
2249         int call_close = 0;
2250
2251         dprintk("%s: begin!\n", __func__);
2252         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2253                 goto out_wait;
2254
2255         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2256         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2257         spin_lock(&state->owner->so_lock);
2258         /* Calculate the change in open mode */
2259         if (state->n_rdwr == 0) {
2260                 if (state->n_rdonly == 0) {
2261                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2262                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2263                         calldata->arg.fmode &= ~FMODE_READ;
2264                 }
2265                 if (state->n_wronly == 0) {
2266                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2267                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2268                         calldata->arg.fmode &= ~FMODE_WRITE;
2269                 }
2270         }
2271         if (!nfs4_valid_open_stateid(state))
2272                 call_close = 0;
2273         spin_unlock(&state->owner->so_lock);
2274
2275         if (!call_close) {
2276                 /* Note: exit _without_ calling nfs4_close_done */
2277                 goto out_no_action;
2278         }
2279
2280         if (calldata->arg.fmode == 0) {
2281                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2282                 if (calldata->roc &&
2283                     pnfs_roc_drain(inode, &calldata->roc_barrier, task))
2284                         goto out_wait;
2285         }
2286
2287         nfs_fattr_init(calldata->res.fattr);
2288         calldata->timestamp = jiffies;
2289         if (nfs4_setup_sequence(NFS_SERVER(inode),
2290                                 &calldata->arg.seq_args,
2291                                 &calldata->res.seq_res,
2292                                 task) != 0)
2293                 nfs_release_seqid(calldata->arg.seqid);
2294         dprintk("%s: done!\n", __func__);
2295         return;
2296 out_no_action:
2297         task->tk_action = NULL;
2298 out_wait:
2299         nfs4_sequence_done(task, &calldata->res.seq_res);
2300 }
2301
2302 static const struct rpc_call_ops nfs4_close_ops = {
2303         .rpc_call_prepare = nfs4_close_prepare,
2304         .rpc_call_done = nfs4_close_done,
2305         .rpc_release = nfs4_free_closedata,
2306 };
2307
2308 /*
2309  * It is possible for data to be read/written from a mem-mapped file
2310  * after the sys_close call (which hits the vfs layer as a flush).
2311  * This means that we can't safely call nfsv4 close on a file until
2312  * the inode is cleared. This in turn means that we are not good
2313  * NFSv4 citizens - we do not indicate to the server to update the file's
2314  * share state even when we are done with one of the three share
2315  * stateid's in the inode.
2316  *
2317  * NOTE: Caller must be holding the sp->so_owner semaphore!
2318  */
2319 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2320 {
2321         struct nfs_server *server = NFS_SERVER(state->inode);
2322         struct nfs4_closedata *calldata;
2323         struct nfs4_state_owner *sp = state->owner;
2324         struct rpc_task *task;
2325         struct rpc_message msg = {
2326                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2327                 .rpc_cred = state->owner->so_cred,
2328         };
2329         struct rpc_task_setup task_setup_data = {
2330                 .rpc_client = server->client,
2331                 .rpc_message = &msg,
2332                 .callback_ops = &nfs4_close_ops,
2333                 .workqueue = nfsiod_workqueue,
2334                 .flags = RPC_TASK_ASYNC,
2335         };
2336         int status = -ENOMEM;
2337
2338         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2339         if (calldata == NULL)
2340                 goto out;
2341         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2342         calldata->inode = state->inode;
2343         calldata->state = state;
2344         calldata->arg.fh = NFS_FH(state->inode);
2345         calldata->arg.stateid = &state->open_stateid;
2346         /* Serialization for the sequence id */
2347         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2348         if (calldata->arg.seqid == NULL)
2349                 goto out_free_calldata;
2350         calldata->arg.fmode = 0;
2351         calldata->arg.bitmask = server->cache_consistency_bitmask;
2352         calldata->res.fattr = &calldata->fattr;
2353         calldata->res.seqid = calldata->arg.seqid;
2354         calldata->res.server = server;
2355         calldata->roc = pnfs_roc(state->inode);
2356         nfs_sb_active(calldata->inode->i_sb);
2357
2358         msg.rpc_argp = &calldata->arg;
2359         msg.rpc_resp = &calldata->res;
2360         task_setup_data.callback_data = calldata;
2361         task = rpc_run_task(&task_setup_data);
2362         if (IS_ERR(task))
2363                 return PTR_ERR(task);
2364         status = 0;
2365         if (wait)
2366                 status = rpc_wait_for_completion_task(task);
2367         rpc_put_task(task);
2368         return status;
2369 out_free_calldata:
2370         kfree(calldata);
2371 out:
2372         nfs4_put_open_state(state);
2373         nfs4_put_state_owner(sp);
2374         return status;
2375 }
2376
2377 static struct inode *
2378 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2379 {
2380         struct nfs4_state *state;
2381
2382         /* Protect against concurrent sillydeletes */
2383         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2384                              ctx->cred, &ctx->mdsthreshold);
2385         if (IS_ERR(state))
2386                 return ERR_CAST(state);
2387         ctx->state = state;
2388         return igrab(state->inode);
2389 }
2390
2391 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2392 {
2393         if (ctx->state == NULL)
2394                 return;
2395         if (is_sync)
2396                 nfs4_close_sync(ctx->state, ctx->mode);
2397         else
2398                 nfs4_close_state(ctx->state, ctx->mode);
2399 }
2400
2401 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2402 {
2403         struct nfs4_server_caps_arg args = {
2404                 .fhandle = fhandle,
2405         };
2406         struct nfs4_server_caps_res res = {};
2407         struct rpc_message msg = {
2408                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2409                 .rpc_argp = &args,
2410                 .rpc_resp = &res,
2411         };
2412         int status;
2413
2414         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2415         if (status == 0) {
2416                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2417                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2418                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2419                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2420                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2421                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2422                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2423                         server->caps |= NFS_CAP_ACLS;
2424                 if (res.has_links != 0)
2425                         server->caps |= NFS_CAP_HARDLINKS;
2426                 if (res.has_symlinks != 0)
2427                         server->caps |= NFS_CAP_SYMLINKS;
2428                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2429                         server->caps |= NFS_CAP_FILEID;
2430                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2431                         server->caps |= NFS_CAP_MODE;
2432                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2433                         server->caps |= NFS_CAP_NLINK;
2434                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2435                         server->caps |= NFS_CAP_OWNER;
2436                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2437                         server->caps |= NFS_CAP_OWNER_GROUP;
2438                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2439                         server->caps |= NFS_CAP_ATIME;
2440                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2441                         server->caps |= NFS_CAP_CTIME;
2442                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2443                         server->caps |= NFS_CAP_MTIME;
2444
2445                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2446                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2447                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2448                 server->acl_bitmask = res.acl_bitmask;
2449                 server->fh_expire_type = res.fh_expire_type;
2450         }
2451
2452         return status;
2453 }
2454
2455 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2456 {
2457         struct nfs4_exception exception = { };
2458         int err;
2459         do {
2460                 err = nfs4_handle_exception(server,
2461                                 _nfs4_server_capabilities(server, fhandle),
2462                                 &exception);
2463         } while (exception.retry);
2464         return err;
2465 }
2466
2467 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2468                 struct nfs_fsinfo *info)
2469 {
2470         struct nfs4_lookup_root_arg args = {
2471                 .bitmask = nfs4_fattr_bitmap,
2472         };
2473         struct nfs4_lookup_res res = {
2474                 .server = server,
2475                 .fattr = info->fattr,
2476                 .fh = fhandle,
2477         };
2478         struct rpc_message msg = {
2479                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2480                 .rpc_argp = &args,
2481                 .rpc_resp = &res,
2482         };
2483
2484         nfs_fattr_init(info->fattr);
2485         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2486 }
2487
2488 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2489                 struct nfs_fsinfo *info)
2490 {
2491         struct nfs4_exception exception = { };
2492         int err;
2493         do {
2494                 err = _nfs4_lookup_root(server, fhandle, info);
2495                 switch (err) {
2496                 case 0:
2497                 case -NFS4ERR_WRONGSEC:
2498                         goto out;
2499                 default:
2500                         err = nfs4_handle_exception(server, err, &exception);
2501                 }
2502         } while (exception.retry);
2503 out:
2504         return err;
2505 }
2506
2507 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2508                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2509 {
2510         struct rpc_auth *auth;
2511         int ret;
2512
2513         auth = rpcauth_create(flavor, server->client);
2514         if (IS_ERR(auth)) {
2515                 ret = -EIO;
2516                 goto out;
2517         }
2518         ret = nfs4_lookup_root(server, fhandle, info);
2519 out:
2520         return ret;
2521 }
2522
2523 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2524                               struct nfs_fsinfo *info)
2525 {
2526         int i, len, status = 0;
2527         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2528
2529         len = rpcauth_list_flavors(flav_array, ARRAY_SIZE(flav_array));
2530         if (len < 0)
2531                 return len;
2532
2533         for (i = 0; i < len; i++) {
2534                 /* AUTH_UNIX is the default flavor if none was specified,
2535                  * thus has already been tried. */
2536                 if (flav_array[i] == RPC_AUTH_UNIX)
2537                         continue;
2538
2539                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2540                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2541                         continue;
2542                 break;
2543         }
2544         /*
2545          * -EACCESS could mean that the user doesn't have correct permissions
2546          * to access the mount.  It could also mean that we tried to mount
2547          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2548          * existing mount programs don't handle -EACCES very well so it should
2549          * be mapped to -EPERM instead.
2550          */
2551         if (status == -EACCES)
2552                 status = -EPERM;
2553         return status;
2554 }
2555
2556 /*
2557  * get the file handle for the "/" directory on the server
2558  */
2559 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2560                          struct nfs_fsinfo *info)
2561 {
2562         int minor_version = server->nfs_client->cl_minorversion;
2563         int status = nfs4_lookup_root(server, fhandle, info);
2564         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2565                 /*
2566                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2567                  * by nfs4_map_errors() as this function exits.
2568                  */
2569                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2570         if (status == 0)
2571                 status = nfs4_server_capabilities(server, fhandle);
2572         if (status == 0)
2573                 status = nfs4_do_fsinfo(server, fhandle, info);
2574         return nfs4_map_errors(status);
2575 }
2576
2577 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2578                               struct nfs_fsinfo *info)
2579 {
2580         int error;
2581         struct nfs_fattr *fattr = info->fattr;
2582
2583         error = nfs4_server_capabilities(server, mntfh);
2584         if (error < 0) {
2585                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2586                 return error;
2587         }
2588
2589         error = nfs4_proc_getattr(server, mntfh, fattr);
2590         if (error < 0) {
2591                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2592                 return error;
2593         }
2594
2595         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2596             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2597                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2598
2599         return error;
2600 }
2601
2602 /*
2603  * Get locations and (maybe) other attributes of a referral.
2604  * Note that we'll actually follow the referral later when
2605  * we detect fsid mismatch in inode revalidation
2606  */
2607 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2608                              const struct qstr *name, struct nfs_fattr *fattr,
2609                              struct nfs_fh *fhandle)
2610 {
2611         int status = -ENOMEM;
2612         struct page *page = NULL;
2613         struct nfs4_fs_locations *locations = NULL;
2614
2615         page = alloc_page(GFP_KERNEL);
2616         if (page == NULL)
2617                 goto out;
2618         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2619         if (locations == NULL)
2620                 goto out;
2621
2622         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2623         if (status != 0)
2624                 goto out;
2625         /* Make sure server returned a different fsid for the referral */
2626         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2627                 dprintk("%s: server did not return a different fsid for"
2628                         " a referral at %s\n", __func__, name->name);
2629                 status = -EIO;
2630                 goto out;
2631         }
2632         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2633         nfs_fixup_referral_attributes(&locations->fattr);
2634
2635         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2636         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2637         memset(fhandle, 0, sizeof(struct nfs_fh));
2638 out:
2639         if (page)
2640                 __free_page(page);
2641         kfree(locations);
2642         return status;
2643 }
2644
2645 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2646 {
2647         struct nfs4_getattr_arg args = {
2648                 .fh = fhandle,
2649                 .bitmask = server->attr_bitmask,
2650         };
2651         struct nfs4_getattr_res res = {
2652                 .fattr = fattr,
2653                 .server = server,
2654         };
2655         struct rpc_message msg = {
2656                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2657                 .rpc_argp = &args,
2658                 .rpc_resp = &res,
2659         };
2660
2661         nfs_fattr_init(fattr);
2662         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2663 }
2664
2665 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2666 {
2667         struct nfs4_exception exception = { };
2668         int err;
2669         do {
2670                 err = nfs4_handle_exception(server,
2671                                 _nfs4_proc_getattr(server, fhandle, fattr),
2672                                 &exception);
2673         } while (exception.retry);
2674         return err;
2675 }
2676
2677 /*
2678  * The file is not closed if it is opened due to the a request to change
2679  * the size of the file. The open call will not be needed once the
2680  * VFS layer lookup-intents are implemented.
2681  *
2682  * Close is called when the inode is destroyed.
2683  * If we haven't opened the file for O_WRONLY, we
2684  * need to in the size_change case to obtain a stateid.
2685  *
2686  * Got race?
2687  * Because OPEN is always done by name in nfsv4, it is
2688  * possible that we opened a different file by the same
2689  * name.  We can recognize this race condition, but we
2690  * can't do anything about it besides returning an error.
2691  *
2692  * This will be fixed with VFS changes (lookup-intent).
2693  */
2694 static int
2695 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2696                   struct iattr *sattr)
2697 {
2698         struct inode *inode = dentry->d_inode;
2699         struct rpc_cred *cred = NULL;
2700         struct nfs4_state *state = NULL;
2701         int status;
2702
2703         if (pnfs_ld_layoutret_on_setattr(inode))
2704                 pnfs_commit_and_return_layout(inode);
2705
2706         nfs_fattr_init(fattr);
2707
2708         /* Deal with open(O_TRUNC) */
2709         if (sattr->ia_valid & ATTR_OPEN)
2710                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2711
2712         /* Optimization: if the end result is no change, don't RPC */
2713         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2714                 return 0;
2715
2716         /* Search for an existing open(O_WRITE) file */
2717         if (sattr->ia_valid & ATTR_FILE) {
2718                 struct nfs_open_context *ctx;
2719
2720                 ctx = nfs_file_open_context(sattr->ia_file);
2721                 if (ctx) {
2722                         cred = ctx->cred;
2723                         state = ctx->state;
2724                 }
2725         }
2726
2727         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2728         if (status == 0)
2729                 nfs_setattr_update_inode(inode, sattr);
2730         return status;
2731 }
2732
2733 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2734                 const struct qstr *name, struct nfs_fh *fhandle,
2735                 struct nfs_fattr *fattr)
2736 {
2737         struct nfs_server *server = NFS_SERVER(dir);
2738         int                    status;
2739         struct nfs4_lookup_arg args = {
2740                 .bitmask = server->attr_bitmask,
2741                 .dir_fh = NFS_FH(dir),
2742                 .name = name,
2743         };
2744         struct nfs4_lookup_res res = {
2745                 .server = server,
2746                 .fattr = fattr,
2747                 .fh = fhandle,
2748         };
2749         struct rpc_message msg = {
2750                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2751                 .rpc_argp = &args,
2752                 .rpc_resp = &res,
2753         };
2754
2755         nfs_fattr_init(fattr);
2756
2757         dprintk("NFS call  lookup %s\n", name->name);
2758         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2759         dprintk("NFS reply lookup: %d\n", status);
2760         return status;
2761 }
2762
2763 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2764 {
2765         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2766                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2767         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2768         fattr->nlink = 2;
2769 }
2770
2771 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2772                                    struct qstr *name, struct nfs_fh *fhandle,
2773                                    struct nfs_fattr *fattr)
2774 {
2775         struct nfs4_exception exception = { };
2776         struct rpc_clnt *client = *clnt;
2777         int err;
2778         do {
2779                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2780                 switch (err) {
2781                 case -NFS4ERR_BADNAME:
2782                         err = -ENOENT;
2783                         goto out;
2784                 case -NFS4ERR_MOVED:
2785                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2786                         goto out;
2787                 case -NFS4ERR_WRONGSEC:
2788                         err = -EPERM;
2789                         if (client != *clnt)
2790                                 goto out;
2791
2792                         client = nfs4_create_sec_client(client, dir, name);
2793                         if (IS_ERR(client))
2794                                 return PTR_ERR(client);
2795
2796                         exception.retry = 1;
2797                         break;
2798                 default:
2799                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2800                 }
2801         } while (exception.retry);
2802
2803 out:
2804         if (err == 0)
2805                 *clnt = client;
2806         else if (client != *clnt)
2807                 rpc_shutdown_client(client);
2808
2809         return err;
2810 }
2811
2812 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2813                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2814 {
2815         int status;
2816         struct rpc_clnt *client = NFS_CLIENT(dir);
2817
2818         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2819         if (client != NFS_CLIENT(dir)) {
2820                 rpc_shutdown_client(client);
2821                 nfs_fixup_secinfo_attributes(fattr);
2822         }
2823         return status;
2824 }
2825
2826 struct rpc_clnt *
2827 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2828                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2829 {
2830         int status;
2831         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2832
2833         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2834         if (status < 0) {
2835                 rpc_shutdown_client(client);
2836                 return ERR_PTR(status);
2837         }
2838         return client;
2839 }
2840
2841 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2842 {
2843         struct nfs_server *server = NFS_SERVER(inode);
2844         struct nfs4_accessargs args = {
2845                 .fh = NFS_FH(inode),
2846                 .bitmask = server->cache_consistency_bitmask,
2847         };
2848         struct nfs4_accessres res = {
2849                 .server = server,
2850         };
2851         struct rpc_message msg = {
2852                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2853                 .rpc_argp = &args,
2854                 .rpc_resp = &res,
2855                 .rpc_cred = entry->cred,
2856         };
2857         int mode = entry->mask;
2858         int status;
2859
2860         /*
2861          * Determine which access bits we want to ask for...
2862          */
2863         if (mode & MAY_READ)
2864                 args.access |= NFS4_ACCESS_READ;
2865         if (S_ISDIR(inode->i_mode)) {
2866                 if (mode & MAY_WRITE)
2867                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2868                 if (mode & MAY_EXEC)
2869                         args.access |= NFS4_ACCESS_LOOKUP;
2870         } else {
2871                 if (mode & MAY_WRITE)
2872                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2873                 if (mode & MAY_EXEC)
2874                         args.access |= NFS4_ACCESS_EXECUTE;
2875         }
2876
2877         res.fattr = nfs_alloc_fattr();
2878         if (res.fattr == NULL)
2879                 return -ENOMEM;
2880
2881         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2882         if (!status) {
2883                 nfs_access_set_mask(entry, res.access);
2884                 nfs_refresh_inode(inode, res.fattr);
2885         }
2886         nfs_free_fattr(res.fattr);
2887         return status;
2888 }
2889
2890 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2891 {
2892         struct nfs4_exception exception = { };
2893         int err;
2894         do {
2895                 err = nfs4_handle_exception(NFS_SERVER(inode),
2896                                 _nfs4_proc_access(inode, entry),
2897                                 &exception);
2898         } while (exception.retry);
2899         return err;
2900 }
2901
2902 /*
2903  * TODO: For the time being, we don't try to get any attributes
2904  * along with any of the zero-copy operations READ, READDIR,
2905  * READLINK, WRITE.
2906  *
2907  * In the case of the first three, we want to put the GETATTR
2908  * after the read-type operation -- this is because it is hard
2909  * to predict the length of a GETATTR response in v4, and thus
2910  * align the READ data correctly.  This means that the GETATTR
2911  * may end up partially falling into the page cache, and we should
2912  * shift it into the 'tail' of the xdr_buf before processing.
2913  * To do this efficiently, we need to know the total length
2914  * of data received, which doesn't seem to be available outside
2915  * of the RPC layer.
2916  *
2917  * In the case of WRITE, we also want to put the GETATTR after
2918  * the operation -- in this case because we want to make sure
2919  * we get the post-operation mtime and size.
2920  *
2921  * Both of these changes to the XDR layer would in fact be quite
2922  * minor, but I decided to leave them for a subsequent patch.
2923  */
2924 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2925                 unsigned int pgbase, unsigned int pglen)
2926 {
2927         struct nfs4_readlink args = {
2928                 .fh       = NFS_FH(inode),
2929                 .pgbase   = pgbase,
2930                 .pglen    = pglen,
2931                 .pages    = &page,
2932         };
2933         struct nfs4_readlink_res res;
2934         struct rpc_message msg = {
2935                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2936                 .rpc_argp = &args,
2937                 .rpc_resp = &res,
2938         };
2939
2940         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2941 }
2942
2943 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2944                 unsigned int pgbase, unsigned int pglen)
2945 {
2946         struct nfs4_exception exception = { };
2947         int err;
2948         do {
2949                 err = nfs4_handle_exception(NFS_SERVER(inode),
2950                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2951                                 &exception);
2952         } while (exception.retry);
2953         return err;
2954 }
2955
2956 /*
2957  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
2958  */
2959 static int
2960 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2961                  int flags)
2962 {
2963         struct nfs_open_context *ctx;
2964         struct nfs4_state *state;
2965         int status = 0;
2966
2967         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
2968         if (IS_ERR(ctx))
2969                 return PTR_ERR(ctx);
2970
2971         sattr->ia_mode &= ~current_umask();
2972         state = nfs4_do_open(dir, dentry, ctx->mode,
2973                         flags, sattr, ctx->cred,
2974                         &ctx->mdsthreshold);
2975         d_drop(dentry);
2976         if (IS_ERR(state)) {
2977                 status = PTR_ERR(state);
2978                 goto out;
2979         }
2980         d_add(dentry, igrab(state->inode));
2981         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2982         ctx->state = state;
2983 out:
2984         put_nfs_open_context(ctx);
2985         return status;
2986 }
2987
2988 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2989 {
2990         struct nfs_server *server = NFS_SERVER(dir);
2991         struct nfs_removeargs args = {
2992                 .fh = NFS_FH(dir),
2993                 .name = *name,
2994         };
2995         struct nfs_removeres res = {
2996                 .server = server,
2997         };
2998         struct rpc_message msg = {
2999                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3000                 .rpc_argp = &args,
3001                 .rpc_resp = &res,
3002         };
3003         int status;
3004
3005         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3006         if (status == 0)
3007                 update_changeattr(dir, &res.cinfo);
3008         return status;
3009 }
3010
3011 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3012 {
3013         struct nfs4_exception exception = { };
3014         int err;
3015         do {
3016                 err = nfs4_handle_exception(NFS_SERVER(dir),
3017                                 _nfs4_proc_remove(dir, name),
3018                                 &exception);
3019         } while (exception.retry);
3020         return err;
3021 }
3022
3023 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3024 {
3025         struct nfs_server *server = NFS_SERVER(dir);
3026         struct nfs_removeargs *args = msg->rpc_argp;
3027         struct nfs_removeres *res = msg->rpc_resp;
3028
3029         res->server = server;
3030         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3031         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3032 }
3033
3034 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3035 {
3036         nfs4_setup_sequence(NFS_SERVER(data->dir),
3037                         &data->args.seq_args,
3038                         &data->res.seq_res,
3039                         task);
3040 }
3041
3042 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3043 {
3044         struct nfs_removeres *res = task->tk_msg.rpc_resp;
3045
3046         if (!nfs4_sequence_done(task, &res->seq_res))
3047                 return 0;
3048         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3049                 return 0;
3050         update_changeattr(dir, &res->cinfo);
3051         return 1;
3052 }
3053
3054 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3055 {
3056         struct nfs_server *server = NFS_SERVER(dir);
3057         struct nfs_renameargs *arg = msg->rpc_argp;
3058         struct nfs_renameres *res = msg->rpc_resp;
3059
3060         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3061         res->server = server;
3062         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3063 }
3064
3065 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3066 {
3067         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3068                         &data->args.seq_args,
3069                         &data->res.seq_res,
3070                         task);
3071 }
3072
3073 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3074                                  struct inode *new_dir)
3075 {
3076         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3077
3078         if (!nfs4_sequence_done(task, &res->seq_res))
3079                 return 0;
3080         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3081                 return 0;
3082
3083         update_changeattr(old_dir, &res->old_cinfo);
3084         update_changeattr(new_dir, &res->new_cinfo);
3085         return 1;
3086 }
3087
3088 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3089                 struct inode *new_dir, struct qstr *new_name)
3090 {
3091         struct nfs_server *server = NFS_SERVER(old_dir);
3092         struct nfs_renameargs arg = {
3093                 .old_dir = NFS_FH(old_dir),
3094                 .new_dir = NFS_FH(new_dir),
3095                 .old_name = old_name,
3096                 .new_name = new_name,
3097         };
3098         struct nfs_renameres res = {
3099                 .server = server,
3100         };
3101         struct rpc_message msg = {
3102                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3103                 .rpc_argp = &arg,
3104                 .rpc_resp = &res,
3105         };
3106         int status = -ENOMEM;
3107
3108         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3109         if (!status) {
3110                 update_changeattr(old_dir, &res.old_cinfo);
3111                 update_changeattr(new_dir, &res.new_cinfo);
3112         }
3113         return status;
3114 }
3115
3116 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3117                 struct inode *new_dir, struct qstr *new_name)
3118 {
3119         struct nfs4_exception exception = { };
3120         int err;
3121         do {
3122                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3123                                 _nfs4_proc_rename(old_dir, old_name,
3124                                         new_dir, new_name),
3125                                 &exception);
3126         } while (exception.retry);
3127         return err;
3128 }
3129
3130 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3131 {
3132         struct nfs_server *server = NFS_SERVER(inode);
3133         struct nfs4_link_arg arg = {
3134                 .fh     = NFS_FH(inode),
3135                 .dir_fh = NFS_FH(dir),
3136                 .name   = name,
3137                 .bitmask = server->attr_bitmask,
3138         };
3139         struct nfs4_link_res res = {
3140                 .server = server,
3141         };
3142         struct rpc_message msg = {
3143                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3144                 .rpc_argp = &arg,
3145                 .rpc_resp = &res,
3146         };
3147         int status = -ENOMEM;
3148
3149         res.fattr = nfs_alloc_fattr();
3150         if (res.fattr == NULL)
3151                 goto out;
3152
3153         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3154         if (!status) {
3155                 update_changeattr(dir, &res.cinfo);
3156                 nfs_post_op_update_inode(inode, res.fattr);
3157         }
3158 out:
3159         nfs_free_fattr(res.fattr);
3160         return status;
3161 }
3162
3163 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3164 {
3165         struct nfs4_exception exception = { };
3166         int err;
3167         do {
3168                 err = nfs4_handle_exception(NFS_SERVER(inode),
3169                                 _nfs4_proc_link(inode, dir, name),
3170                                 &exception);
3171         } while (exception.retry);
3172         return err;
3173 }
3174
3175 struct nfs4_createdata {
3176         struct rpc_message msg;
3177         struct nfs4_create_arg arg;
3178         struct nfs4_create_res res;
3179         struct nfs_fh fh;
3180         struct nfs_fattr fattr;
3181 };
3182
3183 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3184                 struct qstr *name, struct iattr *sattr, u32 ftype)
3185 {
3186         struct nfs4_createdata *data;
3187
3188         data = kzalloc(sizeof(*data), GFP_KERNEL);
3189         if (data != NULL) {
3190                 struct nfs_server *server = NFS_SERVER(dir);
3191
3192                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3193                 data->msg.rpc_argp = &data->arg;
3194                 data->msg.rpc_resp = &data->res;
3195                 data->arg.dir_fh = NFS_FH(dir);
3196                 data->arg.server = server;
3197                 data->arg.name = name;
3198                 data->arg.attrs = sattr;
3199                 data->arg.ftype = ftype;
3200                 data->arg.bitmask = server->attr_bitmask;
3201                 data->res.server = server;
3202                 data->res.fh = &data->fh;
3203                 data->res.fattr = &data->fattr;
3204                 nfs_fattr_init(data->res.fattr);
3205         }
3206         return data;
3207 }
3208
3209 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3210 {
3211         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3212                                     &data->arg.seq_args, &data->res.seq_res, 1);
3213         if (status == 0) {
3214                 update_changeattr(dir, &data->res.dir_cinfo);
3215                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3216         }
3217         return status;
3218 }
3219
3220 static void nfs4_free_createdata(struct nfs4_createdata *data)
3221 {
3222         kfree(data);
3223 }
3224
3225 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3226                 struct page *page, unsigned int len, struct iattr *sattr)
3227 {
3228         struct nfs4_createdata *data;
3229         int status = -ENAMETOOLONG;
3230
3231         if (len > NFS4_MAXPATHLEN)
3232                 goto out;
3233
3234         status = -ENOMEM;
3235         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3236         if (data == NULL)
3237                 goto out;
3238
3239         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3240         data->arg.u.symlink.pages = &page;
3241         data->arg.u.symlink.len = len;
3242
3243         status = nfs4_do_create(dir, dentry, data);
3244
3245         nfs4_free_createdata(data);
3246 out:
3247         return status;
3248 }
3249
3250 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3251                 struct page *page, unsigned int len, struct iattr *sattr)
3252 {
3253         struct nfs4_exception exception = { };
3254         int err;
3255         do {
3256                 err = nfs4_handle_exception(NFS_SERVER(dir),
3257                                 _nfs4_proc_symlink(dir, dentry, page,
3258                                                         len, sattr),
3259                                 &exception);
3260         } while (exception.retry);
3261         return err;
3262 }
3263
3264 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3265                 struct iattr *sattr)
3266 {
3267         struct nfs4_createdata *data;
3268         int status = -ENOMEM;
3269
3270         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3271         if (data == NULL)
3272                 goto out;
3273
3274         status = nfs4_do_create(dir, dentry, data);
3275
3276         nfs4_free_createdata(data);
3277 out:
3278         return status;
3279 }
3280
3281 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3282                 struct iattr *sattr)
3283 {
3284         struct nfs4_exception exception = { };
3285         int err;
3286
3287         sattr->ia_mode &= ~current_umask();
3288         do {
3289                 err = nfs4_handle_exception(NFS_SERVER(dir),
3290                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3291                                 &exception);
3292         } while (exception.retry);
3293         return err;
3294 }
3295
3296 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3297                 u64 cookie, struct page **pages, unsigned int count, int plus)
3298 {
3299         struct inode            *dir = dentry->d_inode;
3300         struct nfs4_readdir_arg args = {
3301                 .fh = NFS_FH(dir),
3302                 .pages = pages,
3303                 .pgbase = 0,
3304                 .count = count,
3305                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3306                 .plus = plus,
3307         };
3308         struct nfs4_readdir_res res;
3309         struct rpc_message msg = {
3310                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3311                 .rpc_argp = &args,
3312                 .rpc_resp = &res,
3313                 .rpc_cred = cred,
3314         };
3315         int                     status;
3316
3317         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3318                         dentry->d_parent->d_name.name,
3319                         dentry->d_name.name,
3320                         (unsigned long long)cookie);
3321         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3322         res.pgbase = args.pgbase;
3323         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3324         if (status >= 0) {
3325                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3326                 status += args.pgbase;
3327         }
3328
3329         nfs_invalidate_atime(dir);
3330
3331         dprintk("%s: returns %d\n", __func__, status);
3332         return status;
3333 }
3334
3335 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3336                 u64 cookie, struct page **pages, unsigned int count, int plus)
3337 {
3338         struct nfs4_exception exception = { };
3339         int err;
3340         do {
3341                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3342                                 _nfs4_proc_readdir(dentry, cred, cookie,
3343                                         pages, count, plus),
3344                                 &exception);
3345         } while (exception.retry);
3346         return err;
3347 }
3348
3349 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3350                 struct iattr *sattr, dev_t rdev)
3351 {
3352         struct nfs4_createdata *data;
3353         int mode = sattr->ia_mode;
3354         int status = -ENOMEM;
3355
3356         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3357         if (data == NULL)
3358                 goto out;
3359
3360         if (S_ISFIFO(mode))
3361                 data->arg.ftype = NF4FIFO;
3362         else if (S_ISBLK(mode)) {
3363                 data->arg.ftype = NF4BLK;
3364                 data->arg.u.device.specdata1 = MAJOR(rdev);
3365                 data->arg.u.device.specdata2 = MINOR(rdev);
3366         }
3367         else if (S_ISCHR(mode)) {
3368                 data->arg.ftype = NF4CHR;
3369                 data->arg.u.device.specdata1 = MAJOR(rdev);
3370                 data->arg.u.device.specdata2 = MINOR(rdev);
3371         } else if (!S_ISSOCK(mode)) {
3372                 status = -EINVAL;
3373                 goto out_free;
3374         }
3375
3376         status = nfs4_do_create(dir, dentry, data);
3377 out_free:
3378         nfs4_free_createdata(data);
3379 out:
3380         return status;
3381 }
3382
3383 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3384                 struct iattr *sattr, dev_t rdev)
3385 {
3386         struct nfs4_exception exception = { };
3387         int err;
3388
3389         sattr->ia_mode &= ~current_umask();
3390         do {
3391                 err = nfs4_handle_exception(NFS_SERVER(dir),
3392                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3393                                 &exception);
3394         } while (exception.retry);
3395         return err;
3396 }
3397
3398 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3399                  struct nfs_fsstat *fsstat)
3400 {
3401         struct nfs4_statfs_arg args = {
3402                 .fh = fhandle,
3403                 .bitmask = server->attr_bitmask,
3404         };
3405         struct nfs4_statfs_res res = {
3406                 .fsstat = fsstat,
3407         };
3408         struct rpc_message msg = {
3409                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3410                 .rpc_argp = &args,
3411                 .rpc_resp = &res,
3412         };
3413
3414         nfs_fattr_init(fsstat->fattr);
3415         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3416 }
3417
3418 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3419 {
3420         struct nfs4_exception exception = { };
3421         int err;
3422         do {
3423                 err = nfs4_handle_exception(server,
3424                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3425                                 &exception);
3426         } while (exception.retry);
3427         return err;
3428 }
3429
3430 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3431                 struct nfs_fsinfo *fsinfo)
3432 {
3433         struct nfs4_fsinfo_arg args = {
3434                 .fh = fhandle,
3435                 .bitmask = server->attr_bitmask,
3436         };
3437         struct nfs4_fsinfo_res res = {
3438                 .fsinfo = fsinfo,
3439         };
3440         struct rpc_message msg = {
3441                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3442                 .rpc_argp = &args,
3443                 .rpc_resp = &res,
3444         };
3445
3446         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3447 }
3448
3449 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3450 {
3451         struct nfs4_exception exception = { };
3452         int err;
3453
3454         do {
3455                 err = nfs4_handle_exception(server,
3456                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3457                                 &exception);
3458         } while (exception.retry);
3459         return err;
3460 }
3461
3462 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3463 {
3464         int error;
3465
3466         nfs_fattr_init(fsinfo->fattr);
3467         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3468         if (error == 0) {
3469                 /* block layout checks this! */
3470                 server->pnfs_blksize = fsinfo->blksize;
3471                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3472         }
3473
3474         return error;
3475 }
3476
3477 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3478                 struct nfs_pathconf *pathconf)
3479 {
3480         struct nfs4_pathconf_arg args = {
3481                 .fh = fhandle,
3482                 .bitmask = server->attr_bitmask,
3483         };
3484         struct nfs4_pathconf_res res = {
3485                 .pathconf = pathconf,
3486         };
3487         struct rpc_message msg = {
3488                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3489                 .rpc_argp = &args,
3490                 .rpc_resp = &res,
3491         };
3492
3493         /* None of the pathconf attributes are mandatory to implement */
3494         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3495                 memset(pathconf, 0, sizeof(*pathconf));
3496                 return 0;
3497         }
3498
3499         nfs_fattr_init(pathconf->fattr);
3500         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3501 }
3502
3503 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3504                 struct nfs_pathconf *pathconf)
3505 {
3506         struct nfs4_exception exception = { };
3507         int err;
3508
3509         do {
3510                 err = nfs4_handle_exception(server,
3511                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3512                                 &exception);
3513         } while (exception.retry);
3514         return err;
3515 }
3516
3517 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
3518                 const struct nfs_open_context *ctx,
3519                 const struct nfs_lock_context *l_ctx,
3520                 fmode_t fmode)
3521 {
3522         const struct nfs_lockowner *lockowner = NULL;
3523
3524         if (l_ctx != NULL)
3525                 lockowner = &l_ctx->lockowner;
3526         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
3527 }
3528 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
3529
3530 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
3531                 const struct nfs_open_context *ctx,
3532                 const struct nfs_lock_context *l_ctx,
3533                 fmode_t fmode)
3534 {
3535         nfs4_stateid current_stateid;
3536
3537         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode))
3538                 return false;
3539         return nfs4_stateid_match(stateid, &current_stateid);
3540 }
3541
3542 static bool nfs4_error_stateid_expired(int err)
3543 {
3544         switch (err) {
3545         case -NFS4ERR_DELEG_REVOKED:
3546         case -NFS4ERR_ADMIN_REVOKED:
3547         case -NFS4ERR_BAD_STATEID:
3548         case -NFS4ERR_STALE_STATEID:
3549         case -NFS4ERR_OLD_STATEID:
3550         case -NFS4ERR_OPENMODE:
3551         case -NFS4ERR_EXPIRED:
3552                 return true;
3553         }
3554         return false;
3555 }
3556
3557 void __nfs4_read_done_cb(struct nfs_read_data *data)
3558 {
3559         nfs_invalidate_atime(data->header->inode);
3560 }
3561
3562 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3563 {
3564         struct nfs_server *server = NFS_SERVER(data->header->inode);
3565
3566         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3567                 rpc_restart_call_prepare(task);
3568                 return -EAGAIN;
3569         }
3570
3571         __nfs4_read_done_cb(data);
3572         if (task->tk_status > 0)
3573                 renew_lease(server, data->timestamp);
3574         return 0;
3575 }
3576
3577 static bool nfs4_read_stateid_changed(struct rpc_task *task,
3578                 struct nfs_readargs *args)
3579 {
3580
3581         if (!nfs4_error_stateid_expired(task->tk_status) ||
3582                 nfs4_stateid_is_current(&args->stateid,
3583                                 args->context,
3584                                 args->lock_context,
3585                                 FMODE_READ))
3586                 return false;
3587         rpc_restart_call_prepare(task);
3588         return true;
3589 }
3590
3591 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3592 {
3593
3594         dprintk("--> %s\n", __func__);
3595
3596         if (!nfs4_sequence_done(task, &data->res.seq_res))
3597                 return -EAGAIN;
3598         if (nfs4_read_stateid_changed(task, &data->args))
3599                 return -EAGAIN;
3600         return data->read_done_cb ? data->read_done_cb(task, data) :
3601                                     nfs4_read_done_cb(task, data);
3602 }
3603
3604 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3605 {
3606         data->timestamp   = jiffies;
3607         data->read_done_cb = nfs4_read_done_cb;
3608         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3609         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3610 }
3611
3612 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3613 {
3614         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3615                         &data->args.seq_args,
3616                         &data->res.seq_res,
3617                         task))
3618                 return;
3619         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3620                         data->args.lock_context, FMODE_READ);
3621 }
3622
3623 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3624 {
3625         struct inode *inode = data->header->inode;
3626
3627         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3628                 rpc_restart_call_prepare(task);
3629                 return -EAGAIN;
3630         }
3631         if (task->tk_status >= 0) {
3632                 renew_lease(NFS_SERVER(inode), data->timestamp);
3633                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3634         }
3635         return 0;
3636 }
3637
3638 static bool nfs4_write_stateid_changed(struct rpc_task *task,
3639                 struct nfs_writeargs *args)
3640 {
3641
3642         if (!nfs4_error_stateid_expired(task->tk_status) ||
3643                 nfs4_stateid_is_current(&args->stateid,
3644                                 args->context,
3645                                 args->lock_context,
3646                                 FMODE_WRITE))
3647                 return false;
3648         rpc_restart_call_prepare(task);
3649         return true;
3650 }
3651
3652 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3653 {
3654         if (!nfs4_sequence_done(task, &data->res.seq_res))
3655                 return -EAGAIN;
3656         if (nfs4_write_stateid_changed(task, &data->args))
3657                 return -EAGAIN;
3658         return data->write_done_cb ? data->write_done_cb(task, data) :
3659                 nfs4_write_done_cb(task, data);
3660 }
3661
3662 static
3663 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3664 {
3665         const struct nfs_pgio_header *hdr = data->header;
3666
3667         /* Don't request attributes for pNFS or O_DIRECT writes */
3668         if (data->ds_clp != NULL || hdr->dreq != NULL)
3669                 return false;
3670         /* Otherwise, request attributes if and only if we don't hold
3671          * a delegation
3672          */
3673         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3674 }
3675
3676 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3677 {
3678         struct nfs_server *server = NFS_SERVER(data->header->inode);
3679
3680         if (!nfs4_write_need_cache_consistency_data(data)) {
3681                 data->args.bitmask = NULL;
3682                 data->res.fattr = NULL;
3683         } else
3684                 data->args.bitmask = server->cache_consistency_bitmask;
3685
3686         if (!data->write_done_cb)
3687                 data->write_done_cb = nfs4_write_done_cb;
3688         data->res.server = server;
3689         data->timestamp   = jiffies;
3690
3691         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3692         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3693 }
3694
3695 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3696 {
3697         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3698                         &data->args.seq_args,
3699                         &data->res.seq_res,
3700                         task))
3701                 return;
3702         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3703                         data->args.lock_context, FMODE_WRITE);
3704 }
3705
3706 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3707 {
3708         nfs4_setup_sequence(NFS_SERVER(data->inode),
3709                         &data->args.seq_args,
3710                         &data->res.seq_res,
3711                         task);
3712 }
3713
3714 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3715 {
3716         struct inode *inode = data->inode;
3717
3718         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3719                 rpc_restart_call_prepare(task);
3720                 return -EAGAIN;
3721         }
3722         return 0;
3723 }
3724
3725 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3726 {
3727         if (!nfs4_sequence_done(task, &data->res.seq_res))
3728                 return -EAGAIN;
3729         return data->commit_done_cb(task, data);
3730 }
3731
3732 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3733 {
3734         struct nfs_server *server = NFS_SERVER(data->inode);
3735
3736         if (data->commit_done_cb == NULL)
3737                 data->commit_done_cb = nfs4_commit_done_cb;
3738         data->res.server = server;
3739         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3740         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3741 }
3742
3743 struct nfs4_renewdata {
3744         struct nfs_client       *client;
3745         unsigned long           timestamp;
3746 };
3747
3748 /*
3749  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3750  * standalone procedure for queueing an asynchronous RENEW.
3751  */
3752 static void nfs4_renew_release(void *calldata)
3753 {
3754         struct nfs4_renewdata *data = calldata;
3755         struct nfs_client *clp = data->client;
3756
3757         if (atomic_read(&clp->cl_count) > 1)
3758                 nfs4_schedule_state_renewal(clp);
3759         nfs_put_client(clp);
3760         kfree(data);
3761 }
3762
3763 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3764 {
3765         struct nfs4_renewdata *data = calldata;
3766         struct nfs_client *clp = data->client;
3767         unsigned long timestamp = data->timestamp;
3768
3769         if (task->tk_status < 0) {
3770                 /* Unless we're shutting down, schedule state recovery! */
3771                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3772                         return;
3773                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3774                         nfs4_schedule_lease_recovery(clp);
3775                         return;
3776                 }
3777                 nfs4_schedule_path_down_recovery(clp);
3778         }
3779         do_renew_lease(clp, timestamp);
3780 }
3781
3782 static const struct rpc_call_ops nfs4_renew_ops = {
3783         .rpc_call_done = nfs4_renew_done,
3784         .rpc_release = nfs4_renew_release,
3785 };
3786
3787 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3788 {
3789         struct rpc_message msg = {
3790                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3791                 .rpc_argp       = clp,
3792                 .rpc_cred       = cred,
3793         };
3794         struct nfs4_renewdata *data;
3795
3796         if (renew_flags == 0)
3797                 return 0;
3798         if (!atomic_inc_not_zero(&clp->cl_count))
3799                 return -EIO;
3800         data = kmalloc(sizeof(*data), GFP_NOFS);
3801         if (data == NULL)
3802                 return -ENOMEM;
3803         data->client = clp;
3804         data->timestamp = jiffies;
3805         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3806                         &nfs4_renew_ops, data);
3807 }
3808
3809 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3810 {
3811         struct rpc_message msg = {
3812                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3813                 .rpc_argp       = clp,
3814                 .rpc_cred       = cred,
3815         };
3816         unsigned long now = jiffies;
3817         int status;
3818
3819         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3820         if (status < 0)
3821                 return status;
3822         do_renew_lease(clp, now);
3823         return 0;
3824 }
3825
3826 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3827 {
3828         return (server->caps & NFS_CAP_ACLS)
3829                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3830                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3831 }
3832
3833 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3834  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3835  * the stack.
3836  */
3837 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3838
3839 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3840                 struct page **pages, unsigned int *pgbase)
3841 {
3842         struct page *newpage, **spages;
3843         int rc = 0;
3844         size_t len;
3845         spages = pages;
3846
3847         do {
3848                 len = min_t(size_t, PAGE_SIZE, buflen);
3849                 newpage = alloc_page(GFP_KERNEL);
3850
3851                 if (newpage == NULL)
3852                         goto unwind;
3853                 memcpy(page_address(newpage), buf, len);
3854                 buf += len;
3855                 buflen -= len;
3856                 *pages++ = newpage;
3857                 rc++;
3858         } while (buflen != 0);
3859
3860         return rc;
3861
3862 unwind:
3863         for(; rc > 0; rc--)
3864                 __free_page(spages[rc-1]);
3865         return -ENOMEM;
3866 }
3867
3868 struct nfs4_cached_acl {
3869         int cached;
3870         size_t len;
3871         char data[0];
3872 };
3873
3874 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3875 {
3876         struct nfs_inode *nfsi = NFS_I(inode);
3877
3878         spin_lock(&inode->i_lock);
3879         kfree(nfsi->nfs4_acl);
3880         nfsi->nfs4_acl = acl;
3881         spin_unlock(&inode->i_lock);
3882 }
3883
3884 static void nfs4_zap_acl_attr(struct inode *inode)
3885 {
3886         nfs4_set_cached_acl(inode, NULL);
3887 }
3888
3889 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3890 {
3891         struct nfs_inode *nfsi = NFS_I(inode);
3892         struct nfs4_cached_acl *acl;
3893         int ret = -ENOENT;
3894
3895         spin_lock(&inode->i_lock);
3896         acl = nfsi->nfs4_acl;
3897         if (acl == NULL)
3898                 goto out;
3899         if (buf == NULL) /* user is just asking for length */
3900                 goto out_len;
3901         if (acl->cached == 0)
3902                 goto out;
3903         ret = -ERANGE; /* see getxattr(2) man page */
3904         if (acl->len > buflen)
3905                 goto out;
3906         memcpy(buf, acl->data, acl->len);
3907 out_len:
3908         ret = acl->len;
3909 out:
3910         spin_unlock(&inode->i_lock);
3911         return ret;
3912 }
3913
3914 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3915 {
3916         struct nfs4_cached_acl *acl;
3917         size_t buflen = sizeof(*acl) + acl_len;
3918
3919         if (buflen <= PAGE_SIZE) {
3920                 acl = kmalloc(buflen, GFP_KERNEL);
3921                 if (acl == NULL)
3922                         goto out;
3923                 acl->cached = 1;
3924                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3925         } else {
3926                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3927                 if (acl == NULL)
3928                         goto out;
3929                 acl->cached = 0;
3930         }
3931         acl->len = acl_len;
3932 out:
3933         nfs4_set_cached_acl(inode, acl);
3934 }
3935
3936 /*
3937  * The getxattr API returns the required buffer length when called with a
3938  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3939  * the required buf.  On a NULL buf, we send a page of data to the server
3940  * guessing that the ACL request can be serviced by a page. If so, we cache
3941  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3942  * the cache. If not so, we throw away the page, and cache the required
3943  * length. The next getxattr call will then produce another round trip to
3944  * the server, this time with the input buf of the required size.
3945  */
3946 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3947 {
3948         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3949         struct nfs_getaclargs args = {
3950                 .fh = NFS_FH(inode),
3951                 .acl_pages = pages,
3952                 .acl_len = buflen,
3953         };
3954         struct nfs_getaclres res = {
3955                 .acl_len = buflen,
3956         };
3957         struct rpc_message msg = {
3958                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3959                 .rpc_argp = &args,
3960                 .rpc_resp = &res,
3961         };
3962         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3963         int ret = -ENOMEM, i;
3964
3965         /* As long as we're doing a round trip to the server anyway,
3966          * let's be prepared for a page of acl data. */
3967         if (npages == 0)
3968                 npages = 1;
3969         if (npages > ARRAY_SIZE(pages))
3970                 return -ERANGE;
3971
3972         for (i = 0; i < npages; i++) {
3973                 pages[i] = alloc_page(GFP_KERNEL);
3974                 if (!pages[i])
3975                         goto out_free;
3976         }
3977
3978         /* for decoding across pages */
3979         res.acl_scratch = alloc_page(GFP_KERNEL);
3980         if (!res.acl_scratch)
3981                 goto out_free;
3982
3983         args.acl_len = npages * PAGE_SIZE;
3984         args.acl_pgbase = 0;
3985
3986         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3987                 __func__, buf, buflen, npages, args.acl_len);
3988         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3989                              &msg, &args.seq_args, &res.seq_res, 0);
3990         if (ret)
3991                 goto out_free;
3992
3993         /* Handle the case where the passed-in buffer is too short */
3994         if (res.acl_flags & NFS4_ACL_TRUNC) {
3995                 /* Did the user only issue a request for the acl length? */
3996                 if (buf == NULL)
3997                         goto out_ok;
3998                 ret = -ERANGE;
3999                 goto out_free;
4000         }
4001         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4002         if (buf) {
4003                 if (res.acl_len > buflen) {
4004                         ret = -ERANGE;
4005                         goto out_free;
4006                 }
4007                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4008         }
4009 out_ok:
4010         ret = res.acl_len;
4011 out_free:
4012         for (i = 0; i < npages; i++)
4013                 if (pages[i])
4014                         __free_page(pages[i]);
4015         if (res.acl_scratch)
4016                 __free_page(res.acl_scratch);
4017         return ret;
4018 }
4019
4020 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4021 {
4022         struct nfs4_exception exception = { };
4023         ssize_t ret;
4024         do {
4025                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4026                 if (ret >= 0)
4027                         break;
4028                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4029         } while (exception.retry);
4030         return ret;
4031 }
4032
4033 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4034 {
4035         struct nfs_server *server = NFS_SERVER(inode);
4036         int ret;
4037
4038         if (!nfs4_server_supports_acls(server))
4039                 return -EOPNOTSUPP;
4040         ret = nfs_revalidate_inode(server, inode);
4041         if (ret < 0)
4042                 return ret;
4043         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4044                 nfs_zap_acl_cache(inode);
4045         ret = nfs4_read_cached_acl(inode, buf, buflen);
4046         if (ret != -ENOENT)
4047                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4048                  * but no cached acl data, just the acl length */
4049                 return ret;
4050         return nfs4_get_acl_uncached(inode, buf, buflen);
4051 }
4052
4053 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4054 {
4055         struct nfs_server *server = NFS_SERVER(inode);
4056         struct page *pages[NFS4ACL_MAXPAGES];
4057         struct nfs_setaclargs arg = {
4058                 .fh             = NFS_FH(inode),
4059                 .acl_pages      = pages,
4060                 .acl_len        = buflen,
4061         };
4062         struct nfs_setaclres res;
4063         struct rpc_message msg = {
4064                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4065                 .rpc_argp       = &arg,
4066                 .rpc_resp       = &res,
4067         };
4068         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4069         int ret, i;
4070
4071         if (!nfs4_server_supports_acls(server))
4072                 return -EOPNOTSUPP;
4073         if (npages > ARRAY_SIZE(pages))
4074                 return -ERANGE;
4075         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4076         if (i < 0)
4077                 return i;
4078         nfs4_inode_return_delegation(inode);
4079         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4080
4081         /*
4082          * Free each page after tx, so the only ref left is
4083          * held by the network stack
4084          */
4085         for (; i > 0; i--)
4086                 put_page(pages[i-1]);
4087
4088         /*
4089          * Acl update can result in inode attribute update.
4090          * so mark the attribute cache invalid.
4091          */
4092         spin_lock(&inode->i_lock);
4093         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4094         spin_unlock(&inode->i_lock);
4095         nfs_access_zap_cache(inode);
4096         nfs_zap_acl_cache(inode);
4097         return ret;
4098 }
4099
4100 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4101 {
4102         struct nfs4_exception exception = { };
4103         int err;
4104         do {
4105                 err = nfs4_handle_exception(NFS_SERVER(inode),
4106                                 __nfs4_proc_set_acl(inode, buf, buflen),
4107                                 &exception);
4108         } while (exception.retry);
4109         return err;
4110 }
4111
4112 static int
4113 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4114 {
4115         struct nfs_client *clp = server->nfs_client;
4116
4117         if (task->tk_status >= 0)
4118                 return 0;
4119         switch(task->tk_status) {
4120                 case -NFS4ERR_DELEG_REVOKED:
4121                 case -NFS4ERR_ADMIN_REVOKED:
4122                 case -NFS4ERR_BAD_STATEID:
4123                         if (state == NULL)
4124                                 break;
4125                         nfs_remove_bad_delegation(state->inode);
4126                 case -NFS4ERR_OPENMODE:
4127                         if (state == NULL)
4128                                 break;
4129                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4130                                 goto stateid_invalid;
4131                         goto wait_on_recovery;
4132                 case -NFS4ERR_EXPIRED:
4133                         if (state != NULL) {
4134                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4135                                         goto stateid_invalid;
4136                         }
4137                 case -NFS4ERR_STALE_STATEID:
4138                 case -NFS4ERR_STALE_CLIENTID:
4139                         nfs4_schedule_lease_recovery(clp);
4140                         goto wait_on_recovery;
4141 #if defined(CONFIG_NFS_V4_1)
4142                 case -NFS4ERR_BADSESSION:
4143                 case -NFS4ERR_BADSLOT:
4144                 case -NFS4ERR_BAD_HIGH_SLOT:
4145                 case -NFS4ERR_DEADSESSION:
4146                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4147                 case -NFS4ERR_SEQ_FALSE_RETRY:
4148                 case -NFS4ERR_SEQ_MISORDERED:
4149                         dprintk("%s ERROR %d, Reset session\n", __func__,
4150                                 task->tk_status);
4151                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4152                         task->tk_status = 0;
4153                         return -EAGAIN;
4154 #endif /* CONFIG_NFS_V4_1 */
4155                 case -NFS4ERR_DELAY:
4156                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4157                 case -NFS4ERR_GRACE:
4158                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4159                         task->tk_status = 0;
4160                         return -EAGAIN;
4161                 case -NFS4ERR_RETRY_UNCACHED_REP:
4162                 case -NFS4ERR_OLD_STATEID:
4163                         task->tk_status = 0;
4164                         return -EAGAIN;
4165         }
4166         task->tk_status = nfs4_map_errors(task->tk_status);
4167         return 0;
4168 stateid_invalid:
4169         task->tk_status = -EIO;
4170         return 0;
4171 wait_on_recovery:
4172         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4173         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4174                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4175         task->tk_status = 0;
4176         return -EAGAIN;
4177 }
4178
4179 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4180                                     nfs4_verifier *bootverf)
4181 {
4182         __be32 verf[2];
4183
4184         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4185                 /* An impossible timestamp guarantees this value
4186                  * will never match a generated boot time. */
4187                 verf[0] = 0;
4188                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4189         } else {
4190                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4191                 verf[0] = (__be32)nn->boot_time.tv_sec;
4192                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4193         }
4194         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4195 }
4196
4197 static unsigned int
4198 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4199                                    char *buf, size_t len)
4200 {
4201         unsigned int result;
4202
4203         rcu_read_lock();
4204         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4205                                 clp->cl_ipaddr,
4206                                 rpc_peeraddr2str(clp->cl_rpcclient,
4207                                                         RPC_DISPLAY_ADDR),
4208                                 rpc_peeraddr2str(clp->cl_rpcclient,
4209                                                         RPC_DISPLAY_PROTO));
4210         rcu_read_unlock();
4211         return result;
4212 }
4213
4214 static unsigned int
4215 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4216                                 char *buf, size_t len)
4217 {
4218         char *nodename = clp->cl_rpcclient->cl_nodename;
4219
4220         if (nfs4_client_id_uniquifier[0] != '\0')
4221                 nodename = nfs4_client_id_uniquifier;
4222         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4223                                 clp->rpc_ops->version, clp->cl_minorversion,
4224                                 nodename);
4225 }
4226
4227 /**
4228  * nfs4_proc_setclientid - Negotiate client ID
4229  * @clp: state data structure
4230  * @program: RPC program for NFSv4 callback service
4231  * @port: IP port number for NFS4 callback service
4232  * @cred: RPC credential to use for this call
4233  * @res: where to place the result
4234  *
4235  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4236  */
4237 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4238                 unsigned short port, struct rpc_cred *cred,
4239                 struct nfs4_setclientid_res *res)
4240 {
4241         nfs4_verifier sc_verifier;
4242         struct nfs4_setclientid setclientid = {
4243                 .sc_verifier = &sc_verifier,
4244                 .sc_prog = program,
4245                 .sc_cb_ident = clp->cl_cb_ident,
4246         };
4247         struct rpc_message msg = {
4248                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4249                 .rpc_argp = &setclientid,
4250                 .rpc_resp = res,
4251                 .rpc_cred = cred,
4252         };
4253         int status;
4254
4255         /* nfs_client_id4 */
4256         nfs4_init_boot_verifier(clp, &sc_verifier);
4257         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4258                 setclientid.sc_name_len =
4259                                 nfs4_init_uniform_client_string(clp,
4260                                                 setclientid.sc_name,
4261                                                 sizeof(setclientid.sc_name));
4262         else
4263                 setclientid.sc_name_len =
4264                                 nfs4_init_nonuniform_client_string(clp,
4265                                                 setclientid.sc_name,
4266                                                 sizeof(setclientid.sc_name));
4267         /* cb_client4 */
4268         rcu_read_lock();
4269         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4270                                 sizeof(setclientid.sc_netid),
4271                                 rpc_peeraddr2str(clp->cl_rpcclient,
4272                                                         RPC_DISPLAY_NETID));
4273         rcu_read_unlock();
4274         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4275                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4276                                 clp->cl_ipaddr, port >> 8, port & 255);
4277
4278         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4279                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4280                 setclientid.sc_name_len, setclientid.sc_name);
4281         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4282         dprintk("NFS reply setclientid: %d\n", status);
4283         return status;
4284 }
4285
4286 /**
4287  * nfs4_proc_setclientid_confirm - Confirm client ID
4288  * @clp: state data structure
4289  * @res: result of a previous SETCLIENTID
4290  * @cred: RPC credential to use for this call
4291  *
4292  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4293  */
4294 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4295                 struct nfs4_setclientid_res *arg,
4296                 struct rpc_cred *cred)
4297 {
4298         struct nfs_fsinfo fsinfo;
4299         struct rpc_message msg = {
4300                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4301                 .rpc_argp = arg,
4302                 .rpc_resp = &fsinfo,
4303                 .rpc_cred = cred,
4304         };
4305         unsigned long now;
4306         int status;
4307
4308         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4309                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4310                 clp->cl_clientid);
4311         now = jiffies;
4312         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4313         if (status == 0) {
4314                 spin_lock(&clp->cl_lock);
4315                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4316                 clp->cl_last_renewal = now;
4317                 spin_unlock(&clp->cl_lock);
4318         }
4319         dprintk("NFS reply setclientid_confirm: %d\n", status);
4320         return status;
4321 }
4322
4323 struct nfs4_delegreturndata {
4324         struct nfs4_delegreturnargs args;
4325         struct nfs4_delegreturnres res;
4326         struct nfs_fh fh;
4327         nfs4_stateid stateid;
4328         unsigned long timestamp;
4329         struct nfs_fattr fattr;
4330         int rpc_status;
4331 };
4332
4333 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4334 {
4335         struct nfs4_delegreturndata *data = calldata;
4336
4337         if (!nfs4_sequence_done(task, &data->res.seq_res))
4338                 return;
4339
4340         switch (task->tk_status) {
4341         case -NFS4ERR_STALE_STATEID:
4342         case -NFS4ERR_EXPIRED:
4343         case 0:
4344                 renew_lease(data->res.server, data->timestamp);
4345                 break;
4346         default:
4347                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4348                                 -EAGAIN) {
4349                         rpc_restart_call_prepare(task);
4350                         return;
4351                 }
4352         }
4353         data->rpc_status = task->tk_status;
4354 }
4355
4356 static void nfs4_delegreturn_release(void *calldata)
4357 {
4358         kfree(calldata);
4359 }
4360
4361 #if defined(CONFIG_NFS_V4_1)
4362 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4363 {
4364         struct nfs4_delegreturndata *d_data;
4365
4366         d_data = (struct nfs4_delegreturndata *)data;
4367
4368         nfs4_setup_sequence(d_data->res.server,
4369                         &d_data->args.seq_args,
4370                         &d_data->res.seq_res,
4371                         task);
4372 }
4373 #endif /* CONFIG_NFS_V4_1 */
4374
4375 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4376 #if defined(CONFIG_NFS_V4_1)
4377         .rpc_call_prepare = nfs4_delegreturn_prepare,
4378 #endif /* CONFIG_NFS_V4_1 */
4379         .rpc_call_done = nfs4_delegreturn_done,
4380         .rpc_release = nfs4_delegreturn_release,
4381 };
4382
4383 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4384 {
4385         struct nfs4_delegreturndata *data;
4386         struct nfs_server *server = NFS_SERVER(inode);
4387         struct rpc_task *task;
4388         struct rpc_message msg = {
4389                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4390                 .rpc_cred = cred,
4391         };
4392         struct rpc_task_setup task_setup_data = {
4393                 .rpc_client = server->client,
4394                 .rpc_message = &msg,
4395                 .callback_ops = &nfs4_delegreturn_ops,
4396                 .flags = RPC_TASK_ASYNC,
4397         };
4398         int status = 0;
4399
4400         data = kzalloc(sizeof(*data), GFP_NOFS);
4401         if (data == NULL)
4402                 return -ENOMEM;
4403         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4404         data->args.fhandle = &data->fh;
4405         data->args.stateid = &data->stateid;
4406         data->args.bitmask = server->cache_consistency_bitmask;
4407         nfs_copy_fh(&data->fh, NFS_FH(inode));
4408         nfs4_stateid_copy(&data->stateid, stateid);
4409         data->res.fattr = &data->fattr;
4410         data->res.server = server;
4411         nfs_fattr_init(data->res.fattr);
4412         data->timestamp = jiffies;
4413         data->rpc_status = 0;
4414
4415         task_setup_data.callback_data = data;
4416         msg.rpc_argp = &data->args;
4417         msg.rpc_resp = &data->res;
4418         task = rpc_run_task(&task_setup_data);
4419         if (IS_ERR(task))
4420                 return PTR_ERR(task);
4421         if (!issync)
4422                 goto out;
4423         status = nfs4_wait_for_completion_rpc_task(task);
4424         if (status != 0)
4425                 goto out;
4426         status = data->rpc_status;
4427         if (status == 0)
4428                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4429         else
4430                 nfs_refresh_inode(inode, &data->fattr);
4431 out:
4432         rpc_put_task(task);
4433         return status;
4434 }
4435
4436 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4437 {
4438         struct nfs_server *server = NFS_SERVER(inode);
4439         struct nfs4_exception exception = { };
4440         int err;
4441         do {
4442                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4443                 switch (err) {
4444                         case -NFS4ERR_STALE_STATEID:
4445                         case -NFS4ERR_EXPIRED:
4446                         case 0:
4447                                 return 0;
4448                 }
4449                 err = nfs4_handle_exception(server, err, &exception);
4450         } while (exception.retry);
4451         return err;
4452 }
4453
4454 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4455 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4456
4457 /*
4458  * sleep, with exponential backoff, and retry the LOCK operation.
4459  */
4460 static unsigned long
4461 nfs4_set_lock_task_retry(unsigned long timeout)
4462 {
4463         freezable_schedule_timeout_killable(timeout);
4464         timeout <<= 1;
4465         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4466                 return NFS4_LOCK_MAXTIMEOUT;
4467         return timeout;
4468 }
4469
4470 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4471 {
4472         struct inode *inode = state->inode;
4473         struct nfs_server *server = NFS_SERVER(inode);
4474         struct nfs_client *clp = server->nfs_client;
4475         struct nfs_lockt_args arg = {
4476                 .fh = NFS_FH(inode),
4477                 .fl = request,
4478         };
4479         struct nfs_lockt_res res = {
4480                 .denied = request,
4481         };
4482         struct rpc_message msg = {
4483                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4484                 .rpc_argp       = &arg,
4485                 .rpc_resp       = &res,
4486                 .rpc_cred       = state->owner->so_cred,
4487         };
4488         struct nfs4_lock_state *lsp;
4489         int status;
4490
4491         arg.lock_owner.clientid = clp->cl_clientid;
4492         status = nfs4_set_lock_state(state, request);
4493         if (status != 0)
4494                 goto out;
4495         lsp = request->fl_u.nfs4_fl.owner;
4496         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4497         arg.lock_owner.s_dev = server->s_dev;
4498         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4499         switch (status) {
4500                 case 0:
4501                         request->fl_type = F_UNLCK;
4502                         break;
4503                 case -NFS4ERR_DENIED:
4504                         status = 0;
4505         }
4506         request->fl_ops->fl_release_private(request);
4507 out:
4508         return status;
4509 }
4510
4511 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4512 {
4513         struct nfs4_exception exception = { };
4514         int err;
4515
4516         do {
4517                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4518                                 _nfs4_proc_getlk(state, cmd, request),
4519                                 &exception);
4520         } while (exception.retry);
4521         return err;
4522 }
4523
4524 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4525 {
4526         int res = 0;
4527         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4528                 case FL_POSIX:
4529                         res = posix_lock_file_wait(file, fl);
4530                         break;
4531                 case FL_FLOCK:
4532                         res = flock_lock_file_wait(file, fl);
4533                         break;
4534                 default:
4535                         BUG();
4536         }
4537         return res;
4538 }
4539
4540 struct nfs4_unlockdata {
4541         struct nfs_locku_args arg;
4542         struct nfs_locku_res res;
4543         struct nfs4_lock_state *lsp;
4544         struct nfs_open_context *ctx;
4545         struct file_lock fl;
4546         const struct nfs_server *server;
4547         unsigned long timestamp;
4548 };
4549
4550 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4551                 struct nfs_open_context *ctx,
4552                 struct nfs4_lock_state *lsp,
4553                 struct nfs_seqid *seqid)
4554 {
4555         struct nfs4_unlockdata *p;
4556         struct inode *inode = lsp->ls_state->inode;
4557
4558         p = kzalloc(sizeof(*p), GFP_NOFS);
4559         if (p == NULL)
4560                 return NULL;
4561         p->arg.fh = NFS_FH(inode);
4562         p->arg.fl = &p->fl;
4563         p->arg.seqid = seqid;
4564         p->res.seqid = seqid;
4565         p->arg.stateid = &lsp->ls_stateid;
4566         p->lsp = lsp;
4567         atomic_inc(&lsp->ls_count);
4568         /* Ensure we don't close file until we're done freeing locks! */
4569         p->ctx = get_nfs_open_context(ctx);
4570         memcpy(&p->fl, fl, sizeof(p->fl));
4571         p->server = NFS_SERVER(inode);
4572         return p;
4573 }
4574
4575 static void nfs4_locku_release_calldata(void *data)
4576 {
4577         struct nfs4_unlockdata *calldata = data;
4578         nfs_free_seqid(calldata->arg.seqid);
4579         nfs4_put_lock_state(calldata->lsp);
4580         put_nfs_open_context(calldata->ctx);
4581         kfree(calldata);
4582 }
4583
4584 static void nfs4_locku_done(struct rpc_task *task, void *data)
4585 {
4586         struct nfs4_unlockdata *calldata = data;
4587
4588         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4589                 return;
4590         switch (task->tk_status) {
4591                 case 0:
4592                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4593                                         &calldata->res.stateid);
4594                         renew_lease(calldata->server, calldata->timestamp);
4595                         break;
4596                 case -NFS4ERR_BAD_STATEID:
4597                 case -NFS4ERR_OLD_STATEID:
4598                 case -NFS4ERR_STALE_STATEID:
4599                 case -NFS4ERR_EXPIRED:
4600                         break;
4601                 default:
4602                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4603                                 rpc_restart_call_prepare(task);
4604         }
4605         nfs_release_seqid(calldata->arg.seqid);
4606 }
4607
4608 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4609 {
4610         struct nfs4_unlockdata *calldata = data;
4611
4612         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4613                 goto out_wait;
4614         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4615                 /* Note: exit _without_ running nfs4_locku_done */
4616                 goto out_no_action;
4617         }
4618         calldata->timestamp = jiffies;
4619         if (nfs4_setup_sequence(calldata->server,
4620                                 &calldata->arg.seq_args,
4621                                 &calldata->res.seq_res,
4622                                 task) != 0)
4623                 nfs_release_seqid(calldata->arg.seqid);
4624         return;
4625 out_no_action:
4626         task->tk_action = NULL;
4627 out_wait:
4628         nfs4_sequence_done(task, &calldata->res.seq_res);
4629 }
4630
4631 static const struct rpc_call_ops nfs4_locku_ops = {
4632         .rpc_call_prepare = nfs4_locku_prepare,
4633         .rpc_call_done = nfs4_locku_done,
4634         .rpc_release = nfs4_locku_release_calldata,
4635 };
4636
4637 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4638                 struct nfs_open_context *ctx,
4639                 struct nfs4_lock_state *lsp,
4640                 struct nfs_seqid *seqid)
4641 {
4642         struct nfs4_unlockdata *data;
4643         struct rpc_message msg = {
4644                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4645                 .rpc_cred = ctx->cred,
4646         };
4647         struct rpc_task_setup task_setup_data = {
4648                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4649                 .rpc_message = &msg,
4650                 .callback_ops = &nfs4_locku_ops,
4651                 .workqueue = nfsiod_workqueue,
4652                 .flags = RPC_TASK_ASYNC,
4653         };
4654
4655         /* Ensure this is an unlock - when canceling a lock, the
4656          * canceled lock is passed in, and it won't be an unlock.
4657          */
4658         fl->fl_type = F_UNLCK;
4659
4660         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4661         if (data == NULL) {
4662                 nfs_free_seqid(seqid);
4663                 return ERR_PTR(-ENOMEM);
4664         }
4665
4666         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4667         msg.rpc_argp = &data->arg;
4668         msg.rpc_resp = &data->res;
4669         task_setup_data.callback_data = data;
4670         return rpc_run_task(&task_setup_data);
4671 }
4672
4673 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4674 {
4675         struct inode *inode = state->inode;
4676         struct nfs4_state_owner *sp = state->owner;
4677         struct nfs_inode *nfsi = NFS_I(inode);
4678         struct nfs_seqid *seqid;
4679         struct nfs4_lock_state *lsp;
4680         struct rpc_task *task;
4681         int status = 0;
4682         unsigned char fl_flags = request->fl_flags;
4683
4684         status = nfs4_set_lock_state(state, request);
4685         /* Unlock _before_ we do the RPC call */
4686         request->fl_flags |= FL_EXISTS;
4687         /* Exclude nfs_delegation_claim_locks() */
4688         mutex_lock(&sp->so_delegreturn_mutex);
4689         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
4690         down_read(&nfsi->rwsem);
4691         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4692                 up_read(&nfsi->rwsem);
4693                 mutex_unlock(&sp->so_delegreturn_mutex);
4694                 goto out;
4695         }
4696         up_read(&nfsi->rwsem);
4697         mutex_unlock(&sp->so_delegreturn_mutex);
4698         if (status != 0)
4699                 goto out;
4700         /* Is this a delegated lock? */
4701         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4702                 goto out;
4703         lsp = request->fl_u.nfs4_fl.owner;
4704         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4705         status = -ENOMEM;
4706         if (seqid == NULL)
4707                 goto out;
4708         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4709         status = PTR_ERR(task);
4710         if (IS_ERR(task))
4711                 goto out;
4712         status = nfs4_wait_for_completion_rpc_task(task);
4713         rpc_put_task(task);
4714 out:
4715         request->fl_flags = fl_flags;
4716         return status;
4717 }
4718
4719 struct nfs4_lockdata {
4720         struct nfs_lock_args arg;
4721         struct nfs_lock_res res;
4722         struct nfs4_lock_state *lsp;
4723         struct nfs_open_context *ctx;
4724         struct file_lock fl;
4725         unsigned long timestamp;
4726         int rpc_status;
4727         int cancelled;
4728         struct nfs_server *server;
4729 };
4730
4731 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4732                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4733                 gfp_t gfp_mask)
4734 {
4735         struct nfs4_lockdata *p;
4736         struct inode *inode = lsp->ls_state->inode;
4737         struct nfs_server *server = NFS_SERVER(inode);
4738
4739         p = kzalloc(sizeof(*p), gfp_mask);
4740         if (p == NULL)
4741                 return NULL;
4742
4743         p->arg.fh = NFS_FH(inode);
4744         p->arg.fl = &p->fl;
4745         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4746         if (p->arg.open_seqid == NULL)
4747                 goto out_free;
4748         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4749         if (p->arg.lock_seqid == NULL)
4750                 goto out_free_seqid;
4751         p->arg.lock_stateid = &lsp->ls_stateid;
4752         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4753         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4754         p->arg.lock_owner.s_dev = server->s_dev;
4755         p->res.lock_seqid = p->arg.lock_seqid;
4756         p->lsp = lsp;
4757         p->server = server;
4758         atomic_inc(&lsp->ls_count);
4759         p->ctx = get_nfs_open_context(ctx);
4760         memcpy(&p->fl, fl, sizeof(p->fl));
4761         return p;
4762 out_free_seqid:
4763         nfs_free_seqid(p->arg.open_seqid);
4764 out_free:
4765         kfree(p);
4766         return NULL;
4767 }
4768
4769 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4770 {
4771         struct nfs4_lockdata *data = calldata;
4772         struct nfs4_state *state = data->lsp->ls_state;
4773
4774         dprintk("%s: begin!\n", __func__);
4775         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4776                 goto out_wait;
4777         /* Do we need to do an open_to_lock_owner? */
4778         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4779                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
4780                         goto out_release_lock_seqid;
4781                 }
4782                 data->arg.open_stateid = &state->stateid;
4783                 data->arg.new_lock_owner = 1;
4784                 data->res.open_seqid = data->arg.open_seqid;
4785         } else
4786                 data->arg.new_lock_owner = 0;
4787         if (!nfs4_valid_open_stateid(state)) {
4788                 data->rpc_status = -EBADF;
4789                 task->tk_action = NULL;
4790                 goto out_release_open_seqid;
4791         }
4792         data->timestamp = jiffies;
4793         if (nfs4_setup_sequence(data->server,
4794                                 &data->arg.seq_args,
4795                                 &data->res.seq_res,
4796                                 task) == 0)
4797                 return;
4798 out_release_open_seqid:
4799         nfs_release_seqid(data->arg.open_seqid);
4800 out_release_lock_seqid:
4801         nfs_release_seqid(data->arg.lock_seqid);
4802 out_wait:
4803         nfs4_sequence_done(task, &data->res.seq_res);
4804         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4805 }
4806
4807 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4808 {
4809         struct nfs4_lockdata *data = calldata;
4810
4811         dprintk("%s: begin!\n", __func__);
4812
4813         if (!nfs4_sequence_done(task, &data->res.seq_res))
4814                 return;
4815
4816         data->rpc_status = task->tk_status;
4817         if (data->arg.new_lock_owner != 0) {
4818                 if (data->rpc_status == 0)
4819                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4820                 else
4821                         goto out;
4822         }
4823         if (data->rpc_status == 0) {
4824                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4825                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4826                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4827         }
4828 out:
4829         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4830 }
4831
4832 static void nfs4_lock_release(void *calldata)
4833 {
4834         struct nfs4_lockdata *data = calldata;
4835
4836         dprintk("%s: begin!\n", __func__);
4837         nfs_free_seqid(data->arg.open_seqid);
4838         if (data->cancelled != 0) {
4839                 struct rpc_task *task;
4840                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4841                                 data->arg.lock_seqid);
4842                 if (!IS_ERR(task))
4843                         rpc_put_task_async(task);
4844                 dprintk("%s: cancelling lock!\n", __func__);
4845         } else
4846                 nfs_free_seqid(data->arg.lock_seqid);
4847         nfs4_put_lock_state(data->lsp);
4848         put_nfs_open_context(data->ctx);
4849         kfree(data);
4850         dprintk("%s: done!\n", __func__);
4851 }
4852
4853 static const struct rpc_call_ops nfs4_lock_ops = {
4854         .rpc_call_prepare = nfs4_lock_prepare,
4855         .rpc_call_done = nfs4_lock_done,
4856         .rpc_release = nfs4_lock_release,
4857 };
4858
4859 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4860 {
4861         switch (error) {
4862         case -NFS4ERR_ADMIN_REVOKED:
4863         case -NFS4ERR_BAD_STATEID:
4864                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4865                 if (new_lock_owner != 0 ||
4866                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4867                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4868                 break;
4869         case -NFS4ERR_STALE_STATEID:
4870                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4871         case -NFS4ERR_EXPIRED:
4872                 nfs4_schedule_lease_recovery(server->nfs_client);
4873         };
4874 }
4875
4876 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4877 {
4878         struct nfs4_lockdata *data;
4879         struct rpc_task *task;
4880         struct rpc_message msg = {
4881                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4882                 .rpc_cred = state->owner->so_cred,
4883         };
4884         struct rpc_task_setup task_setup_data = {
4885                 .rpc_client = NFS_CLIENT(state->inode),
4886                 .rpc_message = &msg,
4887                 .callback_ops = &nfs4_lock_ops,
4888                 .workqueue = nfsiod_workqueue,
4889                 .flags = RPC_TASK_ASYNC,
4890         };
4891         int ret;
4892
4893         dprintk("%s: begin!\n", __func__);
4894         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4895                         fl->fl_u.nfs4_fl.owner,
4896                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4897         if (data == NULL)
4898                 return -ENOMEM;
4899         if (IS_SETLKW(cmd))
4900                 data->arg.block = 1;
4901         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4902         msg.rpc_argp = &data->arg;
4903         msg.rpc_resp = &data->res;
4904         task_setup_data.callback_data = data;
4905         if (recovery_type > NFS_LOCK_NEW) {
4906                 if (recovery_type == NFS_LOCK_RECLAIM)
4907                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4908                 nfs4_set_sequence_privileged(&data->arg.seq_args);
4909         }
4910         task = rpc_run_task(&task_setup_data);
4911         if (IS_ERR(task))
4912                 return PTR_ERR(task);
4913         ret = nfs4_wait_for_completion_rpc_task(task);
4914         if (ret == 0) {
4915                 ret = data->rpc_status;
4916                 if (ret)
4917                         nfs4_handle_setlk_error(data->server, data->lsp,
4918                                         data->arg.new_lock_owner, ret);
4919         } else
4920                 data->cancelled = 1;
4921         rpc_put_task(task);
4922         dprintk("%s: done, ret = %d!\n", __func__, ret);
4923         return ret;
4924 }
4925
4926 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4927 {
4928         struct nfs_server *server = NFS_SERVER(state->inode);
4929         struct nfs4_exception exception = {
4930                 .inode = state->inode,
4931         };
4932         int err;
4933
4934         do {
4935                 /* Cache the lock if possible... */
4936                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4937                         return 0;
4938                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4939                 if (err != -NFS4ERR_DELAY)
4940                         break;
4941                 nfs4_handle_exception(server, err, &exception);
4942         } while (exception.retry);
4943         return err;
4944 }
4945
4946 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4947 {
4948         struct nfs_server *server = NFS_SERVER(state->inode);
4949         struct nfs4_exception exception = {
4950                 .inode = state->inode,
4951         };
4952         int err;
4953
4954         err = nfs4_set_lock_state(state, request);
4955         if (err != 0)
4956                 return err;
4957         do {
4958                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4959                         return 0;
4960                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4961                 switch (err) {
4962                 default:
4963                         goto out;
4964                 case -NFS4ERR_GRACE:
4965                 case -NFS4ERR_DELAY:
4966                         nfs4_handle_exception(server, err, &exception);
4967                         err = 0;
4968                 }
4969         } while (exception.retry);
4970 out:
4971         return err;
4972 }
4973
4974 #if defined(CONFIG_NFS_V4_1)
4975 /**
4976  * nfs41_check_expired_locks - possibly free a lock stateid
4977  *
4978  * @state: NFSv4 state for an inode
4979  *
4980  * Returns NFS_OK if recovery for this stateid is now finished.
4981  * Otherwise a negative NFS4ERR value is returned.
4982  */
4983 static int nfs41_check_expired_locks(struct nfs4_state *state)
4984 {
4985         int status, ret = -NFS4ERR_BAD_STATEID;
4986         struct nfs4_lock_state *lsp;
4987         struct nfs_server *server = NFS_SERVER(state->inode);
4988
4989         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4990                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
4991                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4992                         if (status != NFS_OK) {
4993                                 /* Free the stateid unless the server
4994                                  * informs us the stateid is unrecognized. */
4995                                 if (status != -NFS4ERR_BAD_STATEID)
4996                                         nfs41_free_stateid(server,
4997                                                         &lsp->ls_stateid);
4998                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
4999                                 ret = status;
5000                         }
5001                 }
5002         };
5003
5004         return ret;
5005 }
5006
5007 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5008 {
5009         int status = NFS_OK;
5010
5011         if (test_bit(LK_STATE_IN_USE, &state->flags))
5012                 status = nfs41_check_expired_locks(state);
5013         if (status != NFS_OK)
5014                 status = nfs4_lock_expired(state, request);
5015         return status;
5016 }
5017 #endif
5018
5019 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5020 {
5021         struct nfs4_state_owner *sp = state->owner;
5022         struct nfs_inode *nfsi = NFS_I(state->inode);
5023         unsigned char fl_flags = request->fl_flags;
5024         unsigned int seq;
5025         int status = -ENOLCK;
5026
5027         if ((fl_flags & FL_POSIX) &&
5028                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5029                 goto out;
5030         /* Is this a delegated open? */
5031         status = nfs4_set_lock_state(state, request);
5032         if (status != 0)
5033                 goto out;
5034         request->fl_flags |= FL_ACCESS;
5035         status = do_vfs_lock(request->fl_file, request);
5036         if (status < 0)
5037                 goto out;
5038         down_read(&nfsi->rwsem);
5039         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5040                 /* Yes: cache locks! */
5041                 /* ...but avoid races with delegation recall... */
5042                 request->fl_flags = fl_flags & ~FL_SLEEP;
5043                 status = do_vfs_lock(request->fl_file, request);
5044                 goto out_unlock;
5045         }
5046         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5047         up_read(&nfsi->rwsem);
5048         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5049         if (status != 0)
5050                 goto out;
5051         down_read(&nfsi->rwsem);
5052         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5053                 status = -NFS4ERR_DELAY;
5054                 goto out_unlock;
5055         }
5056         /* Note: we always want to sleep here! */
5057         request->fl_flags = fl_flags | FL_SLEEP;
5058         if (do_vfs_lock(request->fl_file, request) < 0)
5059                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5060                         "manager!\n", __func__);
5061 out_unlock:
5062         up_read(&nfsi->rwsem);
5063 out:
5064         request->fl_flags = fl_flags;
5065         return status;
5066 }
5067
5068 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5069 {
5070         struct nfs4_exception exception = {
5071                 .state = state,
5072                 .inode = state->inode,
5073         };
5074         int err;
5075
5076         do {
5077                 err = _nfs4_proc_setlk(state, cmd, request);
5078                 if (err == -NFS4ERR_DENIED)
5079                         err = -EAGAIN;
5080                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5081                                 err, &exception);
5082         } while (exception.retry);
5083         return err;
5084 }
5085
5086 static int
5087 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5088 {
5089         struct nfs_open_context *ctx;
5090         struct nfs4_state *state;
5091         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5092         int status;
5093
5094         /* verify open state */
5095         ctx = nfs_file_open_context(filp);
5096         state = ctx->state;
5097
5098         if (request->fl_start < 0 || request->fl_end < 0)
5099                 return -EINVAL;
5100
5101         if (IS_GETLK(cmd)) {
5102                 if (state != NULL)
5103                         return nfs4_proc_getlk(state, F_GETLK, request);
5104                 return 0;
5105         }
5106
5107         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5108                 return -EINVAL;
5109
5110         if (request->fl_type == F_UNLCK) {
5111                 if (state != NULL)
5112                         return nfs4_proc_unlck(state, cmd, request);
5113                 return 0;
5114         }
5115
5116         if (state == NULL)
5117                 return -ENOLCK;
5118         /*
5119          * Don't rely on the VFS having checked the file open mode,
5120          * since it won't do this for flock() locks.
5121          */
5122         switch (request->fl_type) {
5123         case F_RDLCK:
5124                 if (!(filp->f_mode & FMODE_READ))
5125                         return -EBADF;
5126                 break;
5127         case F_WRLCK:
5128                 if (!(filp->f_mode & FMODE_WRITE))
5129                         return -EBADF;
5130         }
5131
5132         do {
5133                 status = nfs4_proc_setlk(state, cmd, request);
5134                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5135                         break;
5136                 timeout = nfs4_set_lock_task_retry(timeout);
5137                 status = -ERESTARTSYS;
5138                 if (signalled())
5139                         break;
5140         } while(status < 0);
5141         return status;
5142 }
5143
5144 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
5145 {
5146         struct nfs_server *server = NFS_SERVER(state->inode);
5147         int err;
5148
5149         err = nfs4_set_lock_state(state, fl);
5150         if (err != 0)
5151                 goto out;
5152         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5153         switch (err) {
5154                 default:
5155                         printk(KERN_ERR "NFS: %s: unhandled error "
5156                                 "%d.\n", __func__, err);
5157                 case 0:
5158                 case -ESTALE:
5159                         goto out;
5160                 case -NFS4ERR_STALE_CLIENTID:
5161                 case -NFS4ERR_STALE_STATEID:
5162                         set_bit(NFS_DELEGATED_STATE, &state->flags);
5163                 case -NFS4ERR_EXPIRED:
5164                         nfs4_schedule_lease_recovery(server->nfs_client);
5165                         return -EAGAIN;
5166                 case -NFS4ERR_BADSESSION:
5167                 case -NFS4ERR_BADSLOT:
5168                 case -NFS4ERR_BAD_HIGH_SLOT:
5169                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
5170                 case -NFS4ERR_DEADSESSION:
5171                         set_bit(NFS_DELEGATED_STATE, &state->flags);
5172                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
5173                         return -EAGAIN;
5174                 case -NFS4ERR_DELEG_REVOKED:
5175                 case -NFS4ERR_ADMIN_REVOKED:
5176                 case -NFS4ERR_BAD_STATEID:
5177                 case -NFS4ERR_OPENMODE:
5178                         nfs4_schedule_stateid_recovery(server, state);
5179                         return 0;
5180                 case -NFS4ERR_DELAY:
5181                 case -NFS4ERR_GRACE:
5182                         set_bit(NFS_DELEGATED_STATE, &state->flags);
5183                         ssleep(1);
5184                         return -EAGAIN;
5185                 case -ENOMEM:
5186                 case -NFS4ERR_DENIED:
5187                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
5188                         return 0;
5189         }
5190 out:
5191         return err;
5192 }
5193
5194 struct nfs_release_lockowner_data {
5195         struct nfs4_lock_state *lsp;
5196         struct nfs_server *server;
5197         struct nfs_release_lockowner_args args;
5198 };
5199
5200 static void nfs4_release_lockowner_release(void *calldata)
5201 {
5202         struct nfs_release_lockowner_data *data = calldata;
5203         nfs4_free_lock_state(data->server, data->lsp);
5204         kfree(calldata);
5205 }
5206
5207 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5208         .rpc_release = nfs4_release_lockowner_release,
5209 };
5210
5211 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
5212 {
5213         struct nfs_server *server = lsp->ls_state->owner->so_server;
5214         struct nfs_release_lockowner_data *data;
5215         struct rpc_message msg = {
5216                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5217         };
5218
5219         if (server->nfs_client->cl_mvops->minor_version != 0)
5220                 return -EINVAL;
5221         data = kmalloc(sizeof(*data), GFP_NOFS);
5222         if (!data)
5223                 return -ENOMEM;
5224         data->lsp = lsp;
5225         data->server = server;
5226         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5227         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5228         data->args.lock_owner.s_dev = server->s_dev;
5229         msg.rpc_argp = &data->args;
5230         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5231         return 0;
5232 }
5233
5234 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5235
5236 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5237                                    const void *buf, size_t buflen,
5238                                    int flags, int type)
5239 {
5240         if (strcmp(key, "") != 0)
5241                 return -EINVAL;
5242
5243         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5244 }
5245
5246 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5247                                    void *buf, size_t buflen, int type)
5248 {
5249         if (strcmp(key, "") != 0)
5250                 return -EINVAL;
5251
5252         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5253 }
5254
5255 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5256                                        size_t list_len, const char *name,
5257                                        size_t name_len, int type)
5258 {
5259         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5260
5261         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5262                 return 0;
5263
5264         if (list && len <= list_len)
5265                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5266         return len;
5267 }
5268
5269 /*
5270  * nfs_fhget will use either the mounted_on_fileid or the fileid
5271  */
5272 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5273 {
5274         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5275                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5276               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5277               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5278                 return;
5279
5280         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5281                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5282         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5283         fattr->nlink = 2;
5284 }
5285
5286 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5287                                    const struct qstr *name,
5288                                    struct nfs4_fs_locations *fs_locations,
5289                                    struct page *page)
5290 {
5291         struct nfs_server *server = NFS_SERVER(dir);
5292         u32 bitmask[2] = {
5293                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5294         };
5295         struct nfs4_fs_locations_arg args = {
5296                 .dir_fh = NFS_FH(dir),
5297                 .name = name,
5298                 .page = page,
5299                 .bitmask = bitmask,
5300         };
5301         struct nfs4_fs_locations_res res = {
5302                 .fs_locations = fs_locations,
5303         };
5304         struct rpc_message msg = {
5305                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5306                 .rpc_argp = &args,
5307                 .rpc_resp = &res,
5308         };
5309         int status;
5310
5311         dprintk("%s: start\n", __func__);
5312
5313         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5314          * is not supported */
5315         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5316                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5317         else
5318                 bitmask[0] |= FATTR4_WORD0_FILEID;
5319
5320         nfs_fattr_init(&fs_locations->fattr);
5321         fs_locations->server = server;
5322         fs_locations->nlocations = 0;
5323         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5324         dprintk("%s: returned status = %d\n", __func__, status);
5325         return status;
5326 }
5327
5328 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5329                            const struct qstr *name,
5330                            struct nfs4_fs_locations *fs_locations,
5331                            struct page *page)
5332 {
5333         struct nfs4_exception exception = { };
5334         int err;
5335         do {
5336                 err = nfs4_handle_exception(NFS_SERVER(dir),
5337                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5338                                 &exception);
5339         } while (exception.retry);
5340         return err;
5341 }
5342
5343 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5344 {
5345         int status;
5346         struct nfs4_secinfo_arg args = {
5347                 .dir_fh = NFS_FH(dir),
5348                 .name   = name,
5349         };
5350         struct nfs4_secinfo_res res = {
5351                 .flavors     = flavors,
5352         };
5353         struct rpc_message msg = {
5354                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5355                 .rpc_argp = &args,
5356                 .rpc_resp = &res,
5357         };
5358
5359         dprintk("NFS call  secinfo %s\n", name->name);
5360         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5361         dprintk("NFS reply  secinfo: %d\n", status);
5362         return status;
5363 }
5364
5365 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5366                       struct nfs4_secinfo_flavors *flavors)
5367 {
5368         struct nfs4_exception exception = { };
5369         int err;
5370         do {
5371                 err = nfs4_handle_exception(NFS_SERVER(dir),
5372                                 _nfs4_proc_secinfo(dir, name, flavors),
5373                                 &exception);
5374         } while (exception.retry);
5375         return err;
5376 }
5377
5378 #ifdef CONFIG_NFS_V4_1
5379 /*
5380  * Check the exchange flags returned by the server for invalid flags, having
5381  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5382  * DS flags set.
5383  */
5384 static int nfs4_check_cl_exchange_flags(u32 flags)
5385 {
5386         if (flags & ~EXCHGID4_FLAG_MASK_R)
5387                 goto out_inval;
5388         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5389             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5390                 goto out_inval;
5391         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5392                 goto out_inval;
5393         return NFS_OK;
5394 out_inval:
5395         return -NFS4ERR_INVAL;
5396 }
5397
5398 static bool
5399 nfs41_same_server_scope(struct nfs41_server_scope *a,
5400                         struct nfs41_server_scope *b)
5401 {
5402         if (a->server_scope_sz == b->server_scope_sz &&
5403             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5404                 return true;
5405
5406         return false;
5407 }
5408
5409 /*
5410  * nfs4_proc_bind_conn_to_session()
5411  *
5412  * The 4.1 client currently uses the same TCP connection for the
5413  * fore and backchannel.
5414  */
5415 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5416 {
5417         int status;
5418         struct nfs41_bind_conn_to_session_res res;
5419         struct rpc_message msg = {
5420                 .rpc_proc =
5421                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5422                 .rpc_argp = clp,
5423                 .rpc_resp = &res,
5424                 .rpc_cred = cred,
5425         };
5426
5427         dprintk("--> %s\n", __func__);
5428
5429         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5430         if (unlikely(res.session == NULL)) {
5431                 status = -ENOMEM;
5432                 goto out;
5433         }
5434
5435         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5436         if (status == 0) {
5437                 if (memcmp(res.session->sess_id.data,
5438                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5439                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5440                         status = -EIO;
5441                         goto out_session;
5442                 }
5443                 if (res.dir != NFS4_CDFS4_BOTH) {
5444                         dprintk("NFS: %s: Unexpected direction from server\n",
5445                                 __func__);
5446                         status = -EIO;
5447                         goto out_session;
5448                 }
5449                 if (res.use_conn_in_rdma_mode) {
5450                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5451                                 __func__);
5452                         status = -EIO;
5453                         goto out_session;
5454                 }
5455         }
5456 out_session:
5457         kfree(res.session);
5458 out:
5459         dprintk("<-- %s status= %d\n", __func__, status);
5460         return status;
5461 }
5462
5463 /*
5464  * nfs4_proc_exchange_id()
5465  *
5466  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5467  *
5468  * Since the clientid has expired, all compounds using sessions
5469  * associated with the stale clientid will be returning
5470  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5471  * be in some phase of session reset.
5472  */
5473 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5474 {
5475         nfs4_verifier verifier;
5476         struct nfs41_exchange_id_args args = {
5477                 .verifier = &verifier,
5478                 .client = clp,
5479                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5480         };
5481         struct nfs41_exchange_id_res res = {
5482                 0
5483         };
5484         int status;
5485         struct rpc_message msg = {
5486                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5487                 .rpc_argp = &args,
5488                 .rpc_resp = &res,
5489                 .rpc_cred = cred,
5490         };
5491
5492         nfs4_init_boot_verifier(clp, &verifier);
5493         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5494                                                         sizeof(args.id));
5495         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5496                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5497                 args.id_len, args.id);
5498
5499         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5500                                         GFP_NOFS);
5501         if (unlikely(res.server_owner == NULL)) {
5502                 status = -ENOMEM;
5503                 goto out;
5504         }
5505
5506         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5507                                         GFP_NOFS);
5508         if (unlikely(res.server_scope == NULL)) {
5509                 status = -ENOMEM;
5510                 goto out_server_owner;
5511         }
5512
5513         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5514         if (unlikely(res.impl_id == NULL)) {
5515                 status = -ENOMEM;
5516                 goto out_server_scope;
5517         }
5518
5519         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5520         if (status == 0)
5521                 status = nfs4_check_cl_exchange_flags(res.flags);
5522
5523         if (status == 0) {
5524                 clp->cl_clientid = res.clientid;
5525                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5526                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5527                         clp->cl_seqid = res.seqid;
5528
5529                 kfree(clp->cl_serverowner);
5530                 clp->cl_serverowner = res.server_owner;
5531                 res.server_owner = NULL;
5532
5533                 /* use the most recent implementation id */
5534                 kfree(clp->cl_implid);
5535                 clp->cl_implid = res.impl_id;
5536
5537                 if (clp->cl_serverscope != NULL &&
5538                     !nfs41_same_server_scope(clp->cl_serverscope,
5539                                              res.server_scope)) {
5540                         dprintk("%s: server_scope mismatch detected\n",
5541                                 __func__);
5542                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5543                         kfree(clp->cl_serverscope);
5544                         clp->cl_serverscope = NULL;
5545                 }
5546
5547                 if (clp->cl_serverscope == NULL) {
5548                         clp->cl_serverscope = res.server_scope;
5549                         goto out;
5550                 }
5551         } else
5552                 kfree(res.impl_id);
5553
5554 out_server_owner:
5555         kfree(res.server_owner);
5556 out_server_scope:
5557         kfree(res.server_scope);
5558 out:
5559         if (clp->cl_implid != NULL)
5560                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5561                         "domain: %s, name: %s, date: %llu,%u\n",
5562                         clp->cl_implid->domain, clp->cl_implid->name,
5563                         clp->cl_implid->date.seconds,
5564                         clp->cl_implid->date.nseconds);
5565         dprintk("NFS reply exchange_id: %d\n", status);
5566         return status;
5567 }
5568
5569 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5570                 struct rpc_cred *cred)
5571 {
5572         struct rpc_message msg = {
5573                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5574                 .rpc_argp = clp,
5575                 .rpc_cred = cred,
5576         };
5577         int status;
5578
5579         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5580         if (status)
5581                 dprintk("NFS: Got error %d from the server %s on "
5582                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5583         return status;
5584 }
5585
5586 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5587                 struct rpc_cred *cred)
5588 {
5589         unsigned int loop;
5590         int ret;
5591
5592         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5593                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5594                 switch (ret) {
5595                 case -NFS4ERR_DELAY:
5596                 case -NFS4ERR_CLIENTID_BUSY:
5597                         ssleep(1);
5598                         break;
5599                 default:
5600                         return ret;
5601                 }
5602         }
5603         return 0;
5604 }
5605
5606 int nfs4_destroy_clientid(struct nfs_client *clp)
5607 {
5608         struct rpc_cred *cred;
5609         int ret = 0;
5610
5611         if (clp->cl_mvops->minor_version < 1)
5612                 goto out;
5613         if (clp->cl_exchange_flags == 0)
5614                 goto out;
5615         if (clp->cl_preserve_clid)
5616                 goto out;
5617         cred = nfs4_get_exchange_id_cred(clp);
5618         ret = nfs4_proc_destroy_clientid(clp, cred);
5619         if (cred)
5620                 put_rpccred(cred);
5621         switch (ret) {
5622         case 0:
5623         case -NFS4ERR_STALE_CLIENTID:
5624                 clp->cl_exchange_flags = 0;
5625         }
5626 out:
5627         return ret;
5628 }
5629
5630 struct nfs4_get_lease_time_data {
5631         struct nfs4_get_lease_time_args *args;
5632         struct nfs4_get_lease_time_res *res;
5633         struct nfs_client *clp;
5634 };
5635
5636 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5637                                         void *calldata)
5638 {
5639         struct nfs4_get_lease_time_data *data =
5640                         (struct nfs4_get_lease_time_data *)calldata;
5641
5642         dprintk("--> %s\n", __func__);
5643         /* just setup sequence, do not trigger session recovery
5644            since we're invoked within one */
5645         nfs41_setup_sequence(data->clp->cl_session,
5646                         &data->args->la_seq_args,
5647                         &data->res->lr_seq_res,
5648                         task);
5649         dprintk("<-- %s\n", __func__);
5650 }
5651
5652 /*
5653  * Called from nfs4_state_manager thread for session setup, so don't recover
5654  * from sequence operation or clientid errors.
5655  */
5656 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5657 {
5658         struct nfs4_get_lease_time_data *data =
5659                         (struct nfs4_get_lease_time_data *)calldata;
5660
5661         dprintk("--> %s\n", __func__);
5662         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5663                 return;
5664         switch (task->tk_status) {
5665         case -NFS4ERR_DELAY:
5666         case -NFS4ERR_GRACE:
5667                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5668                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5669                 task->tk_status = 0;
5670                 /* fall through */
5671         case -NFS4ERR_RETRY_UNCACHED_REP:
5672                 rpc_restart_call_prepare(task);
5673                 return;
5674         }
5675         dprintk("<-- %s\n", __func__);
5676 }
5677
5678 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5679         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5680         .rpc_call_done = nfs4_get_lease_time_done,
5681 };
5682
5683 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5684 {
5685         struct rpc_task *task;
5686         struct nfs4_get_lease_time_args args;
5687         struct nfs4_get_lease_time_res res = {
5688                 .lr_fsinfo = fsinfo,
5689         };
5690         struct nfs4_get_lease_time_data data = {
5691                 .args = &args,
5692                 .res = &res,
5693                 .clp = clp,
5694         };
5695         struct rpc_message msg = {
5696                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5697                 .rpc_argp = &args,
5698                 .rpc_resp = &res,
5699         };
5700         struct rpc_task_setup task_setup = {
5701                 .rpc_client = clp->cl_rpcclient,
5702                 .rpc_message = &msg,
5703                 .callback_ops = &nfs4_get_lease_time_ops,
5704                 .callback_data = &data,
5705                 .flags = RPC_TASK_TIMEOUT,
5706         };
5707         int status;
5708
5709         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5710         nfs4_set_sequence_privileged(&args.la_seq_args);
5711         dprintk("--> %s\n", __func__);
5712         task = rpc_run_task(&task_setup);
5713
5714         if (IS_ERR(task))
5715                 status = PTR_ERR(task);
5716         else {
5717                 status = task->tk_status;
5718                 rpc_put_task(task);
5719         }
5720         dprintk("<-- %s return %d\n", __func__, status);
5721
5722         return status;
5723 }
5724
5725 /*
5726  * Initialize the values to be used by the client in CREATE_SESSION
5727  * If nfs4_init_session set the fore channel request and response sizes,
5728  * use them.
5729  *
5730  * Set the back channel max_resp_sz_cached to zero to force the client to
5731  * always set csa_cachethis to FALSE because the current implementation
5732  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5733  */
5734 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5735 {
5736         struct nfs4_session *session = args->client->cl_session;
5737         unsigned int mxrqst_sz = session->fc_target_max_rqst_sz,
5738                      mxresp_sz = session->fc_target_max_resp_sz;
5739
5740         if (mxrqst_sz == 0)
5741                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5742         if (mxresp_sz == 0)
5743                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5744         /* Fore channel attributes */
5745         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5746         args->fc_attrs.max_resp_sz = mxresp_sz;
5747         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5748         args->fc_attrs.max_reqs = max_session_slots;
5749
5750         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5751                 "max_ops=%u max_reqs=%u\n",
5752                 __func__,
5753                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5754                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5755
5756         /* Back channel attributes */
5757         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5758         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5759         args->bc_attrs.max_resp_sz_cached = 0;
5760         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5761         args->bc_attrs.max_reqs = 1;
5762
5763         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5764                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5765                 __func__,
5766                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5767                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5768                 args->bc_attrs.max_reqs);
5769 }
5770
5771 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5772 {
5773         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5774         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5775
5776         if (rcvd->max_resp_sz > sent->max_resp_sz)
5777                 return -EINVAL;
5778         /*
5779          * Our requested max_ops is the minimum we need; we're not
5780          * prepared to break up compounds into smaller pieces than that.
5781          * So, no point even trying to continue if the server won't
5782          * cooperate:
5783          */
5784         if (rcvd->max_ops < sent->max_ops)
5785                 return -EINVAL;
5786         if (rcvd->max_reqs == 0)
5787                 return -EINVAL;
5788         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5789                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5790         return 0;
5791 }
5792
5793 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5794 {
5795         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5796         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5797
5798         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5799                 return -EINVAL;
5800         if (rcvd->max_resp_sz < sent->max_resp_sz)
5801                 return -EINVAL;
5802         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5803                 return -EINVAL;
5804         /* These would render the backchannel useless: */
5805         if (rcvd->max_ops != sent->max_ops)
5806                 return -EINVAL;
5807         if (rcvd->max_reqs != sent->max_reqs)
5808                 return -EINVAL;
5809         return 0;
5810 }
5811
5812 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5813                                      struct nfs4_session *session)
5814 {
5815         int ret;
5816
5817         ret = nfs4_verify_fore_channel_attrs(args, session);
5818         if (ret)
5819                 return ret;
5820         return nfs4_verify_back_channel_attrs(args, session);
5821 }
5822
5823 static int _nfs4_proc_create_session(struct nfs_client *clp,
5824                 struct rpc_cred *cred)
5825 {
5826         struct nfs4_session *session = clp->cl_session;
5827         struct nfs41_create_session_args args = {
5828                 .client = clp,
5829                 .cb_program = NFS4_CALLBACK,
5830         };
5831         struct nfs41_create_session_res res = {
5832                 .client = clp,
5833         };
5834         struct rpc_message msg = {
5835                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5836                 .rpc_argp = &args,
5837                 .rpc_resp = &res,
5838                 .rpc_cred = cred,
5839         };
5840         int status;
5841
5842         nfs4_init_channel_attrs(&args);
5843         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5844
5845         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5846
5847         if (!status) {
5848                 /* Verify the session's negotiated channel_attrs values */
5849                 status = nfs4_verify_channel_attrs(&args, session);
5850                 /* Increment the clientid slot sequence id */
5851                 clp->cl_seqid++;
5852         }
5853
5854         return status;
5855 }
5856
5857 /*
5858  * Issues a CREATE_SESSION operation to the server.
5859  * It is the responsibility of the caller to verify the session is
5860  * expired before calling this routine.
5861  */
5862 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5863 {
5864         int status;
5865         unsigned *ptr;
5866         struct nfs4_session *session = clp->cl_session;
5867
5868         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5869
5870         status = _nfs4_proc_create_session(clp, cred);
5871         if (status)
5872                 goto out;
5873
5874         /* Init or reset the session slot tables */
5875         status = nfs4_setup_session_slot_tables(session);
5876         dprintk("slot table setup returned %d\n", status);
5877         if (status)
5878                 goto out;
5879
5880         ptr = (unsigned *)&session->sess_id.data[0];
5881         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5882                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5883 out:
5884         dprintk("<-- %s\n", __func__);
5885         return status;
5886 }
5887
5888 /*
5889  * Issue the over-the-wire RPC DESTROY_SESSION.
5890  * The caller must serialize access to this routine.
5891  */
5892 int nfs4_proc_destroy_session(struct nfs4_session *session,
5893                 struct rpc_cred *cred)
5894 {
5895         struct rpc_message msg = {
5896                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5897                 .rpc_argp = session,
5898                 .rpc_cred = cred,
5899         };
5900         int status = 0;
5901
5902         dprintk("--> nfs4_proc_destroy_session\n");
5903
5904         /* session is still being setup */
5905         if (session->clp->cl_cons_state != NFS_CS_READY)
5906                 return status;
5907
5908         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5909
5910         if (status)
5911                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5912                         "Session has been destroyed regardless...\n", status);
5913
5914         dprintk("<-- nfs4_proc_destroy_session\n");
5915         return status;
5916 }
5917
5918 /*
5919  * Renew the cl_session lease.
5920  */
5921 struct nfs4_sequence_data {
5922         struct nfs_client *clp;
5923         struct nfs4_sequence_args args;
5924         struct nfs4_sequence_res res;
5925 };
5926
5927 static void nfs41_sequence_release(void *data)
5928 {
5929         struct nfs4_sequence_data *calldata = data;
5930         struct nfs_client *clp = calldata->clp;
5931
5932         if (atomic_read(&clp->cl_count) > 1)
5933                 nfs4_schedule_state_renewal(clp);
5934         nfs_put_client(clp);
5935         kfree(calldata);
5936 }
5937
5938 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5939 {
5940         switch(task->tk_status) {
5941         case -NFS4ERR_DELAY:
5942                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5943                 return -EAGAIN;
5944         default:
5945                 nfs4_schedule_lease_recovery(clp);
5946         }
5947         return 0;
5948 }
5949
5950 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5951 {
5952         struct nfs4_sequence_data *calldata = data;
5953         struct nfs_client *clp = calldata->clp;
5954
5955         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5956                 return;
5957
5958         if (task->tk_status < 0) {
5959                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5960                 if (atomic_read(&clp->cl_count) == 1)
5961                         goto out;
5962
5963                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5964                         rpc_restart_call_prepare(task);
5965                         return;
5966                 }
5967         }
5968         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5969 out:
5970         dprintk("<-- %s\n", __func__);
5971 }
5972
5973 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5974 {
5975         struct nfs4_sequence_data *calldata = data;
5976         struct nfs_client *clp = calldata->clp;
5977         struct nfs4_sequence_args *args;
5978         struct nfs4_sequence_res *res;
5979
5980         args = task->tk_msg.rpc_argp;
5981         res = task->tk_msg.rpc_resp;
5982
5983         nfs41_setup_sequence(clp->cl_session, args, res, task);
5984 }
5985
5986 static const struct rpc_call_ops nfs41_sequence_ops = {
5987         .rpc_call_done = nfs41_sequence_call_done,
5988         .rpc_call_prepare = nfs41_sequence_prepare,
5989         .rpc_release = nfs41_sequence_release,
5990 };
5991
5992 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
5993                 struct rpc_cred *cred,
5994                 bool is_privileged)
5995 {
5996         struct nfs4_sequence_data *calldata;
5997         struct rpc_message msg = {
5998                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5999                 .rpc_cred = cred,
6000         };
6001         struct rpc_task_setup task_setup_data = {
6002                 .rpc_client = clp->cl_rpcclient,
6003                 .rpc_message = &msg,
6004                 .callback_ops = &nfs41_sequence_ops,
6005                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
6006         };
6007
6008         if (!atomic_inc_not_zero(&clp->cl_count))
6009                 return ERR_PTR(-EIO);
6010         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6011         if (calldata == NULL) {
6012                 nfs_put_client(clp);
6013                 return ERR_PTR(-ENOMEM);
6014         }
6015         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6016         if (is_privileged)
6017                 nfs4_set_sequence_privileged(&calldata->args);
6018         msg.rpc_argp = &calldata->args;
6019         msg.rpc_resp = &calldata->res;
6020         calldata->clp = clp;
6021         task_setup_data.callback_data = calldata;
6022
6023         return rpc_run_task(&task_setup_data);
6024 }
6025
6026 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6027 {
6028         struct rpc_task *task;
6029         int ret = 0;
6030
6031         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6032                 return 0;
6033         task = _nfs41_proc_sequence(clp, cred, false);
6034         if (IS_ERR(task))
6035                 ret = PTR_ERR(task);
6036         else
6037                 rpc_put_task_async(task);
6038         dprintk("<-- %s status=%d\n", __func__, ret);
6039         return ret;
6040 }
6041
6042 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6043 {
6044         struct rpc_task *task;
6045         int ret;
6046
6047         task = _nfs41_proc_sequence(clp, cred, true);
6048         if (IS_ERR(task)) {
6049                 ret = PTR_ERR(task);
6050                 goto out;
6051         }
6052         ret = rpc_wait_for_completion_task(task);
6053         if (!ret) {
6054                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6055
6056                 if (task->tk_status == 0)
6057                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6058                 ret = task->tk_status;
6059         }
6060         rpc_put_task(task);
6061 out:
6062         dprintk("<-- %s status=%d\n", __func__, ret);
6063         return ret;
6064 }
6065
6066 struct nfs4_reclaim_complete_data {
6067         struct nfs_client *clp;
6068         struct nfs41_reclaim_complete_args arg;
6069         struct nfs41_reclaim_complete_res res;
6070 };
6071
6072 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6073 {
6074         struct nfs4_reclaim_complete_data *calldata = data;
6075
6076         nfs41_setup_sequence(calldata->clp->cl_session,
6077                         &calldata->arg.seq_args,
6078                         &calldata->res.seq_res,
6079                         task);
6080 }
6081
6082 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6083 {
6084         switch(task->tk_status) {
6085         case 0:
6086         case -NFS4ERR_COMPLETE_ALREADY:
6087         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6088                 break;
6089         case -NFS4ERR_DELAY:
6090                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6091                 /* fall through */
6092         case -NFS4ERR_RETRY_UNCACHED_REP:
6093                 return -EAGAIN;
6094         default:
6095                 nfs4_schedule_lease_recovery(clp);
6096         }
6097         return 0;
6098 }
6099
6100 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6101 {
6102         struct nfs4_reclaim_complete_data *calldata = data;
6103         struct nfs_client *clp = calldata->clp;
6104         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6105
6106         dprintk("--> %s\n", __func__);
6107         if (!nfs41_sequence_done(task, res))
6108                 return;
6109
6110         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6111                 rpc_restart_call_prepare(task);
6112                 return;
6113         }
6114         dprintk("<-- %s\n", __func__);
6115 }
6116
6117 static void nfs4_free_reclaim_complete_data(void *data)
6118 {
6119         struct nfs4_reclaim_complete_data *calldata = data;
6120
6121         kfree(calldata);
6122 }
6123
6124 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6125         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6126         .rpc_call_done = nfs4_reclaim_complete_done,
6127         .rpc_release = nfs4_free_reclaim_complete_data,
6128 };
6129
6130 /*
6131  * Issue a global reclaim complete.
6132  */
6133 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6134 {
6135         struct nfs4_reclaim_complete_data *calldata;
6136         struct rpc_task *task;
6137         struct rpc_message msg = {
6138                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6139         };
6140         struct rpc_task_setup task_setup_data = {
6141                 .rpc_client = clp->cl_rpcclient,
6142                 .rpc_message = &msg,
6143                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6144                 .flags = RPC_TASK_ASYNC,
6145         };
6146         int status = -ENOMEM;
6147
6148         dprintk("--> %s\n", __func__);
6149         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6150         if (calldata == NULL)
6151                 goto out;
6152         calldata->clp = clp;
6153         calldata->arg.one_fs = 0;
6154
6155         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6156         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
6157         msg.rpc_argp = &calldata->arg;
6158         msg.rpc_resp = &calldata->res;
6159         task_setup_data.callback_data = calldata;
6160         task = rpc_run_task(&task_setup_data);
6161         if (IS_ERR(task)) {
6162                 status = PTR_ERR(task);
6163                 goto out;
6164         }
6165         status = nfs4_wait_for_completion_rpc_task(task);
6166         if (status == 0)
6167                 status = task->tk_status;
6168         rpc_put_task(task);
6169         return 0;
6170 out:
6171         dprintk("<-- %s status=%d\n", __func__, status);
6172         return status;
6173 }
6174
6175 static void
6176 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6177 {
6178         struct nfs4_layoutget *lgp = calldata;
6179         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6180         struct nfs4_session *session = nfs4_get_session(server);
6181
6182         dprintk("--> %s\n", __func__);
6183         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6184          * right now covering the LAYOUTGET we are about to send.
6185          * However, that is not so catastrophic, and there seems
6186          * to be no way to prevent it completely.
6187          */
6188         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
6189                                 &lgp->res.seq_res, task))
6190                 return;
6191         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6192                                           NFS_I(lgp->args.inode)->layout,
6193                                           lgp->args.ctx->state)) {
6194                 rpc_exit(task, NFS4_OK);
6195         }
6196 }
6197
6198 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6199 {
6200         struct nfs4_layoutget *lgp = calldata;
6201         struct inode *inode = lgp->args.inode;
6202         struct nfs_server *server = NFS_SERVER(inode);
6203         struct pnfs_layout_hdr *lo;
6204         struct nfs4_state *state = NULL;
6205         unsigned long timeo, giveup;
6206
6207         dprintk("--> %s\n", __func__);
6208
6209         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
6210                 goto out;
6211
6212         switch (task->tk_status) {
6213         case 0:
6214                 goto out;
6215         case -NFS4ERR_LAYOUTTRYLATER:
6216         case -NFS4ERR_RECALLCONFLICT:
6217                 timeo = rpc_get_timeout(task->tk_client);
6218                 giveup = lgp->args.timestamp + timeo;
6219                 if (time_after(giveup, jiffies))
6220                         task->tk_status = -NFS4ERR_DELAY;
6221                 break;
6222         case -NFS4ERR_EXPIRED:
6223         case -NFS4ERR_BAD_STATEID:
6224                 spin_lock(&inode->i_lock);
6225                 lo = NFS_I(inode)->layout;
6226                 if (!lo || list_empty(&lo->plh_segs)) {
6227                         spin_unlock(&inode->i_lock);
6228                         /* If the open stateid was bad, then recover it. */
6229                         state = lgp->args.ctx->state;
6230                 } else {
6231                         LIST_HEAD(head);
6232
6233                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6234                         spin_unlock(&inode->i_lock);
6235                         /* Mark the bad layout state as invalid, then
6236                          * retry using the open stateid. */
6237                         pnfs_free_lseg_list(&head);
6238                 }
6239         }
6240         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6241                 rpc_restart_call_prepare(task);
6242 out:
6243         dprintk("<-- %s\n", __func__);
6244 }
6245
6246 static size_t max_response_pages(struct nfs_server *server)
6247 {
6248         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6249         return nfs_page_array_len(0, max_resp_sz);
6250 }
6251
6252 static void nfs4_free_pages(struct page **pages, size_t size)
6253 {
6254         int i;
6255
6256         if (!pages)
6257                 return;
6258
6259         for (i = 0; i < size; i++) {
6260                 if (!pages[i])
6261                         break;
6262                 __free_page(pages[i]);
6263         }
6264         kfree(pages);
6265 }
6266
6267 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6268 {
6269         struct page **pages;
6270         int i;
6271
6272         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6273         if (!pages) {
6274                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6275                 return NULL;
6276         }
6277
6278         for (i = 0; i < size; i++) {
6279                 pages[i] = alloc_page(gfp_flags);
6280                 if (!pages[i]) {
6281                         dprintk("%s: failed to allocate page\n", __func__);
6282                         nfs4_free_pages(pages, size);
6283                         return NULL;
6284                 }
6285         }
6286
6287         return pages;
6288 }
6289
6290 static void nfs4_layoutget_release(void *calldata)
6291 {
6292         struct nfs4_layoutget *lgp = calldata;
6293         struct inode *inode = lgp->args.inode;
6294         struct nfs_server *server = NFS_SERVER(inode);
6295         size_t max_pages = max_response_pages(server);
6296
6297         dprintk("--> %s\n", __func__);
6298         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6299         pnfs_put_layout_hdr(NFS_I(inode)->layout);
6300         put_nfs_open_context(lgp->args.ctx);
6301         kfree(calldata);
6302         dprintk("<-- %s\n", __func__);
6303 }
6304
6305 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6306         .rpc_call_prepare = nfs4_layoutget_prepare,
6307         .rpc_call_done = nfs4_layoutget_done,
6308         .rpc_release = nfs4_layoutget_release,
6309 };
6310
6311 struct pnfs_layout_segment *
6312 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6313 {
6314         struct inode *inode = lgp->args.inode;
6315         struct nfs_server *server = NFS_SERVER(inode);
6316         size_t max_pages = max_response_pages(server);
6317         struct rpc_task *task;
6318         struct rpc_message msg = {
6319                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6320                 .rpc_argp = &lgp->args,
6321                 .rpc_resp = &lgp->res,
6322         };
6323         struct rpc_task_setup task_setup_data = {
6324                 .rpc_client = server->client,
6325                 .rpc_message = &msg,
6326                 .callback_ops = &nfs4_layoutget_call_ops,
6327                 .callback_data = lgp,
6328                 .flags = RPC_TASK_ASYNC,
6329         };
6330         struct pnfs_layout_segment *lseg = NULL;
6331         int status = 0;
6332
6333         dprintk("--> %s\n", __func__);
6334
6335         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6336         if (!lgp->args.layout.pages) {
6337                 nfs4_layoutget_release(lgp);
6338                 return ERR_PTR(-ENOMEM);
6339         }
6340         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6341         lgp->args.timestamp = jiffies;
6342
6343         lgp->res.layoutp = &lgp->args.layout;
6344         lgp->res.seq_res.sr_slot = NULL;
6345         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6346
6347         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
6348         pnfs_get_layout_hdr(NFS_I(inode)->layout);
6349
6350         task = rpc_run_task(&task_setup_data);
6351         if (IS_ERR(task))
6352                 return ERR_CAST(task);
6353         status = nfs4_wait_for_completion_rpc_task(task);
6354         if (status == 0)
6355                 status = task->tk_status;
6356         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
6357         if (status == 0 && lgp->res.layoutp->len)
6358                 lseg = pnfs_layout_process(lgp);
6359         rpc_put_task(task);
6360         dprintk("<-- %s status=%d\n", __func__, status);
6361         if (status)
6362                 return ERR_PTR(status);
6363         return lseg;
6364 }
6365
6366 static void
6367 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6368 {
6369         struct nfs4_layoutreturn *lrp = calldata;
6370
6371         dprintk("--> %s\n", __func__);
6372         nfs41_setup_sequence(lrp->clp->cl_session,
6373                         &lrp->args.seq_args,
6374                         &lrp->res.seq_res,
6375                         task);
6376 }
6377
6378 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6379 {
6380         struct nfs4_layoutreturn *lrp = calldata;
6381         struct nfs_server *server;
6382
6383         dprintk("--> %s\n", __func__);
6384
6385         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
6386                 return;
6387
6388         server = NFS_SERVER(lrp->args.inode);
6389         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6390                 rpc_restart_call_prepare(task);
6391                 return;
6392         }
6393         dprintk("<-- %s\n", __func__);
6394 }
6395
6396 static void nfs4_layoutreturn_release(void *calldata)
6397 {
6398         struct nfs4_layoutreturn *lrp = calldata;
6399         struct pnfs_layout_hdr *lo = lrp->args.layout;
6400
6401         dprintk("--> %s\n", __func__);
6402         spin_lock(&lo->plh_inode->i_lock);
6403         if (lrp->res.lrs_present)
6404                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6405         lo->plh_block_lgets--;
6406         spin_unlock(&lo->plh_inode->i_lock);
6407         pnfs_put_layout_hdr(lrp->args.layout);
6408         kfree(calldata);
6409         dprintk("<-- %s\n", __func__);
6410 }
6411
6412 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6413         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6414         .rpc_call_done = nfs4_layoutreturn_done,
6415         .rpc_release = nfs4_layoutreturn_release,
6416 };
6417
6418 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6419 {
6420         struct rpc_task *task;
6421         struct rpc_message msg = {
6422                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6423                 .rpc_argp = &lrp->args,
6424                 .rpc_resp = &lrp->res,
6425         };
6426         struct rpc_task_setup task_setup_data = {
6427                 .rpc_client = lrp->clp->cl_rpcclient,
6428                 .rpc_message = &msg,
6429                 .callback_ops = &nfs4_layoutreturn_call_ops,
6430                 .callback_data = lrp,
6431         };
6432         int status;
6433
6434         dprintk("--> %s\n", __func__);
6435         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6436         task = rpc_run_task(&task_setup_data);
6437         if (IS_ERR(task))
6438                 return PTR_ERR(task);
6439         status = task->tk_status;
6440         dprintk("<-- %s status=%d\n", __func__, status);
6441         rpc_put_task(task);
6442         return status;
6443 }
6444
6445 /*
6446  * Retrieve the list of Data Server devices from the MDS.
6447  */
6448 static int _nfs4_getdevicelist(struct nfs_server *server,
6449                                     const struct nfs_fh *fh,
6450                                     struct pnfs_devicelist *devlist)
6451 {
6452         struct nfs4_getdevicelist_args args = {
6453                 .fh = fh,
6454                 .layoutclass = server->pnfs_curr_ld->id,
6455         };
6456         struct nfs4_getdevicelist_res res = {
6457                 .devlist = devlist,
6458         };
6459         struct rpc_message msg = {
6460                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6461                 .rpc_argp = &args,
6462                 .rpc_resp = &res,
6463         };
6464         int status;
6465
6466         dprintk("--> %s\n", __func__);
6467         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6468                                 &res.seq_res, 0);
6469         dprintk("<-- %s status=%d\n", __func__, status);
6470         return status;
6471 }
6472
6473 int nfs4_proc_getdevicelist(struct nfs_server *server,
6474                             const struct nfs_fh *fh,
6475                             struct pnfs_devicelist *devlist)
6476 {
6477         struct nfs4_exception exception = { };
6478         int err;
6479
6480         do {
6481                 err = nfs4_handle_exception(server,
6482                                 _nfs4_getdevicelist(server, fh, devlist),
6483                                 &exception);
6484         } while (exception.retry);
6485
6486         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6487                 err, devlist->num_devs);
6488
6489         return err;
6490 }
6491 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6492
6493 static int
6494 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6495 {
6496         struct nfs4_getdeviceinfo_args args = {
6497                 .pdev = pdev,
6498         };
6499         struct nfs4_getdeviceinfo_res res = {
6500                 .pdev = pdev,
6501         };
6502         struct rpc_message msg = {
6503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6504                 .rpc_argp = &args,
6505                 .rpc_resp = &res,
6506         };
6507         int status;
6508
6509         dprintk("--> %s\n", __func__);
6510         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6511         dprintk("<-- %s status=%d\n", __func__, status);
6512
6513         return status;
6514 }
6515
6516 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6517 {
6518         struct nfs4_exception exception = { };
6519         int err;
6520
6521         do {
6522                 err = nfs4_handle_exception(server,
6523                                         _nfs4_proc_getdeviceinfo(server, pdev),
6524                                         &exception);
6525         } while (exception.retry);
6526         return err;
6527 }
6528 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6529
6530 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6531 {
6532         struct nfs4_layoutcommit_data *data = calldata;
6533         struct nfs_server *server = NFS_SERVER(data->args.inode);
6534         struct nfs4_session *session = nfs4_get_session(server);
6535
6536         nfs41_setup_sequence(session,
6537                         &data->args.seq_args,
6538                         &data->res.seq_res,
6539                         task);
6540 }
6541
6542 static void
6543 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6544 {
6545         struct nfs4_layoutcommit_data *data = calldata;
6546         struct nfs_server *server = NFS_SERVER(data->args.inode);
6547
6548         if (!nfs41_sequence_done(task, &data->res.seq_res))
6549                 return;
6550
6551         switch (task->tk_status) { /* Just ignore these failures */
6552         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6553         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6554         case -NFS4ERR_BADLAYOUT:     /* no layout */
6555         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6556                 task->tk_status = 0;
6557                 break;
6558         case 0:
6559                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6560                                                    data->res.fattr);
6561                 break;
6562         default:
6563                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6564                         rpc_restart_call_prepare(task);
6565                         return;
6566                 }
6567         }
6568 }
6569
6570 static void nfs4_layoutcommit_release(void *calldata)
6571 {
6572         struct nfs4_layoutcommit_data *data = calldata;
6573
6574         pnfs_cleanup_layoutcommit(data);
6575         put_rpccred(data->cred);
6576         kfree(data);
6577 }
6578
6579 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6580         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6581         .rpc_call_done = nfs4_layoutcommit_done,
6582         .rpc_release = nfs4_layoutcommit_release,
6583 };
6584
6585 int
6586 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6587 {
6588         struct rpc_message msg = {
6589                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6590                 .rpc_argp = &data->args,
6591                 .rpc_resp = &data->res,
6592                 .rpc_cred = data->cred,
6593         };
6594         struct rpc_task_setup task_setup_data = {
6595                 .task = &data->task,
6596                 .rpc_client = NFS_CLIENT(data->args.inode),
6597                 .rpc_message = &msg,
6598                 .callback_ops = &nfs4_layoutcommit_ops,
6599                 .callback_data = data,
6600                 .flags = RPC_TASK_ASYNC,
6601         };
6602         struct rpc_task *task;
6603         int status = 0;
6604
6605         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6606                 "lbw: %llu inode %lu\n",
6607                 data->task.tk_pid, sync,
6608                 data->args.lastbytewritten,
6609                 data->args.inode->i_ino);
6610
6611         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6612         task = rpc_run_task(&task_setup_data);
6613         if (IS_ERR(task))
6614                 return PTR_ERR(task);
6615         if (sync == false)
6616                 goto out;
6617         status = nfs4_wait_for_completion_rpc_task(task);
6618         if (status != 0)
6619                 goto out;
6620         status = task->tk_status;
6621 out:
6622         dprintk("%s: status %d\n", __func__, status);
6623         rpc_put_task(task);
6624         return status;
6625 }
6626
6627 static int
6628 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6629                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6630 {
6631         struct nfs41_secinfo_no_name_args args = {
6632                 .style = SECINFO_STYLE_CURRENT_FH,
6633         };
6634         struct nfs4_secinfo_res res = {
6635                 .flavors = flavors,
6636         };
6637         struct rpc_message msg = {
6638                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6639                 .rpc_argp = &args,
6640                 .rpc_resp = &res,
6641         };
6642         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6643 }
6644
6645 static int
6646 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6647                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6648 {
6649         struct nfs4_exception exception = { };
6650         int err;
6651         do {
6652                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6653                 switch (err) {
6654                 case 0:
6655                 case -NFS4ERR_WRONGSEC:
6656                 case -NFS4ERR_NOTSUPP:
6657                         goto out;
6658                 default:
6659                         err = nfs4_handle_exception(server, err, &exception);
6660                 }
6661         } while (exception.retry);
6662 out:
6663         return err;
6664 }
6665
6666 static int
6667 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6668                     struct nfs_fsinfo *info)
6669 {
6670         int err;
6671         struct page *page;
6672         rpc_authflavor_t flavor;
6673         struct nfs4_secinfo_flavors *flavors;
6674
6675         page = alloc_page(GFP_KERNEL);
6676         if (!page) {
6677                 err = -ENOMEM;
6678                 goto out;
6679         }
6680
6681         flavors = page_address(page);
6682         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6683
6684         /*
6685          * Fall back on "guess and check" method if
6686          * the server doesn't support SECINFO_NO_NAME
6687          */
6688         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6689                 err = nfs4_find_root_sec(server, fhandle, info);
6690                 goto out_freepage;
6691         }
6692         if (err)
6693                 goto out_freepage;
6694
6695         flavor = nfs_find_best_sec(flavors);
6696         if (err == 0)
6697                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6698
6699 out_freepage:
6700         put_page(page);
6701         if (err == -EACCES)
6702                 return -EPERM;
6703 out:
6704         return err;
6705 }
6706
6707 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6708 {
6709         int status;
6710         struct nfs41_test_stateid_args args = {
6711                 .stateid = stateid,
6712         };
6713         struct nfs41_test_stateid_res res;
6714         struct rpc_message msg = {
6715                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6716                 .rpc_argp = &args,
6717                 .rpc_resp = &res,
6718         };
6719
6720         dprintk("NFS call  test_stateid %p\n", stateid);
6721         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6722         nfs4_set_sequence_privileged(&args.seq_args);
6723         status = nfs4_call_sync_sequence(server->client, server, &msg,
6724                         &args.seq_args, &res.seq_res);
6725         if (status != NFS_OK) {
6726                 dprintk("NFS reply test_stateid: failed, %d\n", status);
6727                 return status;
6728         }
6729         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
6730         return -res.status;
6731 }
6732
6733 /**
6734  * nfs41_test_stateid - perform a TEST_STATEID operation
6735  *
6736  * @server: server / transport on which to perform the operation
6737  * @stateid: state ID to test
6738  *
6739  * Returns NFS_OK if the server recognizes that "stateid" is valid.
6740  * Otherwise a negative NFS4ERR value is returned if the operation
6741  * failed or the state ID is not currently valid.
6742  */
6743 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6744 {
6745         struct nfs4_exception exception = { };
6746         int err;
6747         do {
6748                 err = _nfs41_test_stateid(server, stateid);
6749                 if (err != -NFS4ERR_DELAY)
6750                         break;
6751                 nfs4_handle_exception(server, err, &exception);
6752         } while (exception.retry);
6753         return err;
6754 }
6755
6756 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6757 {
6758         struct nfs41_free_stateid_args args = {
6759                 .stateid = stateid,
6760         };
6761         struct nfs41_free_stateid_res res;
6762         struct rpc_message msg = {
6763                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6764                 .rpc_argp = &args,
6765                 .rpc_resp = &res,
6766         };
6767         int status;
6768
6769         dprintk("NFS call  free_stateid %p\n", stateid);
6770         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6771         nfs4_set_sequence_privileged(&args.seq_args);
6772         status = nfs4_call_sync_sequence(server->client, server, &msg,
6773                         &args.seq_args, &res.seq_res);
6774         dprintk("NFS reply free_stateid: %d\n", status);
6775         return status;
6776 }
6777
6778 /**
6779  * nfs41_free_stateid - perform a FREE_STATEID operation
6780  *
6781  * @server: server / transport on which to perform the operation
6782  * @stateid: state ID to release
6783  *
6784  * Returns NFS_OK if the server freed "stateid".  Otherwise a
6785  * negative NFS4ERR value is returned.
6786  */
6787 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6788 {
6789         struct nfs4_exception exception = { };
6790         int err;
6791         do {
6792                 err = _nfs4_free_stateid(server, stateid);
6793                 if (err != -NFS4ERR_DELAY)
6794                         break;
6795                 nfs4_handle_exception(server, err, &exception);
6796         } while (exception.retry);
6797         return err;
6798 }
6799
6800 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6801                 const nfs4_stateid *s2)
6802 {
6803         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6804                 return false;
6805
6806         if (s1->seqid == s2->seqid)
6807                 return true;
6808         if (s1->seqid == 0 || s2->seqid == 0)
6809                 return true;
6810
6811         return false;
6812 }
6813
6814 #endif /* CONFIG_NFS_V4_1 */
6815
6816 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6817                 const nfs4_stateid *s2)
6818 {
6819         return nfs4_stateid_match(s1, s2);
6820 }
6821
6822
6823 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6824         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6825         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6826         .recover_open   = nfs4_open_reclaim,
6827         .recover_lock   = nfs4_lock_reclaim,
6828         .establish_clid = nfs4_init_clientid,
6829         .get_clid_cred  = nfs4_get_setclientid_cred,
6830         .detect_trunking = nfs40_discover_server_trunking,
6831 };
6832
6833 #if defined(CONFIG_NFS_V4_1)
6834 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6835         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6836         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6837         .recover_open   = nfs4_open_reclaim,
6838         .recover_lock   = nfs4_lock_reclaim,
6839         .establish_clid = nfs41_init_clientid,
6840         .get_clid_cred  = nfs4_get_exchange_id_cred,
6841         .reclaim_complete = nfs41_proc_reclaim_complete,
6842         .detect_trunking = nfs41_discover_server_trunking,
6843 };
6844 #endif /* CONFIG_NFS_V4_1 */
6845
6846 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6847         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6848         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6849         .recover_open   = nfs4_open_expired,
6850         .recover_lock   = nfs4_lock_expired,
6851         .establish_clid = nfs4_init_clientid,
6852         .get_clid_cred  = nfs4_get_setclientid_cred,
6853 };
6854
6855 #if defined(CONFIG_NFS_V4_1)
6856 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6857         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6858         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6859         .recover_open   = nfs41_open_expired,
6860         .recover_lock   = nfs41_lock_expired,
6861         .establish_clid = nfs41_init_clientid,
6862         .get_clid_cred  = nfs4_get_exchange_id_cred,
6863 };
6864 #endif /* CONFIG_NFS_V4_1 */
6865
6866 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6867         .sched_state_renewal = nfs4_proc_async_renew,
6868         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6869         .renew_lease = nfs4_proc_renew,
6870 };
6871
6872 #if defined(CONFIG_NFS_V4_1)
6873 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6874         .sched_state_renewal = nfs41_proc_async_sequence,
6875         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6876         .renew_lease = nfs4_proc_sequence,
6877 };
6878 #endif
6879
6880 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6881         .minor_version = 0,
6882         .init_caps = NFS_CAP_READDIRPLUS
6883                 | NFS_CAP_ATOMIC_OPEN
6884                 | NFS_CAP_CHANGE_ATTR
6885                 | NFS_CAP_POSIX_LOCK,
6886         .call_sync = _nfs4_call_sync,
6887         .match_stateid = nfs4_match_stateid,
6888         .find_root_sec = nfs4_find_root_sec,
6889         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6890         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6891         .state_renewal_ops = &nfs40_state_renewal_ops,
6892 };
6893
6894 #if defined(CONFIG_NFS_V4_1)
6895 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6896         .minor_version = 1,
6897         .init_caps = NFS_CAP_READDIRPLUS
6898                 | NFS_CAP_ATOMIC_OPEN
6899                 | NFS_CAP_CHANGE_ATTR
6900                 | NFS_CAP_POSIX_LOCK
6901                 | NFS_CAP_STATEID_NFSV41
6902                 | NFS_CAP_ATOMIC_OPEN_V1,
6903         .call_sync = nfs4_call_sync_sequence,
6904         .match_stateid = nfs41_match_stateid,
6905         .find_root_sec = nfs41_find_root_sec,
6906         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6907         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6908         .state_renewal_ops = &nfs41_state_renewal_ops,
6909 };
6910 #endif
6911
6912 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6913         [0] = &nfs_v4_0_minor_ops,
6914 #if defined(CONFIG_NFS_V4_1)
6915         [1] = &nfs_v4_1_minor_ops,
6916 #endif
6917 };
6918
6919 const struct inode_operations nfs4_dir_inode_operations = {
6920         .create         = nfs_create,
6921         .lookup         = nfs_lookup,
6922         .atomic_open    = nfs_atomic_open,
6923         .link           = nfs_link,
6924         .unlink         = nfs_unlink,
6925         .symlink        = nfs_symlink,
6926         .mkdir          = nfs_mkdir,
6927         .rmdir          = nfs_rmdir,
6928         .mknod          = nfs_mknod,
6929         .rename         = nfs_rename,
6930         .permission     = nfs_permission,
6931         .getattr        = nfs_getattr,
6932         .setattr        = nfs_setattr,
6933         .getxattr       = generic_getxattr,
6934         .setxattr       = generic_setxattr,
6935         .listxattr      = generic_listxattr,
6936         .removexattr    = generic_removexattr,
6937 };
6938
6939 static const struct inode_operations nfs4_file_inode_operations = {
6940         .permission     = nfs_permission,
6941         .getattr        = nfs_getattr,
6942         .setattr        = nfs_setattr,
6943         .getxattr       = generic_getxattr,
6944         .setxattr       = generic_setxattr,
6945         .listxattr      = generic_listxattr,
6946         .removexattr    = generic_removexattr,
6947 };
6948
6949 const struct nfs_rpc_ops nfs_v4_clientops = {
6950         .version        = 4,                    /* protocol version */
6951         .dentry_ops     = &nfs4_dentry_operations,
6952         .dir_inode_ops  = &nfs4_dir_inode_operations,
6953         .file_inode_ops = &nfs4_file_inode_operations,
6954         .file_ops       = &nfs4_file_operations,
6955         .getroot        = nfs4_proc_get_root,
6956         .submount       = nfs4_submount,
6957         .try_mount      = nfs4_try_mount,
6958         .getattr        = nfs4_proc_getattr,
6959         .setattr        = nfs4_proc_setattr,
6960         .lookup         = nfs4_proc_lookup,
6961         .access         = nfs4_proc_access,
6962         .readlink       = nfs4_proc_readlink,
6963         .create         = nfs4_proc_create,
6964         .remove         = nfs4_proc_remove,
6965         .unlink_setup   = nfs4_proc_unlink_setup,
6966         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
6967         .unlink_done    = nfs4_proc_unlink_done,
6968         .rename         = nfs4_proc_rename,
6969         .rename_setup   = nfs4_proc_rename_setup,
6970         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
6971         .rename_done    = nfs4_proc_rename_done,
6972         .link           = nfs4_proc_link,
6973         .symlink        = nfs4_proc_symlink,
6974         .mkdir          = nfs4_proc_mkdir,
6975         .rmdir          = nfs4_proc_remove,
6976         .readdir        = nfs4_proc_readdir,
6977         .mknod          = nfs4_proc_mknod,
6978         .statfs         = nfs4_proc_statfs,
6979         .fsinfo         = nfs4_proc_fsinfo,
6980         .pathconf       = nfs4_proc_pathconf,
6981         .set_capabilities = nfs4_server_capabilities,
6982         .decode_dirent  = nfs4_decode_dirent,
6983         .read_setup     = nfs4_proc_read_setup,
6984         .read_pageio_init = pnfs_pageio_init_read,
6985         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
6986         .read_done      = nfs4_read_done,
6987         .write_setup    = nfs4_proc_write_setup,
6988         .write_pageio_init = pnfs_pageio_init_write,
6989         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
6990         .write_done     = nfs4_write_done,
6991         .commit_setup   = nfs4_proc_commit_setup,
6992         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
6993         .commit_done    = nfs4_commit_done,
6994         .lock           = nfs4_proc_lock,
6995         .clear_acl_cache = nfs4_zap_acl_attr,
6996         .close_context  = nfs4_close_context,
6997         .open_context   = nfs4_atomic_open,
6998         .have_delegation = nfs4_have_delegation,
6999         .return_delegation = nfs4_inode_return_delegation,
7000         .alloc_client   = nfs4_alloc_client,
7001         .init_client    = nfs4_init_client,
7002         .free_client    = nfs4_free_client,
7003         .create_server  = nfs4_create_server,
7004         .clone_server   = nfs_clone_server,
7005 };
7006
7007 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7008         .prefix = XATTR_NAME_NFSV4_ACL,
7009         .list   = nfs4_xattr_list_nfs4_acl,
7010         .get    = nfs4_xattr_get_nfs4_acl,
7011         .set    = nfs4_xattr_set_nfs4_acl,
7012 };
7013
7014 const struct xattr_handler *nfs4_xattr_handlers[] = {
7015         &nfs4_xattr_nfs4_acl_handler,
7016         NULL
7017 };
7018
7019 /*
7020  * Local variables:
7021  *  c-basic-offset: 8
7022  * End:
7023  */