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/sunrpc/gss_api.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/nfs_idmap.h>
  56 #include <linux/sunrpc/bc_xprt.h>
  57 #include <linux/xattr.h>
  58 #include <linux/utsname.h>
  59 #include <linux/freezer.h>
  60
  61 #include "nfs4_fs.h"
  62 #include "delegation.h"
  63 #include "internal.h"
  64 #include "iostat.h"
  65 #include "callback.h"
  66 #include "pnfs.h"
  67 #include "netns.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 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  75
  76 static unsigned short max_session_slots = NFS4_DEF_SLOT_TABLE_SIZE;
  77
  78 struct nfs4_opendata;
  79 static int _nfs4_proc_open(struct nfs4_opendata *data);
  80 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  81 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  82 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  83 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  84 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *);
  85 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  86 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  87                             struct nfs_fattr *fattr, struct iattr *sattr,
  88                             struct nfs4_state *state);
  89 #ifdef CONFIG_NFS_V4_1
  90 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
  91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
  92 #endif
  93 /* Prevent leaks of NFSv4 errors into userland */
  94 static int nfs4_map_errors(int err)
  95 {
  96         if (err >= -1000)
  97                 return err;
  98         switch (err) {
  99         case -NFS4ERR_RESOURCE:
 100                 return -EREMOTEIO;
 101         case -NFS4ERR_WRONGSEC:
 102                 return -EPERM;
 103         case -NFS4ERR_BADOWNER:
 104         case -NFS4ERR_BADNAME:
 105                 return -EINVAL;
 106         case -NFS4ERR_SHARE_DENIED:
 107                 return -EACCES;
 108         default:
 109                 dprintk("%s could not handle NFSv4 error %d\n",
 110                                 __func__, -err);
 111                 break;
 112         }
 113         return -EIO;
 114 }
 115
 116 /*
 117  * This is our standard bitmap for GETATTR requests.
 118  */
 119 const u32 nfs4_fattr_bitmap[3] = {
 120         FATTR4_WORD0_TYPE
 121         | FATTR4_WORD0_CHANGE
 122         | FATTR4_WORD0_SIZE
 123         | FATTR4_WORD0_FSID
 124         | FATTR4_WORD0_FILEID,
 125         FATTR4_WORD1_MODE
 126         | FATTR4_WORD1_NUMLINKS
 127         | FATTR4_WORD1_OWNER
 128         | FATTR4_WORD1_OWNER_GROUP
 129         | FATTR4_WORD1_RAWDEV
 130         | FATTR4_WORD1_SPACE_USED
 131         | FATTR4_WORD1_TIME_ACCESS
 132         | FATTR4_WORD1_TIME_METADATA
 133         | FATTR4_WORD1_TIME_MODIFY
 134 };
 135
 136 static const u32 nfs4_pnfs_open_bitmap[3] = {
 137         FATTR4_WORD0_TYPE
 138         | FATTR4_WORD0_CHANGE
 139         | FATTR4_WORD0_SIZE
 140         | FATTR4_WORD0_FSID
 141         | FATTR4_WORD0_FILEID,
 142         FATTR4_WORD1_MODE
 143         | FATTR4_WORD1_NUMLINKS
 144         | FATTR4_WORD1_OWNER
 145         | FATTR4_WORD1_OWNER_GROUP
 146         | FATTR4_WORD1_RAWDEV
 147         | FATTR4_WORD1_SPACE_USED
 148         | FATTR4_WORD1_TIME_ACCESS
 149         | FATTR4_WORD1_TIME_METADATA
 150         | FATTR4_WORD1_TIME_MODIFY,
 151         FATTR4_WORD2_MDSTHRESHOLD
 152 };
 153
 154 const u32 nfs4_statfs_bitmap[2] = {
 155         FATTR4_WORD0_FILES_AVAIL
 156         | FATTR4_WORD0_FILES_FREE
 157         | FATTR4_WORD0_FILES_TOTAL,
 158         FATTR4_WORD1_SPACE_AVAIL
 159         | FATTR4_WORD1_SPACE_FREE
 160         | FATTR4_WORD1_SPACE_TOTAL
 161 };
 162
 163 const u32 nfs4_pathconf_bitmap[2] = {
 164         FATTR4_WORD0_MAXLINK
 165         | FATTR4_WORD0_MAXNAME,
 166         0
 167 };
 168
 169 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 170                         | FATTR4_WORD0_MAXREAD
 171                         | FATTR4_WORD0_MAXWRITE
 172                         | FATTR4_WORD0_LEASE_TIME,
 173                         FATTR4_WORD1_TIME_DELTA
 174                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 175                         FATTR4_WORD2_LAYOUT_BLKSIZE
 176 };
 177
 178 const u32 nfs4_fs_locations_bitmap[2] = {
 179         FATTR4_WORD0_TYPE
 180         | FATTR4_WORD0_CHANGE
 181         | FATTR4_WORD0_SIZE
 182         | FATTR4_WORD0_FSID
 183         | FATTR4_WORD0_FILEID
 184         | FATTR4_WORD0_FS_LOCATIONS,
 185         FATTR4_WORD1_MODE
 186         | FATTR4_WORD1_NUMLINKS
 187         | FATTR4_WORD1_OWNER
 188         | FATTR4_WORD1_OWNER_GROUP
 189         | FATTR4_WORD1_RAWDEV
 190         | FATTR4_WORD1_SPACE_USED
 191         | FATTR4_WORD1_TIME_ACCESS
 192         | FATTR4_WORD1_TIME_METADATA
 193         | FATTR4_WORD1_TIME_MODIFY
 194         | FATTR4_WORD1_MOUNTED_ON_FILEID
 195 };
 196
 197 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 198                 struct nfs4_readdir_arg *readdir)
 199 {
 200         __be32 *start, *p;
 201
 202         BUG_ON(readdir->count < 80);
 203         if (cookie > 2) {
 204                 readdir->cookie = cookie;
 205                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 206                 return;
 207         }
 208
 209         readdir->cookie = 0;
 210         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 211         if (cookie == 2)
 212                 return;
 213
 214         /*
 215          * NFSv4 servers do not return entries for '.' and '..'
 216          * Therefore, we fake these entries here.  We let '.'
 217          * have cookie 0 and '..' have cookie 1.  Note that
 218          * when talking to the server, we always send cookie 0
 219          * instead of 1 or 2.
 220          */
 221         start = p = kmap_atomic(*readdir->pages);
 222
 223         if (cookie == 0) {
 224                 *p++ = xdr_one;                                  /* next */
 225                 *p++ = xdr_zero;                   /* cookie, first word */
 226                 *p++ = xdr_one;                   /* cookie, second word */
 227                 *p++ = xdr_one;                             /* entry len */
 228                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 229                 p++;
 230                 *p++ = xdr_one;                         /* bitmap length */
 231                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 232                 *p++ = htonl(8);              /* attribute buffer length */
 233                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 234         }
 235
 236         *p++ = xdr_one;                                  /* next */
 237         *p++ = xdr_zero;                   /* cookie, first word */
 238         *p++ = xdr_two;                   /* cookie, second word */
 239         *p++ = xdr_two;                             /* entry len */
 240         memcpy(p, "..\0\0", 4);                         /* entry */
 241         p++;
 242         *p++ = xdr_one;                         /* bitmap length */
 243         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 244         *p++ = htonl(8);              /* attribute buffer length */
 245         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 246
 247         readdir->pgbase = (char *)p - (char *)start;
 248         readdir->count -= readdir->pgbase;
 249         kunmap_atomic(start);
 250 }
 251
 252 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 253 {
 254         int res;
 255
 256         might_sleep();
 257
 258         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 259                         nfs_wait_bit_killable, TASK_KILLABLE);
 260         return res;
 261 }
 262
 263 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 264 {
 265         int res = 0;
 266
 267         might_sleep();
 268
 269         if (*timeout <= 0)
 270                 *timeout = NFS4_POLL_RETRY_MIN;
 271         if (*timeout > NFS4_POLL_RETRY_MAX)
 272                 *timeout = NFS4_POLL_RETRY_MAX;
 273         freezable_schedule_timeout_killable(*timeout);
 274         if (fatal_signal_pending(current))
 275                 res = -ERESTARTSYS;
 276         *timeout <<= 1;
 277         return res;
 278 }
 279
 280 /* This is the error handling routine for processes that are allowed
 281  * to sleep.
 282  */
 283 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 284 {
 285         struct nfs_client *clp = server->nfs_client;
 286         struct nfs4_state *state = exception->state;
 287         struct inode *inode = exception->inode;
 288         int ret = errorcode;
 289
 290         exception->retry = 0;
 291         switch(errorcode) {
 292                 case 0:
 293                         return 0;
 294                 case -NFS4ERR_OPENMODE:
 295                         if (inode && nfs_have_delegation(inode, FMODE_READ)) {
 296                                 nfs_inode_return_delegation(inode);
 297                                 exception->retry = 1;
 298                                 return 0;
 299                         }
 300                         if (state == NULL)
 301                                 break;
 302                         nfs4_schedule_stateid_recovery(server, state);
 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                         nfs4_schedule_stateid_recovery(server, state);
 311                         goto wait_on_recovery;
 312                 case -NFS4ERR_EXPIRED:
 313                         if (state != NULL)
 314                                 nfs4_schedule_stateid_recovery(server, state);
 315                 case -NFS4ERR_STALE_STATEID:
 316                 case -NFS4ERR_STALE_CLIENTID:
 317                         nfs4_schedule_lease_recovery(clp);
 318                         goto wait_on_recovery;
 319 #if defined(CONFIG_NFS_V4_1)
 320                 case -NFS4ERR_BADSESSION:
 321                 case -NFS4ERR_BADSLOT:
 322                 case -NFS4ERR_BAD_HIGH_SLOT:
 323                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 324                 case -NFS4ERR_DEADSESSION:
 325                 case -NFS4ERR_SEQ_FALSE_RETRY:
 326                 case -NFS4ERR_SEQ_MISORDERED:
 327                         dprintk("%s ERROR: %d Reset session\n", __func__,
 328                                 errorcode);
 329                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 330                         exception->retry = 1;
 331                         break;
 332 #endif /* defined(CONFIG_NFS_V4_1) */
 333                 case -NFS4ERR_FILE_OPEN:
 334                         if (exception->timeout > HZ) {
 335                                 /* We have retried a decent amount, time to
 336                                  * fail
 337                                  */
 338                                 ret = -EBUSY;
 339                                 break;
 340                         }
 341                 case -NFS4ERR_GRACE:
 342                 case -NFS4ERR_DELAY:
 343                 case -EKEYEXPIRED:
 344                         ret = nfs4_delay(server->client, &exception->timeout);
 345                         if (ret != 0)
 346                                 break;
 347                 case -NFS4ERR_RETRY_UNCACHED_REP:
 348                 case -NFS4ERR_OLD_STATEID:
 349                         exception->retry = 1;
 350                         break;
 351                 case -NFS4ERR_BADOWNER:
 352                         /* The following works around a Linux server bug! */
 353                 case -NFS4ERR_BADNAME:
 354                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 355                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 356                                 exception->retry = 1;
 357                                 printk(KERN_WARNING "NFS: v4 server %s "
 358                                                 "does not accept raw "
 359                                                 "uid/gids. "
 360                                                 "Reenabling the idmapper.\n",
 361                                                 server->nfs_client->cl_hostname);
 362                         }
 363         }
 364         /* We failed to handle the error */
 365         return nfs4_map_errors(ret);
 366 wait_on_recovery:
 367         ret = nfs4_wait_clnt_recover(clp);
 368         if (ret == 0)
 369                 exception->retry = 1;
 370         return ret;
 371 }
 372
 373
 374 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 375 {
 376         spin_lock(&clp->cl_lock);
 377         if (time_before(clp->cl_last_renewal,timestamp))
 378                 clp->cl_last_renewal = timestamp;
 379         spin_unlock(&clp->cl_lock);
 380 }
 381
 382 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 383 {
 384         do_renew_lease(server->nfs_client, timestamp);
 385 }
 386
 387 #if defined(CONFIG_NFS_V4_1)
 388
 389 /*
 390  * nfs4_free_slot - free a slot and efficiently update slot table.
 391  *
 392  * freeing a slot is trivially done by clearing its respective bit
 393  * in the bitmap.
 394  * If the freed slotid equals highest_used_slotid we want to update it
 395  * so that the server would be able to size down the slot table if needed,
 396  * otherwise we know that the highest_used_slotid is still in use.
 397  * When updating highest_used_slotid there may be "holes" in the bitmap
 398  * so we need to scan down from highest_used_slotid to 0 looking for the now
 399  * highest slotid in use.
 400  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
 401  *
 402  * Must be called while holding tbl->slot_tbl_lock
 403  */
 404 static void
 405 nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
 406 {
 407         BUG_ON(slotid >= NFS4_MAX_SLOT_TABLE);
 408         /* clear used bit in bitmap */
 409         __clear_bit(slotid, tbl->used_slots);
 410
 411         /* update highest_used_slotid when it is freed */
 412         if (slotid == tbl->highest_used_slotid) {
 413                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 414                 if (slotid < tbl->max_slots)
 415                         tbl->highest_used_slotid = slotid;
 416                 else
 417                         tbl->highest_used_slotid = NFS4_NO_SLOT;
 418         }
 419         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
 420                 slotid, tbl->highest_used_slotid);
 421 }
 422
 423 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
 424 {
 425         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 426         return true;
 427 }
 428
 429 /*
 430  * Signal state manager thread if session fore channel is drained
 431  */
 432 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 433 {
 434         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 435                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
 436                                 nfs4_set_task_privileged, NULL);
 437                 return;
 438         }
 439
 440         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 441                 return;
 442
 443         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 444         complete(&ses->fc_slot_table.complete);
 445 }
 446
 447 /*
 448  * Signal state manager thread if session back channel is drained
 449  */
 450 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 451 {
 452         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 453             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 454                 return;
 455         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 456         complete(&ses->bc_slot_table.complete);
 457 }
 458
 459 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 460 {
 461         struct nfs4_slot_table *tbl;
 462
 463         tbl = &res->sr_session->fc_slot_table;
 464         if (!res->sr_slot) {
 465                 /* just wake up the next guy waiting since
 466                  * we may have not consumed a slot after all */
 467                 dprintk("%s: No slot\n", __func__);
 468                 return;
 469         }
 470
 471         spin_lock(&tbl->slot_tbl_lock);
 472         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 473         nfs4_check_drain_fc_complete(res->sr_session);
 474         spin_unlock(&tbl->slot_tbl_lock);
 475         res->sr_slot = NULL;
 476 }
 477
 478 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 479 {
 480         unsigned long timestamp;
 481         struct nfs_client *clp;
 482
 483         /*
 484          * sr_status remains 1 if an RPC level error occurred. The server
 485          * may or may not have processed the sequence operation..
 486          * Proceed as if the server received and processed the sequence
 487          * operation.
 488          */
 489         if (res->sr_status == 1)
 490                 res->sr_status = NFS_OK;
 491
 492         /* don't increment the sequence number if the task wasn't sent */
 493         if (!RPC_WAS_SENT(task))
 494                 goto out;
 495
 496         /* Check the SEQUENCE operation status */
 497         switch (res->sr_status) {
 498         case 0:
 499                 /* Update the slot's sequence and clientid lease timer */
 500                 ++res->sr_slot->seq_nr;
 501                 timestamp = res->sr_renewal_time;
 502                 clp = res->sr_session->clp;
 503                 do_renew_lease(clp, timestamp);
 504                 /* Check sequence flags */
 505                 if (res->sr_status_flags != 0)
 506                         nfs4_schedule_lease_recovery(clp);
 507                 break;
 508         case -NFS4ERR_DELAY:
 509                 /* The server detected a resend of the RPC call and
 510                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 511                  * of RFC5661.
 512                  */
 513                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 514                         __func__,
 515                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 516                         res->sr_slot->seq_nr);
 517                 goto out_retry;
 518         default:
 519                 /* Just update the slot sequence no. */
 520                 ++res->sr_slot->seq_nr;
 521         }
 522 out:
 523         /* The session may be reset by one of the error handlers. */
 524         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 525         nfs41_sequence_free_slot(res);
 526         return 1;
 527 out_retry:
 528         if (!rpc_restart_call(task))
 529                 goto out;
 530         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 531         return 0;
 532 }
 533
 534 static int nfs4_sequence_done(struct rpc_task *task,
 535                                struct nfs4_sequence_res *res)
 536 {
 537         if (res->sr_session == NULL)
 538                 return 1;
 539         return nfs41_sequence_done(task, res);
 540 }
 541
 542 /*
 543  * nfs4_find_slot - efficiently look for a free slot
 544  *
 545  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 546  * If found, we mark the slot as used, update the highest_used_slotid,
 547  * and respectively set up the sequence operation args.
 548  * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
 549  *
 550  * Note: must be called with under the slot_tbl_lock.
 551  */
 552 static u32
 553 nfs4_find_slot(struct nfs4_slot_table *tbl)
 554 {
 555         u32 slotid;
 556         u32 ret_id = NFS4_NO_SLOT;
 557
 558         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
 559                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 560                 tbl->max_slots);
 561         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 562         if (slotid >= tbl->max_slots)
 563                 goto out;
 564         __set_bit(slotid, tbl->used_slots);
 565         if (slotid > tbl->highest_used_slotid ||
 566                         tbl->highest_used_slotid == NFS4_NO_SLOT)
 567                 tbl->highest_used_slotid = slotid;
 568         ret_id = slotid;
 569 out:
 570         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 571                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 572         return ret_id;
 573 }
 574
 575 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 576                 struct nfs4_sequence_res *res, int cache_reply)
 577 {
 578         args->sa_session = NULL;
 579         args->sa_cache_this = 0;
 580         if (cache_reply)
 581                 args->sa_cache_this = 1;
 582         res->sr_session = NULL;
 583         res->sr_slot = NULL;
 584 }
 585
 586 int nfs41_setup_sequence(struct nfs4_session *session,
 587                                 struct nfs4_sequence_args *args,
 588                                 struct nfs4_sequence_res *res,
 589                                 struct rpc_task *task)
 590 {
 591         struct nfs4_slot *slot;
 592         struct nfs4_slot_table *tbl;
 593         u32 slotid;
 594
 595         dprintk("--> %s\n", __func__);
 596         /* slot already allocated? */
 597         if (res->sr_slot != NULL)
 598                 return 0;
 599
 600         tbl = &session->fc_slot_table;
 601
 602         spin_lock(&tbl->slot_tbl_lock);
 603         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 604             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 605                 /* The state manager will wait until the slot table is empty */
 606                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 607                 spin_unlock(&tbl->slot_tbl_lock);
 608                 dprintk("%s session is draining\n", __func__);
 609                 return -EAGAIN;
 610         }
 611
 612         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 613             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 614                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 615                 spin_unlock(&tbl->slot_tbl_lock);
 616                 dprintk("%s enforce FIFO order\n", __func__);
 617                 return -EAGAIN;
 618         }
 619
 620         slotid = nfs4_find_slot(tbl);
 621         if (slotid == NFS4_NO_SLOT) {
 622                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 623                 spin_unlock(&tbl->slot_tbl_lock);
 624                 dprintk("<-- %s: no free slots\n", __func__);
 625                 return -EAGAIN;
 626         }
 627         spin_unlock(&tbl->slot_tbl_lock);
 628
 629         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 630         slot = tbl->slots + slotid;
 631         args->sa_session = session;
 632         args->sa_slotid = slotid;
 633
 634         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 635
 636         res->sr_session = session;
 637         res->sr_slot = slot;
 638         res->sr_renewal_time = jiffies;
 639         res->sr_status_flags = 0;
 640         /*
 641          * sr_status is only set in decode_sequence, and so will remain
 642          * set to 1 if an rpc level failure occurs.
 643          */
 644         res->sr_status = 1;
 645         return 0;
 646 }
 647 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 648
 649 int nfs4_setup_sequence(const struct nfs_server *server,
 650                         struct nfs4_sequence_args *args,
 651                         struct nfs4_sequence_res *res,
 652                         struct rpc_task *task)
 653 {
 654         struct nfs4_session *session = nfs4_get_session(server);
 655         int ret = 0;
 656
 657         if (session == NULL)
 658                 goto out;
 659
 660         dprintk("--> %s clp %p session %p sr_slot %td\n",
 661                 __func__, session->clp, session, res->sr_slot ?
 662                         res->sr_slot - session->fc_slot_table.slots : -1);
 663
 664         ret = nfs41_setup_sequence(session, args, res, task);
 665 out:
 666         dprintk("<-- %s status=%d\n", __func__, ret);
 667         return ret;
 668 }
 669
 670 struct nfs41_call_sync_data {
 671         const struct nfs_server *seq_server;
 672         struct nfs4_sequence_args *seq_args;
 673         struct nfs4_sequence_res *seq_res;
 674 };
 675
 676 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 677 {
 678         struct nfs41_call_sync_data *data = calldata;
 679
 680         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 681
 682         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 683                                 data->seq_res, task))
 684                 return;
 685         rpc_call_start(task);
 686 }
 687
 688 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 689 {
 690         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 691         nfs41_call_sync_prepare(task, calldata);
 692 }
 693
 694 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 695 {
 696         struct nfs41_call_sync_data *data = calldata;
 697
 698         nfs41_sequence_done(task, data->seq_res);
 699 }
 700
 701 static const struct rpc_call_ops nfs41_call_sync_ops = {
 702         .rpc_call_prepare = nfs41_call_sync_prepare,
 703         .rpc_call_done = nfs41_call_sync_done,
 704 };
 705
 706 static const struct rpc_call_ops nfs41_call_priv_sync_ops = {
 707         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 708         .rpc_call_done = nfs41_call_sync_done,
 709 };
 710
 711 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 712                                    struct nfs_server *server,
 713                                    struct rpc_message *msg,
 714                                    struct nfs4_sequence_args *args,
 715                                    struct nfs4_sequence_res *res,
 716                                    int privileged)
 717 {
 718         int ret;
 719         struct rpc_task *task;
 720         struct nfs41_call_sync_data data = {
 721                 .seq_server = server,
 722                 .seq_args = args,
 723                 .seq_res = res,
 724         };
 725         struct rpc_task_setup task_setup = {
 726                 .rpc_client = clnt,
 727                 .rpc_message = msg,
 728                 .callback_ops = &nfs41_call_sync_ops,
 729                 .callback_data = &data
 730         };
 731
 732         if (privileged)
 733                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 734         task = rpc_run_task(&task_setup);
 735         if (IS_ERR(task))
 736                 ret = PTR_ERR(task);
 737         else {
 738                 ret = task->tk_status;
 739                 rpc_put_task(task);
 740         }
 741         return ret;
 742 }
 743
 744 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 745                             struct nfs_server *server,
 746                             struct rpc_message *msg,
 747                             struct nfs4_sequence_args *args,
 748                             struct nfs4_sequence_res *res,
 749                             int cache_reply)
 750 {
 751         nfs41_init_sequence(args, res, cache_reply);
 752         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
 753 }
 754
 755 #else
 756 static inline
 757 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 758                 struct nfs4_sequence_res *res, int cache_reply)
 759 {
 760 }
 761
 762 static int nfs4_sequence_done(struct rpc_task *task,
 763                                struct nfs4_sequence_res *res)
 764 {
 765         return 1;
 766 }
 767 #endif /* CONFIG_NFS_V4_1 */
 768
 769 int _nfs4_call_sync(struct rpc_clnt *clnt,
 770                     struct nfs_server *server,
 771                     struct rpc_message *msg,
 772                     struct nfs4_sequence_args *args,
 773                     struct nfs4_sequence_res *res,
 774                     int cache_reply)
 775 {
 776         nfs41_init_sequence(args, res, cache_reply);
 777         return rpc_call_sync(clnt, msg, 0);
 778 }
 779
 780 static inline
 781 int nfs4_call_sync(struct rpc_clnt *clnt,
 782                    struct nfs_server *server,
 783                    struct rpc_message *msg,
 784                    struct nfs4_sequence_args *args,
 785                    struct nfs4_sequence_res *res,
 786                    int cache_reply)
 787 {
 788         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 789                                                 args, res, cache_reply);
 790 }
 791
 792 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 793 {
 794         struct nfs_inode *nfsi = NFS_I(dir);
 795
 796         spin_lock(&dir->i_lock);
 797         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 798         if (!cinfo->atomic || cinfo->before != dir->i_version)
 799                 nfs_force_lookup_revalidate(dir);
 800         dir->i_version = cinfo->after;
 801         spin_unlock(&dir->i_lock);
 802 }
 803
 804 struct nfs4_opendata {
 805         struct kref kref;
 806         struct nfs_openargs o_arg;
 807         struct nfs_openres o_res;
 808         struct nfs_open_confirmargs c_arg;
 809         struct nfs_open_confirmres c_res;
 810         struct nfs4_string owner_name;
 811         struct nfs4_string group_name;
 812         struct nfs_fattr f_attr;
 813         struct dentry *dir;
 814         struct dentry *dentry;
 815         struct nfs4_state_owner *owner;
 816         struct nfs4_state *state;
 817         struct iattr attrs;
 818         unsigned long timestamp;
 819         unsigned int rpc_done : 1;
 820         int rpc_status;
 821         int cancelled;
 822 };
 823
 824
 825 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 826 {
 827         p->o_res.f_attr = &p->f_attr;
 828         p->o_res.seqid = p->o_arg.seqid;
 829         p->c_res.seqid = p->c_arg.seqid;
 830         p->o_res.server = p->o_arg.server;
 831         nfs_fattr_init(&p->f_attr);
 832         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 833 }
 834
 835 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 836                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 837                 const struct iattr *attrs,
 838                 gfp_t gfp_mask)
 839 {
 840         struct dentry *parent = dget_parent(dentry);
 841         struct inode *dir = parent->d_inode;
 842         struct nfs_server *server = NFS_SERVER(dir);
 843         struct nfs4_opendata *p;
 844
 845         p = kzalloc(sizeof(*p), gfp_mask);
 846         if (p == NULL)
 847                 goto err;
 848         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 849         if (p->o_arg.seqid == NULL)
 850                 goto err_free;
 851         nfs_sb_active(dentry->d_sb);
 852         p->dentry = dget(dentry);
 853         p->dir = parent;
 854         p->owner = sp;
 855         atomic_inc(&sp->so_count);
 856         p->o_arg.fh = NFS_FH(dir);
 857         p->o_arg.open_flags = flags;
 858         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 859         p->o_arg.clientid = server->nfs_client->cl_clientid;
 860         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
 861         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 862         p->o_arg.name = &dentry->d_name;
 863         p->o_arg.server = server;
 864         p->o_arg.bitmask = server->attr_bitmask;
 865         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
 866         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 867         if (attrs != NULL && attrs->ia_valid != 0) {
 868                 __be32 verf[2];
 869
 870                 p->o_arg.u.attrs = &p->attrs;
 871                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 872
 873                 verf[0] = jiffies;
 874                 verf[1] = current->pid;
 875                 memcpy(p->o_arg.u.verifier.data, verf,
 876                                 sizeof(p->o_arg.u.verifier.data));
 877         }
 878         p->c_arg.fh = &p->o_res.fh;
 879         p->c_arg.stateid = &p->o_res.stateid;
 880         p->c_arg.seqid = p->o_arg.seqid;
 881         nfs4_init_opendata_res(p);
 882         kref_init(&p->kref);
 883         return p;
 884 err_free:
 885         kfree(p);
 886 err:
 887         dput(parent);
 888         return NULL;
 889 }
 890
 891 static void nfs4_opendata_free(struct kref *kref)
 892 {
 893         struct nfs4_opendata *p = container_of(kref,
 894                         struct nfs4_opendata, kref);
 895         struct super_block *sb = p->dentry->d_sb;
 896
 897         nfs_free_seqid(p->o_arg.seqid);
 898         if (p->state != NULL)
 899                 nfs4_put_open_state(p->state);
 900         nfs4_put_state_owner(p->owner);
 901         dput(p->dir);
 902         dput(p->dentry);
 903         nfs_sb_deactive(sb);
 904         nfs_fattr_free_names(&p->f_attr);
 905         kfree(p);
 906 }
 907
 908 static void nfs4_opendata_put(struct nfs4_opendata *p)
 909 {
 910         if (p != NULL)
 911                 kref_put(&p->kref, nfs4_opendata_free);
 912 }
 913
 914 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 915 {
 916         int ret;
 917
 918         ret = rpc_wait_for_completion_task(task);
 919         return ret;
 920 }
 921
 922 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 923 {
 924         int ret = 0;
 925
 926         if (open_mode & (O_EXCL|O_TRUNC))
 927                 goto out;
 928         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 929                 case FMODE_READ:
 930                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 931                                 && state->n_rdonly != 0;
 932                         break;
 933                 case FMODE_WRITE:
 934                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 935                                 && state->n_wronly != 0;
 936                         break;
 937                 case FMODE_READ|FMODE_WRITE:
 938                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 939                                 && state->n_rdwr != 0;
 940         }
 941 out:
 942         return ret;
 943 }
 944
 945 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 946 {
 947         if (delegation == NULL)
 948                 return 0;
 949         if ((delegation->type & fmode) != fmode)
 950                 return 0;
 951         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &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             (deleg_cur->type & fmode) != fmode)
1031                 goto no_delegation_unlock;
1032
1033         if (delegation == NULL)
1034                 delegation = &deleg_cur->stateid;
1035         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1036                 goto no_delegation_unlock;
1037
1038         nfs_mark_delegation_referenced(deleg_cur);
1039         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1040         ret = 1;
1041 no_delegation_unlock:
1042         spin_unlock(&deleg_cur->lock);
1043 no_delegation:
1044         rcu_read_unlock();
1045
1046         if (!ret && open_stateid != NULL) {
1047                 __update_open_stateid(state, open_stateid, NULL, fmode);
1048                 ret = 1;
1049         }
1050
1051         return ret;
1052 }
1053
1054
1055 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1056 {
1057         struct nfs_delegation *delegation;
1058
1059         rcu_read_lock();
1060         delegation = rcu_dereference(NFS_I(inode)->delegation);
1061         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1062                 rcu_read_unlock();
1063                 return;
1064         }
1065         rcu_read_unlock();
1066         nfs_inode_return_delegation(inode);
1067 }
1068
1069 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1070 {
1071         struct nfs4_state *state = opendata->state;
1072         struct nfs_inode *nfsi = NFS_I(state->inode);
1073         struct nfs_delegation *delegation;
1074         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1075         fmode_t fmode = opendata->o_arg.fmode;
1076         nfs4_stateid stateid;
1077         int ret = -EAGAIN;
1078
1079         for (;;) {
1080                 if (can_open_cached(state, fmode, open_mode)) {
1081                         spin_lock(&state->owner->so_lock);
1082                         if (can_open_cached(state, fmode, open_mode)) {
1083                                 update_open_stateflags(state, fmode);
1084                                 spin_unlock(&state->owner->so_lock);
1085                                 goto out_return_state;
1086                         }
1087                         spin_unlock(&state->owner->so_lock);
1088                 }
1089                 rcu_read_lock();
1090                 delegation = rcu_dereference(nfsi->delegation);
1091                 if (!can_open_delegated(delegation, fmode)) {
1092                         rcu_read_unlock();
1093                         break;
1094                 }
1095                 /* Save the delegation */
1096                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1097                 rcu_read_unlock();
1098                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1099                 if (ret != 0)
1100                         goto out;
1101                 ret = -EAGAIN;
1102
1103                 /* Try to update the stateid using the delegation */
1104                 if (update_open_stateid(state, NULL, &stateid, fmode))
1105                         goto out_return_state;
1106         }
1107 out:
1108         return ERR_PTR(ret);
1109 out_return_state:
1110         atomic_inc(&state->count);
1111         return state;
1112 }
1113
1114 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1115 {
1116         struct inode *inode;
1117         struct nfs4_state *state = NULL;
1118         struct nfs_delegation *delegation;
1119         int ret;
1120
1121         if (!data->rpc_done) {
1122                 state = nfs4_try_open_cached(data);
1123                 goto out;
1124         }
1125
1126         ret = -EAGAIN;
1127         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1128                 goto err;
1129         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1130         ret = PTR_ERR(inode);
1131         if (IS_ERR(inode))
1132                 goto err;
1133         ret = -ENOMEM;
1134         state = nfs4_get_open_state(inode, data->owner);
1135         if (state == NULL)
1136                 goto err_put_inode;
1137         if (data->o_res.delegation_type != 0) {
1138                 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1139                 int delegation_flags = 0;
1140
1141                 rcu_read_lock();
1142                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1143                 if (delegation)
1144                         delegation_flags = delegation->flags;
1145                 rcu_read_unlock();
1146                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1147                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1148                                         "returning a delegation for "
1149                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1150                                         clp->cl_hostname);
1151                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1152                         nfs_inode_set_delegation(state->inode,
1153                                         data->owner->so_cred,
1154                                         &data->o_res);
1155                 else
1156                         nfs_inode_reclaim_delegation(state->inode,
1157                                         data->owner->so_cred,
1158                                         &data->o_res);
1159         }
1160
1161         update_open_stateid(state, &data->o_res.stateid, NULL,
1162                         data->o_arg.fmode);
1163         iput(inode);
1164 out:
1165         return state;
1166 err_put_inode:
1167         iput(inode);
1168 err:
1169         return ERR_PTR(ret);
1170 }
1171
1172 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1173 {
1174         struct nfs_inode *nfsi = NFS_I(state->inode);
1175         struct nfs_open_context *ctx;
1176
1177         spin_lock(&state->inode->i_lock);
1178         list_for_each_entry(ctx, &nfsi->open_files, list) {
1179                 if (ctx->state != state)
1180                         continue;
1181                 get_nfs_open_context(ctx);
1182                 spin_unlock(&state->inode->i_lock);
1183                 return ctx;
1184         }
1185         spin_unlock(&state->inode->i_lock);
1186         return ERR_PTR(-ENOENT);
1187 }
1188
1189 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1190 {
1191         struct nfs4_opendata *opendata;
1192
1193         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1194         if (opendata == NULL)
1195                 return ERR_PTR(-ENOMEM);
1196         opendata->state = state;
1197         atomic_inc(&state->count);
1198         return opendata;
1199 }
1200
1201 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1202 {
1203         struct nfs4_state *newstate;
1204         int ret;
1205
1206         opendata->o_arg.open_flags = 0;
1207         opendata->o_arg.fmode = fmode;
1208         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1209         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1210         nfs4_init_opendata_res(opendata);
1211         ret = _nfs4_recover_proc_open(opendata);
1212         if (ret != 0)
1213                 return ret;
1214         newstate = nfs4_opendata_to_nfs4_state(opendata);
1215         if (IS_ERR(newstate))
1216                 return PTR_ERR(newstate);
1217         nfs4_close_state(newstate, fmode);
1218         *res = newstate;
1219         return 0;
1220 }
1221
1222 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1223 {
1224         struct nfs4_state *newstate;
1225         int ret;
1226
1227         /* memory barrier prior to reading state->n_* */
1228         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1229         smp_rmb();
1230         if (state->n_rdwr != 0) {
1231                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1232                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1233                 if (ret != 0)
1234                         return ret;
1235                 if (newstate != state)
1236                         return -ESTALE;
1237         }
1238         if (state->n_wronly != 0) {
1239                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1240                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1241                 if (ret != 0)
1242                         return ret;
1243                 if (newstate != state)
1244                         return -ESTALE;
1245         }
1246         if (state->n_rdonly != 0) {
1247                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1248                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1249                 if (ret != 0)
1250                         return ret;
1251                 if (newstate != state)
1252                         return -ESTALE;
1253         }
1254         /*
1255          * We may have performed cached opens for all three recoveries.
1256          * Check if we need to update the current stateid.
1257          */
1258         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1259             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1260                 write_seqlock(&state->seqlock);
1261                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1262                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1263                 write_sequnlock(&state->seqlock);
1264         }
1265         return 0;
1266 }
1267
1268 /*
1269  * OPEN_RECLAIM:
1270  *      reclaim state on the server after a reboot.
1271  */
1272 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1273 {
1274         struct nfs_delegation *delegation;
1275         struct nfs4_opendata *opendata;
1276         fmode_t delegation_type = 0;
1277         int status;
1278
1279         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1280         if (IS_ERR(opendata))
1281                 return PTR_ERR(opendata);
1282         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1283         opendata->o_arg.fh = NFS_FH(state->inode);
1284         rcu_read_lock();
1285         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1286         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1287                 delegation_type = delegation->type;
1288         rcu_read_unlock();
1289         opendata->o_arg.u.delegation_type = delegation_type;
1290         status = nfs4_open_recover(opendata, state);
1291         nfs4_opendata_put(opendata);
1292         return status;
1293 }
1294
1295 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1296 {
1297         struct nfs_server *server = NFS_SERVER(state->inode);
1298         struct nfs4_exception exception = { };
1299         int err;
1300         do {
1301                 err = _nfs4_do_open_reclaim(ctx, state);
1302                 if (err != -NFS4ERR_DELAY)
1303                         break;
1304                 nfs4_handle_exception(server, err, &exception);
1305         } while (exception.retry);
1306         return err;
1307 }
1308
1309 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1310 {
1311         struct nfs_open_context *ctx;
1312         int ret;
1313
1314         ctx = nfs4_state_find_open_context(state);
1315         if (IS_ERR(ctx))
1316                 return PTR_ERR(ctx);
1317         ret = nfs4_do_open_reclaim(ctx, state);
1318         put_nfs_open_context(ctx);
1319         return ret;
1320 }
1321
1322 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1323 {
1324         struct nfs4_opendata *opendata;
1325         int ret;
1326
1327         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1328         if (IS_ERR(opendata))
1329                 return PTR_ERR(opendata);
1330         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1331         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1332         ret = nfs4_open_recover(opendata, state);
1333         nfs4_opendata_put(opendata);
1334         return ret;
1335 }
1336
1337 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1338 {
1339         struct nfs4_exception exception = { };
1340         struct nfs_server *server = NFS_SERVER(state->inode);
1341         int err;
1342         do {
1343                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1344                 switch (err) {
1345                         case 0:
1346                         case -ENOENT:
1347                         case -ESTALE:
1348                                 goto out;
1349                         case -NFS4ERR_BADSESSION:
1350                         case -NFS4ERR_BADSLOT:
1351                         case -NFS4ERR_BAD_HIGH_SLOT:
1352                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1353                         case -NFS4ERR_DEADSESSION:
1354                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1355                                 goto out;
1356                         case -NFS4ERR_STALE_CLIENTID:
1357                         case -NFS4ERR_STALE_STATEID:
1358                         case -NFS4ERR_EXPIRED:
1359                                 /* Don't recall a delegation if it was lost */
1360                                 nfs4_schedule_lease_recovery(server->nfs_client);
1361                                 goto out;
1362                         case -ERESTARTSYS:
1363                                 /*
1364                                  * The show must go on: exit, but mark the
1365                                  * stateid as needing recovery.
1366                                  */
1367                         case -NFS4ERR_DELEG_REVOKED:
1368                         case -NFS4ERR_ADMIN_REVOKED:
1369                         case -NFS4ERR_BAD_STATEID:
1370                                 nfs_inode_find_state_and_recover(state->inode,
1371                                                 stateid);
1372                                 nfs4_schedule_stateid_recovery(server, state);
1373                         case -EKEYEXPIRED:
1374                                 /*
1375                                  * User RPCSEC_GSS context has expired.
1376                                  * We cannot recover this stateid now, so
1377                                  * skip it and allow recovery thread to
1378                                  * proceed.
1379                                  */
1380                         case -ENOMEM:
1381                                 err = 0;
1382                                 goto out;
1383                 }
1384                 err = nfs4_handle_exception(server, err, &exception);
1385         } while (exception.retry);
1386 out:
1387         return err;
1388 }
1389
1390 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1391 {
1392         struct nfs4_opendata *data = calldata;
1393
1394         data->rpc_status = task->tk_status;
1395         if (data->rpc_status == 0) {
1396                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1397                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1398                 renew_lease(data->o_res.server, data->timestamp);
1399                 data->rpc_done = 1;
1400         }
1401 }
1402
1403 static void nfs4_open_confirm_release(void *calldata)
1404 {
1405         struct nfs4_opendata *data = calldata;
1406         struct nfs4_state *state = NULL;
1407
1408         /* If this request hasn't been cancelled, do nothing */
1409         if (data->cancelled == 0)
1410                 goto out_free;
1411         /* In case of error, no cleanup! */
1412         if (!data->rpc_done)
1413                 goto out_free;
1414         state = nfs4_opendata_to_nfs4_state(data);
1415         if (!IS_ERR(state))
1416                 nfs4_close_state(state, data->o_arg.fmode);
1417 out_free:
1418         nfs4_opendata_put(data);
1419 }
1420
1421 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1422         .rpc_call_done = nfs4_open_confirm_done,
1423         .rpc_release = nfs4_open_confirm_release,
1424 };
1425
1426 /*
1427  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1428  */
1429 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1430 {
1431         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1432         struct rpc_task *task;
1433         struct  rpc_message msg = {
1434                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1435                 .rpc_argp = &data->c_arg,
1436                 .rpc_resp = &data->c_res,
1437                 .rpc_cred = data->owner->so_cred,
1438         };
1439         struct rpc_task_setup task_setup_data = {
1440                 .rpc_client = server->client,
1441                 .rpc_message = &msg,
1442                 .callback_ops = &nfs4_open_confirm_ops,
1443                 .callback_data = data,
1444                 .workqueue = nfsiod_workqueue,
1445                 .flags = RPC_TASK_ASYNC,
1446         };
1447         int status;
1448
1449         kref_get(&data->kref);
1450         data->rpc_done = 0;
1451         data->rpc_status = 0;
1452         data->timestamp = jiffies;
1453         task = rpc_run_task(&task_setup_data);
1454         if (IS_ERR(task))
1455                 return PTR_ERR(task);
1456         status = nfs4_wait_for_completion_rpc_task(task);
1457         if (status != 0) {
1458                 data->cancelled = 1;
1459                 smp_wmb();
1460         } else
1461                 status = data->rpc_status;
1462         rpc_put_task(task);
1463         return status;
1464 }
1465
1466 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1467 {
1468         struct nfs4_opendata *data = calldata;
1469         struct nfs4_state_owner *sp = data->owner;
1470
1471         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1472                 return;
1473         /*
1474          * Check if we still need to send an OPEN call, or if we can use
1475          * a delegation instead.
1476          */
1477         if (data->state != NULL) {
1478                 struct nfs_delegation *delegation;
1479
1480                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1481                         goto out_no_action;
1482                 rcu_read_lock();
1483                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1484                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1485                     can_open_delegated(delegation, data->o_arg.fmode))
1486                         goto unlock_no_action;
1487                 rcu_read_unlock();
1488         }
1489         /* Update client id. */
1490         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1491         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1492                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1493                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1494         }
1495         data->timestamp = jiffies;
1496         if (nfs4_setup_sequence(data->o_arg.server,
1497                                 &data->o_arg.seq_args,
1498                                 &data->o_res.seq_res, task))
1499                 return;
1500         rpc_call_start(task);
1501         return;
1502 unlock_no_action:
1503         rcu_read_unlock();
1504 out_no_action:
1505         task->tk_action = NULL;
1506
1507 }
1508
1509 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1510 {
1511         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1512         nfs4_open_prepare(task, calldata);
1513 }
1514
1515 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1516 {
1517         struct nfs4_opendata *data = calldata;
1518
1519         data->rpc_status = task->tk_status;
1520
1521         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1522                 return;
1523
1524         if (task->tk_status == 0) {
1525                 switch (data->o_res.f_attr->mode & S_IFMT) {
1526                         case S_IFREG:
1527                                 break;
1528                         case S_IFLNK:
1529                                 data->rpc_status = -ELOOP;
1530                                 break;
1531                         case S_IFDIR:
1532                                 data->rpc_status = -EISDIR;
1533                                 break;
1534                         default:
1535                                 data->rpc_status = -ENOTDIR;
1536                 }
1537                 renew_lease(data->o_res.server, data->timestamp);
1538                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1539                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1540         }
1541         data->rpc_done = 1;
1542 }
1543
1544 static void nfs4_open_release(void *calldata)
1545 {
1546         struct nfs4_opendata *data = calldata;
1547         struct nfs4_state *state = NULL;
1548
1549         /* If this request hasn't been cancelled, do nothing */
1550         if (data->cancelled == 0)
1551                 goto out_free;
1552         /* In case of error, no cleanup! */
1553         if (data->rpc_status != 0 || !data->rpc_done)
1554                 goto out_free;
1555         /* In case we need an open_confirm, no cleanup! */
1556         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1557                 goto out_free;
1558         state = nfs4_opendata_to_nfs4_state(data);
1559         if (!IS_ERR(state))
1560                 nfs4_close_state(state, data->o_arg.fmode);
1561 out_free:
1562         nfs4_opendata_put(data);
1563 }
1564
1565 static const struct rpc_call_ops nfs4_open_ops = {
1566         .rpc_call_prepare = nfs4_open_prepare,
1567         .rpc_call_done = nfs4_open_done,
1568         .rpc_release = nfs4_open_release,
1569 };
1570
1571 static const struct rpc_call_ops nfs4_recover_open_ops = {
1572         .rpc_call_prepare = nfs4_recover_open_prepare,
1573         .rpc_call_done = nfs4_open_done,
1574         .rpc_release = nfs4_open_release,
1575 };
1576
1577 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1578 {
1579         struct inode *dir = data->dir->d_inode;
1580         struct nfs_server *server = NFS_SERVER(dir);
1581         struct nfs_openargs *o_arg = &data->o_arg;
1582         struct nfs_openres *o_res = &data->o_res;
1583         struct rpc_task *task;
1584         struct rpc_message msg = {
1585                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1586                 .rpc_argp = o_arg,
1587                 .rpc_resp = o_res,
1588                 .rpc_cred = data->owner->so_cred,
1589         };
1590         struct rpc_task_setup task_setup_data = {
1591                 .rpc_client = server->client,
1592                 .rpc_message = &msg,
1593                 .callback_ops = &nfs4_open_ops,
1594                 .callback_data = data,
1595                 .workqueue = nfsiod_workqueue,
1596                 .flags = RPC_TASK_ASYNC,
1597         };
1598         int status;
1599
1600         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1601         kref_get(&data->kref);
1602         data->rpc_done = 0;
1603         data->rpc_status = 0;
1604         data->cancelled = 0;
1605         if (isrecover)
1606                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1607         task = rpc_run_task(&task_setup_data);
1608         if (IS_ERR(task))
1609                 return PTR_ERR(task);
1610         status = nfs4_wait_for_completion_rpc_task(task);
1611         if (status != 0) {
1612                 data->cancelled = 1;
1613                 smp_wmb();
1614         } else
1615                 status = data->rpc_status;
1616         rpc_put_task(task);
1617
1618         return status;
1619 }
1620
1621 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1622 {
1623         struct inode *dir = data->dir->d_inode;
1624         struct nfs_openres *o_res = &data->o_res;
1625         int status;
1626
1627         status = nfs4_run_open_task(data, 1);
1628         if (status != 0 || !data->rpc_done)
1629                 return status;
1630
1631         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1632
1633         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1634                 status = _nfs4_proc_open_confirm(data);
1635                 if (status != 0)
1636                         return status;
1637         }
1638
1639         return status;
1640 }
1641
1642 /*
1643  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1644  */
1645 static int _nfs4_proc_open(struct nfs4_opendata *data)
1646 {
1647         struct inode *dir = data->dir->d_inode;
1648         struct nfs_server *server = NFS_SERVER(dir);
1649         struct nfs_openargs *o_arg = &data->o_arg;
1650         struct nfs_openres *o_res = &data->o_res;
1651         int status;
1652
1653         status = nfs4_run_open_task(data, 0);
1654         if (!data->rpc_done)
1655                 return status;
1656         if (status != 0) {
1657                 if (status == -NFS4ERR_BADNAME &&
1658                                 !(o_arg->open_flags & O_CREAT))
1659                         return -ENOENT;
1660                 return status;
1661         }
1662
1663         nfs_fattr_map_and_free_names(server, &data->f_attr);
1664
1665         if (o_arg->open_flags & O_CREAT)
1666                 update_changeattr(dir, &o_res->cinfo);
1667         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1668                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1669         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1670                 status = _nfs4_proc_open_confirm(data);
1671                 if (status != 0)
1672                         return status;
1673         }
1674         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1675                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1676         return 0;
1677 }
1678
1679 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1680 {
1681         unsigned int loop;
1682         int ret;
1683
1684         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1685                 ret = nfs4_wait_clnt_recover(clp);
1686                 if (ret != 0)
1687                         break;
1688                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1689                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1690                         break;
1691                 nfs4_schedule_state_manager(clp);
1692                 ret = -EIO;
1693         }
1694         return ret;
1695 }
1696
1697 static int nfs4_recover_expired_lease(struct nfs_server *server)
1698 {
1699         return nfs4_client_recover_expired_lease(server->nfs_client);
1700 }
1701
1702 /*
1703  * OPEN_EXPIRED:
1704  *      reclaim state on the server after a network partition.
1705  *      Assumes caller holds the appropriate lock
1706  */
1707 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1708 {
1709         struct nfs4_opendata *opendata;
1710         int ret;
1711
1712         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1713         if (IS_ERR(opendata))
1714                 return PTR_ERR(opendata);
1715         ret = nfs4_open_recover(opendata, state);
1716         if (ret == -ESTALE)
1717                 d_drop(ctx->dentry);
1718         nfs4_opendata_put(opendata);
1719         return ret;
1720 }
1721
1722 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1723 {
1724         struct nfs_server *server = NFS_SERVER(state->inode);
1725         struct nfs4_exception exception = { };
1726         int err;
1727
1728         do {
1729                 err = _nfs4_open_expired(ctx, state);
1730                 switch (err) {
1731                 default:
1732                         goto out;
1733                 case -NFS4ERR_GRACE:
1734                 case -NFS4ERR_DELAY:
1735                         nfs4_handle_exception(server, err, &exception);
1736                         err = 0;
1737                 }
1738         } while (exception.retry);
1739 out:
1740         return err;
1741 }
1742
1743 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1744 {
1745         struct nfs_open_context *ctx;
1746         int ret;
1747
1748         ctx = nfs4_state_find_open_context(state);
1749         if (IS_ERR(ctx))
1750                 return PTR_ERR(ctx);
1751         ret = nfs4_do_open_expired(ctx, state);
1752         put_nfs_open_context(ctx);
1753         return ret;
1754 }
1755
1756 #if defined(CONFIG_NFS_V4_1)
1757 static int nfs41_check_expired_stateid(struct nfs4_state *state, nfs4_stateid *stateid, unsigned int flags)
1758 {
1759         int status = NFS_OK;
1760         struct nfs_server *server = NFS_SERVER(state->inode);
1761
1762         if (state->flags & flags) {
1763                 status = nfs41_test_stateid(server, stateid);
1764                 if (status != NFS_OK) {
1765                         nfs41_free_stateid(server, stateid);
1766                         state->flags &= ~flags;
1767                 }
1768         }
1769         return status;
1770 }
1771
1772 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1773 {
1774         int deleg_status, open_status;
1775         int deleg_flags = 1 << NFS_DELEGATED_STATE;
1776         int open_flags = (1 << NFS_O_RDONLY_STATE) | (1 << NFS_O_WRONLY_STATE) | (1 << NFS_O_RDWR_STATE);
1777
1778         deleg_status = nfs41_check_expired_stateid(state, &state->stateid, deleg_flags);
1779         open_status = nfs41_check_expired_stateid(state,  &state->open_stateid, open_flags);
1780
1781         if ((deleg_status == NFS_OK) && (open_status == NFS_OK))
1782                 return NFS_OK;
1783         return nfs4_open_expired(sp, state);
1784 }
1785 #endif
1786
1787 /*
1788  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1789  * fields corresponding to attributes that were used to store the verifier.
1790  * Make sure we clobber those fields in the later setattr call
1791  */
1792 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1793 {
1794         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1795             !(sattr->ia_valid & ATTR_ATIME_SET))
1796                 sattr->ia_valid |= ATTR_ATIME;
1797
1798         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1799             !(sattr->ia_valid & ATTR_MTIME_SET))
1800                 sattr->ia_valid |= ATTR_MTIME;
1801 }
1802
1803 /*
1804  * Returns a referenced nfs4_state
1805  */
1806 static int _nfs4_do_open(struct inode *dir,
1807                         struct dentry *dentry,
1808                         fmode_t fmode,
1809                         int flags,
1810                         struct iattr *sattr,
1811                         struct rpc_cred *cred,
1812                         struct nfs4_state **res,
1813                         struct nfs4_threshold **ctx_th)
1814 {
1815         struct nfs4_state_owner  *sp;
1816         struct nfs4_state     *state = NULL;
1817         struct nfs_server       *server = NFS_SERVER(dir);
1818         struct nfs4_opendata *opendata;
1819         int status;
1820
1821         /* Protect against reboot recovery conflicts */
1822         status = -ENOMEM;
1823         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1824         if (sp == NULL) {
1825                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1826                 goto out_err;
1827         }
1828         status = nfs4_recover_expired_lease(server);
1829         if (status != 0)
1830                 goto err_put_state_owner;
1831         if (dentry->d_inode != NULL)
1832                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1833         status = -ENOMEM;
1834         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1835         if (opendata == NULL)
1836                 goto err_put_state_owner;
1837
1838         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1839                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1840                 if (!opendata->f_attr.mdsthreshold)
1841                         goto err_opendata_put;
1842                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
1843         }
1844         if (dentry->d_inode != NULL)
1845                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1846
1847         status = _nfs4_proc_open(opendata);
1848         if (status != 0)
1849                 goto err_opendata_put;
1850
1851         state = nfs4_opendata_to_nfs4_state(opendata);
1852         status = PTR_ERR(state);
1853         if (IS_ERR(state))
1854                 goto err_opendata_put;
1855         if (server->caps & NFS_CAP_POSIX_LOCK)
1856                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1857
1858         if (opendata->o_arg.open_flags & O_EXCL) {
1859                 nfs4_exclusive_attrset(opendata, sattr);
1860
1861                 nfs_fattr_init(opendata->o_res.f_attr);
1862                 status = nfs4_do_setattr(state->inode, cred,
1863                                 opendata->o_res.f_attr, sattr,
1864                                 state);
1865                 if (status == 0)
1866                         nfs_setattr_update_inode(state->inode, sattr);
1867                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1868         }
1869
1870         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
1871                 *ctx_th = opendata->f_attr.mdsthreshold;
1872         else
1873                 kfree(opendata->f_attr.mdsthreshold);
1874         opendata->f_attr.mdsthreshold = NULL;
1875
1876         nfs4_opendata_put(opendata);
1877         nfs4_put_state_owner(sp);
1878         *res = state;
1879         return 0;
1880 err_opendata_put:
1881         kfree(opendata->f_attr.mdsthreshold);
1882         nfs4_opendata_put(opendata);
1883 err_put_state_owner:
1884         nfs4_put_state_owner(sp);
1885 out_err:
1886         *res = NULL;
1887         return status;
1888 }
1889
1890
1891 static struct nfs4_state *nfs4_do_open(struct inode *dir,
1892                                         struct dentry *dentry,
1893                                         fmode_t fmode,
1894                                         int flags,
1895                                         struct iattr *sattr,
1896                                         struct rpc_cred *cred,
1897                                         struct nfs4_threshold **ctx_th)
1898 {
1899         struct nfs4_exception exception = { };
1900         struct nfs4_state *res;
1901         int status;
1902
1903         do {
1904                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
1905                                        &res, ctx_th);
1906                 if (status == 0)
1907                         break;
1908                 /* NOTE: BAD_SEQID means the server and client disagree about the
1909                  * book-keeping w.r.t. state-changing operations
1910                  * (OPEN/CLOSE/LOCK/LOCKU...)
1911                  * It is actually a sign of a bug on the client or on the server.
1912                  *
1913                  * If we receive a BAD_SEQID error in the particular case of
1914                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1915                  * have unhashed the old state_owner for us, and that we can
1916                  * therefore safely retry using a new one. We should still warn
1917                  * the user though...
1918                  */
1919                 if (status == -NFS4ERR_BAD_SEQID) {
1920                         pr_warn_ratelimited("NFS: v4 server %s "
1921                                         " returned a bad sequence-id error!\n",
1922                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1923                         exception.retry = 1;
1924                         continue;
1925                 }
1926                 /*
1927                  * BAD_STATEID on OPEN means that the server cancelled our
1928                  * state before it received the OPEN_CONFIRM.
1929                  * Recover by retrying the request as per the discussion
1930                  * on Page 181 of RFC3530.
1931                  */
1932                 if (status == -NFS4ERR_BAD_STATEID) {
1933                         exception.retry = 1;
1934                         continue;
1935                 }
1936                 if (status == -EAGAIN) {
1937                         /* We must have found a delegation */
1938                         exception.retry = 1;
1939                         continue;
1940                 }
1941                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1942                                         status, &exception));
1943         } while (exception.retry);
1944         return res;
1945 }
1946
1947 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1948                             struct nfs_fattr *fattr, struct iattr *sattr,
1949                             struct nfs4_state *state)
1950 {
1951         struct nfs_server *server = NFS_SERVER(inode);
1952         struct nfs_setattrargs  arg = {
1953                 .fh             = NFS_FH(inode),
1954                 .iap            = sattr,
1955                 .server         = server,
1956                 .bitmask = server->attr_bitmask,
1957         };
1958         struct nfs_setattrres  res = {
1959                 .fattr          = fattr,
1960                 .server         = server,
1961         };
1962         struct rpc_message msg = {
1963                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1964                 .rpc_argp       = &arg,
1965                 .rpc_resp       = &res,
1966                 .rpc_cred       = cred,
1967         };
1968         unsigned long timestamp = jiffies;
1969         int status;
1970
1971         nfs_fattr_init(fattr);
1972
1973         if (state != NULL) {
1974                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
1975                                 current->files, current->tgid);
1976         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
1977                                 FMODE_WRITE)) {
1978                 /* Use that stateid */
1979         } else
1980                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
1981
1982         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1983         if (status == 0 && state != NULL)
1984                 renew_lease(server, timestamp);
1985         return status;
1986 }
1987
1988 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1989                            struct nfs_fattr *fattr, struct iattr *sattr,
1990                            struct nfs4_state *state)
1991 {
1992         struct nfs_server *server = NFS_SERVER(inode);
1993         struct nfs4_exception exception = {
1994                 .state = state,
1995                 .inode = inode,
1996         };
1997         int err;
1998         do {
1999                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2000                 switch (err) {
2001                 case -NFS4ERR_OPENMODE:
2002                         if (state && !(state->state & FMODE_WRITE)) {
2003                                 err = -EBADF;
2004                                 if (sattr->ia_valid & ATTR_OPEN)
2005                                         err = -EACCES;
2006                                 goto out;
2007                         }
2008                 }
2009                 err = nfs4_handle_exception(server, err, &exception);
2010         } while (exception.retry);
2011 out:
2012         return err;
2013 }
2014
2015 struct nfs4_closedata {
2016         struct inode *inode;
2017         struct nfs4_state *state;
2018         struct nfs_closeargs arg;
2019         struct nfs_closeres res;
2020         struct nfs_fattr fattr;
2021         unsigned long timestamp;
2022         bool roc;
2023         u32 roc_barrier;
2024 };
2025
2026 static void nfs4_free_closedata(void *data)
2027 {
2028         struct nfs4_closedata *calldata = data;
2029         struct nfs4_state_owner *sp = calldata->state->owner;
2030         struct super_block *sb = calldata->state->inode->i_sb;
2031
2032         if (calldata->roc)
2033                 pnfs_roc_release(calldata->state->inode);
2034         nfs4_put_open_state(calldata->state);
2035         nfs_free_seqid(calldata->arg.seqid);
2036         nfs4_put_state_owner(sp);
2037         nfs_sb_deactive(sb);
2038         kfree(calldata);
2039 }
2040
2041 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2042                 fmode_t fmode)
2043 {
2044         spin_lock(&state->owner->so_lock);
2045         if (!(fmode & FMODE_READ))
2046                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2047         if (!(fmode & FMODE_WRITE))
2048                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2049         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2050         spin_unlock(&state->owner->so_lock);
2051 }
2052
2053 static void nfs4_close_done(struct rpc_task *task, void *data)
2054 {
2055         struct nfs4_closedata *calldata = data;
2056         struct nfs4_state *state = calldata->state;
2057         struct nfs_server *server = NFS_SERVER(calldata->inode);
2058
2059         dprintk("%s: begin!\n", __func__);
2060         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2061                 return;
2062         /* hmm. we are done with the inode, and in the process of freeing
2063          * the state_owner. we keep this around to process errors
2064          */
2065         switch (task->tk_status) {
2066                 case 0:
2067                         if (calldata->roc)
2068                                 pnfs_roc_set_barrier(state->inode,
2069                                                      calldata->roc_barrier);
2070                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2071                         renew_lease(server, calldata->timestamp);
2072                         nfs4_close_clear_stateid_flags(state,
2073                                         calldata->arg.fmode);
2074                         break;
2075                 case -NFS4ERR_STALE_STATEID:
2076                 case -NFS4ERR_OLD_STATEID:
2077                 case -NFS4ERR_BAD_STATEID:
2078                 case -NFS4ERR_EXPIRED:
2079                         if (calldata->arg.fmode == 0)
2080                                 break;
2081                 default:
2082                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2083                                 rpc_restart_call_prepare(task);
2084         }
2085         nfs_release_seqid(calldata->arg.seqid);
2086         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2087         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2088 }
2089
2090 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2091 {
2092         struct nfs4_closedata *calldata = data;
2093         struct nfs4_state *state = calldata->state;
2094         int call_close = 0;
2095
2096         dprintk("%s: begin!\n", __func__);
2097         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2098                 return;
2099
2100         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2101         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2102         spin_lock(&state->owner->so_lock);
2103         /* Calculate the change in open mode */
2104         if (state->n_rdwr == 0) {
2105                 if (state->n_rdonly == 0) {
2106                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2107                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2108                         calldata->arg.fmode &= ~FMODE_READ;
2109                 }
2110                 if (state->n_wronly == 0) {
2111                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2112                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2113                         calldata->arg.fmode &= ~FMODE_WRITE;
2114                 }
2115         }
2116         spin_unlock(&state->owner->so_lock);
2117
2118         if (!call_close) {
2119                 /* Note: exit _without_ calling nfs4_close_done */
2120                 task->tk_action = NULL;
2121                 goto out;
2122         }
2123
2124         if (calldata->arg.fmode == 0) {
2125                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2126                 if (calldata->roc &&
2127                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2128                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2129                                      task, NULL);
2130                         goto out;
2131                 }
2132         }
2133
2134         nfs_fattr_init(calldata->res.fattr);
2135         calldata->timestamp = jiffies;
2136         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2137                                 &calldata->arg.seq_args,
2138                                 &calldata->res.seq_res,
2139                                 task))
2140                 goto out;
2141         rpc_call_start(task);
2142 out:
2143         dprintk("%s: done!\n", __func__);
2144 }
2145
2146 static const struct rpc_call_ops nfs4_close_ops = {
2147         .rpc_call_prepare = nfs4_close_prepare,
2148         .rpc_call_done = nfs4_close_done,
2149         .rpc_release = nfs4_free_closedata,
2150 };
2151
2152 /*
2153  * It is possible for data to be read/written from a mem-mapped file
2154  * after the sys_close call (which hits the vfs layer as a flush).
2155  * This means that we can't safely call nfsv4 close on a file until
2156  * the inode is cleared. This in turn means that we are not good
2157  * NFSv4 citizens - we do not indicate to the server to update the file's
2158  * share state even when we are done with one of the three share
2159  * stateid's in the inode.
2160  *
2161  * NOTE: Caller must be holding the sp->so_owner semaphore!
2162  */
2163 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2164 {
2165         struct nfs_server *server = NFS_SERVER(state->inode);
2166         struct nfs4_closedata *calldata;
2167         struct nfs4_state_owner *sp = state->owner;
2168         struct rpc_task *task;
2169         struct rpc_message msg = {
2170                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2171                 .rpc_cred = state->owner->so_cred,
2172         };
2173         struct rpc_task_setup task_setup_data = {
2174                 .rpc_client = server->client,
2175                 .rpc_message = &msg,
2176                 .callback_ops = &nfs4_close_ops,
2177                 .workqueue = nfsiod_workqueue,
2178                 .flags = RPC_TASK_ASYNC,
2179         };
2180         int status = -ENOMEM;
2181
2182         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2183         if (calldata == NULL)
2184                 goto out;
2185         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2186         calldata->inode = state->inode;
2187         calldata->state = state;
2188         calldata->arg.fh = NFS_FH(state->inode);
2189         calldata->arg.stateid = &state->open_stateid;
2190         /* Serialization for the sequence id */
2191         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2192         if (calldata->arg.seqid == NULL)
2193                 goto out_free_calldata;
2194         calldata->arg.fmode = 0;
2195         calldata->arg.bitmask = server->cache_consistency_bitmask;
2196         calldata->res.fattr = &calldata->fattr;
2197         calldata->res.seqid = calldata->arg.seqid;
2198         calldata->res.server = server;
2199         calldata->roc = roc;
2200         nfs_sb_active(calldata->inode->i_sb);
2201
2202         msg.rpc_argp = &calldata->arg;
2203         msg.rpc_resp = &calldata->res;
2204         task_setup_data.callback_data = calldata;
2205         task = rpc_run_task(&task_setup_data);
2206         if (IS_ERR(task))
2207                 return PTR_ERR(task);
2208         status = 0;
2209         if (wait)
2210                 status = rpc_wait_for_completion_task(task);
2211         rpc_put_task(task);
2212         return status;
2213 out_free_calldata:
2214         kfree(calldata);
2215 out:
2216         if (roc)
2217                 pnfs_roc_release(state->inode);
2218         nfs4_put_open_state(state);
2219         nfs4_put_state_owner(sp);
2220         return status;
2221 }
2222
2223 static struct inode *
2224 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2225 {
2226         struct nfs4_state *state;
2227
2228         /* Protect against concurrent sillydeletes */
2229         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2230                              ctx->cred, &ctx->mdsthreshold);
2231         if (IS_ERR(state))
2232                 return ERR_CAST(state);
2233         ctx->state = state;
2234         return igrab(state->inode);
2235 }
2236
2237 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2238 {
2239         if (ctx->state == NULL)
2240                 return;
2241         if (is_sync)
2242                 nfs4_close_sync(ctx->state, ctx->mode);
2243         else
2244                 nfs4_close_state(ctx->state, ctx->mode);
2245 }
2246
2247 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2248 {
2249         struct nfs4_server_caps_arg args = {
2250                 .fhandle = fhandle,
2251         };
2252         struct nfs4_server_caps_res res = {};
2253         struct rpc_message msg = {
2254                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2255                 .rpc_argp = &args,
2256                 .rpc_resp = &res,
2257         };
2258         int status;
2259
2260         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2261         if (status == 0) {
2262                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2263                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2264                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2265                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2266                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2267                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2268                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2269                         server->caps |= NFS_CAP_ACLS;
2270                 if (res.has_links != 0)
2271                         server->caps |= NFS_CAP_HARDLINKS;
2272                 if (res.has_symlinks != 0)
2273                         server->caps |= NFS_CAP_SYMLINKS;
2274                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2275                         server->caps |= NFS_CAP_FILEID;
2276                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2277                         server->caps |= NFS_CAP_MODE;
2278                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2279                         server->caps |= NFS_CAP_NLINK;
2280                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2281                         server->caps |= NFS_CAP_OWNER;
2282                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2283                         server->caps |= NFS_CAP_OWNER_GROUP;
2284                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2285                         server->caps |= NFS_CAP_ATIME;
2286                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2287                         server->caps |= NFS_CAP_CTIME;
2288                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2289                         server->caps |= NFS_CAP_MTIME;
2290
2291                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2292                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2293                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2294                 server->acl_bitmask = res.acl_bitmask;
2295                 server->fh_expire_type = res.fh_expire_type;
2296         }
2297
2298         return status;
2299 }
2300
2301 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2302 {
2303         struct nfs4_exception exception = { };
2304         int err;
2305         do {
2306                 err = nfs4_handle_exception(server,
2307                                 _nfs4_server_capabilities(server, fhandle),
2308                                 &exception);
2309         } while (exception.retry);
2310         return err;
2311 }
2312
2313 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2314                 struct nfs_fsinfo *info)
2315 {
2316         struct nfs4_lookup_root_arg args = {
2317                 .bitmask = nfs4_fattr_bitmap,
2318         };
2319         struct nfs4_lookup_res res = {
2320                 .server = server,
2321                 .fattr = info->fattr,
2322                 .fh = fhandle,
2323         };
2324         struct rpc_message msg = {
2325                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2326                 .rpc_argp = &args,
2327                 .rpc_resp = &res,
2328         };
2329
2330         nfs_fattr_init(info->fattr);
2331         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2332 }
2333
2334 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2335                 struct nfs_fsinfo *info)
2336 {
2337         struct nfs4_exception exception = { };
2338         int err;
2339         do {
2340                 err = _nfs4_lookup_root(server, fhandle, info);
2341                 switch (err) {
2342                 case 0:
2343                 case -NFS4ERR_WRONGSEC:
2344                         goto out;
2345                 default:
2346                         err = nfs4_handle_exception(server, err, &exception);
2347                 }
2348         } while (exception.retry);
2349 out:
2350         return err;
2351 }
2352
2353 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2354                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2355 {
2356         struct rpc_auth *auth;
2357         int ret;
2358
2359         auth = rpcauth_create(flavor, server->client);
2360         if (!auth) {
2361                 ret = -EIO;
2362                 goto out;
2363         }
2364         ret = nfs4_lookup_root(server, fhandle, info);
2365 out:
2366         return ret;
2367 }
2368
2369 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2370                               struct nfs_fsinfo *info)
2371 {
2372         int i, len, status = 0;
2373         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2374
2375         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2376         flav_array[len] = RPC_AUTH_NULL;
2377         len += 1;
2378
2379         for (i = 0; i < len; i++) {
2380                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2381                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2382                         continue;
2383                 break;
2384         }
2385         /*
2386          * -EACCESS could mean that the user doesn't have correct permissions
2387          * to access the mount.  It could also mean that we tried to mount
2388          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2389          * existing mount programs don't handle -EACCES very well so it should
2390          * be mapped to -EPERM instead.
2391          */
2392         if (status == -EACCES)
2393                 status = -EPERM;
2394         return status;
2395 }
2396
2397 /*
2398  * get the file handle for the "/" directory on the server
2399  */
2400 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2401                          struct nfs_fsinfo *info)
2402 {
2403         int minor_version = server->nfs_client->cl_minorversion;
2404         int status = nfs4_lookup_root(server, fhandle, info);
2405         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2406                 /*
2407                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2408                  * by nfs4_map_errors() as this function exits.
2409                  */
2410                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2411         if (status == 0)
2412                 status = nfs4_server_capabilities(server, fhandle);
2413         if (status == 0)
2414                 status = nfs4_do_fsinfo(server, fhandle, info);
2415         return nfs4_map_errors(status);
2416 }
2417
2418 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2419                               struct nfs_fsinfo *info)
2420 {
2421         int error;
2422         struct nfs_fattr *fattr = info->fattr;
2423
2424         error = nfs4_server_capabilities(server, mntfh);
2425         if (error < 0) {
2426                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2427                 return error;
2428         }
2429
2430         error = nfs4_proc_getattr(server, mntfh, fattr);
2431         if (error < 0) {
2432                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2433                 return error;
2434         }
2435
2436         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2437             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2438                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2439
2440         return error;
2441 }
2442
2443 /*
2444  * Get locations and (maybe) other attributes of a referral.
2445  * Note that we'll actually follow the referral later when
2446  * we detect fsid mismatch in inode revalidation
2447  */
2448 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2449                              const struct qstr *name, struct nfs_fattr *fattr,
2450                              struct nfs_fh *fhandle)
2451 {
2452         int status = -ENOMEM;
2453         struct page *page = NULL;
2454         struct nfs4_fs_locations *locations = NULL;
2455
2456         page = alloc_page(GFP_KERNEL);
2457         if (page == NULL)
2458                 goto out;
2459         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2460         if (locations == NULL)
2461                 goto out;
2462
2463         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2464         if (status != 0)
2465                 goto out;
2466         /* Make sure server returned a different fsid for the referral */
2467         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2468                 dprintk("%s: server did not return a different fsid for"
2469                         " a referral at %s\n", __func__, name->name);
2470                 status = -EIO;
2471                 goto out;
2472         }
2473         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2474         nfs_fixup_referral_attributes(&locations->fattr);
2475
2476         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2477         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2478         memset(fhandle, 0, sizeof(struct nfs_fh));
2479 out:
2480         if (page)
2481                 __free_page(page);
2482         kfree(locations);
2483         return status;
2484 }
2485
2486 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2487 {
2488         struct nfs4_getattr_arg args = {
2489                 .fh = fhandle,
2490                 .bitmask = server->attr_bitmask,
2491         };
2492         struct nfs4_getattr_res res = {
2493                 .fattr = fattr,
2494                 .server = server,
2495         };
2496         struct rpc_message msg = {
2497                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2498                 .rpc_argp = &args,
2499                 .rpc_resp = &res,
2500         };
2501
2502         nfs_fattr_init(fattr);
2503         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2504 }
2505
2506 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2507 {
2508         struct nfs4_exception exception = { };
2509         int err;
2510         do {
2511                 err = nfs4_handle_exception(server,
2512                                 _nfs4_proc_getattr(server, fhandle, fattr),
2513                                 &exception);
2514         } while (exception.retry);
2515         return err;
2516 }
2517
2518 /*
2519  * The file is not closed if it is opened due to the a request to change
2520  * the size of the file. The open call will not be needed once the
2521  * VFS layer lookup-intents are implemented.
2522  *
2523  * Close is called when the inode is destroyed.
2524  * If we haven't opened the file for O_WRONLY, we
2525  * need to in the size_change case to obtain a stateid.
2526  *
2527  * Got race?
2528  * Because OPEN is always done by name in nfsv4, it is
2529  * possible that we opened a different file by the same
2530  * name.  We can recognize this race condition, but we
2531  * can't do anything about it besides returning an error.
2532  *
2533  * This will be fixed with VFS changes (lookup-intent).
2534  */
2535 static int
2536 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2537                   struct iattr *sattr)
2538 {
2539         struct inode *inode = dentry->d_inode;
2540         struct rpc_cred *cred = NULL;
2541         struct nfs4_state *state = NULL;
2542         int status;
2543
2544         if (pnfs_ld_layoutret_on_setattr(inode))
2545                 pnfs_return_layout(inode);
2546
2547         nfs_fattr_init(fattr);
2548
2549         /* Search for an existing open(O_WRITE) file */
2550         if (sattr->ia_valid & ATTR_FILE) {
2551                 struct nfs_open_context *ctx;
2552
2553                 ctx = nfs_file_open_context(sattr->ia_file);
2554                 if (ctx) {
2555                         cred = ctx->cred;
2556                         state = ctx->state;
2557                 }
2558         }
2559
2560         /* Deal with open(O_TRUNC) */
2561         if (sattr->ia_valid & ATTR_OPEN)
2562                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2563
2564         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2565         if (status == 0)
2566                 nfs_setattr_update_inode(inode, sattr);
2567         return status;
2568 }
2569
2570 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2571                 const struct qstr *name, struct nfs_fh *fhandle,
2572                 struct nfs_fattr *fattr)
2573 {
2574         struct nfs_server *server = NFS_SERVER(dir);
2575         int                    status;
2576         struct nfs4_lookup_arg args = {
2577                 .bitmask = server->attr_bitmask,
2578                 .dir_fh = NFS_FH(dir),
2579                 .name = name,
2580         };
2581         struct nfs4_lookup_res res = {
2582                 .server = server,
2583                 .fattr = fattr,
2584                 .fh = fhandle,
2585         };
2586         struct rpc_message msg = {
2587                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2588                 .rpc_argp = &args,
2589                 .rpc_resp = &res,
2590         };
2591
2592         nfs_fattr_init(fattr);
2593
2594         dprintk("NFS call  lookup %s\n", name->name);
2595         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2596         dprintk("NFS reply lookup: %d\n", status);
2597         return status;
2598 }
2599
2600 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2601 {
2602         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2603                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2604         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2605         fattr->nlink = 2;
2606 }
2607
2608 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2609                                    struct qstr *name, struct nfs_fh *fhandle,
2610                                    struct nfs_fattr *fattr)
2611 {
2612         struct nfs4_exception exception = { };
2613         struct rpc_clnt *client = *clnt;
2614         int err;
2615         do {
2616                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2617                 switch (err) {
2618                 case -NFS4ERR_BADNAME:
2619                         err = -ENOENT;
2620                         goto out;
2621                 case -NFS4ERR_MOVED:
2622                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2623                         goto out;
2624                 case -NFS4ERR_WRONGSEC:
2625                         err = -EPERM;
2626                         if (client != *clnt)
2627                                 goto out;
2628
2629                         client = nfs4_create_sec_client(client, dir, name);
2630                         if (IS_ERR(client))
2631                                 return PTR_ERR(client);
2632
2633                         exception.retry = 1;
2634                         break;
2635                 default:
2636                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2637                 }
2638         } while (exception.retry);
2639
2640 out:
2641         if (err == 0)
2642                 *clnt = client;
2643         else if (client != *clnt)
2644                 rpc_shutdown_client(client);
2645
2646         return err;
2647 }
2648
2649 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2650                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2651 {
2652         int status;
2653         struct rpc_clnt *client = NFS_CLIENT(dir);
2654
2655         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2656         if (client != NFS_CLIENT(dir)) {
2657                 rpc_shutdown_client(client);
2658                 nfs_fixup_secinfo_attributes(fattr);
2659         }
2660         return status;
2661 }
2662
2663 struct rpc_clnt *
2664 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2665                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2666 {
2667         int status;
2668         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2669
2670         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2671         if (status < 0) {
2672                 rpc_shutdown_client(client);
2673                 return ERR_PTR(status);
2674         }
2675         return client;
2676 }
2677
2678 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2679 {
2680         struct nfs_server *server = NFS_SERVER(inode);
2681         struct nfs4_accessargs args = {
2682                 .fh = NFS_FH(inode),
2683                 .bitmask = server->cache_consistency_bitmask,
2684         };
2685         struct nfs4_accessres res = {
2686                 .server = server,
2687         };
2688         struct rpc_message msg = {
2689                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2690                 .rpc_argp = &args,
2691                 .rpc_resp = &res,
2692                 .rpc_cred = entry->cred,
2693         };
2694         int mode = entry->mask;
2695         int status;
2696
2697         /*
2698          * Determine which access bits we want to ask for...
2699          */
2700         if (mode & MAY_READ)
2701                 args.access |= NFS4_ACCESS_READ;
2702         if (S_ISDIR(inode->i_mode)) {
2703                 if (mode & MAY_WRITE)
2704                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2705                 if (mode & MAY_EXEC)
2706                         args.access |= NFS4_ACCESS_LOOKUP;
2707         } else {
2708                 if (mode & MAY_WRITE)
2709                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2710                 if (mode & MAY_EXEC)
2711                         args.access |= NFS4_ACCESS_EXECUTE;
2712         }
2713
2714         res.fattr = nfs_alloc_fattr();
2715         if (res.fattr == NULL)
2716                 return -ENOMEM;
2717
2718         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2719         if (!status) {
2720                 entry->mask = 0;
2721                 if (res.access & NFS4_ACCESS_READ)
2722                         entry->mask |= MAY_READ;
2723                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2724                         entry->mask |= MAY_WRITE;
2725                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2726                         entry->mask |= MAY_EXEC;
2727                 nfs_refresh_inode(inode, res.fattr);
2728         }
2729         nfs_free_fattr(res.fattr);
2730         return status;
2731 }
2732
2733 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2734 {
2735         struct nfs4_exception exception = { };
2736         int err;
2737         do {
2738                 err = nfs4_handle_exception(NFS_SERVER(inode),
2739                                 _nfs4_proc_access(inode, entry),
2740                                 &exception);
2741         } while (exception.retry);
2742         return err;
2743 }
2744
2745 /*
2746  * TODO: For the time being, we don't try to get any attributes
2747  * along with any of the zero-copy operations READ, READDIR,
2748  * READLINK, WRITE.
2749  *
2750  * In the case of the first three, we want to put the GETATTR
2751  * after the read-type operation -- this is because it is hard
2752  * to predict the length of a GETATTR response in v4, and thus
2753  * align the READ data correctly.  This means that the GETATTR
2754  * may end up partially falling into the page cache, and we should
2755  * shift it into the 'tail' of the xdr_buf before processing.
2756  * To do this efficiently, we need to know the total length
2757  * of data received, which doesn't seem to be available outside
2758  * of the RPC layer.
2759  *
2760  * In the case of WRITE, we also want to put the GETATTR after
2761  * the operation -- in this case because we want to make sure
2762  * we get the post-operation mtime and size.  This means that
2763  * we can't use xdr_encode_pages() as written: we need a variant
2764  * of it which would leave room in the 'tail' iovec.
2765  *
2766  * Both of these changes to the XDR layer would in fact be quite
2767  * minor, but I decided to leave them for a subsequent patch.
2768  */
2769 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2770                 unsigned int pgbase, unsigned int pglen)
2771 {
2772         struct nfs4_readlink args = {
2773                 .fh       = NFS_FH(inode),
2774                 .pgbase   = pgbase,
2775                 .pglen    = pglen,
2776                 .pages    = &page,
2777         };
2778         struct nfs4_readlink_res res;
2779         struct rpc_message msg = {
2780                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2781                 .rpc_argp = &args,
2782                 .rpc_resp = &res,
2783         };
2784
2785         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2786 }
2787
2788 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2789                 unsigned int pgbase, unsigned int pglen)
2790 {
2791         struct nfs4_exception exception = { };
2792         int err;
2793         do {
2794                 err = nfs4_handle_exception(NFS_SERVER(inode),
2795                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2796                                 &exception);
2797         } while (exception.retry);
2798         return err;
2799 }
2800
2801 /*
2802  * Got race?
2803  * We will need to arrange for the VFS layer to provide an atomic open.
2804  * Until then, this create/open method is prone to inefficiency and race
2805  * conditions due to the lookup, create, and open VFS calls from sys_open()
2806  * placed on the wire.
2807  *
2808  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2809  * The file will be opened again in the subsequent VFS open call
2810  * (nfs4_proc_file_open).
2811  *
2812  * The open for read will just hang around to be used by any process that
2813  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2814  */
2815
2816 static int
2817 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2818                  int flags, struct nfs_open_context *ctx)
2819 {
2820         struct dentry *de = dentry;
2821         struct nfs4_state *state;
2822         struct rpc_cred *cred = NULL;
2823         fmode_t fmode = 0;
2824         int status = 0;
2825
2826         if (ctx != NULL) {
2827                 cred = ctx->cred;
2828                 de = ctx->dentry;
2829                 fmode = ctx->mode;
2830         }
2831         sattr->ia_mode &= ~current_umask();
2832         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred, NULL);
2833         d_drop(dentry);
2834         if (IS_ERR(state)) {
2835                 status = PTR_ERR(state);
2836                 goto out;
2837         }
2838         d_add(dentry, igrab(state->inode));
2839         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2840         if (ctx != NULL)
2841                 ctx->state = state;
2842         else
2843                 nfs4_close_sync(state, fmode);
2844 out:
2845         return status;
2846 }
2847
2848 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2849 {
2850         struct nfs_server *server = NFS_SERVER(dir);
2851         struct nfs_removeargs args = {
2852                 .fh = NFS_FH(dir),
2853                 .name = *name,
2854         };
2855         struct nfs_removeres res = {
2856                 .server = server,
2857         };
2858         struct rpc_message msg = {
2859                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2860                 .rpc_argp = &args,
2861                 .rpc_resp = &res,
2862         };
2863         int status;
2864
2865         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2866         if (status == 0)
2867                 update_changeattr(dir, &res.cinfo);
2868         return status;
2869 }
2870
2871 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2872 {
2873         struct nfs4_exception exception = { };
2874         int err;
2875         do {
2876                 err = nfs4_handle_exception(NFS_SERVER(dir),
2877                                 _nfs4_proc_remove(dir, name),
2878                                 &exception);
2879         } while (exception.retry);
2880         return err;
2881 }
2882
2883 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2884 {
2885         struct nfs_server *server = NFS_SERVER(dir);
2886         struct nfs_removeargs *args = msg->rpc_argp;
2887         struct nfs_removeres *res = msg->rpc_resp;
2888
2889         res->server = server;
2890         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2891         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2892 }
2893
2894 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
2895 {
2896         if (nfs4_setup_sequence(NFS_SERVER(data->dir),
2897                                 &data->args.seq_args,
2898                                 &data->res.seq_res,
2899                                 task))
2900                 return;
2901         rpc_call_start(task);
2902 }
2903
2904 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2905 {
2906         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2907
2908         if (!nfs4_sequence_done(task, &res->seq_res))
2909                 return 0;
2910         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2911                 return 0;
2912         update_changeattr(dir, &res->cinfo);
2913         return 1;
2914 }
2915
2916 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2917 {
2918         struct nfs_server *server = NFS_SERVER(dir);
2919         struct nfs_renameargs *arg = msg->rpc_argp;
2920         struct nfs_renameres *res = msg->rpc_resp;
2921
2922         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2923         res->server = server;
2924         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2925 }
2926
2927 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
2928 {
2929         if (nfs4_setup_sequence(NFS_SERVER(data->old_dir),
2930                                 &data->args.seq_args,
2931                                 &data->res.seq_res,
2932                                 task))
2933                 return;
2934         rpc_call_start(task);
2935 }
2936
2937 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2938                                  struct inode *new_dir)
2939 {
2940         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2941
2942         if (!nfs4_sequence_done(task, &res->seq_res))
2943                 return 0;
2944         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2945                 return 0;
2946
2947         update_changeattr(old_dir, &res->old_cinfo);
2948         update_changeattr(new_dir, &res->new_cinfo);
2949         return 1;
2950 }
2951
2952 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2953                 struct inode *new_dir, struct qstr *new_name)
2954 {
2955         struct nfs_server *server = NFS_SERVER(old_dir);
2956         struct nfs_renameargs arg = {
2957                 .old_dir = NFS_FH(old_dir),
2958                 .new_dir = NFS_FH(new_dir),
2959                 .old_name = old_name,
2960                 .new_name = new_name,
2961         };
2962         struct nfs_renameres res = {
2963                 .server = server,
2964         };
2965         struct rpc_message msg = {
2966                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2967                 .rpc_argp = &arg,
2968                 .rpc_resp = &res,
2969         };
2970         int status = -ENOMEM;
2971
2972         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2973         if (!status) {
2974                 update_changeattr(old_dir, &res.old_cinfo);
2975                 update_changeattr(new_dir, &res.new_cinfo);
2976         }
2977         return status;
2978 }
2979
2980 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2981                 struct inode *new_dir, struct qstr *new_name)
2982 {
2983         struct nfs4_exception exception = { };
2984         int err;
2985         do {
2986                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2987                                 _nfs4_proc_rename(old_dir, old_name,
2988                                         new_dir, new_name),
2989                                 &exception);
2990         } while (exception.retry);
2991         return err;
2992 }
2993
2994 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2995 {
2996         struct nfs_server *server = NFS_SERVER(inode);
2997         struct nfs4_link_arg arg = {
2998                 .fh     = NFS_FH(inode),
2999                 .dir_fh = NFS_FH(dir),
3000                 .name   = name,
3001                 .bitmask = server->attr_bitmask,
3002         };
3003         struct nfs4_link_res res = {
3004                 .server = server,
3005         };
3006         struct rpc_message msg = {
3007                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3008                 .rpc_argp = &arg,
3009                 .rpc_resp = &res,
3010         };
3011         int status = -ENOMEM;
3012
3013         res.fattr = nfs_alloc_fattr();
3014         if (res.fattr == NULL)
3015                 goto out;
3016
3017         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3018         if (!status) {
3019                 update_changeattr(dir, &res.cinfo);
3020                 nfs_post_op_update_inode(inode, res.fattr);
3021         }
3022 out:
3023         nfs_free_fattr(res.fattr);
3024         return status;
3025 }
3026
3027 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3028 {
3029         struct nfs4_exception exception = { };
3030         int err;
3031         do {
3032                 err = nfs4_handle_exception(NFS_SERVER(inode),
3033                                 _nfs4_proc_link(inode, dir, name),
3034                                 &exception);
3035         } while (exception.retry);
3036         return err;
3037 }
3038
3039 struct nfs4_createdata {
3040         struct rpc_message msg;
3041         struct nfs4_create_arg arg;
3042         struct nfs4_create_res res;
3043         struct nfs_fh fh;
3044         struct nfs_fattr fattr;
3045 };
3046
3047 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3048                 struct qstr *name, struct iattr *sattr, u32 ftype)
3049 {
3050         struct nfs4_createdata *data;
3051
3052         data = kzalloc(sizeof(*data), GFP_KERNEL);
3053         if (data != NULL) {
3054                 struct nfs_server *server = NFS_SERVER(dir);
3055
3056                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3057                 data->msg.rpc_argp = &data->arg;
3058                 data->msg.rpc_resp = &data->res;
3059                 data->arg.dir_fh = NFS_FH(dir);
3060                 data->arg.server = server;
3061                 data->arg.name = name;
3062                 data->arg.attrs = sattr;
3063                 data->arg.ftype = ftype;
3064                 data->arg.bitmask = server->attr_bitmask;
3065                 data->res.server = server;
3066                 data->res.fh = &data->fh;
3067                 data->res.fattr = &data->fattr;
3068                 nfs_fattr_init(data->res.fattr);
3069         }
3070         return data;
3071 }
3072
3073 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3074 {
3075         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3076                                     &data->arg.seq_args, &data->res.seq_res, 1);
3077         if (status == 0) {
3078                 update_changeattr(dir, &data->res.dir_cinfo);
3079                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3080         }
3081         return status;
3082 }
3083
3084 static void nfs4_free_createdata(struct nfs4_createdata *data)
3085 {
3086         kfree(data);
3087 }
3088
3089 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3090                 struct page *page, unsigned int len, struct iattr *sattr)
3091 {
3092         struct nfs4_createdata *data;
3093         int status = -ENAMETOOLONG;
3094
3095         if (len > NFS4_MAXPATHLEN)
3096                 goto out;
3097
3098         status = -ENOMEM;
3099         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3100         if (data == NULL)
3101                 goto out;
3102
3103         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3104         data->arg.u.symlink.pages = &page;
3105         data->arg.u.symlink.len = len;
3106
3107         status = nfs4_do_create(dir, dentry, data);
3108
3109         nfs4_free_createdata(data);
3110 out:
3111         return status;
3112 }
3113
3114 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3115                 struct page *page, unsigned int len, struct iattr *sattr)
3116 {
3117         struct nfs4_exception exception = { };
3118         int err;
3119         do {
3120                 err = nfs4_handle_exception(NFS_SERVER(dir),
3121                                 _nfs4_proc_symlink(dir, dentry, page,
3122                                                         len, sattr),
3123                                 &exception);
3124         } while (exception.retry);
3125         return err;
3126 }
3127
3128 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3129                 struct iattr *sattr)
3130 {
3131         struct nfs4_createdata *data;
3132         int status = -ENOMEM;
3133
3134         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3135         if (data == NULL)
3136                 goto out;
3137
3138         status = nfs4_do_create(dir, dentry, data);
3139
3140         nfs4_free_createdata(data);
3141 out:
3142         return status;
3143 }
3144
3145 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3146                 struct iattr *sattr)
3147 {
3148         struct nfs4_exception exception = { };
3149         int err;
3150
3151         sattr->ia_mode &= ~current_umask();
3152         do {
3153                 err = nfs4_handle_exception(NFS_SERVER(dir),
3154                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3155                                 &exception);
3156         } while (exception.retry);
3157         return err;
3158 }
3159
3160 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3161                 u64 cookie, struct page **pages, unsigned int count, int plus)
3162 {
3163         struct inode            *dir = dentry->d_inode;
3164         struct nfs4_readdir_arg args = {
3165                 .fh = NFS_FH(dir),
3166                 .pages = pages,
3167                 .pgbase = 0,
3168                 .count = count,
3169                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3170                 .plus = plus,
3171         };
3172         struct nfs4_readdir_res res;
3173         struct rpc_message msg = {
3174                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3175                 .rpc_argp = &args,
3176                 .rpc_resp = &res,
3177                 .rpc_cred = cred,
3178         };
3179         int                     status;
3180
3181         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3182                         dentry->d_parent->d_name.name,
3183                         dentry->d_name.name,
3184                         (unsigned long long)cookie);
3185         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3186         res.pgbase = args.pgbase;
3187         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3188         if (status >= 0) {
3189                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3190                 status += args.pgbase;
3191         }
3192
3193         nfs_invalidate_atime(dir);
3194
3195         dprintk("%s: returns %d\n", __func__, status);
3196         return status;
3197 }
3198
3199 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3200                 u64 cookie, struct page **pages, unsigned int count, int plus)
3201 {
3202         struct nfs4_exception exception = { };
3203         int err;
3204         do {
3205                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3206                                 _nfs4_proc_readdir(dentry, cred, cookie,
3207                                         pages, count, plus),
3208                                 &exception);
3209         } while (exception.retry);
3210         return err;
3211 }
3212
3213 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3214                 struct iattr *sattr, dev_t rdev)
3215 {
3216         struct nfs4_createdata *data;
3217         int mode = sattr->ia_mode;
3218         int status = -ENOMEM;
3219
3220         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3221         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3222
3223         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3224         if (data == NULL)
3225                 goto out;
3226
3227         if (S_ISFIFO(mode))
3228                 data->arg.ftype = NF4FIFO;
3229         else if (S_ISBLK(mode)) {
3230                 data->arg.ftype = NF4BLK;
3231                 data->arg.u.device.specdata1 = MAJOR(rdev);
3232                 data->arg.u.device.specdata2 = MINOR(rdev);
3233         }
3234         else if (S_ISCHR(mode)) {
3235                 data->arg.ftype = NF4CHR;
3236                 data->arg.u.device.specdata1 = MAJOR(rdev);
3237                 data->arg.u.device.specdata2 = MINOR(rdev);
3238         }
3239
3240         status = nfs4_do_create(dir, dentry, data);
3241
3242         nfs4_free_createdata(data);
3243 out:
3244         return status;
3245 }
3246
3247 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3248                 struct iattr *sattr, dev_t rdev)
3249 {
3250         struct nfs4_exception exception = { };
3251         int err;
3252
3253         sattr->ia_mode &= ~current_umask();
3254         do {
3255                 err = nfs4_handle_exception(NFS_SERVER(dir),
3256                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3257                                 &exception);
3258         } while (exception.retry);
3259         return err;
3260 }
3261
3262 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3263                  struct nfs_fsstat *fsstat)
3264 {
3265         struct nfs4_statfs_arg args = {
3266                 .fh = fhandle,
3267                 .bitmask = server->attr_bitmask,
3268         };
3269         struct nfs4_statfs_res res = {
3270                 .fsstat = fsstat,
3271         };
3272         struct rpc_message msg = {
3273                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3274                 .rpc_argp = &args,
3275                 .rpc_resp = &res,
3276         };
3277
3278         nfs_fattr_init(fsstat->fattr);
3279         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3280 }
3281
3282 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3283 {
3284         struct nfs4_exception exception = { };
3285         int err;
3286         do {
3287                 err = nfs4_handle_exception(server,
3288                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3289                                 &exception);
3290         } while (exception.retry);
3291         return err;
3292 }
3293
3294 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3295                 struct nfs_fsinfo *fsinfo)
3296 {
3297         struct nfs4_fsinfo_arg args = {
3298                 .fh = fhandle,
3299                 .bitmask = server->attr_bitmask,
3300         };
3301         struct nfs4_fsinfo_res res = {
3302                 .fsinfo = fsinfo,
3303         };
3304         struct rpc_message msg = {
3305                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3306                 .rpc_argp = &args,
3307                 .rpc_resp = &res,
3308         };
3309
3310         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3311 }
3312
3313 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3314 {
3315         struct nfs4_exception exception = { };
3316         int err;
3317
3318         do {
3319                 err = nfs4_handle_exception(server,
3320                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3321                                 &exception);
3322         } while (exception.retry);
3323         return err;
3324 }
3325
3326 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3327 {
3328         nfs_fattr_init(fsinfo->fattr);
3329         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3330 }
3331
3332 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3333                 struct nfs_pathconf *pathconf)
3334 {
3335         struct nfs4_pathconf_arg args = {
3336                 .fh = fhandle,
3337                 .bitmask = server->attr_bitmask,
3338         };
3339         struct nfs4_pathconf_res res = {
3340                 .pathconf = pathconf,
3341         };
3342         struct rpc_message msg = {
3343                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3344                 .rpc_argp = &args,
3345                 .rpc_resp = &res,
3346         };
3347
3348         /* None of the pathconf attributes are mandatory to implement */
3349         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3350                 memset(pathconf, 0, sizeof(*pathconf));
3351                 return 0;
3352         }
3353
3354         nfs_fattr_init(pathconf->fattr);
3355         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3356 }
3357
3358 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3359                 struct nfs_pathconf *pathconf)
3360 {
3361         struct nfs4_exception exception = { };
3362         int err;
3363
3364         do {
3365                 err = nfs4_handle_exception(server,
3366                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3367                                 &exception);
3368         } while (exception.retry);
3369         return err;
3370 }
3371
3372 void __nfs4_read_done_cb(struct nfs_read_data *data)
3373 {
3374         nfs_invalidate_atime(data->header->inode);
3375 }
3376
3377 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3378 {
3379         struct nfs_server *server = NFS_SERVER(data->header->inode);
3380
3381         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3382                 rpc_restart_call_prepare(task);
3383                 return -EAGAIN;
3384         }
3385
3386         __nfs4_read_done_cb(data);
3387         if (task->tk_status > 0)
3388                 renew_lease(server, data->timestamp);
3389         return 0;
3390 }
3391
3392 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3393 {
3394
3395         dprintk("--> %s\n", __func__);
3396
3397         if (!nfs4_sequence_done(task, &data->res.seq_res))
3398                 return -EAGAIN;
3399
3400         return data->read_done_cb ? data->read_done_cb(task, data) :
3401                                     nfs4_read_done_cb(task, data);
3402 }
3403
3404 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3405 {
3406         data->timestamp   = jiffies;
3407         data->read_done_cb = nfs4_read_done_cb;
3408         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3409         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3410 }
3411
3412 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3413 {
3414         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3415                                 &data->args.seq_args,
3416                                 &data->res.seq_res,
3417                                 task))
3418                 return;
3419         rpc_call_start(task);
3420 }
3421
3422 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3423 {
3424         struct inode *inode = data->header->inode;
3425
3426         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3427                 rpc_restart_call_prepare(task);
3428                 return -EAGAIN;
3429         }
3430         if (task->tk_status >= 0) {
3431                 renew_lease(NFS_SERVER(inode), data->timestamp);
3432                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3433         }
3434         return 0;
3435 }
3436
3437 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3438 {
3439         if (!nfs4_sequence_done(task, &data->res.seq_res))
3440                 return -EAGAIN;
3441         return data->write_done_cb ? data->write_done_cb(task, data) :
3442                 nfs4_write_done_cb(task, data);
3443 }
3444
3445 static
3446 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3447 {
3448         const struct nfs_pgio_header *hdr = data->header;
3449
3450         /* Don't request attributes for pNFS or O_DIRECT writes */
3451         if (data->ds_clp != NULL || hdr->dreq != NULL)
3452                 return false;
3453         /* Otherwise, request attributes if and only if we don't hold
3454          * a delegation
3455          */
3456         return nfs_have_delegation(hdr->inode, FMODE_READ) == 0;
3457 }
3458
3459 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3460 {
3461         struct nfs_server *server = NFS_SERVER(data->header->inode);
3462
3463         if (!nfs4_write_need_cache_consistency_data(data)) {
3464                 data->args.bitmask = NULL;
3465                 data->res.fattr = NULL;
3466         } else
3467                 data->args.bitmask = server->cache_consistency_bitmask;
3468
3469         if (!data->write_done_cb)
3470                 data->write_done_cb = nfs4_write_done_cb;
3471         data->res.server = server;
3472         data->timestamp   = jiffies;
3473
3474         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3475         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3476 }
3477
3478 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3479 {
3480         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3481                                 &data->args.seq_args,
3482                                 &data->res.seq_res,
3483                                 task))
3484                 return;
3485         rpc_call_start(task);
3486 }
3487
3488 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3489 {
3490         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3491                                 &data->args.seq_args,
3492                                 &data->res.seq_res,
3493                                 task))
3494                 return;
3495         rpc_call_start(task);
3496 }
3497
3498 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3499 {
3500         struct inode *inode = data->inode;
3501
3502         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3503                 rpc_restart_call_prepare(task);
3504                 return -EAGAIN;
3505         }
3506         return 0;
3507 }
3508
3509 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3510 {
3511         if (!nfs4_sequence_done(task, &data->res.seq_res))
3512                 return -EAGAIN;
3513         return data->commit_done_cb(task, data);
3514 }
3515
3516 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3517 {
3518         struct nfs_server *server = NFS_SERVER(data->inode);
3519
3520         if (data->commit_done_cb == NULL)
3521                 data->commit_done_cb = nfs4_commit_done_cb;
3522         data->res.server = server;
3523         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3524         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3525 }
3526
3527 struct nfs4_renewdata {
3528         struct nfs_client       *client;
3529         unsigned long           timestamp;
3530 };
3531
3532 /*
3533  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3534  * standalone procedure for queueing an asynchronous RENEW.
3535  */
3536 static void nfs4_renew_release(void *calldata)
3537 {
3538         struct nfs4_renewdata *data = calldata;
3539         struct nfs_client *clp = data->client;
3540
3541         if (atomic_read(&clp->cl_count) > 1)
3542                 nfs4_schedule_state_renewal(clp);
3543         nfs_put_client(clp);
3544         kfree(data);
3545 }
3546
3547 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3548 {
3549         struct nfs4_renewdata *data = calldata;
3550         struct nfs_client *clp = data->client;
3551         unsigned long timestamp = data->timestamp;
3552
3553         if (task->tk_status < 0) {
3554                 /* Unless we're shutting down, schedule state recovery! */
3555                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3556                         return;
3557                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3558                         nfs4_schedule_lease_recovery(clp);
3559                         return;
3560                 }
3561                 nfs4_schedule_path_down_recovery(clp);
3562         }
3563         do_renew_lease(clp, timestamp);
3564 }
3565
3566 static const struct rpc_call_ops nfs4_renew_ops = {
3567         .rpc_call_done = nfs4_renew_done,
3568         .rpc_release = nfs4_renew_release,
3569 };
3570
3571 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3572 {
3573         struct rpc_message msg = {
3574                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3575                 .rpc_argp       = clp,
3576                 .rpc_cred       = cred,
3577         };
3578         struct nfs4_renewdata *data;
3579
3580         if (renew_flags == 0)
3581                 return 0;
3582         if (!atomic_inc_not_zero(&clp->cl_count))
3583                 return -EIO;
3584         data = kmalloc(sizeof(*data), GFP_NOFS);
3585         if (data == NULL)
3586                 return -ENOMEM;
3587         data->client = clp;
3588         data->timestamp = jiffies;
3589         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3590                         &nfs4_renew_ops, data);
3591 }
3592
3593 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3594 {
3595         struct rpc_message msg = {
3596                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3597                 .rpc_argp       = clp,
3598                 .rpc_cred       = cred,
3599         };
3600         unsigned long now = jiffies;
3601         int status;
3602
3603         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3604         if (status < 0)
3605                 return status;
3606         do_renew_lease(clp, now);
3607         return 0;
3608 }
3609
3610 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3611 {
3612         return (server->caps & NFS_CAP_ACLS)
3613                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3614                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3615 }
3616
3617 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3618  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3619  * the stack.
3620  */
3621 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3622
3623 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3624                 struct page **pages, unsigned int *pgbase)
3625 {
3626         struct page *newpage, **spages;
3627         int rc = 0;
3628         size_t len;
3629         spages = pages;
3630
3631         do {
3632                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3633                 newpage = alloc_page(GFP_KERNEL);
3634
3635                 if (newpage == NULL)
3636                         goto unwind;
3637                 memcpy(page_address(newpage), buf, len);
3638                 buf += len;
3639                 buflen -= len;
3640                 *pages++ = newpage;
3641                 rc++;
3642         } while (buflen != 0);
3643
3644         return rc;
3645
3646 unwind:
3647         for(; rc > 0; rc--)
3648                 __free_page(spages[rc-1]);
3649         return -ENOMEM;
3650 }
3651
3652 struct nfs4_cached_acl {
3653         int cached;
3654         size_t len;
3655         char data[0];
3656 };
3657
3658 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3659 {
3660         struct nfs_inode *nfsi = NFS_I(inode);
3661
3662         spin_lock(&inode->i_lock);
3663         kfree(nfsi->nfs4_acl);
3664         nfsi->nfs4_acl = acl;
3665         spin_unlock(&inode->i_lock);
3666 }
3667
3668 static void nfs4_zap_acl_attr(struct inode *inode)
3669 {
3670         nfs4_set_cached_acl(inode, NULL);
3671 }
3672
3673 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3674 {
3675         struct nfs_inode *nfsi = NFS_I(inode);
3676         struct nfs4_cached_acl *acl;
3677         int ret = -ENOENT;
3678
3679         spin_lock(&inode->i_lock);
3680         acl = nfsi->nfs4_acl;
3681         if (acl == NULL)
3682                 goto out;
3683         if (buf == NULL) /* user is just asking for length */
3684                 goto out_len;
3685         if (acl->cached == 0)
3686                 goto out;
3687         ret = -ERANGE; /* see getxattr(2) man page */
3688         if (acl->len > buflen)
3689                 goto out;
3690         memcpy(buf, acl->data, acl->len);
3691 out_len:
3692         ret = acl->len;
3693 out:
3694         spin_unlock(&inode->i_lock);
3695         return ret;
3696 }
3697
3698 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3699 {
3700         struct nfs4_cached_acl *acl;
3701
3702         if (pages && acl_len <= PAGE_SIZE) {
3703                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3704                 if (acl == NULL)
3705                         goto out;
3706                 acl->cached = 1;
3707                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3708         } else {
3709                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3710                 if (acl == NULL)
3711                         goto out;
3712                 acl->cached = 0;
3713         }
3714         acl->len = acl_len;
3715 out:
3716         nfs4_set_cached_acl(inode, acl);
3717 }
3718
3719 /*
3720  * The getxattr API returns the required buffer length when called with a
3721  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3722  * the required buf.  On a NULL buf, we send a page of data to the server
3723  * guessing that the ACL request can be serviced by a page. If so, we cache
3724  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3725  * the cache. If not so, we throw away the page, and cache the required
3726  * length. The next getxattr call will then produce another round trip to
3727  * the server, this time with the input buf of the required size.
3728  */
3729 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3730 {
3731         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3732         struct nfs_getaclargs args = {
3733                 .fh = NFS_FH(inode),
3734                 .acl_pages = pages,
3735                 .acl_len = buflen,
3736         };
3737         struct nfs_getaclres res = {
3738                 .acl_len = buflen,
3739         };
3740         struct rpc_message msg = {
3741                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3742                 .rpc_argp = &args,
3743                 .rpc_resp = &res,
3744         };
3745         int ret = -ENOMEM, npages, i, acl_len = 0;
3746
3747         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3748         /* As long as we're doing a round trip to the server anyway,
3749          * let's be prepared for a page of acl data. */
3750         if (npages == 0)
3751                 npages = 1;
3752
3753         /* Add an extra page to handle the bitmap returned */
3754         npages++;
3755
3756         for (i = 0; i < npages; i++) {
3757                 pages[i] = alloc_page(GFP_KERNEL);
3758                 if (!pages[i])
3759                         goto out_free;
3760         }
3761
3762         /* for decoding across pages */
3763         res.acl_scratch = alloc_page(GFP_KERNEL);
3764         if (!res.acl_scratch)
3765                 goto out_free;
3766
3767         args.acl_len = npages * PAGE_SIZE;
3768         args.acl_pgbase = 0;
3769
3770         /* Let decode_getfacl know not to fail if the ACL data is larger than
3771          * the page we send as a guess */
3772         if (buf == NULL)
3773                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3774
3775         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3776                 __func__, buf, buflen, npages, args.acl_len);
3777         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3778                              &msg, &args.seq_args, &res.seq_res, 0);
3779         if (ret)
3780                 goto out_free;
3781
3782         acl_len = res.acl_len - res.acl_data_offset;
3783         if (acl_len > args.acl_len)
3784                 nfs4_write_cached_acl(inode, NULL, 0, acl_len);
3785         else
3786                 nfs4_write_cached_acl(inode, pages, res.acl_data_offset,
3787                                       acl_len);
3788         if (buf) {
3789                 ret = -ERANGE;
3790                 if (acl_len > buflen)
3791                         goto out_free;
3792                 _copy_from_pages(buf, pages, res.acl_data_offset,
3793                                 acl_len);
3794         }
3795         ret = acl_len;
3796 out_free:
3797         for (i = 0; i < npages; i++)
3798                 if (pages[i])
3799                         __free_page(pages[i]);
3800         if (res.acl_scratch)
3801                 __free_page(res.acl_scratch);
3802         return ret;
3803 }
3804
3805 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3806 {
3807         struct nfs4_exception exception = { };
3808         ssize_t ret;
3809         do {
3810                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3811                 if (ret >= 0)
3812                         break;
3813                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3814         } while (exception.retry);
3815         return ret;
3816 }
3817
3818 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3819 {
3820         struct nfs_server *server = NFS_SERVER(inode);
3821         int ret;
3822
3823         if (!nfs4_server_supports_acls(server))
3824                 return -EOPNOTSUPP;
3825         ret = nfs_revalidate_inode(server, inode);
3826         if (ret < 0)
3827                 return ret;
3828         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3829                 nfs_zap_acl_cache(inode);
3830         ret = nfs4_read_cached_acl(inode, buf, buflen);
3831         if (ret != -ENOENT)
3832                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3833                  * but no cached acl data, just the acl length */
3834                 return ret;
3835         return nfs4_get_acl_uncached(inode, buf, buflen);
3836 }
3837
3838 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3839 {
3840         struct nfs_server *server = NFS_SERVER(inode);
3841         struct page *pages[NFS4ACL_MAXPAGES];
3842         struct nfs_setaclargs arg = {
3843                 .fh             = NFS_FH(inode),
3844                 .acl_pages      = pages,
3845                 .acl_len        = buflen,
3846         };
3847         struct nfs_setaclres res;
3848         struct rpc_message msg = {
3849                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3850                 .rpc_argp       = &arg,
3851                 .rpc_resp       = &res,
3852         };
3853         int ret, i;
3854
3855         if (!nfs4_server_supports_acls(server))
3856                 return -EOPNOTSUPP;
3857         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3858         if (i < 0)
3859                 return i;
3860         nfs_inode_return_delegation(inode);
3861         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3862
3863         /*
3864          * Free each page after tx, so the only ref left is
3865          * held by the network stack
3866          */
3867         for (; i > 0; i--)
3868                 put_page(pages[i-1]);
3869
3870         /*
3871          * Acl update can result in inode attribute update.
3872          * so mark the attribute cache invalid.
3873          */
3874         spin_lock(&inode->i_lock);
3875         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3876         spin_unlock(&inode->i_lock);
3877         nfs_access_zap_cache(inode);
3878         nfs_zap_acl_cache(inode);
3879         return ret;
3880 }
3881
3882 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3883 {
3884         struct nfs4_exception exception = { };
3885         int err;
3886         do {
3887                 err = nfs4_handle_exception(NFS_SERVER(inode),
3888                                 __nfs4_proc_set_acl(inode, buf, buflen),
3889                                 &exception);
3890         } while (exception.retry);
3891         return err;
3892 }
3893
3894 static int
3895 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3896 {
3897         struct nfs_client *clp = server->nfs_client;
3898
3899         if (task->tk_status >= 0)
3900                 return 0;
3901         switch(task->tk_status) {
3902                 case -NFS4ERR_DELEG_REVOKED:
3903                 case -NFS4ERR_ADMIN_REVOKED:
3904                 case -NFS4ERR_BAD_STATEID:
3905                         if (state == NULL)
3906                                 break;
3907                         nfs_remove_bad_delegation(state->inode);
3908                 case -NFS4ERR_OPENMODE:
3909                         if (state == NULL)
3910                                 break;
3911                         nfs4_schedule_stateid_recovery(server, state);
3912                         goto wait_on_recovery;
3913                 case -NFS4ERR_EXPIRED:
3914                         if (state != NULL)
3915                                 nfs4_schedule_stateid_recovery(server, state);
3916                 case -NFS4ERR_STALE_STATEID:
3917                 case -NFS4ERR_STALE_CLIENTID:
3918                         nfs4_schedule_lease_recovery(clp);
3919                         goto wait_on_recovery;
3920 #if defined(CONFIG_NFS_V4_1)
3921                 case -NFS4ERR_BADSESSION:
3922                 case -NFS4ERR_BADSLOT:
3923                 case -NFS4ERR_BAD_HIGH_SLOT:
3924                 case -NFS4ERR_DEADSESSION:
3925                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3926                 case -NFS4ERR_SEQ_FALSE_RETRY:
3927                 case -NFS4ERR_SEQ_MISORDERED:
3928                         dprintk("%s ERROR %d, Reset session\n", __func__,
3929                                 task->tk_status);
3930                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
3931                         task->tk_status = 0;
3932                         return -EAGAIN;
3933 #endif /* CONFIG_NFS_V4_1 */
3934                 case -NFS4ERR_DELAY:
3935                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3936                 case -NFS4ERR_GRACE:
3937                 case -EKEYEXPIRED:
3938                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3939                         task->tk_status = 0;
3940                         return -EAGAIN;
3941                 case -NFS4ERR_RETRY_UNCACHED_REP:
3942                 case -NFS4ERR_OLD_STATEID:
3943                         task->tk_status = 0;
3944                         return -EAGAIN;
3945         }
3946         task->tk_status = nfs4_map_errors(task->tk_status);
3947         return 0;
3948 wait_on_recovery:
3949         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3950         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3951                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3952         task->tk_status = 0;
3953         return -EAGAIN;
3954 }
3955
3956 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
3957                                     nfs4_verifier *bootverf)
3958 {
3959         __be32 verf[2];
3960
3961         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
3962                 /* An impossible timestamp guarantees this value
3963                  * will never match a generated boot time. */
3964                 verf[0] = 0;
3965                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
3966         } else {
3967                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
3968                 verf[0] = (__be32)nn->boot_time.tv_sec;
3969                 verf[1] = (__be32)nn->boot_time.tv_nsec;
3970         }
3971         memcpy(bootverf->data, verf, sizeof(bootverf->data));
3972 }
3973
3974 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3975                 unsigned short port, struct rpc_cred *cred,
3976                 struct nfs4_setclientid_res *res)
3977 {
3978         nfs4_verifier sc_verifier;
3979         struct nfs4_setclientid setclientid = {
3980                 .sc_verifier = &sc_verifier,
3981                 .sc_prog = program,
3982                 .sc_cb_ident = clp->cl_cb_ident,
3983         };
3984         struct rpc_message msg = {
3985                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3986                 .rpc_argp = &setclientid,
3987                 .rpc_resp = res,
3988                 .rpc_cred = cred,
3989         };
3990         int loop = 0;
3991         int status;
3992
3993         nfs4_init_boot_verifier(clp, &sc_verifier);
3994
3995         for(;;) {
3996                 rcu_read_lock();
3997                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3998                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3999                                 clp->cl_ipaddr,
4000                                 rpc_peeraddr2str(clp->cl_rpcclient,
4001                                                         RPC_DISPLAY_ADDR),
4002                                 rpc_peeraddr2str(clp->cl_rpcclient,
4003                                                         RPC_DISPLAY_PROTO),
4004                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4005                                 clp->cl_id_uniquifier);
4006                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4007                                 sizeof(setclientid.sc_netid),
4008                                 rpc_peeraddr2str(clp->cl_rpcclient,
4009                                                         RPC_DISPLAY_NETID));
4010                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4011                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4012                                 clp->cl_ipaddr, port >> 8, port & 255);
4013                 rcu_read_unlock();
4014
4015                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4016                 if (status != -NFS4ERR_CLID_INUSE)
4017                         break;
4018                 if (loop != 0) {
4019                         ++clp->cl_id_uniquifier;
4020                         break;
4021                 }
4022                 ++loop;
4023                 ssleep(clp->cl_lease_time / HZ + 1);
4024         }
4025         return status;
4026 }
4027
4028 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4029                 struct nfs4_setclientid_res *arg,
4030                 struct rpc_cred *cred)
4031 {
4032         struct nfs_fsinfo fsinfo;
4033         struct rpc_message msg = {
4034                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4035                 .rpc_argp = arg,
4036                 .rpc_resp = &fsinfo,
4037                 .rpc_cred = cred,
4038         };
4039         unsigned long now;
4040         int status;
4041
4042         now = jiffies;
4043         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4044         if (status == 0) {
4045                 spin_lock(&clp->cl_lock);
4046                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4047                 clp->cl_last_renewal = now;
4048                 spin_unlock(&clp->cl_lock);
4049         }
4050         return status;
4051 }
4052
4053 struct nfs4_delegreturndata {
4054         struct nfs4_delegreturnargs args;
4055         struct nfs4_delegreturnres res;
4056         struct nfs_fh fh;
4057         nfs4_stateid stateid;
4058         unsigned long timestamp;
4059         struct nfs_fattr fattr;
4060         int rpc_status;
4061 };
4062
4063 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4064 {
4065         struct nfs4_delegreturndata *data = calldata;
4066
4067         if (!nfs4_sequence_done(task, &data->res.seq_res))
4068                 return;
4069
4070         switch (task->tk_status) {
4071         case -NFS4ERR_STALE_STATEID:
4072         case -NFS4ERR_EXPIRED:
4073         case 0:
4074                 renew_lease(data->res.server, data->timestamp);
4075                 break;
4076         default:
4077                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4078                                 -EAGAIN) {
4079                         rpc_restart_call_prepare(task);
4080                         return;
4081                 }
4082         }
4083         data->rpc_status = task->tk_status;
4084 }
4085
4086 static void nfs4_delegreturn_release(void *calldata)
4087 {
4088         kfree(calldata);
4089 }
4090
4091 #if defined(CONFIG_NFS_V4_1)
4092 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4093 {
4094         struct nfs4_delegreturndata *d_data;
4095
4096         d_data = (struct nfs4_delegreturndata *)data;
4097
4098         if (nfs4_setup_sequence(d_data->res.server,
4099                                 &d_data->args.seq_args,
4100                                 &d_data->res.seq_res, task))
4101                 return;
4102         rpc_call_start(task);
4103 }
4104 #endif /* CONFIG_NFS_V4_1 */
4105
4106 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4107 #if defined(CONFIG_NFS_V4_1)
4108         .rpc_call_prepare = nfs4_delegreturn_prepare,
4109 #endif /* CONFIG_NFS_V4_1 */
4110         .rpc_call_done = nfs4_delegreturn_done,
4111         .rpc_release = nfs4_delegreturn_release,
4112 };
4113
4114 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4115 {
4116         struct nfs4_delegreturndata *data;
4117         struct nfs_server *server = NFS_SERVER(inode);
4118         struct rpc_task *task;
4119         struct rpc_message msg = {
4120                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4121                 .rpc_cred = cred,
4122         };
4123         struct rpc_task_setup task_setup_data = {
4124                 .rpc_client = server->client,
4125                 .rpc_message = &msg,
4126                 .callback_ops = &nfs4_delegreturn_ops,
4127                 .flags = RPC_TASK_ASYNC,
4128         };
4129         int status = 0;
4130
4131         data = kzalloc(sizeof(*data), GFP_NOFS);
4132         if (data == NULL)
4133                 return -ENOMEM;
4134         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4135         data->args.fhandle = &data->fh;
4136         data->args.stateid = &data->stateid;
4137         data->args.bitmask = server->cache_consistency_bitmask;
4138         nfs_copy_fh(&data->fh, NFS_FH(inode));
4139         nfs4_stateid_copy(&data->stateid, stateid);
4140         data->res.fattr = &data->fattr;
4141         data->res.server = server;
4142         nfs_fattr_init(data->res.fattr);
4143         data->timestamp = jiffies;
4144         data->rpc_status = 0;
4145
4146         task_setup_data.callback_data = data;
4147         msg.rpc_argp = &data->args;
4148         msg.rpc_resp = &data->res;
4149         task = rpc_run_task(&task_setup_data);
4150         if (IS_ERR(task))
4151                 return PTR_ERR(task);
4152         if (!issync)
4153                 goto out;
4154         status = nfs4_wait_for_completion_rpc_task(task);
4155         if (status != 0)
4156                 goto out;
4157         status = data->rpc_status;
4158         if (status == 0)
4159                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4160         else
4161                 nfs_refresh_inode(inode, &data->fattr);
4162 out:
4163         rpc_put_task(task);
4164         return status;
4165 }
4166
4167 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4168 {
4169         struct nfs_server *server = NFS_SERVER(inode);
4170         struct nfs4_exception exception = { };
4171         int err;
4172         do {
4173                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4174                 switch (err) {
4175                         case -NFS4ERR_STALE_STATEID:
4176                         case -NFS4ERR_EXPIRED:
4177                         case 0:
4178                                 return 0;
4179                 }
4180                 err = nfs4_handle_exception(server, err, &exception);
4181         } while (exception.retry);
4182         return err;
4183 }
4184
4185 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4186 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4187
4188 /*
4189  * sleep, with exponential backoff, and retry the LOCK operation.
4190  */
4191 static unsigned long
4192 nfs4_set_lock_task_retry(unsigned long timeout)
4193 {
4194         freezable_schedule_timeout_killable(timeout);
4195         timeout <<= 1;
4196         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4197                 return NFS4_LOCK_MAXTIMEOUT;
4198         return timeout;
4199 }
4200
4201 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4202 {
4203         struct inode *inode = state->inode;
4204         struct nfs_server *server = NFS_SERVER(inode);
4205         struct nfs_client *clp = server->nfs_client;
4206         struct nfs_lockt_args arg = {
4207                 .fh = NFS_FH(inode),
4208                 .fl = request,
4209         };
4210         struct nfs_lockt_res res = {
4211                 .denied = request,
4212         };
4213         struct rpc_message msg = {
4214                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4215                 .rpc_argp       = &arg,
4216                 .rpc_resp       = &res,
4217                 .rpc_cred       = state->owner->so_cred,
4218         };
4219         struct nfs4_lock_state *lsp;
4220         int status;
4221
4222         arg.lock_owner.clientid = clp->cl_clientid;
4223         status = nfs4_set_lock_state(state, request);
4224         if (status != 0)
4225                 goto out;
4226         lsp = request->fl_u.nfs4_fl.owner;
4227         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4228         arg.lock_owner.s_dev = server->s_dev;
4229         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4230         switch (status) {
4231                 case 0:
4232                         request->fl_type = F_UNLCK;
4233                         break;
4234                 case -NFS4ERR_DENIED:
4235                         status = 0;
4236         }
4237         request->fl_ops->fl_release_private(request);
4238 out:
4239         return status;
4240 }
4241
4242 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4243 {
4244         struct nfs4_exception exception = { };
4245         int err;
4246
4247         do {
4248                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4249                                 _nfs4_proc_getlk(state, cmd, request),
4250                                 &exception);
4251         } while (exception.retry);
4252         return err;
4253 }
4254
4255 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4256 {
4257         int res = 0;
4258         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4259                 case FL_POSIX:
4260                         res = posix_lock_file_wait(file, fl);
4261                         break;
4262                 case FL_FLOCK:
4263                         res = flock_lock_file_wait(file, fl);
4264                         break;
4265                 default:
4266                         BUG();
4267         }
4268         return res;
4269 }
4270
4271 struct nfs4_unlockdata {
4272         struct nfs_locku_args arg;
4273         struct nfs_locku_res res;
4274         struct nfs4_lock_state *lsp;
4275         struct nfs_open_context *ctx;
4276         struct file_lock fl;
4277         const struct nfs_server *server;
4278         unsigned long timestamp;
4279 };
4280
4281 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4282                 struct nfs_open_context *ctx,
4283                 struct nfs4_lock_state *lsp,
4284                 struct nfs_seqid *seqid)
4285 {
4286         struct nfs4_unlockdata *p;
4287         struct inode *inode = lsp->ls_state->inode;
4288
4289         p = kzalloc(sizeof(*p), GFP_NOFS);
4290         if (p == NULL)
4291                 return NULL;
4292         p->arg.fh = NFS_FH(inode);
4293         p->arg.fl = &p->fl;
4294         p->arg.seqid = seqid;
4295         p->res.seqid = seqid;
4296         p->arg.stateid = &lsp->ls_stateid;
4297         p->lsp = lsp;
4298         atomic_inc(&lsp->ls_count);
4299         /* Ensure we don't close file until we're done freeing locks! */
4300         p->ctx = get_nfs_open_context(ctx);
4301         memcpy(&p->fl, fl, sizeof(p->fl));
4302         p->server = NFS_SERVER(inode);
4303         return p;
4304 }
4305
4306 static void nfs4_locku_release_calldata(void *data)
4307 {
4308         struct nfs4_unlockdata *calldata = data;
4309         nfs_free_seqid(calldata->arg.seqid);
4310         nfs4_put_lock_state(calldata->lsp);
4311         put_nfs_open_context(calldata->ctx);
4312         kfree(calldata);
4313 }
4314
4315 static void nfs4_locku_done(struct rpc_task *task, void *data)
4316 {
4317         struct nfs4_unlockdata *calldata = data;
4318
4319         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4320                 return;
4321         switch (task->tk_status) {
4322                 case 0:
4323                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4324                                         &calldata->res.stateid);
4325                         renew_lease(calldata->server, calldata->timestamp);
4326                         break;
4327                 case -NFS4ERR_BAD_STATEID:
4328                 case -NFS4ERR_OLD_STATEID:
4329                 case -NFS4ERR_STALE_STATEID:
4330                 case -NFS4ERR_EXPIRED:
4331                         break;
4332                 default:
4333                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4334                                 rpc_restart_call_prepare(task);
4335         }
4336 }
4337
4338 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4339 {
4340         struct nfs4_unlockdata *calldata = data;
4341
4342         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4343                 return;
4344         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4345                 /* Note: exit _without_ running nfs4_locku_done */
4346                 task->tk_action = NULL;
4347                 return;
4348         }
4349         calldata->timestamp = jiffies;
4350         if (nfs4_setup_sequence(calldata->server,
4351                                 &calldata->arg.seq_args,
4352                                 &calldata->res.seq_res, task))
4353                 return;
4354         rpc_call_start(task);
4355 }
4356
4357 static const struct rpc_call_ops nfs4_locku_ops = {
4358         .rpc_call_prepare = nfs4_locku_prepare,
4359         .rpc_call_done = nfs4_locku_done,
4360         .rpc_release = nfs4_locku_release_calldata,
4361 };
4362
4363 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4364                 struct nfs_open_context *ctx,
4365                 struct nfs4_lock_state *lsp,
4366                 struct nfs_seqid *seqid)
4367 {
4368         struct nfs4_unlockdata *data;
4369         struct rpc_message msg = {
4370                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4371                 .rpc_cred = ctx->cred,
4372         };
4373         struct rpc_task_setup task_setup_data = {
4374                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4375                 .rpc_message = &msg,
4376                 .callback_ops = &nfs4_locku_ops,
4377                 .workqueue = nfsiod_workqueue,
4378                 .flags = RPC_TASK_ASYNC,
4379         };
4380
4381         /* Ensure this is an unlock - when canceling a lock, the
4382          * canceled lock is passed in, and it won't be an unlock.
4383          */
4384         fl->fl_type = F_UNLCK;
4385
4386         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4387         if (data == NULL) {
4388                 nfs_free_seqid(seqid);
4389                 return ERR_PTR(-ENOMEM);
4390         }
4391
4392         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4393         msg.rpc_argp = &data->arg;
4394         msg.rpc_resp = &data->res;
4395         task_setup_data.callback_data = data;
4396         return rpc_run_task(&task_setup_data);
4397 }
4398
4399 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4400 {
4401         struct nfs_inode *nfsi = NFS_I(state->inode);
4402         struct nfs_seqid *seqid;
4403         struct nfs4_lock_state *lsp;
4404         struct rpc_task *task;
4405         int status = 0;
4406         unsigned char fl_flags = request->fl_flags;
4407
4408         status = nfs4_set_lock_state(state, request);
4409         /* Unlock _before_ we do the RPC call */
4410         request->fl_flags |= FL_EXISTS;
4411         down_read(&nfsi->rwsem);
4412         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4413                 up_read(&nfsi->rwsem);
4414                 goto out;
4415         }
4416         up_read(&nfsi->rwsem);
4417         if (status != 0)
4418                 goto out;
4419         /* Is this a delegated lock? */
4420         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4421                 goto out;
4422         lsp = request->fl_u.nfs4_fl.owner;
4423         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4424         status = -ENOMEM;
4425         if (seqid == NULL)
4426                 goto out;
4427         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4428         status = PTR_ERR(task);
4429         if (IS_ERR(task))
4430                 goto out;
4431         status = nfs4_wait_for_completion_rpc_task(task);
4432         rpc_put_task(task);
4433 out:
4434         request->fl_flags = fl_flags;
4435         return status;
4436 }
4437
4438 struct nfs4_lockdata {
4439         struct nfs_lock_args arg;
4440         struct nfs_lock_res res;
4441         struct nfs4_lock_state *lsp;
4442         struct nfs_open_context *ctx;
4443         struct file_lock fl;
4444         unsigned long timestamp;
4445         int rpc_status;
4446         int cancelled;
4447         struct nfs_server *server;
4448 };
4449
4450 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4451                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4452                 gfp_t gfp_mask)
4453 {
4454         struct nfs4_lockdata *p;
4455         struct inode *inode = lsp->ls_state->inode;
4456         struct nfs_server *server = NFS_SERVER(inode);
4457
4458         p = kzalloc(sizeof(*p), gfp_mask);
4459         if (p == NULL)
4460                 return NULL;
4461
4462         p->arg.fh = NFS_FH(inode);
4463         p->arg.fl = &p->fl;
4464         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4465         if (p->arg.open_seqid == NULL)
4466                 goto out_free;
4467         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4468         if (p->arg.lock_seqid == NULL)
4469                 goto out_free_seqid;
4470         p->arg.lock_stateid = &lsp->ls_stateid;
4471         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4472         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4473         p->arg.lock_owner.s_dev = server->s_dev;
4474         p->res.lock_seqid = p->arg.lock_seqid;
4475         p->lsp = lsp;
4476         p->server = server;
4477         atomic_inc(&lsp->ls_count);
4478         p->ctx = get_nfs_open_context(ctx);
4479         memcpy(&p->fl, fl, sizeof(p->fl));
4480         return p;
4481 out_free_seqid:
4482         nfs_free_seqid(p->arg.open_seqid);
4483 out_free:
4484         kfree(p);
4485         return NULL;
4486 }
4487
4488 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4489 {
4490         struct nfs4_lockdata *data = calldata;
4491         struct nfs4_state *state = data->lsp->ls_state;
4492
4493         dprintk("%s: begin!\n", __func__);
4494         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4495                 return;
4496         /* Do we need to do an open_to_lock_owner? */
4497         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4498                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4499                         return;
4500                 data->arg.open_stateid = &state->stateid;
4501                 data->arg.new_lock_owner = 1;
4502                 data->res.open_seqid = data->arg.open_seqid;
4503         } else
4504                 data->arg.new_lock_owner = 0;
4505         data->timestamp = jiffies;
4506         if (nfs4_setup_sequence(data->server,
4507                                 &data->arg.seq_args,
4508                                 &data->res.seq_res, task))
4509                 return;
4510         rpc_call_start(task);
4511         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4512 }
4513
4514 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4515 {
4516         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4517         nfs4_lock_prepare(task, calldata);
4518 }
4519
4520 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4521 {
4522         struct nfs4_lockdata *data = calldata;
4523
4524         dprintk("%s: begin!\n", __func__);
4525
4526         if (!nfs4_sequence_done(task, &data->res.seq_res))
4527                 return;
4528
4529         data->rpc_status = task->tk_status;
4530         if (data->arg.new_lock_owner != 0) {
4531                 if (data->rpc_status == 0)
4532                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4533                 else
4534                         goto out;
4535         }
4536         if (data->rpc_status == 0) {
4537                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4538                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4539                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4540         }
4541 out:
4542         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4543 }
4544
4545 static void nfs4_lock_release(void *calldata)
4546 {
4547         struct nfs4_lockdata *data = calldata;
4548
4549         dprintk("%s: begin!\n", __func__);
4550         nfs_free_seqid(data->arg.open_seqid);
4551         if (data->cancelled != 0) {
4552                 struct rpc_task *task;
4553                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4554                                 data->arg.lock_seqid);
4555                 if (!IS_ERR(task))
4556                         rpc_put_task_async(task);
4557                 dprintk("%s: cancelling lock!\n", __func__);
4558         } else
4559                 nfs_free_seqid(data->arg.lock_seqid);
4560         nfs4_put_lock_state(data->lsp);
4561         put_nfs_open_context(data->ctx);
4562         kfree(data);
4563         dprintk("%s: done!\n", __func__);
4564 }
4565
4566 static const struct rpc_call_ops nfs4_lock_ops = {
4567         .rpc_call_prepare = nfs4_lock_prepare,
4568         .rpc_call_done = nfs4_lock_done,
4569         .rpc_release = nfs4_lock_release,
4570 };
4571
4572 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4573         .rpc_call_prepare = nfs4_recover_lock_prepare,
4574         .rpc_call_done = nfs4_lock_done,
4575         .rpc_release = nfs4_lock_release,
4576 };
4577
4578 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4579 {
4580         switch (error) {
4581         case -NFS4ERR_ADMIN_REVOKED:
4582         case -NFS4ERR_BAD_STATEID:
4583                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4584                 if (new_lock_owner != 0 ||
4585                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4586                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4587                 break;
4588         case -NFS4ERR_STALE_STATEID:
4589                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4590         case -NFS4ERR_EXPIRED:
4591                 nfs4_schedule_lease_recovery(server->nfs_client);
4592         };
4593 }
4594
4595 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4596 {
4597         struct nfs4_lockdata *data;
4598         struct rpc_task *task;
4599         struct rpc_message msg = {
4600                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4601                 .rpc_cred = state->owner->so_cred,
4602         };
4603         struct rpc_task_setup task_setup_data = {
4604                 .rpc_client = NFS_CLIENT(state->inode),
4605                 .rpc_message = &msg,
4606                 .callback_ops = &nfs4_lock_ops,
4607                 .workqueue = nfsiod_workqueue,
4608                 .flags = RPC_TASK_ASYNC,
4609         };
4610         int ret;
4611
4612         dprintk("%s: begin!\n", __func__);
4613         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4614                         fl->fl_u.nfs4_fl.owner,
4615                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4616         if (data == NULL)
4617                 return -ENOMEM;
4618         if (IS_SETLKW(cmd))
4619                 data->arg.block = 1;
4620         if (recovery_type > NFS_LOCK_NEW) {
4621                 if (recovery_type == NFS_LOCK_RECLAIM)
4622                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4623                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4624         }
4625         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4626         msg.rpc_argp = &data->arg;
4627         msg.rpc_resp = &data->res;
4628         task_setup_data.callback_data = data;
4629         task = rpc_run_task(&task_setup_data);
4630         if (IS_ERR(task))
4631                 return PTR_ERR(task);
4632         ret = nfs4_wait_for_completion_rpc_task(task);
4633         if (ret == 0) {
4634                 ret = data->rpc_status;
4635                 if (ret)
4636                         nfs4_handle_setlk_error(data->server, data->lsp,
4637                                         data->arg.new_lock_owner, ret);
4638         } else
4639                 data->cancelled = 1;
4640         rpc_put_task(task);
4641         dprintk("%s: done, ret = %d!\n", __func__, ret);
4642         return ret;
4643 }
4644
4645 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4646 {
4647         struct nfs_server *server = NFS_SERVER(state->inode);
4648         struct nfs4_exception exception = {
4649                 .inode = state->inode,
4650         };
4651         int err;
4652
4653         do {
4654                 /* Cache the lock if possible... */
4655                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4656                         return 0;
4657                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4658                 if (err != -NFS4ERR_DELAY)
4659                         break;
4660                 nfs4_handle_exception(server, err, &exception);
4661         } while (exception.retry);
4662         return err;
4663 }
4664
4665 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4666 {
4667         struct nfs_server *server = NFS_SERVER(state->inode);
4668         struct nfs4_exception exception = {
4669                 .inode = state->inode,
4670         };
4671         int err;
4672
4673         err = nfs4_set_lock_state(state, request);
4674         if (err != 0)
4675                 return err;
4676         do {
4677                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4678                         return 0;
4679                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4680                 switch (err) {
4681                 default:
4682                         goto out;
4683                 case -NFS4ERR_GRACE:
4684                 case -NFS4ERR_DELAY:
4685                         nfs4_handle_exception(server, err, &exception);
4686                         err = 0;
4687                 }
4688         } while (exception.retry);
4689 out:
4690         return err;
4691 }
4692
4693 #if defined(CONFIG_NFS_V4_1)
4694 static int nfs41_check_expired_locks(struct nfs4_state *state)
4695 {
4696         int status, ret = NFS_OK;
4697         struct nfs4_lock_state *lsp;
4698         struct nfs_server *server = NFS_SERVER(state->inode);
4699
4700         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4701                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4702                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4703                         if (status != NFS_OK) {
4704                                 nfs41_free_stateid(server, &lsp->ls_stateid);
4705                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4706                                 ret = status;
4707                         }
4708                 }
4709         };
4710
4711         return ret;
4712 }
4713
4714 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4715 {
4716         int status = NFS_OK;
4717
4718         if (test_bit(LK_STATE_IN_USE, &state->flags))
4719                 status = nfs41_check_expired_locks(state);
4720         if (status == NFS_OK)
4721                 return status;
4722         return nfs4_lock_expired(state, request);
4723 }
4724 #endif
4725
4726 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4727 {
4728         struct nfs_inode *nfsi = NFS_I(state->inode);
4729         unsigned char fl_flags = request->fl_flags;
4730         int status = -ENOLCK;
4731
4732         if ((fl_flags & FL_POSIX) &&
4733                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4734                 goto out;
4735         /* Is this a delegated open? */
4736         status = nfs4_set_lock_state(state, request);
4737         if (status != 0)
4738                 goto out;
4739         request->fl_flags |= FL_ACCESS;
4740         status = do_vfs_lock(request->fl_file, request);
4741         if (status < 0)
4742                 goto out;
4743         down_read(&nfsi->rwsem);
4744         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4745                 /* Yes: cache locks! */
4746                 /* ...but avoid races with delegation recall... */
4747                 request->fl_flags = fl_flags & ~FL_SLEEP;
4748                 status = do_vfs_lock(request->fl_file, request);
4749                 goto out_unlock;
4750         }
4751         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4752         if (status != 0)
4753                 goto out_unlock;
4754         /* Note: we always want to sleep here! */
4755         request->fl_flags = fl_flags | FL_SLEEP;
4756         if (do_vfs_lock(request->fl_file, request) < 0)
4757                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4758                         "manager!\n", __func__);
4759 out_unlock:
4760         up_read(&nfsi->rwsem);
4761 out:
4762         request->fl_flags = fl_flags;
4763         return status;
4764 }
4765
4766 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4767 {
4768         struct nfs4_exception exception = {
4769                 .state = state,
4770                 .inode = state->inode,
4771         };
4772         int err;
4773
4774         do {
4775                 err = _nfs4_proc_setlk(state, cmd, request);
4776                 if (err == -NFS4ERR_DENIED)
4777                         err = -EAGAIN;
4778                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4779                                 err, &exception);
4780         } while (exception.retry);
4781         return err;
4782 }
4783
4784 static int
4785 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4786 {
4787         struct nfs_open_context *ctx;
4788         struct nfs4_state *state;
4789         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4790         int status;
4791
4792         /* verify open state */
4793         ctx = nfs_file_open_context(filp);
4794         state = ctx->state;
4795
4796         if (request->fl_start < 0 || request->fl_end < 0)
4797                 return -EINVAL;
4798
4799         if (IS_GETLK(cmd)) {
4800                 if (state != NULL)
4801                         return nfs4_proc_getlk(state, F_GETLK, request);
4802                 return 0;
4803         }
4804
4805         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4806                 return -EINVAL;
4807
4808         if (request->fl_type == F_UNLCK) {
4809                 if (state != NULL)
4810                         return nfs4_proc_unlck(state, cmd, request);
4811                 return 0;
4812         }
4813
4814         if (state == NULL)
4815                 return -ENOLCK;
4816         /*
4817          * Don't rely on the VFS having checked the file open mode,
4818          * since it won't do this for flock() locks.
4819          */
4820         switch (request->fl_type & (F_RDLCK|F_WRLCK|F_UNLCK)) {
4821         case F_RDLCK:
4822                 if (!(filp->f_mode & FMODE_READ))
4823                         return -EBADF;
4824                 break;
4825         case F_WRLCK:
4826                 if (!(filp->f_mode & FMODE_WRITE))
4827                         return -EBADF;
4828         }
4829
4830         do {
4831                 status = nfs4_proc_setlk(state, cmd, request);
4832                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4833                         break;
4834                 timeout = nfs4_set_lock_task_retry(timeout);
4835                 status = -ERESTARTSYS;
4836                 if (signalled())
4837                         break;
4838         } while(status < 0);
4839         return status;
4840 }
4841
4842 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4843 {
4844         struct nfs_server *server = NFS_SERVER(state->inode);
4845         struct nfs4_exception exception = { };
4846         int err;
4847
4848         err = nfs4_set_lock_state(state, fl);
4849         if (err != 0)
4850                 goto out;
4851         do {
4852                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4853                 switch (err) {
4854                         default:
4855                                 printk(KERN_ERR "NFS: %s: unhandled error "
4856                                         "%d.\n", __func__, err);
4857                         case 0:
4858                         case -ESTALE:
4859                                 goto out;
4860                         case -NFS4ERR_EXPIRED:
4861                                 nfs4_schedule_stateid_recovery(server, state);
4862                         case -NFS4ERR_STALE_CLIENTID:
4863                         case -NFS4ERR_STALE_STATEID:
4864                                 nfs4_schedule_lease_recovery(server->nfs_client);
4865                                 goto out;
4866                         case -NFS4ERR_BADSESSION:
4867                         case -NFS4ERR_BADSLOT:
4868                         case -NFS4ERR_BAD_HIGH_SLOT:
4869                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4870                         case -NFS4ERR_DEADSESSION:
4871                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
4872                                 goto out;
4873                         case -ERESTARTSYS:
4874                                 /*
4875                                  * The show must go on: exit, but mark the
4876                                  * stateid as needing recovery.
4877                                  */
4878                         case -NFS4ERR_DELEG_REVOKED:
4879                         case -NFS4ERR_ADMIN_REVOKED:
4880                         case -NFS4ERR_BAD_STATEID:
4881                         case -NFS4ERR_OPENMODE:
4882                                 nfs4_schedule_stateid_recovery(server, state);
4883                                 err = 0;
4884                                 goto out;
4885                         case -EKEYEXPIRED:
4886                                 /*
4887                                  * User RPCSEC_GSS context has expired.
4888                                  * We cannot recover this stateid now, so
4889                                  * skip it and allow recovery thread to
4890                                  * proceed.
4891                                  */
4892                                 err = 0;
4893                                 goto out;
4894                         case -ENOMEM:
4895                         case -NFS4ERR_DENIED:
4896                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4897                                 err = 0;
4898                                 goto out;
4899                         case -NFS4ERR_DELAY:
4900                                 break;
4901                 }
4902                 err = nfs4_handle_exception(server, err, &exception);
4903         } while (exception.retry);
4904 out:
4905         return err;
4906 }
4907
4908 struct nfs_release_lockowner_data {
4909         struct nfs4_lock_state *lsp;
4910         struct nfs_server *server;
4911         struct nfs_release_lockowner_args args;
4912 };
4913
4914 static void nfs4_release_lockowner_release(void *calldata)
4915 {
4916         struct nfs_release_lockowner_data *data = calldata;
4917         nfs4_free_lock_state(data->server, data->lsp);
4918         kfree(calldata);
4919 }
4920
4921 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
4922         .rpc_release = nfs4_release_lockowner_release,
4923 };
4924
4925 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
4926 {
4927         struct nfs_server *server = lsp->ls_state->owner->so_server;
4928         struct nfs_release_lockowner_data *data;
4929         struct rpc_message msg = {
4930                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4931         };
4932
4933         if (server->nfs_client->cl_mvops->minor_version != 0)
4934                 return -EINVAL;
4935         data = kmalloc(sizeof(*data), GFP_NOFS);
4936         if (!data)
4937                 return -ENOMEM;
4938         data->lsp = lsp;
4939         data->server = server;
4940         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
4941         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
4942         data->args.lock_owner.s_dev = server->s_dev;
4943         msg.rpc_argp = &data->args;
4944         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
4945         return 0;
4946 }
4947
4948 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4949
4950 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4951                                    const void *buf, size_t buflen,
4952                                    int flags, int type)
4953 {
4954         if (strcmp(key, "") != 0)
4955                 return -EINVAL;
4956
4957         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4958 }
4959
4960 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4961                                    void *buf, size_t buflen, int type)
4962 {
4963         if (strcmp(key, "") != 0)
4964                 return -EINVAL;
4965
4966         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4967 }
4968
4969 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4970                                        size_t list_len, const char *name,
4971                                        size_t name_len, int type)
4972 {
4973         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4974
4975         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4976                 return 0;
4977
4978         if (list && len <= list_len)
4979                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4980         return len;
4981 }
4982
4983 /*
4984  * nfs_fhget will use either the mounted_on_fileid or the fileid
4985  */
4986 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4987 {
4988         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4989                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4990               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4991               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
4992                 return;
4993
4994         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4995                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
4996         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4997         fattr->nlink = 2;
4998 }
4999
5000 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5001                                    const struct qstr *name,
5002                                    struct nfs4_fs_locations *fs_locations,
5003                                    struct page *page)
5004 {
5005         struct nfs_server *server = NFS_SERVER(dir);
5006         u32 bitmask[2] = {
5007                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5008         };
5009         struct nfs4_fs_locations_arg args = {
5010                 .dir_fh = NFS_FH(dir),
5011                 .name = name,
5012                 .page = page,
5013                 .bitmask = bitmask,
5014         };
5015         struct nfs4_fs_locations_res res = {
5016                 .fs_locations = fs_locations,
5017         };
5018         struct rpc_message msg = {
5019                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5020                 .rpc_argp = &args,
5021                 .rpc_resp = &res,
5022         };
5023         int status;
5024
5025         dprintk("%s: start\n", __func__);
5026
5027         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5028          * is not supported */
5029         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5030                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5031         else
5032                 bitmask[0] |= FATTR4_WORD0_FILEID;
5033
5034         nfs_fattr_init(&fs_locations->fattr);
5035         fs_locations->server = server;
5036         fs_locations->nlocations = 0;
5037         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5038         dprintk("%s: returned status = %d\n", __func__, status);
5039         return status;
5040 }
5041
5042 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5043                            const struct qstr *name,
5044                            struct nfs4_fs_locations *fs_locations,
5045                            struct page *page)
5046 {
5047         struct nfs4_exception exception = { };
5048         int err;
5049         do {
5050                 err = nfs4_handle_exception(NFS_SERVER(dir),
5051                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5052                                 &exception);
5053         } while (exception.retry);
5054         return err;
5055 }
5056
5057 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5058 {
5059         int status;
5060         struct nfs4_secinfo_arg args = {
5061                 .dir_fh = NFS_FH(dir),
5062                 .name   = name,
5063         };
5064         struct nfs4_secinfo_res res = {
5065                 .flavors     = flavors,
5066         };
5067         struct rpc_message msg = {
5068                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5069                 .rpc_argp = &args,
5070                 .rpc_resp = &res,
5071         };
5072
5073         dprintk("NFS call  secinfo %s\n", name->name);
5074         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5075         dprintk("NFS reply  secinfo: %d\n", status);
5076         return status;
5077 }
5078
5079 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5080                       struct nfs4_secinfo_flavors *flavors)
5081 {
5082         struct nfs4_exception exception = { };
5083         int err;
5084         do {
5085                 err = nfs4_handle_exception(NFS_SERVER(dir),
5086                                 _nfs4_proc_secinfo(dir, name, flavors),
5087                                 &exception);
5088         } while (exception.retry);
5089         return err;
5090 }
5091
5092 #ifdef CONFIG_NFS_V4_1
5093 /*
5094  * Check the exchange flags returned by the server for invalid flags, having
5095  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5096  * DS flags set.
5097  */
5098 static int nfs4_check_cl_exchange_flags(u32 flags)
5099 {
5100         if (flags & ~EXCHGID4_FLAG_MASK_R)
5101                 goto out_inval;
5102         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5103             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5104                 goto out_inval;
5105         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5106                 goto out_inval;
5107         return NFS_OK;
5108 out_inval:
5109         return -NFS4ERR_INVAL;
5110 }
5111
5112 static bool
5113 nfs41_same_server_scope(struct nfs41_server_scope *a,
5114                         struct nfs41_server_scope *b)
5115 {
5116         if (a->server_scope_sz == b->server_scope_sz &&
5117             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5118                 return true;
5119
5120         return false;
5121 }
5122
5123 /*
5124  * nfs4_proc_bind_conn_to_session()
5125  *
5126  * The 4.1 client currently uses the same TCP connection for the
5127  * fore and backchannel.
5128  */
5129 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5130 {
5131         int status;
5132         struct nfs41_bind_conn_to_session_res res;
5133         struct rpc_message msg = {
5134                 .rpc_proc =
5135                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5136                 .rpc_argp = clp,
5137                 .rpc_resp = &res,
5138                 .rpc_cred = cred,
5139         };
5140
5141         dprintk("--> %s\n", __func__);
5142         BUG_ON(clp == NULL);
5143
5144         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5145         if (unlikely(res.session == NULL)) {
5146                 status = -ENOMEM;
5147                 goto out;
5148         }
5149
5150         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5151         if (status == 0) {
5152                 if (memcmp(res.session->sess_id.data,
5153                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5154                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5155                         status = -EIO;
5156                         goto out_session;
5157                 }
5158                 if (res.dir != NFS4_CDFS4_BOTH) {
5159                         dprintk("NFS: %s: Unexpected direction from server\n",
5160                                 __func__);
5161                         status = -EIO;
5162                         goto out_session;
5163                 }
5164                 if (res.use_conn_in_rdma_mode) {
5165                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5166                                 __func__);
5167                         status = -EIO;
5168                         goto out_session;
5169                 }
5170         }
5171 out_session:
5172         kfree(res.session);
5173 out:
5174         dprintk("<-- %s status= %d\n", __func__, status);
5175         return status;
5176 }
5177
5178 /*
5179  * nfs4_proc_exchange_id()
5180  *
5181  * Since the clientid has expired, all compounds using sessions
5182  * associated with the stale clientid will be returning
5183  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5184  * be in some phase of session reset.
5185  */
5186 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5187 {
5188         nfs4_verifier verifier;
5189         struct nfs41_exchange_id_args args = {
5190                 .verifier = &verifier,
5191                 .client = clp,
5192                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5193         };
5194         struct nfs41_exchange_id_res res = {
5195                 0
5196         };
5197         int status;
5198         struct rpc_message msg = {
5199                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5200                 .rpc_argp = &args,
5201                 .rpc_resp = &res,
5202                 .rpc_cred = cred,
5203         };
5204
5205         dprintk("--> %s\n", __func__);
5206         BUG_ON(clp == NULL);
5207
5208         nfs4_init_boot_verifier(clp, &verifier);
5209
5210         args.id_len = scnprintf(args.id, sizeof(args.id),
5211                                 "%s/%s/%u",
5212                                 clp->cl_ipaddr,
5213                                 clp->cl_rpcclient->cl_nodename,
5214                                 clp->cl_rpcclient->cl_auth->au_flavor);
5215
5216         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5217                                         GFP_NOFS);
5218         if (unlikely(res.server_owner == NULL)) {
5219                 status = -ENOMEM;
5220                 goto out;
5221         }
5222
5223         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5224                                         GFP_NOFS);
5225         if (unlikely(res.server_scope == NULL)) {
5226                 status = -ENOMEM;
5227                 goto out_server_owner;
5228         }
5229
5230         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5231         if (unlikely(res.impl_id == NULL)) {
5232                 status = -ENOMEM;
5233                 goto out_server_scope;
5234         }
5235
5236         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5237         if (status == 0)
5238                 status = nfs4_check_cl_exchange_flags(res.flags);
5239
5240         if (status == 0) {
5241                 clp->cl_clientid = res.clientid;
5242                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5243                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5244                         clp->cl_seqid = res.seqid;
5245
5246                 kfree(clp->cl_serverowner);
5247                 clp->cl_serverowner = res.server_owner;
5248                 res.server_owner = NULL;
5249
5250                 /* use the most recent implementation id */
5251                 kfree(clp->cl_implid);
5252                 clp->cl_implid = res.impl_id;
5253
5254                 if (clp->cl_serverscope != NULL &&
5255                     !nfs41_same_server_scope(clp->cl_serverscope,
5256                                              res.server_scope)) {
5257                         dprintk("%s: server_scope mismatch detected\n",
5258                                 __func__);
5259                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5260                         kfree(clp->cl_serverscope);
5261                         clp->cl_serverscope = NULL;
5262                 }
5263
5264                 if (clp->cl_serverscope == NULL) {
5265                         clp->cl_serverscope = res.server_scope;
5266                         goto out;
5267                 }
5268         } else
5269                 kfree(res.impl_id);
5270
5271 out_server_owner:
5272         kfree(res.server_owner);
5273 out_server_scope:
5274         kfree(res.server_scope);
5275 out:
5276         if (clp->cl_implid != NULL)
5277                 dprintk("%s: Server Implementation ID: "
5278                         "domain: %s, name: %s, date: %llu,%u\n",
5279                         __func__, clp->cl_implid->domain, clp->cl_implid->name,
5280                         clp->cl_implid->date.seconds,
5281                         clp->cl_implid->date.nseconds);
5282         dprintk("<-- %s status= %d\n", __func__, status);
5283         return status;
5284 }
5285
5286 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5287                 struct rpc_cred *cred)
5288 {
5289         struct rpc_message msg = {
5290                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5291                 .rpc_argp = clp,
5292                 .rpc_cred = cred,
5293         };
5294         int status;
5295
5296         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5297         if (status)
5298                 pr_warn("NFS: Got error %d from the server %s on "
5299                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5300         return status;
5301 }
5302
5303 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5304                 struct rpc_cred *cred)
5305 {
5306         unsigned int loop;
5307         int ret;
5308
5309         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5310                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5311                 switch (ret) {
5312                 case -NFS4ERR_DELAY:
5313                 case -NFS4ERR_CLIENTID_BUSY:
5314                         ssleep(1);
5315                         break;
5316                 default:
5317                         return ret;
5318                 }
5319         }
5320         return 0;
5321 }
5322
5323 int nfs4_destroy_clientid(struct nfs_client *clp)
5324 {
5325         struct rpc_cred *cred;
5326         int ret = 0;
5327
5328         if (clp->cl_mvops->minor_version < 1)
5329                 goto out;
5330         if (clp->cl_exchange_flags == 0)
5331                 goto out;
5332         cred = nfs4_get_exchange_id_cred(clp);
5333         ret = nfs4_proc_destroy_clientid(clp, cred);
5334         if (cred)
5335                 put_rpccred(cred);
5336         switch (ret) {
5337         case 0:
5338         case -NFS4ERR_STALE_CLIENTID:
5339                 clp->cl_exchange_flags = 0;
5340         }
5341 out:
5342         return ret;
5343 }
5344
5345 struct nfs4_get_lease_time_data {
5346         struct nfs4_get_lease_time_args *args;
5347         struct nfs4_get_lease_time_res *res;
5348         struct nfs_client *clp;
5349 };
5350
5351 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5352                                         void *calldata)
5353 {
5354         int ret;
5355         struct nfs4_get_lease_time_data *data =
5356                         (struct nfs4_get_lease_time_data *)calldata;
5357
5358         dprintk("--> %s\n", __func__);
5359         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5360         /* just setup sequence, do not trigger session recovery
5361            since we're invoked within one */
5362         ret = nfs41_setup_sequence(data->clp->cl_session,
5363                                    &data->args->la_seq_args,
5364                                    &data->res->lr_seq_res, task);
5365
5366         BUG_ON(ret == -EAGAIN);
5367         rpc_call_start(task);
5368         dprintk("<-- %s\n", __func__);
5369 }
5370
5371 /*
5372  * Called from nfs4_state_manager thread for session setup, so don't recover
5373  * from sequence operation or clientid errors.
5374  */
5375 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5376 {
5377         struct nfs4_get_lease_time_data *data =
5378                         (struct nfs4_get_lease_time_data *)calldata;
5379
5380         dprintk("--> %s\n", __func__);
5381         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5382                 return;
5383         switch (task->tk_status) {
5384         case -NFS4ERR_DELAY:
5385         case -NFS4ERR_GRACE:
5386                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5387                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5388                 task->tk_status = 0;
5389                 /* fall through */
5390         case -NFS4ERR_RETRY_UNCACHED_REP:
5391                 rpc_restart_call_prepare(task);
5392                 return;
5393         }
5394         dprintk("<-- %s\n", __func__);
5395 }
5396
5397 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5398         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5399         .rpc_call_done = nfs4_get_lease_time_done,
5400 };
5401
5402 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5403 {
5404         struct rpc_task *task;
5405         struct nfs4_get_lease_time_args args;
5406         struct nfs4_get_lease_time_res res = {
5407                 .lr_fsinfo = fsinfo,
5408         };
5409         struct nfs4_get_lease_time_data data = {
5410                 .args = &args,
5411                 .res = &res,
5412                 .clp = clp,
5413         };
5414         struct rpc_message msg = {
5415                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5416                 .rpc_argp = &args,
5417                 .rpc_resp = &res,
5418         };
5419         struct rpc_task_setup task_setup = {
5420                 .rpc_client = clp->cl_rpcclient,
5421                 .rpc_message = &msg,
5422                 .callback_ops = &nfs4_get_lease_time_ops,
5423                 .callback_data = &data,
5424                 .flags = RPC_TASK_TIMEOUT,
5425         };
5426         int status;
5427
5428         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5429         dprintk("--> %s\n", __func__);
5430         task = rpc_run_task(&task_setup);
5431
5432         if (IS_ERR(task))
5433                 status = PTR_ERR(task);
5434         else {
5435                 status = task->tk_status;
5436                 rpc_put_task(task);
5437         }
5438         dprintk("<-- %s return %d\n", __func__, status);
5439
5440         return status;
5441 }
5442
5443 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5444 {
5445         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5446 }
5447
5448 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5449                 struct nfs4_slot *new,
5450                 u32 max_slots,
5451                 u32 ivalue)
5452 {
5453         struct nfs4_slot *old = NULL;
5454         u32 i;
5455
5456         spin_lock(&tbl->slot_tbl_lock);
5457         if (new) {
5458                 old = tbl->slots;
5459                 tbl->slots = new;
5460                 tbl->max_slots = max_slots;
5461         }
5462         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5463         for (i = 0; i < tbl->max_slots; i++)
5464                 tbl->slots[i].seq_nr = ivalue;
5465         spin_unlock(&tbl->slot_tbl_lock);
5466         kfree(old);
5467 }
5468
5469 /*
5470  * (re)Initialise a slot table
5471  */
5472 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5473                                  u32 ivalue)
5474 {
5475         struct nfs4_slot *new = NULL;
5476         int ret = -ENOMEM;
5477
5478         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5479                 max_reqs, tbl->max_slots);
5480
5481         /* Does the newly negotiated max_reqs match the existing slot table? */
5482         if (max_reqs != tbl->max_slots) {
5483                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5484                 if (!new)
5485                         goto out;
5486         }
5487         ret = 0;
5488
5489         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5490         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5491                 tbl, tbl->slots, tbl->max_slots);
5492 out:
5493         dprintk("<-- %s: return %d\n", __func__, ret);
5494         return ret;
5495 }
5496
5497 /* Destroy the slot table */
5498 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5499 {
5500         if (session->fc_slot_table.slots != NULL) {
5501                 kfree(session->fc_slot_table.slots);
5502                 session->fc_slot_table.slots = NULL;
5503         }
5504         if (session->bc_slot_table.slots != NULL) {
5505                 kfree(session->bc_slot_table.slots);
5506                 session->bc_slot_table.slots = NULL;
5507         }
5508         return;
5509 }
5510
5511 /*
5512  * Initialize or reset the forechannel and backchannel tables
5513  */
5514 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5515 {
5516         struct nfs4_slot_table *tbl;
5517         int status;
5518
5519         dprintk("--> %s\n", __func__);
5520         /* Fore channel */
5521         tbl = &ses->fc_slot_table;
5522         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5523         if (status) /* -ENOMEM */
5524                 return status;
5525         /* Back channel */
5526         tbl = &ses->bc_slot_table;
5527         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5528         if (status && tbl->slots == NULL)
5529                 /* Fore and back channel share a connection so get
5530                  * both slot tables or neither */
5531                 nfs4_destroy_slot_tables(ses);
5532         return status;
5533 }
5534
5535 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5536 {
5537         struct nfs4_session *session;
5538         struct nfs4_slot_table *tbl;
5539
5540         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5541         if (!session)
5542                 return NULL;
5543
5544         tbl = &session->fc_slot_table;
5545         tbl->highest_used_slotid = NFS4_NO_SLOT;
5546         spin_lock_init(&tbl->slot_tbl_lock);
5547         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5548         init_completion(&tbl->complete);
5549
5550         tbl = &session->bc_slot_table;
5551         tbl->highest_used_slotid = NFS4_NO_SLOT;
5552         spin_lock_init(&tbl->slot_tbl_lock);
5553         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5554         init_completion(&tbl->complete);
5555
5556         session->session_state = 1<<NFS4_SESSION_INITING;
5557
5558         session->clp = clp;
5559         return session;
5560 }
5561
5562 void nfs4_destroy_session(struct nfs4_session *session)
5563 {
5564         struct rpc_xprt *xprt;
5565         struct rpc_cred *cred;
5566
5567         cred = nfs4_get_exchange_id_cred(session->clp);
5568         nfs4_proc_destroy_session(session, cred);
5569         if (cred)
5570                 put_rpccred(cred);
5571
5572         rcu_read_lock();
5573         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5574         rcu_read_unlock();
5575         dprintk("%s Destroy backchannel for xprt %p\n",
5576                 __func__, xprt);
5577         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5578         nfs4_destroy_slot_tables(session);
5579         kfree(session);
5580 }
5581
5582 /*
5583  * Initialize the values to be used by the client in CREATE_SESSION
5584  * If nfs4_init_session set the fore channel request and response sizes,
5585  * use them.
5586  *
5587  * Set the back channel max_resp_sz_cached to zero to force the client to
5588  * always set csa_cachethis to FALSE because the current implementation
5589  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5590  */
5591 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5592 {
5593         struct nfs4_session *session = args->client->cl_session;
5594         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5595                      mxresp_sz = session->fc_attrs.max_resp_sz;
5596
5597         if (mxrqst_sz == 0)
5598                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5599         if (mxresp_sz == 0)
5600                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5601         /* Fore channel attributes */
5602         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5603         args->fc_attrs.max_resp_sz = mxresp_sz;
5604         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5605         args->fc_attrs.max_reqs = max_session_slots;
5606
5607         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5608                 "max_ops=%u max_reqs=%u\n",
5609                 __func__,
5610                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5611                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5612
5613         /* Back channel attributes */
5614         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5615         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5616         args->bc_attrs.max_resp_sz_cached = 0;
5617         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5618         args->bc_attrs.max_reqs = 1;
5619
5620         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5621                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5622                 __func__,
5623                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5624                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5625                 args->bc_attrs.max_reqs);
5626 }
5627
5628 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5629 {
5630         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5631         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5632
5633         if (rcvd->max_resp_sz > sent->max_resp_sz)
5634                 return -EINVAL;
5635         /*
5636          * Our requested max_ops is the minimum we need; we're not
5637          * prepared to break up compounds into smaller pieces than that.
5638          * So, no point even trying to continue if the server won't
5639          * cooperate:
5640          */
5641         if (rcvd->max_ops < sent->max_ops)
5642                 return -EINVAL;
5643         if (rcvd->max_reqs == 0)
5644                 return -EINVAL;
5645         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5646                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5647         return 0;
5648 }
5649
5650 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5651 {
5652         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5653         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5654
5655         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5656                 return -EINVAL;
5657         if (rcvd->max_resp_sz < sent->max_resp_sz)
5658                 return -EINVAL;
5659         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5660                 return -EINVAL;
5661         /* These would render the backchannel useless: */
5662         if (rcvd->max_ops != sent->max_ops)
5663                 return -EINVAL;
5664         if (rcvd->max_reqs != sent->max_reqs)
5665                 return -EINVAL;
5666         return 0;
5667 }
5668
5669 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5670                                      struct nfs4_session *session)
5671 {
5672         int ret;
5673
5674         ret = nfs4_verify_fore_channel_attrs(args, session);
5675         if (ret)
5676                 return ret;
5677         return nfs4_verify_back_channel_attrs(args, session);
5678 }
5679
5680 static int _nfs4_proc_create_session(struct nfs_client *clp,
5681                 struct rpc_cred *cred)
5682 {
5683         struct nfs4_session *session = clp->cl_session;
5684         struct nfs41_create_session_args args = {
5685                 .client = clp,
5686                 .cb_program = NFS4_CALLBACK,
5687         };
5688         struct nfs41_create_session_res res = {
5689                 .client = clp,
5690         };
5691         struct rpc_message msg = {
5692                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5693                 .rpc_argp = &args,
5694                 .rpc_resp = &res,
5695                 .rpc_cred = cred,
5696         };
5697         int status;
5698
5699         nfs4_init_channel_attrs(&args);
5700         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5701
5702         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5703
5704         if (!status)
5705                 /* Verify the session's negotiated channel_attrs values */
5706                 status = nfs4_verify_channel_attrs(&args, session);
5707         if (!status) {
5708                 /* Increment the clientid slot sequence id */
5709                 clp->cl_seqid++;
5710         }
5711
5712         return status;
5713 }
5714
5715 /*
5716  * Issues a CREATE_SESSION operation to the server.
5717  * It is the responsibility of the caller to verify the session is
5718  * expired before calling this routine.
5719  */
5720 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5721 {
5722         int status;
5723         unsigned *ptr;
5724         struct nfs4_session *session = clp->cl_session;
5725
5726         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5727
5728         status = _nfs4_proc_create_session(clp, cred);
5729         if (status)
5730                 goto out;
5731
5732         /* Init or reset the session slot tables */
5733         status = nfs4_setup_session_slot_tables(session);
5734         dprintk("slot table setup returned %d\n", status);
5735         if (status)
5736                 goto out;
5737
5738         ptr = (unsigned *)&session->sess_id.data[0];
5739         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5740                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5741 out:
5742         dprintk("<-- %s\n", __func__);
5743         return status;
5744 }
5745
5746 /*
5747  * Issue the over-the-wire RPC DESTROY_SESSION.
5748  * The caller must serialize access to this routine.
5749  */
5750 int nfs4_proc_destroy_session(struct nfs4_session *session,
5751                 struct rpc_cred *cred)
5752 {
5753         struct rpc_message msg = {
5754                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5755                 .rpc_argp = session,
5756                 .rpc_cred = cred,
5757         };
5758         int status = 0;
5759
5760         dprintk("--> nfs4_proc_destroy_session\n");
5761
5762         /* session is still being setup */
5763         if (session->clp->cl_cons_state != NFS_CS_READY)
5764                 return status;
5765
5766         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5767
5768         if (status)
5769                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5770                         "Session has been destroyed regardless...\n", status);
5771
5772         dprintk("<-- nfs4_proc_destroy_session\n");
5773         return status;
5774 }
5775
5776 /*
5777  * With sessions, the client is not marked ready until after a
5778  * successful EXCHANGE_ID and CREATE_SESSION.
5779  *
5780  * Map errors cl_cons_state errors to EPROTONOSUPPORT to indicate
5781  * other versions of NFS can be tried.
5782  */
5783 static int nfs41_check_session_ready(struct nfs_client *clp)
5784 {
5785         int ret;
5786
5787         if (clp->cl_cons_state == NFS_CS_SESSION_INITING) {
5788                 ret = nfs4_client_recover_expired_lease(clp);
5789                 if (ret)
5790                         return ret;
5791         }
5792         if (clp->cl_cons_state < NFS_CS_READY)
5793                 return -EPROTONOSUPPORT;
5794         smp_rmb();
5795         return 0;
5796 }
5797
5798 int nfs4_init_session(struct nfs_server *server)
5799 {
5800         struct nfs_client *clp = server->nfs_client;
5801         struct nfs4_session *session;
5802         unsigned int rsize, wsize;
5803
5804         if (!nfs4_has_session(clp))
5805                 return 0;
5806
5807         session = clp->cl_session;
5808         spin_lock(&clp->cl_lock);
5809         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
5810
5811                 rsize = server->rsize;
5812                 if (rsize == 0)
5813                         rsize = NFS_MAX_FILE_IO_SIZE;
5814                 wsize = server->wsize;
5815                 if (wsize == 0)
5816                         wsize = NFS_MAX_FILE_IO_SIZE;
5817
5818                 session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5819                 session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5820         }
5821         spin_unlock(&clp->cl_lock);
5822
5823         return nfs41_check_session_ready(clp);
5824 }
5825
5826 int nfs4_init_ds_session(struct nfs_client *clp, unsigned long lease_time)
5827 {
5828         struct nfs4_session *session = clp->cl_session;
5829         int ret;
5830
5831         spin_lock(&clp->cl_lock);
5832         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
5833                 /*
5834                  * Do not set NFS_CS_CHECK_LEASE_TIME instead set the
5835                  * DS lease to be equal to the MDS lease.
5836                  */
5837                 clp->cl_lease_time = lease_time;
5838                 clp->cl_last_renewal = jiffies;
5839         }
5840         spin_unlock(&clp->cl_lock);
5841
5842         ret = nfs41_check_session_ready(clp);
5843         if (ret)
5844                 return ret;
5845         /* Test for the DS role */
5846         if (!is_ds_client(clp))
5847                 return -ENODEV;
5848         return 0;
5849 }
5850 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5851
5852
5853 /*
5854  * Renew the cl_session lease.
5855  */
5856 struct nfs4_sequence_data {
5857         struct nfs_client *clp;
5858         struct nfs4_sequence_args args;
5859         struct nfs4_sequence_res res;
5860 };
5861
5862 static void nfs41_sequence_release(void *data)
5863 {
5864         struct nfs4_sequence_data *calldata = data;
5865         struct nfs_client *clp = calldata->clp;
5866
5867         if (atomic_read(&clp->cl_count) > 1)
5868                 nfs4_schedule_state_renewal(clp);
5869         nfs_put_client(clp);
5870         kfree(calldata);
5871 }
5872
5873 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5874 {
5875         switch(task->tk_status) {
5876         case -NFS4ERR_DELAY:
5877                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5878                 return -EAGAIN;
5879         default:
5880                 nfs4_schedule_lease_recovery(clp);
5881         }
5882         return 0;
5883 }
5884
5885 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5886 {
5887         struct nfs4_sequence_data *calldata = data;
5888         struct nfs_client *clp = calldata->clp;
5889
5890         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5891                 return;
5892
5893         if (task->tk_status < 0) {
5894                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5895                 if (atomic_read(&clp->cl_count) == 1)
5896                         goto out;
5897
5898                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5899                         rpc_restart_call_prepare(task);
5900                         return;
5901                 }
5902         }
5903         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5904 out:
5905         dprintk("<-- %s\n", __func__);
5906 }
5907
5908 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5909 {
5910         struct nfs4_sequence_data *calldata = data;
5911         struct nfs_client *clp = calldata->clp;
5912         struct nfs4_sequence_args *args;
5913         struct nfs4_sequence_res *res;
5914
5915         args = task->tk_msg.rpc_argp;
5916         res = task->tk_msg.rpc_resp;
5917
5918         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5919                 return;
5920         rpc_call_start(task);
5921 }
5922
5923 static const struct rpc_call_ops nfs41_sequence_ops = {
5924         .rpc_call_done = nfs41_sequence_call_done,
5925         .rpc_call_prepare = nfs41_sequence_prepare,
5926         .rpc_release = nfs41_sequence_release,
5927 };
5928
5929 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5930 {
5931         struct nfs4_sequence_data *calldata;
5932         struct rpc_message msg = {
5933                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5934                 .rpc_cred = cred,
5935         };
5936         struct rpc_task_setup task_setup_data = {
5937                 .rpc_client = clp->cl_rpcclient,
5938                 .rpc_message = &msg,
5939                 .callback_ops = &nfs41_sequence_ops,
5940                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5941         };
5942
5943         if (!atomic_inc_not_zero(&clp->cl_count))
5944                 return ERR_PTR(-EIO);
5945         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5946         if (calldata == NULL) {
5947                 nfs_put_client(clp);
5948                 return ERR_PTR(-ENOMEM);
5949         }
5950         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5951         msg.rpc_argp = &calldata->args;
5952         msg.rpc_resp = &calldata->res;
5953         calldata->clp = clp;
5954         task_setup_data.callback_data = calldata;
5955
5956         return rpc_run_task(&task_setup_data);
5957 }
5958
5959 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5960 {
5961         struct rpc_task *task;
5962         int ret = 0;
5963
5964         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5965                 return 0;
5966         task = _nfs41_proc_sequence(clp, cred);
5967         if (IS_ERR(task))
5968                 ret = PTR_ERR(task);
5969         else
5970                 rpc_put_task_async(task);
5971         dprintk("<-- %s status=%d\n", __func__, ret);
5972         return ret;
5973 }
5974
5975 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5976 {
5977         struct rpc_task *task;
5978         int ret;
5979
5980         task = _nfs41_proc_sequence(clp, cred);
5981         if (IS_ERR(task)) {
5982                 ret = PTR_ERR(task);
5983                 goto out;
5984         }
5985         ret = rpc_wait_for_completion_task(task);
5986         if (!ret) {
5987                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5988
5989                 if (task->tk_status == 0)
5990                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5991                 ret = task->tk_status;
5992         }
5993         rpc_put_task(task);
5994 out:
5995         dprintk("<-- %s status=%d\n", __func__, ret);
5996         return ret;
5997 }
5998
5999 struct nfs4_reclaim_complete_data {
6000         struct nfs_client *clp;
6001         struct nfs41_reclaim_complete_args arg;
6002         struct nfs41_reclaim_complete_res res;
6003 };
6004
6005 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6006 {
6007         struct nfs4_reclaim_complete_data *calldata = data;
6008
6009         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
6010         if (nfs41_setup_sequence(calldata->clp->cl_session,
6011                                 &calldata->arg.seq_args,
6012                                 &calldata->res.seq_res, task))
6013                 return;
6014
6015         rpc_call_start(task);
6016 }
6017
6018 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6019 {
6020         switch(task->tk_status) {
6021         case 0:
6022         case -NFS4ERR_COMPLETE_ALREADY:
6023         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6024                 break;
6025         case -NFS4ERR_DELAY:
6026                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6027                 /* fall through */
6028         case -NFS4ERR_RETRY_UNCACHED_REP:
6029                 return -EAGAIN;
6030         default:
6031                 nfs4_schedule_lease_recovery(clp);
6032         }
6033         return 0;
6034 }
6035
6036 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6037 {
6038         struct nfs4_reclaim_complete_data *calldata = data;
6039         struct nfs_client *clp = calldata->clp;
6040         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6041
6042         dprintk("--> %s\n", __func__);
6043         if (!nfs41_sequence_done(task, res))
6044                 return;
6045
6046         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6047                 rpc_restart_call_prepare(task);
6048                 return;
6049         }
6050         dprintk("<-- %s\n", __func__);
6051 }
6052
6053 static void nfs4_free_reclaim_complete_data(void *data)
6054 {
6055         struct nfs4_reclaim_complete_data *calldata = data;
6056
6057         kfree(calldata);
6058 }
6059
6060 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6061         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6062         .rpc_call_done = nfs4_reclaim_complete_done,
6063         .rpc_release = nfs4_free_reclaim_complete_data,
6064 };
6065
6066 /*
6067  * Issue a global reclaim complete.
6068  */
6069 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6070 {
6071         struct nfs4_reclaim_complete_data *calldata;
6072         struct rpc_task *task;
6073         struct rpc_message msg = {
6074                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6075         };
6076         struct rpc_task_setup task_setup_data = {
6077                 .rpc_client = clp->cl_rpcclient,
6078                 .rpc_message = &msg,
6079                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6080                 .flags = RPC_TASK_ASYNC,
6081         };
6082         int status = -ENOMEM;
6083
6084         dprintk("--> %s\n", __func__);
6085         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6086         if (calldata == NULL)
6087                 goto out;
6088         calldata->clp = clp;
6089         calldata->arg.one_fs = 0;
6090
6091         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6092         msg.rpc_argp = &calldata->arg;
6093         msg.rpc_resp = &calldata->res;
6094         task_setup_data.callback_data = calldata;
6095         task = rpc_run_task(&task_setup_data);
6096         if (IS_ERR(task)) {
6097                 status = PTR_ERR(task);
6098                 goto out;
6099         }
6100         status = nfs4_wait_for_completion_rpc_task(task);
6101         if (status == 0)
6102                 status = task->tk_status;
6103         rpc_put_task(task);
6104         return 0;
6105 out:
6106         dprintk("<-- %s status=%d\n", __func__, status);
6107         return status;
6108 }
6109
6110 static void
6111 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6112 {
6113         struct nfs4_layoutget *lgp = calldata;
6114         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6115
6116         dprintk("--> %s\n", __func__);
6117         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6118          * right now covering the LAYOUTGET we are about to send.
6119          * However, that is not so catastrophic, and there seems
6120          * to be no way to prevent it completely.
6121          */
6122         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
6123                                 &lgp->res.seq_res, task))
6124                 return;
6125         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6126                                           NFS_I(lgp->args.inode)->layout,
6127                                           lgp->args.ctx->state)) {
6128                 rpc_exit(task, NFS4_OK);
6129                 return;
6130         }
6131         rpc_call_start(task);
6132 }
6133
6134 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6135 {
6136         struct nfs4_layoutget *lgp = calldata;
6137         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6138
6139         dprintk("--> %s\n", __func__);
6140
6141         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
6142                 return;
6143
6144         switch (task->tk_status) {
6145         case 0:
6146                 break;
6147         case -NFS4ERR_LAYOUTTRYLATER:
6148         case -NFS4ERR_RECALLCONFLICT:
6149                 task->tk_status = -NFS4ERR_DELAY;
6150                 /* Fall through */
6151         default:
6152                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6153                         rpc_restart_call_prepare(task);
6154                         return;
6155                 }
6156         }
6157         dprintk("<-- %s\n", __func__);
6158 }
6159
6160 static void nfs4_layoutget_release(void *calldata)
6161 {
6162         struct nfs4_layoutget *lgp = calldata;
6163
6164         dprintk("--> %s\n", __func__);
6165         put_nfs_open_context(lgp->args.ctx);
6166         kfree(calldata);
6167         dprintk("<-- %s\n", __func__);
6168 }
6169
6170 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6171         .rpc_call_prepare = nfs4_layoutget_prepare,
6172         .rpc_call_done = nfs4_layoutget_done,
6173         .rpc_release = nfs4_layoutget_release,
6174 };
6175
6176 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
6177 {
6178         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6179         struct rpc_task *task;
6180         struct rpc_message msg = {
6181                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6182                 .rpc_argp = &lgp->args,
6183                 .rpc_resp = &lgp->res,
6184         };
6185         struct rpc_task_setup task_setup_data = {
6186                 .rpc_client = server->client,
6187                 .rpc_message = &msg,
6188                 .callback_ops = &nfs4_layoutget_call_ops,
6189                 .callback_data = lgp,
6190                 .flags = RPC_TASK_ASYNC,
6191         };
6192         int status = 0;
6193
6194         dprintk("--> %s\n", __func__);
6195
6196         lgp->res.layoutp = &lgp->args.layout;
6197         lgp->res.seq_res.sr_slot = NULL;
6198         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6199         task = rpc_run_task(&task_setup_data);
6200         if (IS_ERR(task))
6201                 return PTR_ERR(task);
6202         status = nfs4_wait_for_completion_rpc_task(task);
6203         if (status == 0)
6204                 status = task->tk_status;
6205         if (status == 0)
6206                 status = pnfs_layout_process(lgp);
6207         rpc_put_task(task);
6208         dprintk("<-- %s status=%d\n", __func__, status);
6209         return status;
6210 }
6211
6212 static void
6213 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6214 {
6215         struct nfs4_layoutreturn *lrp = calldata;
6216
6217         dprintk("--> %s\n", __func__);
6218         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
6219                                 &lrp->res.seq_res, task))
6220                 return;
6221         rpc_call_start(task);
6222 }
6223
6224 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6225 {
6226         struct nfs4_layoutreturn *lrp = calldata;
6227         struct nfs_server *server;
6228         struct pnfs_layout_hdr *lo = lrp->args.layout;
6229
6230         dprintk("--> %s\n", __func__);
6231
6232         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
6233                 return;
6234
6235         server = NFS_SERVER(lrp->args.inode);
6236         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6237                 rpc_restart_call_prepare(task);
6238                 return;
6239         }
6240         spin_lock(&lo->plh_inode->i_lock);
6241         if (task->tk_status == 0) {
6242                 if (lrp->res.lrs_present) {
6243                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6244                 } else
6245                         BUG_ON(!list_empty(&lo->plh_segs));
6246         }
6247         lo->plh_block_lgets--;
6248         spin_unlock(&lo->plh_inode->i_lock);
6249         dprintk("<-- %s\n", __func__);
6250 }
6251
6252 static void nfs4_layoutreturn_release(void *calldata)
6253 {
6254         struct nfs4_layoutreturn *lrp = calldata;
6255
6256         dprintk("--> %s\n", __func__);
6257         put_layout_hdr(lrp->args.layout);
6258         kfree(calldata);
6259         dprintk("<-- %s\n", __func__);
6260 }
6261
6262 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6263         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6264         .rpc_call_done = nfs4_layoutreturn_done,
6265         .rpc_release = nfs4_layoutreturn_release,
6266 };
6267
6268 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6269 {
6270         struct rpc_task *task;
6271         struct rpc_message msg = {
6272                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6273                 .rpc_argp = &lrp->args,
6274                 .rpc_resp = &lrp->res,
6275         };
6276         struct rpc_task_setup task_setup_data = {
6277                 .rpc_client = lrp->clp->cl_rpcclient,
6278                 .rpc_message = &msg,
6279                 .callback_ops = &nfs4_layoutreturn_call_ops,
6280                 .callback_data = lrp,
6281         };
6282         int status;
6283
6284         dprintk("--> %s\n", __func__);
6285         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6286         task = rpc_run_task(&task_setup_data);
6287         if (IS_ERR(task))
6288                 return PTR_ERR(task);
6289         status = task->tk_status;
6290         dprintk("<-- %s status=%d\n", __func__, status);
6291         rpc_put_task(task);
6292         return status;
6293 }
6294
6295 /*
6296  * Retrieve the list of Data Server devices from the MDS.
6297  */
6298 static int _nfs4_getdevicelist(struct nfs_server *server,
6299                                     const struct nfs_fh *fh,
6300                                     struct pnfs_devicelist *devlist)
6301 {
6302         struct nfs4_getdevicelist_args args = {
6303                 .fh = fh,
6304                 .layoutclass = server->pnfs_curr_ld->id,
6305         };
6306         struct nfs4_getdevicelist_res res = {
6307                 .devlist = devlist,
6308         };
6309         struct rpc_message msg = {
6310                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6311                 .rpc_argp = &args,
6312                 .rpc_resp = &res,
6313         };
6314         int status;
6315
6316         dprintk("--> %s\n", __func__);
6317         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6318                                 &res.seq_res, 0);
6319         dprintk("<-- %s status=%d\n", __func__, status);
6320         return status;
6321 }
6322
6323 int nfs4_proc_getdevicelist(struct nfs_server *server,
6324                             const struct nfs_fh *fh,
6325                             struct pnfs_devicelist *devlist)
6326 {
6327         struct nfs4_exception exception = { };
6328         int err;
6329
6330         do {
6331                 err = nfs4_handle_exception(server,
6332                                 _nfs4_getdevicelist(server, fh, devlist),
6333                                 &exception);
6334         } while (exception.retry);
6335
6336         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6337                 err, devlist->num_devs);
6338
6339         return err;
6340 }
6341 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6342
6343 static int
6344 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6345 {
6346         struct nfs4_getdeviceinfo_args args = {
6347                 .pdev = pdev,
6348         };
6349         struct nfs4_getdeviceinfo_res res = {
6350                 .pdev = pdev,
6351         };
6352         struct rpc_message msg = {
6353                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6354                 .rpc_argp = &args,
6355                 .rpc_resp = &res,
6356         };
6357         int status;
6358
6359         dprintk("--> %s\n", __func__);
6360         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6361         dprintk("<-- %s status=%d\n", __func__, status);
6362
6363         return status;
6364 }
6365
6366 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6367 {
6368         struct nfs4_exception exception = { };
6369         int err;
6370
6371         do {
6372                 err = nfs4_handle_exception(server,
6373                                         _nfs4_proc_getdeviceinfo(server, pdev),
6374                                         &exception);
6375         } while (exception.retry);
6376         return err;
6377 }
6378 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6379
6380 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6381 {
6382         struct nfs4_layoutcommit_data *data = calldata;
6383         struct nfs_server *server = NFS_SERVER(data->args.inode);
6384
6385         if (nfs4_setup_sequence(server, &data->args.seq_args,
6386                                 &data->res.seq_res, task))
6387                 return;
6388         rpc_call_start(task);
6389 }
6390
6391 static void
6392 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6393 {
6394         struct nfs4_layoutcommit_data *data = calldata;
6395         struct nfs_server *server = NFS_SERVER(data->args.inode);
6396
6397         if (!nfs4_sequence_done(task, &data->res.seq_res))
6398                 return;
6399
6400         switch (task->tk_status) { /* Just ignore these failures */
6401         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6402         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6403         case -NFS4ERR_BADLAYOUT:     /* no layout */
6404         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6405                 task->tk_status = 0;
6406                 break;
6407         case 0:
6408                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6409                                                    data->res.fattr);
6410                 break;
6411         default:
6412                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6413                         rpc_restart_call_prepare(task);
6414                         return;
6415                 }
6416         }
6417 }
6418
6419 static void nfs4_layoutcommit_release(void *calldata)
6420 {
6421         struct nfs4_layoutcommit_data *data = calldata;
6422         struct pnfs_layout_segment *lseg, *tmp;
6423         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6424
6425         pnfs_cleanup_layoutcommit(data);
6426         /* Matched by references in pnfs_set_layoutcommit */
6427         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6428                 list_del_init(&lseg->pls_lc_list);
6429                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6430                                        &lseg->pls_flags))
6431                         put_lseg(lseg);
6432         }
6433
6434         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6435         smp_mb__after_clear_bit();
6436         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6437
6438         put_rpccred(data->cred);
6439         kfree(data);
6440 }
6441
6442 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6443         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6444         .rpc_call_done = nfs4_layoutcommit_done,
6445         .rpc_release = nfs4_layoutcommit_release,
6446 };
6447
6448 int
6449 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6450 {
6451         struct rpc_message msg = {
6452                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6453                 .rpc_argp = &data->args,
6454                 .rpc_resp = &data->res,
6455                 .rpc_cred = data->cred,
6456         };
6457         struct rpc_task_setup task_setup_data = {
6458                 .task = &data->task,
6459                 .rpc_client = NFS_CLIENT(data->args.inode),
6460                 .rpc_message = &msg,
6461                 .callback_ops = &nfs4_layoutcommit_ops,
6462                 .callback_data = data,
6463                 .flags = RPC_TASK_ASYNC,
6464         };
6465         struct rpc_task *task;
6466         int status = 0;
6467
6468         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6469                 "lbw: %llu inode %lu\n",
6470                 data->task.tk_pid, sync,
6471                 data->args.lastbytewritten,
6472                 data->args.inode->i_ino);
6473
6474         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6475         task = rpc_run_task(&task_setup_data);
6476         if (IS_ERR(task))
6477                 return PTR_ERR(task);
6478         if (sync == false)
6479                 goto out;
6480         status = nfs4_wait_for_completion_rpc_task(task);
6481         if (status != 0)
6482                 goto out;
6483         status = task->tk_status;
6484 out:
6485         dprintk("%s: status %d\n", __func__, status);
6486         rpc_put_task(task);
6487         return status;
6488 }
6489
6490 static int
6491 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6492                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6493 {
6494         struct nfs41_secinfo_no_name_args args = {
6495                 .style = SECINFO_STYLE_CURRENT_FH,
6496         };
6497         struct nfs4_secinfo_res res = {
6498                 .flavors = flavors,
6499         };
6500         struct rpc_message msg = {
6501                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6502                 .rpc_argp = &args,
6503                 .rpc_resp = &res,
6504         };
6505         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6506 }
6507
6508 static int
6509 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6510                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6511 {
6512         struct nfs4_exception exception = { };
6513         int err;
6514         do {
6515                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6516                 switch (err) {
6517                 case 0:
6518                 case -NFS4ERR_WRONGSEC:
6519                 case -NFS4ERR_NOTSUPP:
6520                         goto out;
6521                 default:
6522                         err = nfs4_handle_exception(server, err, &exception);
6523                 }
6524         } while (exception.retry);
6525 out:
6526         return err;
6527 }
6528
6529 static int
6530 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6531                     struct nfs_fsinfo *info)
6532 {
6533         int err;
6534         struct page *page;
6535         rpc_authflavor_t flavor;
6536         struct nfs4_secinfo_flavors *flavors;
6537
6538         page = alloc_page(GFP_KERNEL);
6539         if (!page) {
6540                 err = -ENOMEM;
6541                 goto out;
6542         }
6543
6544         flavors = page_address(page);
6545         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6546
6547         /*
6548          * Fall back on "guess and check" method if
6549          * the server doesn't support SECINFO_NO_NAME
6550          */
6551         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6552                 err = nfs4_find_root_sec(server, fhandle, info);
6553                 goto out_freepage;
6554         }
6555         if (err)
6556                 goto out_freepage;
6557
6558         flavor = nfs_find_best_sec(flavors);
6559         if (err == 0)
6560                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6561
6562 out_freepage:
6563         put_page(page);
6564         if (err == -EACCES)
6565                 return -EPERM;
6566 out:
6567         return err;
6568 }
6569
6570 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6571 {
6572         int status;
6573         struct nfs41_test_stateid_args args = {
6574                 .stateid = stateid,
6575         };
6576         struct nfs41_test_stateid_res res;
6577         struct rpc_message msg = {
6578                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6579                 .rpc_argp = &args,
6580                 .rpc_resp = &res,
6581         };
6582
6583         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6584         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6585
6586         if (status == NFS_OK)
6587                 return res.status;
6588         return status;
6589 }
6590
6591 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6592 {
6593         struct nfs4_exception exception = { };
6594         int err;
6595         do {
6596                 err = nfs4_handle_exception(server,
6597                                 _nfs41_test_stateid(server, stateid),
6598                                 &exception);
6599         } while (exception.retry);
6600         return err;
6601 }
6602
6603 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6604 {
6605         struct nfs41_free_stateid_args args = {
6606                 .stateid = stateid,
6607         };
6608         struct nfs41_free_stateid_res res;
6609         struct rpc_message msg = {
6610                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6611                 .rpc_argp = &args,
6612                 .rpc_resp = &res,
6613         };
6614
6615         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6616         return nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6617 }
6618
6619 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6620 {
6621         struct nfs4_exception exception = { };
6622         int err;
6623         do {
6624                 err = nfs4_handle_exception(server,
6625                                 _nfs4_free_stateid(server, stateid),
6626                                 &exception);
6627         } while (exception.retry);
6628         return err;
6629 }
6630
6631 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6632                 const nfs4_stateid *s2)
6633 {
6634         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6635                 return false;
6636
6637         if (s1->seqid == s2->seqid)
6638                 return true;
6639         if (s1->seqid == 0 || s2->seqid == 0)
6640                 return true;
6641
6642         return false;
6643 }
6644
6645 #endif /* CONFIG_NFS_V4_1 */
6646
6647 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6648                 const nfs4_stateid *s2)
6649 {
6650         return nfs4_stateid_match(s1, s2);
6651 }
6652
6653
6654 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6655         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6656         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6657         .recover_open   = nfs4_open_reclaim,
6658         .recover_lock   = nfs4_lock_reclaim,
6659         .establish_clid = nfs4_init_clientid,
6660         .get_clid_cred  = nfs4_get_setclientid_cred,
6661 };
6662
6663 #if defined(CONFIG_NFS_V4_1)
6664 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6665         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6666         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6667         .recover_open   = nfs4_open_reclaim,
6668         .recover_lock   = nfs4_lock_reclaim,
6669         .establish_clid = nfs41_init_clientid,
6670         .get_clid_cred  = nfs4_get_exchange_id_cred,
6671         .reclaim_complete = nfs41_proc_reclaim_complete,
6672 };
6673 #endif /* CONFIG_NFS_V4_1 */
6674
6675 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6676         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6677         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6678         .recover_open   = nfs4_open_expired,
6679         .recover_lock   = nfs4_lock_expired,
6680         .establish_clid = nfs4_init_clientid,
6681         .get_clid_cred  = nfs4_get_setclientid_cred,
6682 };
6683
6684 #if defined(CONFIG_NFS_V4_1)
6685 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6686         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6687         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6688         .recover_open   = nfs41_open_expired,
6689         .recover_lock   = nfs41_lock_expired,
6690         .establish_clid = nfs41_init_clientid,
6691         .get_clid_cred  = nfs4_get_exchange_id_cred,
6692 };
6693 #endif /* CONFIG_NFS_V4_1 */
6694
6695 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6696         .sched_state_renewal = nfs4_proc_async_renew,
6697         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6698         .renew_lease = nfs4_proc_renew,
6699 };
6700
6701 #if defined(CONFIG_NFS_V4_1)
6702 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6703         .sched_state_renewal = nfs41_proc_async_sequence,
6704         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6705         .renew_lease = nfs4_proc_sequence,
6706 };
6707 #endif
6708
6709 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6710         .minor_version = 0,
6711         .call_sync = _nfs4_call_sync,
6712         .match_stateid = nfs4_match_stateid,
6713         .find_root_sec = nfs4_find_root_sec,
6714         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6715         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6716         .state_renewal_ops = &nfs40_state_renewal_ops,
6717 };
6718
6719 #if defined(CONFIG_NFS_V4_1)
6720 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6721         .minor_version = 1,
6722         .call_sync = _nfs4_call_sync_session,
6723         .match_stateid = nfs41_match_stateid,
6724         .find_root_sec = nfs41_find_root_sec,
6725         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6726         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6727         .state_renewal_ops = &nfs41_state_renewal_ops,
6728 };
6729 #endif
6730
6731 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6732         [0] = &nfs_v4_0_minor_ops,
6733 #if defined(CONFIG_NFS_V4_1)
6734         [1] = &nfs_v4_1_minor_ops,
6735 #endif
6736 };
6737
6738 static const struct inode_operations nfs4_file_inode_operations = {
6739         .permission     = nfs_permission,
6740         .getattr        = nfs_getattr,
6741         .setattr        = nfs_setattr,
6742         .getxattr       = generic_getxattr,
6743         .setxattr       = generic_setxattr,
6744         .listxattr      = generic_listxattr,
6745         .removexattr    = generic_removexattr,
6746 };
6747
6748 const struct nfs_rpc_ops nfs_v4_clientops = {
6749         .version        = 4,                    /* protocol version */
6750         .dentry_ops     = &nfs4_dentry_operations,
6751         .dir_inode_ops  = &nfs4_dir_inode_operations,
6752         .file_inode_ops = &nfs4_file_inode_operations,
6753         .file_ops       = &nfs4_file_operations,
6754         .getroot        = nfs4_proc_get_root,
6755         .submount       = nfs4_submount,
6756         .getattr        = nfs4_proc_getattr,
6757         .setattr        = nfs4_proc_setattr,
6758         .lookup         = nfs4_proc_lookup,
6759         .access         = nfs4_proc_access,
6760         .readlink       = nfs4_proc_readlink,
6761         .create         = nfs4_proc_create,
6762         .remove         = nfs4_proc_remove,
6763         .unlink_setup   = nfs4_proc_unlink_setup,
6764         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
6765         .unlink_done    = nfs4_proc_unlink_done,
6766         .rename         = nfs4_proc_rename,
6767         .rename_setup   = nfs4_proc_rename_setup,
6768         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
6769         .rename_done    = nfs4_proc_rename_done,
6770         .link           = nfs4_proc_link,
6771         .symlink        = nfs4_proc_symlink,
6772         .mkdir          = nfs4_proc_mkdir,
6773         .rmdir          = nfs4_proc_remove,
6774         .readdir        = nfs4_proc_readdir,
6775         .mknod          = nfs4_proc_mknod,
6776         .statfs         = nfs4_proc_statfs,
6777         .fsinfo         = nfs4_proc_fsinfo,
6778         .pathconf       = nfs4_proc_pathconf,
6779         .set_capabilities = nfs4_server_capabilities,
6780         .decode_dirent  = nfs4_decode_dirent,
6781         .read_setup     = nfs4_proc_read_setup,
6782         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
6783         .read_done      = nfs4_read_done,
6784         .write_setup    = nfs4_proc_write_setup,
6785         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
6786         .write_done     = nfs4_write_done,
6787         .commit_setup   = nfs4_proc_commit_setup,
6788         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
6789         .commit_done    = nfs4_commit_done,
6790         .lock           = nfs4_proc_lock,
6791         .clear_acl_cache = nfs4_zap_acl_attr,
6792         .close_context  = nfs4_close_context,
6793         .open_context   = nfs4_atomic_open,
6794         .init_client    = nfs4_init_client,
6795 };
6796
6797 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6798         .prefix = XATTR_NAME_NFSV4_ACL,
6799         .list   = nfs4_xattr_list_nfs4_acl,
6800         .get    = nfs4_xattr_get_nfs4_acl,
6801         .set    = nfs4_xattr_set_nfs4_acl,
6802 };
6803
6804 const struct xattr_handler *nfs4_xattr_handlers[] = {
6805         &nfs4_xattr_nfs4_acl_handler,
6806         NULL
6807 };
6808
6809 module_param(max_session_slots, ushort, 0644);
6810 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
6811                 "requests the client will negotiate");
6812
6813 /*
6814  * Local variables:
6815  *  c-basic-offset: 8
6816  * End:
6817  */