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         case -NFS4ERR_MINOR_VERS_MISMATCH:
 109                 return -EPROTONOSUPPORT;
 110         default:
 111                 dprintk("%s could not handle NFSv4 error %d\n",
 112                                 __func__, -err);
 113                 break;
 114         }
 115         return -EIO;
 116 }
 117
 118 /*
 119  * This is our standard bitmap for GETATTR requests.
 120  */
 121 const u32 nfs4_fattr_bitmap[3] = {
 122         FATTR4_WORD0_TYPE
 123         | FATTR4_WORD0_CHANGE
 124         | FATTR4_WORD0_SIZE
 125         | FATTR4_WORD0_FSID
 126         | FATTR4_WORD0_FILEID,
 127         FATTR4_WORD1_MODE
 128         | FATTR4_WORD1_NUMLINKS
 129         | FATTR4_WORD1_OWNER
 130         | FATTR4_WORD1_OWNER_GROUP
 131         | FATTR4_WORD1_RAWDEV
 132         | FATTR4_WORD1_SPACE_USED
 133         | FATTR4_WORD1_TIME_ACCESS
 134         | FATTR4_WORD1_TIME_METADATA
 135         | FATTR4_WORD1_TIME_MODIFY
 136 };
 137
 138 static const u32 nfs4_pnfs_open_bitmap[3] = {
 139         FATTR4_WORD0_TYPE
 140         | FATTR4_WORD0_CHANGE
 141         | FATTR4_WORD0_SIZE
 142         | FATTR4_WORD0_FSID
 143         | FATTR4_WORD0_FILEID,
 144         FATTR4_WORD1_MODE
 145         | FATTR4_WORD1_NUMLINKS
 146         | FATTR4_WORD1_OWNER
 147         | FATTR4_WORD1_OWNER_GROUP
 148         | FATTR4_WORD1_RAWDEV
 149         | FATTR4_WORD1_SPACE_USED
 150         | FATTR4_WORD1_TIME_ACCESS
 151         | FATTR4_WORD1_TIME_METADATA
 152         | FATTR4_WORD1_TIME_MODIFY,
 153         FATTR4_WORD2_MDSTHRESHOLD
 154 };
 155
 156 const u32 nfs4_statfs_bitmap[2] = {
 157         FATTR4_WORD0_FILES_AVAIL
 158         | FATTR4_WORD0_FILES_FREE
 159         | FATTR4_WORD0_FILES_TOTAL,
 160         FATTR4_WORD1_SPACE_AVAIL
 161         | FATTR4_WORD1_SPACE_FREE
 162         | FATTR4_WORD1_SPACE_TOTAL
 163 };
 164
 165 const u32 nfs4_pathconf_bitmap[2] = {
 166         FATTR4_WORD0_MAXLINK
 167         | FATTR4_WORD0_MAXNAME,
 168         0
 169 };
 170
 171 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 172                         | FATTR4_WORD0_MAXREAD
 173                         | FATTR4_WORD0_MAXWRITE
 174                         | FATTR4_WORD0_LEASE_TIME,
 175                         FATTR4_WORD1_TIME_DELTA
 176                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 177                         FATTR4_WORD2_LAYOUT_BLKSIZE
 178 };
 179
 180 const u32 nfs4_fs_locations_bitmap[2] = {
 181         FATTR4_WORD0_TYPE
 182         | FATTR4_WORD0_CHANGE
 183         | FATTR4_WORD0_SIZE
 184         | FATTR4_WORD0_FSID
 185         | FATTR4_WORD0_FILEID
 186         | FATTR4_WORD0_FS_LOCATIONS,
 187         FATTR4_WORD1_MODE
 188         | FATTR4_WORD1_NUMLINKS
 189         | FATTR4_WORD1_OWNER
 190         | FATTR4_WORD1_OWNER_GROUP
 191         | FATTR4_WORD1_RAWDEV
 192         | FATTR4_WORD1_SPACE_USED
 193         | FATTR4_WORD1_TIME_ACCESS
 194         | FATTR4_WORD1_TIME_METADATA
 195         | FATTR4_WORD1_TIME_MODIFY
 196         | FATTR4_WORD1_MOUNTED_ON_FILEID
 197 };
 198
 199 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 200                 struct nfs4_readdir_arg *readdir)
 201 {
 202         __be32 *start, *p;
 203
 204         BUG_ON(readdir->count < 80);
 205         if (cookie > 2) {
 206                 readdir->cookie = cookie;
 207                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 208                 return;
 209         }
 210
 211         readdir->cookie = 0;
 212         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 213         if (cookie == 2)
 214                 return;
 215
 216         /*
 217          * NFSv4 servers do not return entries for '.' and '..'
 218          * Therefore, we fake these entries here.  We let '.'
 219          * have cookie 0 and '..' have cookie 1.  Note that
 220          * when talking to the server, we always send cookie 0
 221          * instead of 1 or 2.
 222          */
 223         start = p = kmap_atomic(*readdir->pages);
 224
 225         if (cookie == 0) {
 226                 *p++ = xdr_one;                                  /* next */
 227                 *p++ = xdr_zero;                   /* cookie, first word */
 228                 *p++ = xdr_one;                   /* cookie, second word */
 229                 *p++ = xdr_one;                             /* entry len */
 230                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 231                 p++;
 232                 *p++ = xdr_one;                         /* bitmap length */
 233                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 234                 *p++ = htonl(8);              /* attribute buffer length */
 235                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 236         }
 237
 238         *p++ = xdr_one;                                  /* next */
 239         *p++ = xdr_zero;                   /* cookie, first word */
 240         *p++ = xdr_two;                   /* cookie, second word */
 241         *p++ = xdr_two;                             /* entry len */
 242         memcpy(p, "..\0\0", 4);                         /* entry */
 243         p++;
 244         *p++ = xdr_one;                         /* bitmap length */
 245         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 246         *p++ = htonl(8);              /* attribute buffer length */
 247         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 248
 249         readdir->pgbase = (char *)p - (char *)start;
 250         readdir->count -= readdir->pgbase;
 251         kunmap_atomic(start);
 252 }
 253
 254 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 255 {
 256         int res;
 257
 258         might_sleep();
 259
 260         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 261                         nfs_wait_bit_killable, TASK_KILLABLE);
 262         return res;
 263 }
 264
 265 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 266 {
 267         int res = 0;
 268
 269         might_sleep();
 270
 271         if (*timeout <= 0)
 272                 *timeout = NFS4_POLL_RETRY_MIN;
 273         if (*timeout > NFS4_POLL_RETRY_MAX)
 274                 *timeout = NFS4_POLL_RETRY_MAX;
 275         freezable_schedule_timeout_killable(*timeout);
 276         if (fatal_signal_pending(current))
 277                 res = -ERESTARTSYS;
 278         *timeout <<= 1;
 279         return res;
 280 }
 281
 282 /* This is the error handling routine for processes that are allowed
 283  * to sleep.
 284  */
 285 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 286 {
 287         struct nfs_client *clp = server->nfs_client;
 288         struct nfs4_state *state = exception->state;
 289         struct inode *inode = exception->inode;
 290         int ret = errorcode;
 291
 292         exception->retry = 0;
 293         switch(errorcode) {
 294                 case 0:
 295                         return 0;
 296                 case -NFS4ERR_OPENMODE:
 297                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 298                                 nfs4_inode_return_delegation(inode);
 299                                 exception->retry = 1;
 300                                 return 0;
 301                         }
 302                         if (state == NULL)
 303                                 break;
 304                         nfs4_schedule_stateid_recovery(server, state);
 305                         goto wait_on_recovery;
 306                 case -NFS4ERR_DELEG_REVOKED:
 307                 case -NFS4ERR_ADMIN_REVOKED:
 308                 case -NFS4ERR_BAD_STATEID:
 309                         if (state == NULL)
 310                                 break;
 311                         nfs_remove_bad_delegation(state->inode);
 312                         nfs4_schedule_stateid_recovery(server, state);
 313                         goto wait_on_recovery;
 314                 case -NFS4ERR_EXPIRED:
 315                         if (state != NULL)
 316                                 nfs4_schedule_stateid_recovery(server, state);
 317                 case -NFS4ERR_STALE_STATEID:
 318                 case -NFS4ERR_STALE_CLIENTID:
 319                         nfs4_schedule_lease_recovery(clp);
 320                         goto wait_on_recovery;
 321 #if defined(CONFIG_NFS_V4_1)
 322                 case -NFS4ERR_BADSESSION:
 323                 case -NFS4ERR_BADSLOT:
 324                 case -NFS4ERR_BAD_HIGH_SLOT:
 325                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 326                 case -NFS4ERR_DEADSESSION:
 327                 case -NFS4ERR_SEQ_FALSE_RETRY:
 328                 case -NFS4ERR_SEQ_MISORDERED:
 329                         dprintk("%s ERROR: %d Reset session\n", __func__,
 330                                 errorcode);
 331                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 332                         exception->retry = 1;
 333                         break;
 334 #endif /* defined(CONFIG_NFS_V4_1) */
 335                 case -NFS4ERR_FILE_OPEN:
 336                         if (exception->timeout > HZ) {
 337                                 /* We have retried a decent amount, time to
 338                                  * fail
 339                                  */
 340                                 ret = -EBUSY;
 341                                 break;
 342                         }
 343                 case -NFS4ERR_GRACE:
 344                 case -NFS4ERR_DELAY:
 345                 case -EKEYEXPIRED:
 346                         ret = nfs4_delay(server->client, &exception->timeout);
 347                         if (ret != 0)
 348                                 break;
 349                 case -NFS4ERR_RETRY_UNCACHED_REP:
 350                 case -NFS4ERR_OLD_STATEID:
 351                         exception->retry = 1;
 352                         break;
 353                 case -NFS4ERR_BADOWNER:
 354                         /* The following works around a Linux server bug! */
 355                 case -NFS4ERR_BADNAME:
 356                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 357                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 358                                 exception->retry = 1;
 359                                 printk(KERN_WARNING "NFS: v4 server %s "
 360                                                 "does not accept raw "
 361                                                 "uid/gids. "
 362                                                 "Reenabling the idmapper.\n",
 363                                                 server->nfs_client->cl_hostname);
 364                         }
 365         }
 366         /* We failed to handle the error */
 367         return nfs4_map_errors(ret);
 368 wait_on_recovery:
 369         ret = nfs4_wait_clnt_recover(clp);
 370         if (ret == 0)
 371                 exception->retry = 1;
 372         return ret;
 373 }
 374
 375
 376 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 377 {
 378         spin_lock(&clp->cl_lock);
 379         if (time_before(clp->cl_last_renewal,timestamp))
 380                 clp->cl_last_renewal = timestamp;
 381         spin_unlock(&clp->cl_lock);
 382 }
 383
 384 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 385 {
 386         do_renew_lease(server->nfs_client, timestamp);
 387 }
 388
 389 #if defined(CONFIG_NFS_V4_1)
 390
 391 /*
 392  * nfs4_free_slot - free a slot and efficiently update slot table.
 393  *
 394  * freeing a slot is trivially done by clearing its respective bit
 395  * in the bitmap.
 396  * If the freed slotid equals highest_used_slotid we want to update it
 397  * so that the server would be able to size down the slot table if needed,
 398  * otherwise we know that the highest_used_slotid is still in use.
 399  * When updating highest_used_slotid there may be "holes" in the bitmap
 400  * so we need to scan down from highest_used_slotid to 0 looking for the now
 401  * highest slotid in use.
 402  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
 403  *
 404  * Must be called while holding tbl->slot_tbl_lock
 405  */
 406 static void
 407 nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
 408 {
 409         BUG_ON(slotid >= NFS4_MAX_SLOT_TABLE);
 410         /* clear used bit in bitmap */
 411         __clear_bit(slotid, tbl->used_slots);
 412
 413         /* update highest_used_slotid when it is freed */
 414         if (slotid == tbl->highest_used_slotid) {
 415                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 416                 if (slotid < tbl->max_slots)
 417                         tbl->highest_used_slotid = slotid;
 418                 else
 419                         tbl->highest_used_slotid = NFS4_NO_SLOT;
 420         }
 421         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
 422                 slotid, tbl->highest_used_slotid);
 423 }
 424
 425 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
 426 {
 427         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 428         return true;
 429 }
 430
 431 /*
 432  * Signal state manager thread if session fore channel is drained
 433  */
 434 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 435 {
 436         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 437                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
 438                                 nfs4_set_task_privileged, NULL);
 439                 return;
 440         }
 441
 442         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 443                 return;
 444
 445         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 446         complete(&ses->fc_slot_table.complete);
 447 }
 448
 449 /*
 450  * Signal state manager thread if session back channel is drained
 451  */
 452 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 453 {
 454         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 455             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 456                 return;
 457         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 458         complete(&ses->bc_slot_table.complete);
 459 }
 460
 461 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 462 {
 463         struct nfs4_slot_table *tbl;
 464
 465         tbl = &res->sr_session->fc_slot_table;
 466         if (!res->sr_slot) {
 467                 /* just wake up the next guy waiting since
 468                  * we may have not consumed a slot after all */
 469                 dprintk("%s: No slot\n", __func__);
 470                 return;
 471         }
 472
 473         spin_lock(&tbl->slot_tbl_lock);
 474         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 475         nfs4_check_drain_fc_complete(res->sr_session);
 476         spin_unlock(&tbl->slot_tbl_lock);
 477         res->sr_slot = NULL;
 478 }
 479
 480 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 481 {
 482         unsigned long timestamp;
 483         struct nfs_client *clp;
 484
 485         /*
 486          * sr_status remains 1 if an RPC level error occurred. The server
 487          * may or may not have processed the sequence operation..
 488          * Proceed as if the server received and processed the sequence
 489          * operation.
 490          */
 491         if (res->sr_status == 1)
 492                 res->sr_status = NFS_OK;
 493
 494         /* don't increment the sequence number if the task wasn't sent */
 495         if (!RPC_WAS_SENT(task))
 496                 goto out;
 497
 498         /* Check the SEQUENCE operation status */
 499         switch (res->sr_status) {
 500         case 0:
 501                 /* Update the slot's sequence and clientid lease timer */
 502                 ++res->sr_slot->seq_nr;
 503                 timestamp = res->sr_renewal_time;
 504                 clp = res->sr_session->clp;
 505                 do_renew_lease(clp, timestamp);
 506                 /* Check sequence flags */
 507                 if (res->sr_status_flags != 0)
 508                         nfs4_schedule_lease_recovery(clp);
 509                 break;
 510         case -NFS4ERR_DELAY:
 511                 /* The server detected a resend of the RPC call and
 512                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 513                  * of RFC5661.
 514                  */
 515                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 516                         __func__,
 517                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 518                         res->sr_slot->seq_nr);
 519                 goto out_retry;
 520         default:
 521                 /* Just update the slot sequence no. */
 522                 ++res->sr_slot->seq_nr;
 523         }
 524 out:
 525         /* The session may be reset by one of the error handlers. */
 526         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 527         nfs41_sequence_free_slot(res);
 528         return 1;
 529 out_retry:
 530         if (!rpc_restart_call(task))
 531                 goto out;
 532         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 533         return 0;
 534 }
 535
 536 static int nfs4_sequence_done(struct rpc_task *task,
 537                                struct nfs4_sequence_res *res)
 538 {
 539         if (res->sr_session == NULL)
 540                 return 1;
 541         return nfs41_sequence_done(task, res);
 542 }
 543
 544 /*
 545  * nfs4_find_slot - efficiently look for a free slot
 546  *
 547  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 548  * If found, we mark the slot as used, update the highest_used_slotid,
 549  * and respectively set up the sequence operation args.
 550  * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
 551  *
 552  * Note: must be called with under the slot_tbl_lock.
 553  */
 554 static u32
 555 nfs4_find_slot(struct nfs4_slot_table *tbl)
 556 {
 557         u32 slotid;
 558         u32 ret_id = NFS4_NO_SLOT;
 559
 560         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
 561                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 562                 tbl->max_slots);
 563         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 564         if (slotid >= tbl->max_slots)
 565                 goto out;
 566         __set_bit(slotid, tbl->used_slots);
 567         if (slotid > tbl->highest_used_slotid ||
 568                         tbl->highest_used_slotid == NFS4_NO_SLOT)
 569                 tbl->highest_used_slotid = slotid;
 570         ret_id = slotid;
 571 out:
 572         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 573                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 574         return ret_id;
 575 }
 576
 577 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 578                 struct nfs4_sequence_res *res, int cache_reply)
 579 {
 580         args->sa_session = NULL;
 581         args->sa_cache_this = 0;
 582         if (cache_reply)
 583                 args->sa_cache_this = 1;
 584         res->sr_session = NULL;
 585         res->sr_slot = NULL;
 586 }
 587
 588 int nfs41_setup_sequence(struct nfs4_session *session,
 589                                 struct nfs4_sequence_args *args,
 590                                 struct nfs4_sequence_res *res,
 591                                 struct rpc_task *task)
 592 {
 593         struct nfs4_slot *slot;
 594         struct nfs4_slot_table *tbl;
 595         u32 slotid;
 596
 597         dprintk("--> %s\n", __func__);
 598         /* slot already allocated? */
 599         if (res->sr_slot != NULL)
 600                 return 0;
 601
 602         tbl = &session->fc_slot_table;
 603
 604         spin_lock(&tbl->slot_tbl_lock);
 605         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 606             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 607                 /* The state manager will wait until the slot table is empty */
 608                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 609                 spin_unlock(&tbl->slot_tbl_lock);
 610                 dprintk("%s session is draining\n", __func__);
 611                 return -EAGAIN;
 612         }
 613
 614         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 615             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 616                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 617                 spin_unlock(&tbl->slot_tbl_lock);
 618                 dprintk("%s enforce FIFO order\n", __func__);
 619                 return -EAGAIN;
 620         }
 621
 622         slotid = nfs4_find_slot(tbl);
 623         if (slotid == NFS4_NO_SLOT) {
 624                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 625                 spin_unlock(&tbl->slot_tbl_lock);
 626                 dprintk("<-- %s: no free slots\n", __func__);
 627                 return -EAGAIN;
 628         }
 629         spin_unlock(&tbl->slot_tbl_lock);
 630
 631         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 632         slot = tbl->slots + slotid;
 633         args->sa_session = session;
 634         args->sa_slotid = slotid;
 635
 636         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 637
 638         res->sr_session = session;
 639         res->sr_slot = slot;
 640         res->sr_renewal_time = jiffies;
 641         res->sr_status_flags = 0;
 642         /*
 643          * sr_status is only set in decode_sequence, and so will remain
 644          * set to 1 if an rpc level failure occurs.
 645          */
 646         res->sr_status = 1;
 647         return 0;
 648 }
 649 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 650
 651 int nfs4_setup_sequence(const struct nfs_server *server,
 652                         struct nfs4_sequence_args *args,
 653                         struct nfs4_sequence_res *res,
 654                         struct rpc_task *task)
 655 {
 656         struct nfs4_session *session = nfs4_get_session(server);
 657         int ret = 0;
 658
 659         if (session == NULL)
 660                 goto out;
 661
 662         dprintk("--> %s clp %p session %p sr_slot %td\n",
 663                 __func__, session->clp, session, res->sr_slot ?
 664                         res->sr_slot - session->fc_slot_table.slots : -1);
 665
 666         ret = nfs41_setup_sequence(session, args, res, task);
 667 out:
 668         dprintk("<-- %s status=%d\n", __func__, ret);
 669         return ret;
 670 }
 671
 672 struct nfs41_call_sync_data {
 673         const struct nfs_server *seq_server;
 674         struct nfs4_sequence_args *seq_args;
 675         struct nfs4_sequence_res *seq_res;
 676 };
 677
 678 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 679 {
 680         struct nfs41_call_sync_data *data = calldata;
 681
 682         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 683
 684         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 685                                 data->seq_res, task))
 686                 return;
 687         rpc_call_start(task);
 688 }
 689
 690 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 691 {
 692         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 693         nfs41_call_sync_prepare(task, calldata);
 694 }
 695
 696 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 697 {
 698         struct nfs41_call_sync_data *data = calldata;
 699
 700         nfs41_sequence_done(task, data->seq_res);
 701 }
 702
 703 static const struct rpc_call_ops nfs41_call_sync_ops = {
 704         .rpc_call_prepare = nfs41_call_sync_prepare,
 705         .rpc_call_done = nfs41_call_sync_done,
 706 };
 707
 708 static const struct rpc_call_ops nfs41_call_priv_sync_ops = {
 709         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 710         .rpc_call_done = nfs41_call_sync_done,
 711 };
 712
 713 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 714                                    struct nfs_server *server,
 715                                    struct rpc_message *msg,
 716                                    struct nfs4_sequence_args *args,
 717                                    struct nfs4_sequence_res *res,
 718                                    int privileged)
 719 {
 720         int ret;
 721         struct rpc_task *task;
 722         struct nfs41_call_sync_data data = {
 723                 .seq_server = server,
 724                 .seq_args = args,
 725                 .seq_res = res,
 726         };
 727         struct rpc_task_setup task_setup = {
 728                 .rpc_client = clnt,
 729                 .rpc_message = msg,
 730                 .callback_ops = &nfs41_call_sync_ops,
 731                 .callback_data = &data
 732         };
 733
 734         if (privileged)
 735                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 736         task = rpc_run_task(&task_setup);
 737         if (IS_ERR(task))
 738                 ret = PTR_ERR(task);
 739         else {
 740                 ret = task->tk_status;
 741                 rpc_put_task(task);
 742         }
 743         return ret;
 744 }
 745
 746 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 747                             struct nfs_server *server,
 748                             struct rpc_message *msg,
 749                             struct nfs4_sequence_args *args,
 750                             struct nfs4_sequence_res *res,
 751                             int cache_reply)
 752 {
 753         nfs41_init_sequence(args, res, cache_reply);
 754         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
 755 }
 756
 757 #else
 758 static inline
 759 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 760                 struct nfs4_sequence_res *res, int cache_reply)
 761 {
 762 }
 763
 764 static int nfs4_sequence_done(struct rpc_task *task,
 765                                struct nfs4_sequence_res *res)
 766 {
 767         return 1;
 768 }
 769 #endif /* CONFIG_NFS_V4_1 */
 770
 771 int _nfs4_call_sync(struct rpc_clnt *clnt,
 772                     struct nfs_server *server,
 773                     struct rpc_message *msg,
 774                     struct nfs4_sequence_args *args,
 775                     struct nfs4_sequence_res *res,
 776                     int cache_reply)
 777 {
 778         nfs41_init_sequence(args, res, cache_reply);
 779         return rpc_call_sync(clnt, msg, 0);
 780 }
 781
 782 static inline
 783 int nfs4_call_sync(struct rpc_clnt *clnt,
 784                    struct nfs_server *server,
 785                    struct rpc_message *msg,
 786                    struct nfs4_sequence_args *args,
 787                    struct nfs4_sequence_res *res,
 788                    int cache_reply)
 789 {
 790         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 791                                                 args, res, cache_reply);
 792 }
 793
 794 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 795 {
 796         struct nfs_inode *nfsi = NFS_I(dir);
 797
 798         spin_lock(&dir->i_lock);
 799         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 800         if (!cinfo->atomic || cinfo->before != dir->i_version)
 801                 nfs_force_lookup_revalidate(dir);
 802         dir->i_version = cinfo->after;
 803         spin_unlock(&dir->i_lock);
 804 }
 805
 806 struct nfs4_opendata {
 807         struct kref kref;
 808         struct nfs_openargs o_arg;
 809         struct nfs_openres o_res;
 810         struct nfs_open_confirmargs c_arg;
 811         struct nfs_open_confirmres c_res;
 812         struct nfs4_string owner_name;
 813         struct nfs4_string group_name;
 814         struct nfs_fattr f_attr;
 815         struct dentry *dir;
 816         struct dentry *dentry;
 817         struct nfs4_state_owner *owner;
 818         struct nfs4_state *state;
 819         struct iattr attrs;
 820         unsigned long timestamp;
 821         unsigned int rpc_done : 1;
 822         int rpc_status;
 823         int cancelled;
 824 };
 825
 826
 827 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 828 {
 829         p->o_res.f_attr = &p->f_attr;
 830         p->o_res.seqid = p->o_arg.seqid;
 831         p->c_res.seqid = p->c_arg.seqid;
 832         p->o_res.server = p->o_arg.server;
 833         nfs_fattr_init(&p->f_attr);
 834         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 835 }
 836
 837 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 838                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 839                 const struct iattr *attrs,
 840                 gfp_t gfp_mask)
 841 {
 842         struct dentry *parent = dget_parent(dentry);
 843         struct inode *dir = parent->d_inode;
 844         struct nfs_server *server = NFS_SERVER(dir);
 845         struct nfs4_opendata *p;
 846
 847         p = kzalloc(sizeof(*p), gfp_mask);
 848         if (p == NULL)
 849                 goto err;
 850         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 851         if (p->o_arg.seqid == NULL)
 852                 goto err_free;
 853         nfs_sb_active(dentry->d_sb);
 854         p->dentry = dget(dentry);
 855         p->dir = parent;
 856         p->owner = sp;
 857         atomic_inc(&sp->so_count);
 858         p->o_arg.fh = NFS_FH(dir);
 859         p->o_arg.open_flags = flags;
 860         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 861         p->o_arg.clientid = server->nfs_client->cl_clientid;
 862         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
 863         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 864         p->o_arg.name = &dentry->d_name;
 865         p->o_arg.server = server;
 866         p->o_arg.bitmask = server->attr_bitmask;
 867         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
 868         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 869         if (attrs != NULL && attrs->ia_valid != 0) {
 870                 __be32 verf[2];
 871
 872                 p->o_arg.u.attrs = &p->attrs;
 873                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 874
 875                 verf[0] = jiffies;
 876                 verf[1] = current->pid;
 877                 memcpy(p->o_arg.u.verifier.data, verf,
 878                                 sizeof(p->o_arg.u.verifier.data));
 879         }
 880         p->c_arg.fh = &p->o_res.fh;
 881         p->c_arg.stateid = &p->o_res.stateid;
 882         p->c_arg.seqid = p->o_arg.seqid;
 883         nfs4_init_opendata_res(p);
 884         kref_init(&p->kref);
 885         return p;
 886 err_free:
 887         kfree(p);
 888 err:
 889         dput(parent);
 890         return NULL;
 891 }
 892
 893 static void nfs4_opendata_free(struct kref *kref)
 894 {
 895         struct nfs4_opendata *p = container_of(kref,
 896                         struct nfs4_opendata, kref);
 897         struct super_block *sb = p->dentry->d_sb;
 898
 899         nfs_free_seqid(p->o_arg.seqid);
 900         if (p->state != NULL)
 901                 nfs4_put_open_state(p->state);
 902         nfs4_put_state_owner(p->owner);
 903         dput(p->dir);
 904         dput(p->dentry);
 905         nfs_sb_deactive(sb);
 906         nfs_fattr_free_names(&p->f_attr);
 907         kfree(p);
 908 }
 909
 910 static void nfs4_opendata_put(struct nfs4_opendata *p)
 911 {
 912         if (p != NULL)
 913                 kref_put(&p->kref, nfs4_opendata_free);
 914 }
 915
 916 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 917 {
 918         int ret;
 919
 920         ret = rpc_wait_for_completion_task(task);
 921         return ret;
 922 }
 923
 924 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 925 {
 926         int ret = 0;
 927
 928         if (open_mode & (O_EXCL|O_TRUNC))
 929                 goto out;
 930         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 931                 case FMODE_READ:
 932                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 933                                 && state->n_rdonly != 0;
 934                         break;
 935                 case FMODE_WRITE:
 936                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 937                                 && state->n_wronly != 0;
 938                         break;
 939                 case FMODE_READ|FMODE_WRITE:
 940                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 941                                 && state->n_rdwr != 0;
 942         }
 943 out:
 944         return ret;
 945 }
 946
 947 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 948 {
 949         if (delegation == NULL)
 950                 return 0;
 951         if ((delegation->type & fmode) != fmode)
 952                 return 0;
 953         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 954                 return 0;
 955         nfs_mark_delegation_referenced(delegation);
 956         return 1;
 957 }
 958
 959 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 960 {
 961         switch (fmode) {
 962                 case FMODE_WRITE:
 963                         state->n_wronly++;
 964                         break;
 965                 case FMODE_READ:
 966                         state->n_rdonly++;
 967                         break;
 968                 case FMODE_READ|FMODE_WRITE:
 969                         state->n_rdwr++;
 970         }
 971         nfs4_state_set_mode_locked(state, state->state | fmode);
 972 }
 973
 974 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 975 {
 976         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 977                 nfs4_stateid_copy(&state->stateid, stateid);
 978         nfs4_stateid_copy(&state->open_stateid, stateid);
 979         switch (fmode) {
 980                 case FMODE_READ:
 981                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 982                         break;
 983                 case FMODE_WRITE:
 984                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 985                         break;
 986                 case FMODE_READ|FMODE_WRITE:
 987                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 988         }
 989 }
 990
 991 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 992 {
 993         write_seqlock(&state->seqlock);
 994         nfs_set_open_stateid_locked(state, stateid, fmode);
 995         write_sequnlock(&state->seqlock);
 996 }
 997
 998 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 999 {
1000         /*
1001          * Protect the call to nfs4_state_set_mode_locked and
1002          * serialise the stateid update
1003          */
1004         write_seqlock(&state->seqlock);
1005         if (deleg_stateid != NULL) {
1006                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1007                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1008         }
1009         if (open_stateid != NULL)
1010                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1011         write_sequnlock(&state->seqlock);
1012         spin_lock(&state->owner->so_lock);
1013         update_open_stateflags(state, fmode);
1014         spin_unlock(&state->owner->so_lock);
1015 }
1016
1017 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1018 {
1019         struct nfs_inode *nfsi = NFS_I(state->inode);
1020         struct nfs_delegation *deleg_cur;
1021         int ret = 0;
1022
1023         fmode &= (FMODE_READ|FMODE_WRITE);
1024
1025         rcu_read_lock();
1026         deleg_cur = rcu_dereference(nfsi->delegation);
1027         if (deleg_cur == NULL)
1028                 goto no_delegation;
1029
1030         spin_lock(&deleg_cur->lock);
1031         if (nfsi->delegation != deleg_cur ||
1032             (deleg_cur->type & fmode) != fmode)
1033                 goto no_delegation_unlock;
1034
1035         if (delegation == NULL)
1036                 delegation = &deleg_cur->stateid;
1037         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1038                 goto no_delegation_unlock;
1039
1040         nfs_mark_delegation_referenced(deleg_cur);
1041         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1042         ret = 1;
1043 no_delegation_unlock:
1044         spin_unlock(&deleg_cur->lock);
1045 no_delegation:
1046         rcu_read_unlock();
1047
1048         if (!ret && open_stateid != NULL) {
1049                 __update_open_stateid(state, open_stateid, NULL, fmode);
1050                 ret = 1;
1051         }
1052
1053         return ret;
1054 }
1055
1056
1057 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1058 {
1059         struct nfs_delegation *delegation;
1060
1061         rcu_read_lock();
1062         delegation = rcu_dereference(NFS_I(inode)->delegation);
1063         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1064                 rcu_read_unlock();
1065                 return;
1066         }
1067         rcu_read_unlock();
1068         nfs4_inode_return_delegation(inode);
1069 }
1070
1071 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1072 {
1073         struct nfs4_state *state = opendata->state;
1074         struct nfs_inode *nfsi = NFS_I(state->inode);
1075         struct nfs_delegation *delegation;
1076         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1077         fmode_t fmode = opendata->o_arg.fmode;
1078         nfs4_stateid stateid;
1079         int ret = -EAGAIN;
1080
1081         for (;;) {
1082                 if (can_open_cached(state, fmode, open_mode)) {
1083                         spin_lock(&state->owner->so_lock);
1084                         if (can_open_cached(state, fmode, open_mode)) {
1085                                 update_open_stateflags(state, fmode);
1086                                 spin_unlock(&state->owner->so_lock);
1087                                 goto out_return_state;
1088                         }
1089                         spin_unlock(&state->owner->so_lock);
1090                 }
1091                 rcu_read_lock();
1092                 delegation = rcu_dereference(nfsi->delegation);
1093                 if (!can_open_delegated(delegation, fmode)) {
1094                         rcu_read_unlock();
1095                         break;
1096                 }
1097                 /* Save the delegation */
1098                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1099                 rcu_read_unlock();
1100                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1101                 if (ret != 0)
1102                         goto out;
1103                 ret = -EAGAIN;
1104
1105                 /* Try to update the stateid using the delegation */
1106                 if (update_open_stateid(state, NULL, &stateid, fmode))
1107                         goto out_return_state;
1108         }
1109 out:
1110         return ERR_PTR(ret);
1111 out_return_state:
1112         atomic_inc(&state->count);
1113         return state;
1114 }
1115
1116 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1117 {
1118         struct inode *inode;
1119         struct nfs4_state *state = NULL;
1120         struct nfs_delegation *delegation;
1121         int ret;
1122
1123         if (!data->rpc_done) {
1124                 state = nfs4_try_open_cached(data);
1125                 goto out;
1126         }
1127
1128         ret = -EAGAIN;
1129         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1130                 goto err;
1131         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1132         ret = PTR_ERR(inode);
1133         if (IS_ERR(inode))
1134                 goto err;
1135         ret = -ENOMEM;
1136         state = nfs4_get_open_state(inode, data->owner);
1137         if (state == NULL)
1138                 goto err_put_inode;
1139         if (data->o_res.delegation_type != 0) {
1140                 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1141                 int delegation_flags = 0;
1142
1143                 rcu_read_lock();
1144                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1145                 if (delegation)
1146                         delegation_flags = delegation->flags;
1147                 rcu_read_unlock();
1148                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1149                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1150                                         "returning a delegation for "
1151                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1152                                         clp->cl_hostname);
1153                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1154                         nfs_inode_set_delegation(state->inode,
1155                                         data->owner->so_cred,
1156                                         &data->o_res);
1157                 else
1158                         nfs_inode_reclaim_delegation(state->inode,
1159                                         data->owner->so_cred,
1160                                         &data->o_res);
1161         }
1162
1163         update_open_stateid(state, &data->o_res.stateid, NULL,
1164                         data->o_arg.fmode);
1165         iput(inode);
1166 out:
1167         return state;
1168 err_put_inode:
1169         iput(inode);
1170 err:
1171         return ERR_PTR(ret);
1172 }
1173
1174 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1175 {
1176         struct nfs_inode *nfsi = NFS_I(state->inode);
1177         struct nfs_open_context *ctx;
1178
1179         spin_lock(&state->inode->i_lock);
1180         list_for_each_entry(ctx, &nfsi->open_files, list) {
1181                 if (ctx->state != state)
1182                         continue;
1183                 get_nfs_open_context(ctx);
1184                 spin_unlock(&state->inode->i_lock);
1185                 return ctx;
1186         }
1187         spin_unlock(&state->inode->i_lock);
1188         return ERR_PTR(-ENOENT);
1189 }
1190
1191 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1192 {
1193         struct nfs4_opendata *opendata;
1194
1195         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1196         if (opendata == NULL)
1197                 return ERR_PTR(-ENOMEM);
1198         opendata->state = state;
1199         atomic_inc(&state->count);
1200         return opendata;
1201 }
1202
1203 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1204 {
1205         struct nfs4_state *newstate;
1206         int ret;
1207
1208         opendata->o_arg.open_flags = 0;
1209         opendata->o_arg.fmode = fmode;
1210         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1211         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1212         nfs4_init_opendata_res(opendata);
1213         ret = _nfs4_recover_proc_open(opendata);
1214         if (ret != 0)
1215                 return ret;
1216         newstate = nfs4_opendata_to_nfs4_state(opendata);
1217         if (IS_ERR(newstate))
1218                 return PTR_ERR(newstate);
1219         nfs4_close_state(newstate, fmode);
1220         *res = newstate;
1221         return 0;
1222 }
1223
1224 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1225 {
1226         struct nfs4_state *newstate;
1227         int ret;
1228
1229         /* memory barrier prior to reading state->n_* */
1230         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1231         smp_rmb();
1232         if (state->n_rdwr != 0) {
1233                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1234                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1235                 if (ret != 0)
1236                         return ret;
1237                 if (newstate != state)
1238                         return -ESTALE;
1239         }
1240         if (state->n_wronly != 0) {
1241                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1242                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1243                 if (ret != 0)
1244                         return ret;
1245                 if (newstate != state)
1246                         return -ESTALE;
1247         }
1248         if (state->n_rdonly != 0) {
1249                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1250                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1251                 if (ret != 0)
1252                         return ret;
1253                 if (newstate != state)
1254                         return -ESTALE;
1255         }
1256         /*
1257          * We may have performed cached opens for all three recoveries.
1258          * Check if we need to update the current stateid.
1259          */
1260         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1261             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1262                 write_seqlock(&state->seqlock);
1263                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1264                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1265                 write_sequnlock(&state->seqlock);
1266         }
1267         return 0;
1268 }
1269
1270 /*
1271  * OPEN_RECLAIM:
1272  *      reclaim state on the server after a reboot.
1273  */
1274 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1275 {
1276         struct nfs_delegation *delegation;
1277         struct nfs4_opendata *opendata;
1278         fmode_t delegation_type = 0;
1279         int status;
1280
1281         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1282         if (IS_ERR(opendata))
1283                 return PTR_ERR(opendata);
1284         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1285         opendata->o_arg.fh = NFS_FH(state->inode);
1286         rcu_read_lock();
1287         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1288         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1289                 delegation_type = delegation->type;
1290         rcu_read_unlock();
1291         opendata->o_arg.u.delegation_type = delegation_type;
1292         status = nfs4_open_recover(opendata, state);
1293         nfs4_opendata_put(opendata);
1294         return status;
1295 }
1296
1297 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1298 {
1299         struct nfs_server *server = NFS_SERVER(state->inode);
1300         struct nfs4_exception exception = { };
1301         int err;
1302         do {
1303                 err = _nfs4_do_open_reclaim(ctx, state);
1304                 if (err != -NFS4ERR_DELAY)
1305                         break;
1306                 nfs4_handle_exception(server, err, &exception);
1307         } while (exception.retry);
1308         return err;
1309 }
1310
1311 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1312 {
1313         struct nfs_open_context *ctx;
1314         int ret;
1315
1316         ctx = nfs4_state_find_open_context(state);
1317         if (IS_ERR(ctx))
1318                 return PTR_ERR(ctx);
1319         ret = nfs4_do_open_reclaim(ctx, state);
1320         put_nfs_open_context(ctx);
1321         return ret;
1322 }
1323
1324 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1325 {
1326         struct nfs4_opendata *opendata;
1327         int ret;
1328
1329         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1330         if (IS_ERR(opendata))
1331                 return PTR_ERR(opendata);
1332         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1333         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1334         ret = nfs4_open_recover(opendata, state);
1335         nfs4_opendata_put(opendata);
1336         return ret;
1337 }
1338
1339 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1340 {
1341         struct nfs4_exception exception = { };
1342         struct nfs_server *server = NFS_SERVER(state->inode);
1343         int err;
1344         do {
1345                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1346                 switch (err) {
1347                         case 0:
1348                         case -ENOENT:
1349                         case -ESTALE:
1350                                 goto out;
1351                         case -NFS4ERR_BADSESSION:
1352                         case -NFS4ERR_BADSLOT:
1353                         case -NFS4ERR_BAD_HIGH_SLOT:
1354                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1355                         case -NFS4ERR_DEADSESSION:
1356                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1357                                 goto out;
1358                         case -NFS4ERR_STALE_CLIENTID:
1359                         case -NFS4ERR_STALE_STATEID:
1360                         case -NFS4ERR_EXPIRED:
1361                                 /* Don't recall a delegation if it was lost */
1362                                 nfs4_schedule_lease_recovery(server->nfs_client);
1363                                 goto out;
1364                         case -ERESTARTSYS:
1365                                 /*
1366                                  * The show must go on: exit, but mark the
1367                                  * stateid as needing recovery.
1368                                  */
1369                         case -NFS4ERR_DELEG_REVOKED:
1370                         case -NFS4ERR_ADMIN_REVOKED:
1371                         case -NFS4ERR_BAD_STATEID:
1372                                 nfs_inode_find_state_and_recover(state->inode,
1373                                                 stateid);
1374                                 nfs4_schedule_stateid_recovery(server, state);
1375                         case -EKEYEXPIRED:
1376                                 /*
1377                                  * User RPCSEC_GSS context has expired.
1378                                  * We cannot recover this stateid now, so
1379                                  * skip it and allow recovery thread to
1380                                  * proceed.
1381                                  */
1382                         case -ENOMEM:
1383                                 err = 0;
1384                                 goto out;
1385                 }
1386                 err = nfs4_handle_exception(server, err, &exception);
1387         } while (exception.retry);
1388 out:
1389         return err;
1390 }
1391
1392 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1393 {
1394         struct nfs4_opendata *data = calldata;
1395
1396         data->rpc_status = task->tk_status;
1397         if (data->rpc_status == 0) {
1398                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1399                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1400                 renew_lease(data->o_res.server, data->timestamp);
1401                 data->rpc_done = 1;
1402         }
1403 }
1404
1405 static void nfs4_open_confirm_release(void *calldata)
1406 {
1407         struct nfs4_opendata *data = calldata;
1408         struct nfs4_state *state = NULL;
1409
1410         /* If this request hasn't been cancelled, do nothing */
1411         if (data->cancelled == 0)
1412                 goto out_free;
1413         /* In case of error, no cleanup! */
1414         if (!data->rpc_done)
1415                 goto out_free;
1416         state = nfs4_opendata_to_nfs4_state(data);
1417         if (!IS_ERR(state))
1418                 nfs4_close_state(state, data->o_arg.fmode);
1419 out_free:
1420         nfs4_opendata_put(data);
1421 }
1422
1423 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1424         .rpc_call_done = nfs4_open_confirm_done,
1425         .rpc_release = nfs4_open_confirm_release,
1426 };
1427
1428 /*
1429  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1430  */
1431 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1432 {
1433         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1434         struct rpc_task *task;
1435         struct  rpc_message msg = {
1436                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1437                 .rpc_argp = &data->c_arg,
1438                 .rpc_resp = &data->c_res,
1439                 .rpc_cred = data->owner->so_cred,
1440         };
1441         struct rpc_task_setup task_setup_data = {
1442                 .rpc_client = server->client,
1443                 .rpc_message = &msg,
1444                 .callback_ops = &nfs4_open_confirm_ops,
1445                 .callback_data = data,
1446                 .workqueue = nfsiod_workqueue,
1447                 .flags = RPC_TASK_ASYNC,
1448         };
1449         int status;
1450
1451         kref_get(&data->kref);
1452         data->rpc_done = 0;
1453         data->rpc_status = 0;
1454         data->timestamp = jiffies;
1455         task = rpc_run_task(&task_setup_data);
1456         if (IS_ERR(task))
1457                 return PTR_ERR(task);
1458         status = nfs4_wait_for_completion_rpc_task(task);
1459         if (status != 0) {
1460                 data->cancelled = 1;
1461                 smp_wmb();
1462         } else
1463                 status = data->rpc_status;
1464         rpc_put_task(task);
1465         return status;
1466 }
1467
1468 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1469 {
1470         struct nfs4_opendata *data = calldata;
1471         struct nfs4_state_owner *sp = data->owner;
1472
1473         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1474                 return;
1475         /*
1476          * Check if we still need to send an OPEN call, or if we can use
1477          * a delegation instead.
1478          */
1479         if (data->state != NULL) {
1480                 struct nfs_delegation *delegation;
1481
1482                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1483                         goto out_no_action;
1484                 rcu_read_lock();
1485                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1486                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1487                     can_open_delegated(delegation, data->o_arg.fmode))
1488                         goto unlock_no_action;
1489                 rcu_read_unlock();
1490         }
1491         /* Update client id. */
1492         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1493         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1494                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1495                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1496         }
1497         data->timestamp = jiffies;
1498         if (nfs4_setup_sequence(data->o_arg.server,
1499                                 &data->o_arg.seq_args,
1500                                 &data->o_res.seq_res, task))
1501                 return;
1502         rpc_call_start(task);
1503         return;
1504 unlock_no_action:
1505         rcu_read_unlock();
1506 out_no_action:
1507         task->tk_action = NULL;
1508
1509 }
1510
1511 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1512 {
1513         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1514         nfs4_open_prepare(task, calldata);
1515 }
1516
1517 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1518 {
1519         struct nfs4_opendata *data = calldata;
1520
1521         data->rpc_status = task->tk_status;
1522
1523         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1524                 return;
1525
1526         if (task->tk_status == 0) {
1527                 switch (data->o_res.f_attr->mode & S_IFMT) {
1528                         case S_IFREG:
1529                                 break;
1530                         case S_IFLNK:
1531                                 data->rpc_status = -ELOOP;
1532                                 break;
1533                         case S_IFDIR:
1534                                 data->rpc_status = -EISDIR;
1535                                 break;
1536                         default:
1537                                 data->rpc_status = -ENOTDIR;
1538                 }
1539                 renew_lease(data->o_res.server, data->timestamp);
1540                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1541                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1542         }
1543         data->rpc_done = 1;
1544 }
1545
1546 static void nfs4_open_release(void *calldata)
1547 {
1548         struct nfs4_opendata *data = calldata;
1549         struct nfs4_state *state = NULL;
1550
1551         /* If this request hasn't been cancelled, do nothing */
1552         if (data->cancelled == 0)
1553                 goto out_free;
1554         /* In case of error, no cleanup! */
1555         if (data->rpc_status != 0 || !data->rpc_done)
1556                 goto out_free;
1557         /* In case we need an open_confirm, no cleanup! */
1558         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1559                 goto out_free;
1560         state = nfs4_opendata_to_nfs4_state(data);
1561         if (!IS_ERR(state))
1562                 nfs4_close_state(state, data->o_arg.fmode);
1563 out_free:
1564         nfs4_opendata_put(data);
1565 }
1566
1567 static const struct rpc_call_ops nfs4_open_ops = {
1568         .rpc_call_prepare = nfs4_open_prepare,
1569         .rpc_call_done = nfs4_open_done,
1570         .rpc_release = nfs4_open_release,
1571 };
1572
1573 static const struct rpc_call_ops nfs4_recover_open_ops = {
1574         .rpc_call_prepare = nfs4_recover_open_prepare,
1575         .rpc_call_done = nfs4_open_done,
1576         .rpc_release = nfs4_open_release,
1577 };
1578
1579 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1580 {
1581         struct inode *dir = data->dir->d_inode;
1582         struct nfs_server *server = NFS_SERVER(dir);
1583         struct nfs_openargs *o_arg = &data->o_arg;
1584         struct nfs_openres *o_res = &data->o_res;
1585         struct rpc_task *task;
1586         struct rpc_message msg = {
1587                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1588                 .rpc_argp = o_arg,
1589                 .rpc_resp = o_res,
1590                 .rpc_cred = data->owner->so_cred,
1591         };
1592         struct rpc_task_setup task_setup_data = {
1593                 .rpc_client = server->client,
1594                 .rpc_message = &msg,
1595                 .callback_ops = &nfs4_open_ops,
1596                 .callback_data = data,
1597                 .workqueue = nfsiod_workqueue,
1598                 .flags = RPC_TASK_ASYNC,
1599         };
1600         int status;
1601
1602         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1603         kref_get(&data->kref);
1604         data->rpc_done = 0;
1605         data->rpc_status = 0;
1606         data->cancelled = 0;
1607         if (isrecover)
1608                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1609         task = rpc_run_task(&task_setup_data);
1610         if (IS_ERR(task))
1611                 return PTR_ERR(task);
1612         status = nfs4_wait_for_completion_rpc_task(task);
1613         if (status != 0) {
1614                 data->cancelled = 1;
1615                 smp_wmb();
1616         } else
1617                 status = data->rpc_status;
1618         rpc_put_task(task);
1619
1620         return status;
1621 }
1622
1623 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1624 {
1625         struct inode *dir = data->dir->d_inode;
1626         struct nfs_openres *o_res = &data->o_res;
1627         int status;
1628
1629         status = nfs4_run_open_task(data, 1);
1630         if (status != 0 || !data->rpc_done)
1631                 return status;
1632
1633         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1634
1635         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1636                 status = _nfs4_proc_open_confirm(data);
1637                 if (status != 0)
1638                         return status;
1639         }
1640
1641         return status;
1642 }
1643
1644 /*
1645  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1646  */
1647 static int _nfs4_proc_open(struct nfs4_opendata *data)
1648 {
1649         struct inode *dir = data->dir->d_inode;
1650         struct nfs_server *server = NFS_SERVER(dir);
1651         struct nfs_openargs *o_arg = &data->o_arg;
1652         struct nfs_openres *o_res = &data->o_res;
1653         int status;
1654
1655         status = nfs4_run_open_task(data, 0);
1656         if (!data->rpc_done)
1657                 return status;
1658         if (status != 0) {
1659                 if (status == -NFS4ERR_BADNAME &&
1660                                 !(o_arg->open_flags & O_CREAT))
1661                         return -ENOENT;
1662                 return status;
1663         }
1664
1665         nfs_fattr_map_and_free_names(server, &data->f_attr);
1666
1667         if (o_arg->open_flags & O_CREAT)
1668                 update_changeattr(dir, &o_res->cinfo);
1669         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1670                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1671         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1672                 status = _nfs4_proc_open_confirm(data);
1673                 if (status != 0)
1674                         return status;
1675         }
1676         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1677                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1678         return 0;
1679 }
1680
1681 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1682 {
1683         unsigned int loop;
1684         int ret;
1685
1686         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1687                 ret = nfs4_wait_clnt_recover(clp);
1688                 if (ret != 0)
1689                         break;
1690                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1691                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1692                         break;
1693                 nfs4_schedule_state_manager(clp);
1694                 ret = -EIO;
1695         }
1696         return ret;
1697 }
1698
1699 static int nfs4_recover_expired_lease(struct nfs_server *server)
1700 {
1701         return nfs4_client_recover_expired_lease(server->nfs_client);
1702 }
1703
1704 /*
1705  * OPEN_EXPIRED:
1706  *      reclaim state on the server after a network partition.
1707  *      Assumes caller holds the appropriate lock
1708  */
1709 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1710 {
1711         struct nfs4_opendata *opendata;
1712         int ret;
1713
1714         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1715         if (IS_ERR(opendata))
1716                 return PTR_ERR(opendata);
1717         ret = nfs4_open_recover(opendata, state);
1718         if (ret == -ESTALE)
1719                 d_drop(ctx->dentry);
1720         nfs4_opendata_put(opendata);
1721         return ret;
1722 }
1723
1724 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1725 {
1726         struct nfs_server *server = NFS_SERVER(state->inode);
1727         struct nfs4_exception exception = { };
1728         int err;
1729
1730         do {
1731                 err = _nfs4_open_expired(ctx, state);
1732                 switch (err) {
1733                 default:
1734                         goto out;
1735                 case -NFS4ERR_GRACE:
1736                 case -NFS4ERR_DELAY:
1737                         nfs4_handle_exception(server, err, &exception);
1738                         err = 0;
1739                 }
1740         } while (exception.retry);
1741 out:
1742         return err;
1743 }
1744
1745 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1746 {
1747         struct nfs_open_context *ctx;
1748         int ret;
1749
1750         ctx = nfs4_state_find_open_context(state);
1751         if (IS_ERR(ctx))
1752                 return PTR_ERR(ctx);
1753         ret = nfs4_do_open_expired(ctx, state);
1754         put_nfs_open_context(ctx);
1755         return ret;
1756 }
1757
1758 #if defined(CONFIG_NFS_V4_1)
1759 static int nfs41_check_expired_stateid(struct nfs4_state *state, nfs4_stateid *stateid, unsigned int flags)
1760 {
1761         int status = NFS_OK;
1762         struct nfs_server *server = NFS_SERVER(state->inode);
1763
1764         if (state->flags & flags) {
1765                 status = nfs41_test_stateid(server, stateid);
1766                 if (status != NFS_OK) {
1767                         if (status != -NFS4ERR_BAD_STATEID)
1768                                 nfs41_free_stateid(server, stateid);
1769                         state->flags &= ~flags;
1770                 }
1771         }
1772         return status;
1773 }
1774
1775 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1776 {
1777         int deleg_status, open_status;
1778         int deleg_flags = 1 << NFS_DELEGATED_STATE;
1779         int open_flags = (1 << NFS_O_RDONLY_STATE) | (1 << NFS_O_WRONLY_STATE) | (1 << NFS_O_RDWR_STATE);
1780
1781         deleg_status = nfs41_check_expired_stateid(state, &state->stateid, deleg_flags);
1782         open_status = nfs41_check_expired_stateid(state,  &state->open_stateid, open_flags);
1783
1784         if ((deleg_status == NFS_OK) && (open_status == NFS_OK))
1785                 return NFS_OK;
1786         return nfs4_open_expired(sp, state);
1787 }
1788 #endif
1789
1790 /*
1791  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1792  * fields corresponding to attributes that were used to store the verifier.
1793  * Make sure we clobber those fields in the later setattr call
1794  */
1795 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1796 {
1797         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1798             !(sattr->ia_valid & ATTR_ATIME_SET))
1799                 sattr->ia_valid |= ATTR_ATIME;
1800
1801         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1802             !(sattr->ia_valid & ATTR_MTIME_SET))
1803                 sattr->ia_valid |= ATTR_MTIME;
1804 }
1805
1806 /*
1807  * Returns a referenced nfs4_state
1808  */
1809 static int _nfs4_do_open(struct inode *dir,
1810                         struct dentry *dentry,
1811                         fmode_t fmode,
1812                         int flags,
1813                         struct iattr *sattr,
1814                         struct rpc_cred *cred,
1815                         struct nfs4_state **res,
1816                         struct nfs4_threshold **ctx_th)
1817 {
1818         struct nfs4_state_owner  *sp;
1819         struct nfs4_state     *state = NULL;
1820         struct nfs_server       *server = NFS_SERVER(dir);
1821         struct nfs4_opendata *opendata;
1822         int status;
1823
1824         /* Protect against reboot recovery conflicts */
1825         status = -ENOMEM;
1826         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1827         if (sp == NULL) {
1828                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1829                 goto out_err;
1830         }
1831         status = nfs4_recover_expired_lease(server);
1832         if (status != 0)
1833                 goto err_put_state_owner;
1834         if (dentry->d_inode != NULL)
1835                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1836         status = -ENOMEM;
1837         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1838         if (opendata == NULL)
1839                 goto err_put_state_owner;
1840
1841         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1842                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1843                 if (!opendata->f_attr.mdsthreshold)
1844                         goto err_opendata_put;
1845                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
1846         }
1847         if (dentry->d_inode != NULL)
1848                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1849
1850         status = _nfs4_proc_open(opendata);
1851         if (status != 0)
1852                 goto err_opendata_put;
1853
1854         state = nfs4_opendata_to_nfs4_state(opendata);
1855         status = PTR_ERR(state);
1856         if (IS_ERR(state))
1857                 goto err_opendata_put;
1858         if (server->caps & NFS_CAP_POSIX_LOCK)
1859                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1860
1861         if (opendata->o_arg.open_flags & O_EXCL) {
1862                 nfs4_exclusive_attrset(opendata, sattr);
1863
1864                 nfs_fattr_init(opendata->o_res.f_attr);
1865                 status = nfs4_do_setattr(state->inode, cred,
1866                                 opendata->o_res.f_attr, sattr,
1867                                 state);
1868                 if (status == 0)
1869                         nfs_setattr_update_inode(state->inode, sattr);
1870                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1871         }
1872
1873         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
1874                 *ctx_th = opendata->f_attr.mdsthreshold;
1875         else
1876                 kfree(opendata->f_attr.mdsthreshold);
1877         opendata->f_attr.mdsthreshold = NULL;
1878
1879         nfs4_opendata_put(opendata);
1880         nfs4_put_state_owner(sp);
1881         *res = state;
1882         return 0;
1883 err_opendata_put:
1884         kfree(opendata->f_attr.mdsthreshold);
1885         nfs4_opendata_put(opendata);
1886 err_put_state_owner:
1887         nfs4_put_state_owner(sp);
1888 out_err:
1889         *res = NULL;
1890         return status;
1891 }
1892
1893
1894 static struct nfs4_state *nfs4_do_open(struct inode *dir,
1895                                         struct dentry *dentry,
1896                                         fmode_t fmode,
1897                                         int flags,
1898                                         struct iattr *sattr,
1899                                         struct rpc_cred *cred,
1900                                         struct nfs4_threshold **ctx_th)
1901 {
1902         struct nfs4_exception exception = { };
1903         struct nfs4_state *res;
1904         int status;
1905
1906         fmode &= FMODE_READ|FMODE_WRITE;
1907         do {
1908                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
1909                                        &res, ctx_th);
1910                 if (status == 0)
1911                         break;
1912                 /* NOTE: BAD_SEQID means the server and client disagree about the
1913                  * book-keeping w.r.t. state-changing operations
1914                  * (OPEN/CLOSE/LOCK/LOCKU...)
1915                  * It is actually a sign of a bug on the client or on the server.
1916                  *
1917                  * If we receive a BAD_SEQID error in the particular case of
1918                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1919                  * have unhashed the old state_owner for us, and that we can
1920                  * therefore safely retry using a new one. We should still warn
1921                  * the user though...
1922                  */
1923                 if (status == -NFS4ERR_BAD_SEQID) {
1924                         pr_warn_ratelimited("NFS: v4 server %s "
1925                                         " returned a bad sequence-id error!\n",
1926                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1927                         exception.retry = 1;
1928                         continue;
1929                 }
1930                 /*
1931                  * BAD_STATEID on OPEN means that the server cancelled our
1932                  * state before it received the OPEN_CONFIRM.
1933                  * Recover by retrying the request as per the discussion
1934                  * on Page 181 of RFC3530.
1935                  */
1936                 if (status == -NFS4ERR_BAD_STATEID) {
1937                         exception.retry = 1;
1938                         continue;
1939                 }
1940                 if (status == -EAGAIN) {
1941                         /* We must have found a delegation */
1942                         exception.retry = 1;
1943                         continue;
1944                 }
1945                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1946                                         status, &exception));
1947         } while (exception.retry);
1948         return res;
1949 }
1950
1951 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1952                             struct nfs_fattr *fattr, struct iattr *sattr,
1953                             struct nfs4_state *state)
1954 {
1955         struct nfs_server *server = NFS_SERVER(inode);
1956         struct nfs_setattrargs  arg = {
1957                 .fh             = NFS_FH(inode),
1958                 .iap            = sattr,
1959                 .server         = server,
1960                 .bitmask = server->attr_bitmask,
1961         };
1962         struct nfs_setattrres  res = {
1963                 .fattr          = fattr,
1964                 .server         = server,
1965         };
1966         struct rpc_message msg = {
1967                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1968                 .rpc_argp       = &arg,
1969                 .rpc_resp       = &res,
1970                 .rpc_cred       = cred,
1971         };
1972         unsigned long timestamp = jiffies;
1973         int status;
1974
1975         nfs_fattr_init(fattr);
1976
1977         if (state != NULL) {
1978                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
1979                                 current->files, current->tgid);
1980         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
1981                                 FMODE_WRITE)) {
1982                 /* Use that stateid */
1983         } else
1984                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
1985
1986         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1987         if (status == 0 && state != NULL)
1988                 renew_lease(server, timestamp);
1989         return status;
1990 }
1991
1992 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1993                            struct nfs_fattr *fattr, struct iattr *sattr,
1994                            struct nfs4_state *state)
1995 {
1996         struct nfs_server *server = NFS_SERVER(inode);
1997         struct nfs4_exception exception = {
1998                 .state = state,
1999                 .inode = inode,
2000         };
2001         int err;
2002         do {
2003                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2004                 switch (err) {
2005                 case -NFS4ERR_OPENMODE:
2006                         if (state && !(state->state & FMODE_WRITE)) {
2007                                 err = -EBADF;
2008                                 if (sattr->ia_valid & ATTR_OPEN)
2009                                         err = -EACCES;
2010                                 goto out;
2011                         }
2012                 }
2013                 err = nfs4_handle_exception(server, err, &exception);
2014         } while (exception.retry);
2015 out:
2016         return err;
2017 }
2018
2019 struct nfs4_closedata {
2020         struct inode *inode;
2021         struct nfs4_state *state;
2022         struct nfs_closeargs arg;
2023         struct nfs_closeres res;
2024         struct nfs_fattr fattr;
2025         unsigned long timestamp;
2026         bool roc;
2027         u32 roc_barrier;
2028 };
2029
2030 static void nfs4_free_closedata(void *data)
2031 {
2032         struct nfs4_closedata *calldata = data;
2033         struct nfs4_state_owner *sp = calldata->state->owner;
2034         struct super_block *sb = calldata->state->inode->i_sb;
2035
2036         if (calldata->roc)
2037                 pnfs_roc_release(calldata->state->inode);
2038         nfs4_put_open_state(calldata->state);
2039         nfs_free_seqid(calldata->arg.seqid);
2040         nfs4_put_state_owner(sp);
2041         nfs_sb_deactive(sb);
2042         kfree(calldata);
2043 }
2044
2045 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2046                 fmode_t fmode)
2047 {
2048         spin_lock(&state->owner->so_lock);
2049         if (!(fmode & FMODE_READ))
2050                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2051         if (!(fmode & FMODE_WRITE))
2052                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2053         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2054         spin_unlock(&state->owner->so_lock);
2055 }
2056
2057 static void nfs4_close_done(struct rpc_task *task, void *data)
2058 {
2059         struct nfs4_closedata *calldata = data;
2060         struct nfs4_state *state = calldata->state;
2061         struct nfs_server *server = NFS_SERVER(calldata->inode);
2062
2063         dprintk("%s: begin!\n", __func__);
2064         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2065                 return;
2066         /* hmm. we are done with the inode, and in the process of freeing
2067          * the state_owner. we keep this around to process errors
2068          */
2069         switch (task->tk_status) {
2070                 case 0:
2071                         if (calldata->roc)
2072                                 pnfs_roc_set_barrier(state->inode,
2073                                                      calldata->roc_barrier);
2074                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2075                         renew_lease(server, calldata->timestamp);
2076                         nfs4_close_clear_stateid_flags(state,
2077                                         calldata->arg.fmode);
2078                         break;
2079                 case -NFS4ERR_STALE_STATEID:
2080                 case -NFS4ERR_OLD_STATEID:
2081                 case -NFS4ERR_BAD_STATEID:
2082                 case -NFS4ERR_EXPIRED:
2083                         if (calldata->arg.fmode == 0)
2084                                 break;
2085                 default:
2086                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2087                                 rpc_restart_call_prepare(task);
2088         }
2089         nfs_release_seqid(calldata->arg.seqid);
2090         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2091         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2092 }
2093
2094 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2095 {
2096         struct nfs4_closedata *calldata = data;
2097         struct nfs4_state *state = calldata->state;
2098         int call_close = 0;
2099
2100         dprintk("%s: begin!\n", __func__);
2101         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2102                 return;
2103
2104         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2105         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2106         spin_lock(&state->owner->so_lock);
2107         /* Calculate the change in open mode */
2108         if (state->n_rdwr == 0) {
2109                 if (state->n_rdonly == 0) {
2110                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2111                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2112                         calldata->arg.fmode &= ~FMODE_READ;
2113                 }
2114                 if (state->n_wronly == 0) {
2115                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2116                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2117                         calldata->arg.fmode &= ~FMODE_WRITE;
2118                 }
2119         }
2120         spin_unlock(&state->owner->so_lock);
2121
2122         if (!call_close) {
2123                 /* Note: exit _without_ calling nfs4_close_done */
2124                 task->tk_action = NULL;
2125                 goto out;
2126         }
2127
2128         if (calldata->arg.fmode == 0) {
2129                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2130                 if (calldata->roc &&
2131                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2132                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2133                                      task, NULL);
2134                         goto out;
2135                 }
2136         }
2137
2138         nfs_fattr_init(calldata->res.fattr);
2139         calldata->timestamp = jiffies;
2140         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2141                                 &calldata->arg.seq_args,
2142                                 &calldata->res.seq_res,
2143                                 task))
2144                 goto out;
2145         rpc_call_start(task);
2146 out:
2147         dprintk("%s: done!\n", __func__);
2148 }
2149
2150 static const struct rpc_call_ops nfs4_close_ops = {
2151         .rpc_call_prepare = nfs4_close_prepare,
2152         .rpc_call_done = nfs4_close_done,
2153         .rpc_release = nfs4_free_closedata,
2154 };
2155
2156 /*
2157  * It is possible for data to be read/written from a mem-mapped file
2158  * after the sys_close call (which hits the vfs layer as a flush).
2159  * This means that we can't safely call nfsv4 close on a file until
2160  * the inode is cleared. This in turn means that we are not good
2161  * NFSv4 citizens - we do not indicate to the server to update the file's
2162  * share state even when we are done with one of the three share
2163  * stateid's in the inode.
2164  *
2165  * NOTE: Caller must be holding the sp->so_owner semaphore!
2166  */
2167 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2168 {
2169         struct nfs_server *server = NFS_SERVER(state->inode);
2170         struct nfs4_closedata *calldata;
2171         struct nfs4_state_owner *sp = state->owner;
2172         struct rpc_task *task;
2173         struct rpc_message msg = {
2174                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2175                 .rpc_cred = state->owner->so_cred,
2176         };
2177         struct rpc_task_setup task_setup_data = {
2178                 .rpc_client = server->client,
2179                 .rpc_message = &msg,
2180                 .callback_ops = &nfs4_close_ops,
2181                 .workqueue = nfsiod_workqueue,
2182                 .flags = RPC_TASK_ASYNC,
2183         };
2184         int status = -ENOMEM;
2185
2186         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2187         if (calldata == NULL)
2188                 goto out;
2189         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2190         calldata->inode = state->inode;
2191         calldata->state = state;
2192         calldata->arg.fh = NFS_FH(state->inode);
2193         calldata->arg.stateid = &state->open_stateid;
2194         /* Serialization for the sequence id */
2195         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2196         if (calldata->arg.seqid == NULL)
2197                 goto out_free_calldata;
2198         calldata->arg.fmode = 0;
2199         calldata->arg.bitmask = server->cache_consistency_bitmask;
2200         calldata->res.fattr = &calldata->fattr;
2201         calldata->res.seqid = calldata->arg.seqid;
2202         calldata->res.server = server;
2203         calldata->roc = roc;
2204         nfs_sb_active(calldata->inode->i_sb);
2205
2206         msg.rpc_argp = &calldata->arg;
2207         msg.rpc_resp = &calldata->res;
2208         task_setup_data.callback_data = calldata;
2209         task = rpc_run_task(&task_setup_data);
2210         if (IS_ERR(task))
2211                 return PTR_ERR(task);
2212         status = 0;
2213         if (wait)
2214                 status = rpc_wait_for_completion_task(task);
2215         rpc_put_task(task);
2216         return status;
2217 out_free_calldata:
2218         kfree(calldata);
2219 out:
2220         if (roc)
2221                 pnfs_roc_release(state->inode);
2222         nfs4_put_open_state(state);
2223         nfs4_put_state_owner(sp);
2224         return status;
2225 }
2226
2227 static struct inode *
2228 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2229 {
2230         struct nfs4_state *state;
2231
2232         /* Protect against concurrent sillydeletes */
2233         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2234                              ctx->cred, &ctx->mdsthreshold);
2235         if (IS_ERR(state))
2236                 return ERR_CAST(state);
2237         ctx->state = state;
2238         return igrab(state->inode);
2239 }
2240
2241 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2242 {
2243         if (ctx->state == NULL)
2244                 return;
2245         if (is_sync)
2246                 nfs4_close_sync(ctx->state, ctx->mode);
2247         else
2248                 nfs4_close_state(ctx->state, ctx->mode);
2249 }
2250
2251 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2252 {
2253         struct nfs4_server_caps_arg args = {
2254                 .fhandle = fhandle,
2255         };
2256         struct nfs4_server_caps_res res = {};
2257         struct rpc_message msg = {
2258                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2259                 .rpc_argp = &args,
2260                 .rpc_resp = &res,
2261         };
2262         int status;
2263
2264         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2265         if (status == 0) {
2266                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2267                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2268                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2269                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2270                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2271                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2272                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2273                         server->caps |= NFS_CAP_ACLS;
2274                 if (res.has_links != 0)
2275                         server->caps |= NFS_CAP_HARDLINKS;
2276                 if (res.has_symlinks != 0)
2277                         server->caps |= NFS_CAP_SYMLINKS;
2278                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2279                         server->caps |= NFS_CAP_FILEID;
2280                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2281                         server->caps |= NFS_CAP_MODE;
2282                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2283                         server->caps |= NFS_CAP_NLINK;
2284                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2285                         server->caps |= NFS_CAP_OWNER;
2286                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2287                         server->caps |= NFS_CAP_OWNER_GROUP;
2288                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2289                         server->caps |= NFS_CAP_ATIME;
2290                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2291                         server->caps |= NFS_CAP_CTIME;
2292                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2293                         server->caps |= NFS_CAP_MTIME;
2294
2295                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2296                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2297                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2298                 server->acl_bitmask = res.acl_bitmask;
2299                 server->fh_expire_type = res.fh_expire_type;
2300         }
2301
2302         return status;
2303 }
2304
2305 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2306 {
2307         struct nfs4_exception exception = { };
2308         int err;
2309         do {
2310                 err = nfs4_handle_exception(server,
2311                                 _nfs4_server_capabilities(server, fhandle),
2312                                 &exception);
2313         } while (exception.retry);
2314         return err;
2315 }
2316
2317 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2318                 struct nfs_fsinfo *info)
2319 {
2320         struct nfs4_lookup_root_arg args = {
2321                 .bitmask = nfs4_fattr_bitmap,
2322         };
2323         struct nfs4_lookup_res res = {
2324                 .server = server,
2325                 .fattr = info->fattr,
2326                 .fh = fhandle,
2327         };
2328         struct rpc_message msg = {
2329                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2330                 .rpc_argp = &args,
2331                 .rpc_resp = &res,
2332         };
2333
2334         nfs_fattr_init(info->fattr);
2335         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2336 }
2337
2338 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2339                 struct nfs_fsinfo *info)
2340 {
2341         struct nfs4_exception exception = { };
2342         int err;
2343         do {
2344                 err = _nfs4_lookup_root(server, fhandle, info);
2345                 switch (err) {
2346                 case 0:
2347                 case -NFS4ERR_WRONGSEC:
2348                         goto out;
2349                 default:
2350                         err = nfs4_handle_exception(server, err, &exception);
2351                 }
2352         } while (exception.retry);
2353 out:
2354         return err;
2355 }
2356
2357 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2358                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2359 {
2360         struct rpc_auth *auth;
2361         int ret;
2362
2363         auth = rpcauth_create(flavor, server->client);
2364         if (!auth) {
2365                 ret = -EIO;
2366                 goto out;
2367         }
2368         ret = nfs4_lookup_root(server, fhandle, info);
2369 out:
2370         return ret;
2371 }
2372
2373 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2374                               struct nfs_fsinfo *info)
2375 {
2376         int i, len, status = 0;
2377         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2378
2379         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2380         flav_array[len] = RPC_AUTH_NULL;
2381         len += 1;
2382
2383         for (i = 0; i < len; i++) {
2384                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2385                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2386                         continue;
2387                 break;
2388         }
2389         /*
2390          * -EACCESS could mean that the user doesn't have correct permissions
2391          * to access the mount.  It could also mean that we tried to mount
2392          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2393          * existing mount programs don't handle -EACCES very well so it should
2394          * be mapped to -EPERM instead.
2395          */
2396         if (status == -EACCES)
2397                 status = -EPERM;
2398         return status;
2399 }
2400
2401 /*
2402  * get the file handle for the "/" directory on the server
2403  */
2404 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2405                          struct nfs_fsinfo *info)
2406 {
2407         int minor_version = server->nfs_client->cl_minorversion;
2408         int status = nfs4_lookup_root(server, fhandle, info);
2409         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2410                 /*
2411                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2412                  * by nfs4_map_errors() as this function exits.
2413                  */
2414                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2415         if (status == 0)
2416                 status = nfs4_server_capabilities(server, fhandle);
2417         if (status == 0)
2418                 status = nfs4_do_fsinfo(server, fhandle, info);
2419         return nfs4_map_errors(status);
2420 }
2421
2422 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2423                               struct nfs_fsinfo *info)
2424 {
2425         int error;
2426         struct nfs_fattr *fattr = info->fattr;
2427
2428         error = nfs4_server_capabilities(server, mntfh);
2429         if (error < 0) {
2430                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2431                 return error;
2432         }
2433
2434         error = nfs4_proc_getattr(server, mntfh, fattr);
2435         if (error < 0) {
2436                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2437                 return error;
2438         }
2439
2440         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2441             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2442                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2443
2444         return error;
2445 }
2446
2447 /*
2448  * Get locations and (maybe) other attributes of a referral.
2449  * Note that we'll actually follow the referral later when
2450  * we detect fsid mismatch in inode revalidation
2451  */
2452 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2453                              const struct qstr *name, struct nfs_fattr *fattr,
2454                              struct nfs_fh *fhandle)
2455 {
2456         int status = -ENOMEM;
2457         struct page *page = NULL;
2458         struct nfs4_fs_locations *locations = NULL;
2459
2460         page = alloc_page(GFP_KERNEL);
2461         if (page == NULL)
2462                 goto out;
2463         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2464         if (locations == NULL)
2465                 goto out;
2466
2467         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2468         if (status != 0)
2469                 goto out;
2470         /* Make sure server returned a different fsid for the referral */
2471         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2472                 dprintk("%s: server did not return a different fsid for"
2473                         " a referral at %s\n", __func__, name->name);
2474                 status = -EIO;
2475                 goto out;
2476         }
2477         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2478         nfs_fixup_referral_attributes(&locations->fattr);
2479
2480         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2481         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2482         memset(fhandle, 0, sizeof(struct nfs_fh));
2483 out:
2484         if (page)
2485                 __free_page(page);
2486         kfree(locations);
2487         return status;
2488 }
2489
2490 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2491 {
2492         struct nfs4_getattr_arg args = {
2493                 .fh = fhandle,
2494                 .bitmask = server->attr_bitmask,
2495         };
2496         struct nfs4_getattr_res res = {
2497                 .fattr = fattr,
2498                 .server = server,
2499         };
2500         struct rpc_message msg = {
2501                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2502                 .rpc_argp = &args,
2503                 .rpc_resp = &res,
2504         };
2505
2506         nfs_fattr_init(fattr);
2507         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2508 }
2509
2510 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2511 {
2512         struct nfs4_exception exception = { };
2513         int err;
2514         do {
2515                 err = nfs4_handle_exception(server,
2516                                 _nfs4_proc_getattr(server, fhandle, fattr),
2517                                 &exception);
2518         } while (exception.retry);
2519         return err;
2520 }
2521
2522 /*
2523  * The file is not closed if it is opened due to the a request to change
2524  * the size of the file. The open call will not be needed once the
2525  * VFS layer lookup-intents are implemented.
2526  *
2527  * Close is called when the inode is destroyed.
2528  * If we haven't opened the file for O_WRONLY, we
2529  * need to in the size_change case to obtain a stateid.
2530  *
2531  * Got race?
2532  * Because OPEN is always done by name in nfsv4, it is
2533  * possible that we opened a different file by the same
2534  * name.  We can recognize this race condition, but we
2535  * can't do anything about it besides returning an error.
2536  *
2537  * This will be fixed with VFS changes (lookup-intent).
2538  */
2539 static int
2540 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2541                   struct iattr *sattr)
2542 {
2543         struct inode *inode = dentry->d_inode;
2544         struct rpc_cred *cred = NULL;
2545         struct nfs4_state *state = NULL;
2546         int status;
2547
2548         if (pnfs_ld_layoutret_on_setattr(inode))
2549                 pnfs_return_layout(inode);
2550
2551         nfs_fattr_init(fattr);
2552
2553         /* Deal with open(O_TRUNC) */
2554         if (sattr->ia_valid & ATTR_OPEN)
2555                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2556
2557         /* Optimization: if the end result is no change, don't RPC */
2558         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2559                 return 0;
2560
2561         /* Search for an existing open(O_WRITE) file */
2562         if (sattr->ia_valid & ATTR_FILE) {
2563                 struct nfs_open_context *ctx;
2564
2565                 ctx = nfs_file_open_context(sattr->ia_file);
2566                 if (ctx) {
2567                         cred = ctx->cred;
2568                         state = ctx->state;
2569                 }
2570         }
2571
2572         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2573         if (status == 0)
2574                 nfs_setattr_update_inode(inode, sattr);
2575         return status;
2576 }
2577
2578 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2579                 const struct qstr *name, struct nfs_fh *fhandle,
2580                 struct nfs_fattr *fattr)
2581 {
2582         struct nfs_server *server = NFS_SERVER(dir);
2583         int                    status;
2584         struct nfs4_lookup_arg args = {
2585                 .bitmask = server->attr_bitmask,
2586                 .dir_fh = NFS_FH(dir),
2587                 .name = name,
2588         };
2589         struct nfs4_lookup_res res = {
2590                 .server = server,
2591                 .fattr = fattr,
2592                 .fh = fhandle,
2593         };
2594         struct rpc_message msg = {
2595                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2596                 .rpc_argp = &args,
2597                 .rpc_resp = &res,
2598         };
2599
2600         nfs_fattr_init(fattr);
2601
2602         dprintk("NFS call  lookup %s\n", name->name);
2603         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2604         dprintk("NFS reply lookup: %d\n", status);
2605         return status;
2606 }
2607
2608 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2609 {
2610         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2611                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2612         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2613         fattr->nlink = 2;
2614 }
2615
2616 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2617                                    struct qstr *name, struct nfs_fh *fhandle,
2618                                    struct nfs_fattr *fattr)
2619 {
2620         struct nfs4_exception exception = { };
2621         struct rpc_clnt *client = *clnt;
2622         int err;
2623         do {
2624                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2625                 switch (err) {
2626                 case -NFS4ERR_BADNAME:
2627                         err = -ENOENT;
2628                         goto out;
2629                 case -NFS4ERR_MOVED:
2630                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2631                         goto out;
2632                 case -NFS4ERR_WRONGSEC:
2633                         err = -EPERM;
2634                         if (client != *clnt)
2635                                 goto out;
2636
2637                         client = nfs4_create_sec_client(client, dir, name);
2638                         if (IS_ERR(client))
2639                                 return PTR_ERR(client);
2640
2641                         exception.retry = 1;
2642                         break;
2643                 default:
2644                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2645                 }
2646         } while (exception.retry);
2647
2648 out:
2649         if (err == 0)
2650                 *clnt = client;
2651         else if (client != *clnt)
2652                 rpc_shutdown_client(client);
2653
2654         return err;
2655 }
2656
2657 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2658                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2659 {
2660         int status;
2661         struct rpc_clnt *client = NFS_CLIENT(dir);
2662
2663         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2664         if (client != NFS_CLIENT(dir)) {
2665                 rpc_shutdown_client(client);
2666                 nfs_fixup_secinfo_attributes(fattr);
2667         }
2668         return status;
2669 }
2670
2671 struct rpc_clnt *
2672 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2673                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2674 {
2675         int status;
2676         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2677
2678         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2679         if (status < 0) {
2680                 rpc_shutdown_client(client);
2681                 return ERR_PTR(status);
2682         }
2683         return client;
2684 }
2685
2686 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2687 {
2688         struct nfs_server *server = NFS_SERVER(inode);
2689         struct nfs4_accessargs args = {
2690                 .fh = NFS_FH(inode),
2691                 .bitmask = server->cache_consistency_bitmask,
2692         };
2693         struct nfs4_accessres res = {
2694                 .server = server,
2695         };
2696         struct rpc_message msg = {
2697                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2698                 .rpc_argp = &args,
2699                 .rpc_resp = &res,
2700                 .rpc_cred = entry->cred,
2701         };
2702         int mode = entry->mask;
2703         int status;
2704
2705         /*
2706          * Determine which access bits we want to ask for...
2707          */
2708         if (mode & MAY_READ)
2709                 args.access |= NFS4_ACCESS_READ;
2710         if (S_ISDIR(inode->i_mode)) {
2711                 if (mode & MAY_WRITE)
2712                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2713                 if (mode & MAY_EXEC)
2714                         args.access |= NFS4_ACCESS_LOOKUP;
2715         } else {
2716                 if (mode & MAY_WRITE)
2717                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2718                 if (mode & MAY_EXEC)
2719                         args.access |= NFS4_ACCESS_EXECUTE;
2720         }
2721
2722         res.fattr = nfs_alloc_fattr();
2723         if (res.fattr == NULL)
2724                 return -ENOMEM;
2725
2726         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2727         if (!status) {
2728                 entry->mask = 0;
2729                 if (res.access & NFS4_ACCESS_READ)
2730                         entry->mask |= MAY_READ;
2731                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2732                         entry->mask |= MAY_WRITE;
2733                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2734                         entry->mask |= MAY_EXEC;
2735                 nfs_refresh_inode(inode, res.fattr);
2736         }
2737         nfs_free_fattr(res.fattr);
2738         return status;
2739 }
2740
2741 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2742 {
2743         struct nfs4_exception exception = { };
2744         int err;
2745         do {
2746                 err = nfs4_handle_exception(NFS_SERVER(inode),
2747                                 _nfs4_proc_access(inode, entry),
2748                                 &exception);
2749         } while (exception.retry);
2750         return err;
2751 }
2752
2753 /*
2754  * TODO: For the time being, we don't try to get any attributes
2755  * along with any of the zero-copy operations READ, READDIR,
2756  * READLINK, WRITE.
2757  *
2758  * In the case of the first three, we want to put the GETATTR
2759  * after the read-type operation -- this is because it is hard
2760  * to predict the length of a GETATTR response in v4, and thus
2761  * align the READ data correctly.  This means that the GETATTR
2762  * may end up partially falling into the page cache, and we should
2763  * shift it into the 'tail' of the xdr_buf before processing.
2764  * To do this efficiently, we need to know the total length
2765  * of data received, which doesn't seem to be available outside
2766  * of the RPC layer.
2767  *
2768  * In the case of WRITE, we also want to put the GETATTR after
2769  * the operation -- in this case because we want to make sure
2770  * we get the post-operation mtime and size.
2771  *
2772  * Both of these changes to the XDR layer would in fact be quite
2773  * minor, but I decided to leave them for a subsequent patch.
2774  */
2775 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2776                 unsigned int pgbase, unsigned int pglen)
2777 {
2778         struct nfs4_readlink args = {
2779                 .fh       = NFS_FH(inode),
2780                 .pgbase   = pgbase,
2781                 .pglen    = pglen,
2782                 .pages    = &page,
2783         };
2784         struct nfs4_readlink_res res;
2785         struct rpc_message msg = {
2786                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2787                 .rpc_argp = &args,
2788                 .rpc_resp = &res,
2789         };
2790
2791         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2792 }
2793
2794 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2795                 unsigned int pgbase, unsigned int pglen)
2796 {
2797         struct nfs4_exception exception = { };
2798         int err;
2799         do {
2800                 err = nfs4_handle_exception(NFS_SERVER(inode),
2801                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2802                                 &exception);
2803         } while (exception.retry);
2804         return err;
2805 }
2806
2807 /*
2808  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
2809  */
2810 static int
2811 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2812                  int flags)
2813 {
2814         struct nfs_open_context *ctx;
2815         struct nfs4_state *state;
2816         int status = 0;
2817
2818         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
2819         if (IS_ERR(ctx))
2820                 return PTR_ERR(ctx);
2821
2822         sattr->ia_mode &= ~current_umask();
2823         state = nfs4_do_open(dir, dentry, ctx->mode,
2824                         flags, sattr, ctx->cred,
2825                         &ctx->mdsthreshold);
2826         d_drop(dentry);
2827         if (IS_ERR(state)) {
2828                 status = PTR_ERR(state);
2829                 goto out;
2830         }
2831         d_add(dentry, igrab(state->inode));
2832         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2833         ctx->state = state;
2834 out:
2835         put_nfs_open_context(ctx);
2836         return status;
2837 }
2838
2839 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2840 {
2841         struct nfs_server *server = NFS_SERVER(dir);
2842         struct nfs_removeargs args = {
2843                 .fh = NFS_FH(dir),
2844                 .name = *name,
2845         };
2846         struct nfs_removeres res = {
2847                 .server = server,
2848         };
2849         struct rpc_message msg = {
2850                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2851                 .rpc_argp = &args,
2852                 .rpc_resp = &res,
2853         };
2854         int status;
2855
2856         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2857         if (status == 0)
2858                 update_changeattr(dir, &res.cinfo);
2859         return status;
2860 }
2861
2862 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2863 {
2864         struct nfs4_exception exception = { };
2865         int err;
2866         do {
2867                 err = nfs4_handle_exception(NFS_SERVER(dir),
2868                                 _nfs4_proc_remove(dir, name),
2869                                 &exception);
2870         } while (exception.retry);
2871         return err;
2872 }
2873
2874 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2875 {
2876         struct nfs_server *server = NFS_SERVER(dir);
2877         struct nfs_removeargs *args = msg->rpc_argp;
2878         struct nfs_removeres *res = msg->rpc_resp;
2879
2880         res->server = server;
2881         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2882         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2883 }
2884
2885 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
2886 {
2887         if (nfs4_setup_sequence(NFS_SERVER(data->dir),
2888                                 &data->args.seq_args,
2889                                 &data->res.seq_res,
2890                                 task))
2891                 return;
2892         rpc_call_start(task);
2893 }
2894
2895 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2896 {
2897         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2898
2899         if (!nfs4_sequence_done(task, &res->seq_res))
2900                 return 0;
2901         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2902                 return 0;
2903         update_changeattr(dir, &res->cinfo);
2904         return 1;
2905 }
2906
2907 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2908 {
2909         struct nfs_server *server = NFS_SERVER(dir);
2910         struct nfs_renameargs *arg = msg->rpc_argp;
2911         struct nfs_renameres *res = msg->rpc_resp;
2912
2913         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2914         res->server = server;
2915         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2916 }
2917
2918 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
2919 {
2920         if (nfs4_setup_sequence(NFS_SERVER(data->old_dir),
2921                                 &data->args.seq_args,
2922                                 &data->res.seq_res,
2923                                 task))
2924                 return;
2925         rpc_call_start(task);
2926 }
2927
2928 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2929                                  struct inode *new_dir)
2930 {
2931         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2932
2933         if (!nfs4_sequence_done(task, &res->seq_res))
2934                 return 0;
2935         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2936                 return 0;
2937
2938         update_changeattr(old_dir, &res->old_cinfo);
2939         update_changeattr(new_dir, &res->new_cinfo);
2940         return 1;
2941 }
2942
2943 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2944                 struct inode *new_dir, struct qstr *new_name)
2945 {
2946         struct nfs_server *server = NFS_SERVER(old_dir);
2947         struct nfs_renameargs arg = {
2948                 .old_dir = NFS_FH(old_dir),
2949                 .new_dir = NFS_FH(new_dir),
2950                 .old_name = old_name,
2951                 .new_name = new_name,
2952         };
2953         struct nfs_renameres res = {
2954                 .server = server,
2955         };
2956         struct rpc_message msg = {
2957                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2958                 .rpc_argp = &arg,
2959                 .rpc_resp = &res,
2960         };
2961         int status = -ENOMEM;
2962
2963         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2964         if (!status) {
2965                 update_changeattr(old_dir, &res.old_cinfo);
2966                 update_changeattr(new_dir, &res.new_cinfo);
2967         }
2968         return status;
2969 }
2970
2971 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2972                 struct inode *new_dir, struct qstr *new_name)
2973 {
2974         struct nfs4_exception exception = { };
2975         int err;
2976         do {
2977                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2978                                 _nfs4_proc_rename(old_dir, old_name,
2979                                         new_dir, new_name),
2980                                 &exception);
2981         } while (exception.retry);
2982         return err;
2983 }
2984
2985 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2986 {
2987         struct nfs_server *server = NFS_SERVER(inode);
2988         struct nfs4_link_arg arg = {
2989                 .fh     = NFS_FH(inode),
2990                 .dir_fh = NFS_FH(dir),
2991                 .name   = name,
2992                 .bitmask = server->attr_bitmask,
2993         };
2994         struct nfs4_link_res res = {
2995                 .server = server,
2996         };
2997         struct rpc_message msg = {
2998                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2999                 .rpc_argp = &arg,
3000                 .rpc_resp = &res,
3001         };
3002         int status = -ENOMEM;
3003
3004         res.fattr = nfs_alloc_fattr();
3005         if (res.fattr == NULL)
3006                 goto out;
3007
3008         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3009         if (!status) {
3010                 update_changeattr(dir, &res.cinfo);
3011                 nfs_post_op_update_inode(inode, res.fattr);
3012         }
3013 out:
3014         nfs_free_fattr(res.fattr);
3015         return status;
3016 }
3017
3018 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3019 {
3020         struct nfs4_exception exception = { };
3021         int err;
3022         do {
3023                 err = nfs4_handle_exception(NFS_SERVER(inode),
3024                                 _nfs4_proc_link(inode, dir, name),
3025                                 &exception);
3026         } while (exception.retry);
3027         return err;
3028 }
3029
3030 struct nfs4_createdata {
3031         struct rpc_message msg;
3032         struct nfs4_create_arg arg;
3033         struct nfs4_create_res res;
3034         struct nfs_fh fh;
3035         struct nfs_fattr fattr;
3036 };
3037
3038 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3039                 struct qstr *name, struct iattr *sattr, u32 ftype)
3040 {
3041         struct nfs4_createdata *data;
3042
3043         data = kzalloc(sizeof(*data), GFP_KERNEL);
3044         if (data != NULL) {
3045                 struct nfs_server *server = NFS_SERVER(dir);
3046
3047                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3048                 data->msg.rpc_argp = &data->arg;
3049                 data->msg.rpc_resp = &data->res;
3050                 data->arg.dir_fh = NFS_FH(dir);
3051                 data->arg.server = server;
3052                 data->arg.name = name;
3053                 data->arg.attrs = sattr;
3054                 data->arg.ftype = ftype;
3055                 data->arg.bitmask = server->attr_bitmask;
3056                 data->res.server = server;
3057                 data->res.fh = &data->fh;
3058                 data->res.fattr = &data->fattr;
3059                 nfs_fattr_init(data->res.fattr);
3060         }
3061         return data;
3062 }
3063
3064 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3065 {
3066         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3067                                     &data->arg.seq_args, &data->res.seq_res, 1);
3068         if (status == 0) {
3069                 update_changeattr(dir, &data->res.dir_cinfo);
3070                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3071         }
3072         return status;
3073 }
3074
3075 static void nfs4_free_createdata(struct nfs4_createdata *data)
3076 {
3077         kfree(data);
3078 }
3079
3080 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3081                 struct page *page, unsigned int len, struct iattr *sattr)
3082 {
3083         struct nfs4_createdata *data;
3084         int status = -ENAMETOOLONG;
3085
3086         if (len > NFS4_MAXPATHLEN)
3087                 goto out;
3088
3089         status = -ENOMEM;
3090         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3091         if (data == NULL)
3092                 goto out;
3093
3094         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3095         data->arg.u.symlink.pages = &page;
3096         data->arg.u.symlink.len = len;
3097
3098         status = nfs4_do_create(dir, dentry, data);
3099
3100         nfs4_free_createdata(data);
3101 out:
3102         return status;
3103 }
3104
3105 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3106                 struct page *page, unsigned int len, struct iattr *sattr)
3107 {
3108         struct nfs4_exception exception = { };
3109         int err;
3110         do {
3111                 err = nfs4_handle_exception(NFS_SERVER(dir),
3112                                 _nfs4_proc_symlink(dir, dentry, page,
3113                                                         len, sattr),
3114                                 &exception);
3115         } while (exception.retry);
3116         return err;
3117 }
3118
3119 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3120                 struct iattr *sattr)
3121 {
3122         struct nfs4_createdata *data;
3123         int status = -ENOMEM;
3124
3125         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3126         if (data == NULL)
3127                 goto out;
3128
3129         status = nfs4_do_create(dir, dentry, data);
3130
3131         nfs4_free_createdata(data);
3132 out:
3133         return status;
3134 }
3135
3136 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3137                 struct iattr *sattr)
3138 {
3139         struct nfs4_exception exception = { };
3140         int err;
3141
3142         sattr->ia_mode &= ~current_umask();
3143         do {
3144                 err = nfs4_handle_exception(NFS_SERVER(dir),
3145                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3146                                 &exception);
3147         } while (exception.retry);
3148         return err;
3149 }
3150
3151 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3152                 u64 cookie, struct page **pages, unsigned int count, int plus)
3153 {
3154         struct inode            *dir = dentry->d_inode;
3155         struct nfs4_readdir_arg args = {
3156                 .fh = NFS_FH(dir),
3157                 .pages = pages,
3158                 .pgbase = 0,
3159                 .count = count,
3160                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3161                 .plus = plus,
3162         };
3163         struct nfs4_readdir_res res;
3164         struct rpc_message msg = {
3165                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3166                 .rpc_argp = &args,
3167                 .rpc_resp = &res,
3168                 .rpc_cred = cred,
3169         };
3170         int                     status;
3171
3172         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3173                         dentry->d_parent->d_name.name,
3174                         dentry->d_name.name,
3175                         (unsigned long long)cookie);
3176         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3177         res.pgbase = args.pgbase;
3178         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3179         if (status >= 0) {
3180                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3181                 status += args.pgbase;
3182         }
3183
3184         nfs_invalidate_atime(dir);
3185
3186         dprintk("%s: returns %d\n", __func__, status);
3187         return status;
3188 }
3189
3190 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3191                 u64 cookie, struct page **pages, unsigned int count, int plus)
3192 {
3193         struct nfs4_exception exception = { };
3194         int err;
3195         do {
3196                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3197                                 _nfs4_proc_readdir(dentry, cred, cookie,
3198                                         pages, count, plus),
3199                                 &exception);
3200         } while (exception.retry);
3201         return err;
3202 }
3203
3204 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3205                 struct iattr *sattr, dev_t rdev)
3206 {
3207         struct nfs4_createdata *data;
3208         int mode = sattr->ia_mode;
3209         int status = -ENOMEM;
3210
3211         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3212         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3213
3214         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3215         if (data == NULL)
3216                 goto out;
3217
3218         if (S_ISFIFO(mode))
3219                 data->arg.ftype = NF4FIFO;
3220         else if (S_ISBLK(mode)) {
3221                 data->arg.ftype = NF4BLK;
3222                 data->arg.u.device.specdata1 = MAJOR(rdev);
3223                 data->arg.u.device.specdata2 = MINOR(rdev);
3224         }
3225         else if (S_ISCHR(mode)) {
3226                 data->arg.ftype = NF4CHR;
3227                 data->arg.u.device.specdata1 = MAJOR(rdev);
3228                 data->arg.u.device.specdata2 = MINOR(rdev);
3229         }
3230
3231         status = nfs4_do_create(dir, dentry, data);
3232
3233         nfs4_free_createdata(data);
3234 out:
3235         return status;
3236 }
3237
3238 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3239                 struct iattr *sattr, dev_t rdev)
3240 {
3241         struct nfs4_exception exception = { };
3242         int err;
3243
3244         sattr->ia_mode &= ~current_umask();
3245         do {
3246                 err = nfs4_handle_exception(NFS_SERVER(dir),
3247                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3248                                 &exception);
3249         } while (exception.retry);
3250         return err;
3251 }
3252
3253 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3254                  struct nfs_fsstat *fsstat)
3255 {
3256         struct nfs4_statfs_arg args = {
3257                 .fh = fhandle,
3258                 .bitmask = server->attr_bitmask,
3259         };
3260         struct nfs4_statfs_res res = {
3261                 .fsstat = fsstat,
3262         };
3263         struct rpc_message msg = {
3264                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3265                 .rpc_argp = &args,
3266                 .rpc_resp = &res,
3267         };
3268
3269         nfs_fattr_init(fsstat->fattr);
3270         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3271 }
3272
3273 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3274 {
3275         struct nfs4_exception exception = { };
3276         int err;
3277         do {
3278                 err = nfs4_handle_exception(server,
3279                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3280                                 &exception);
3281         } while (exception.retry);
3282         return err;
3283 }
3284
3285 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3286                 struct nfs_fsinfo *fsinfo)
3287 {
3288         struct nfs4_fsinfo_arg args = {
3289                 .fh = fhandle,
3290                 .bitmask = server->attr_bitmask,
3291         };
3292         struct nfs4_fsinfo_res res = {
3293                 .fsinfo = fsinfo,
3294         };
3295         struct rpc_message msg = {
3296                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3297                 .rpc_argp = &args,
3298                 .rpc_resp = &res,
3299         };
3300
3301         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3302 }
3303
3304 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3305 {
3306         struct nfs4_exception exception = { };
3307         int err;
3308
3309         do {
3310                 err = nfs4_handle_exception(server,
3311                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3312                                 &exception);
3313         } while (exception.retry);
3314         return err;
3315 }
3316
3317 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3318 {
3319         int error;
3320
3321         nfs_fattr_init(fsinfo->fattr);
3322         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3323         if (error == 0)
3324                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3325
3326         return error;
3327 }
3328
3329 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3330                 struct nfs_pathconf *pathconf)
3331 {
3332         struct nfs4_pathconf_arg args = {
3333                 .fh = fhandle,
3334                 .bitmask = server->attr_bitmask,
3335         };
3336         struct nfs4_pathconf_res res = {
3337                 .pathconf = pathconf,
3338         };
3339         struct rpc_message msg = {
3340                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3341                 .rpc_argp = &args,
3342                 .rpc_resp = &res,
3343         };
3344
3345         /* None of the pathconf attributes are mandatory to implement */
3346         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3347                 memset(pathconf, 0, sizeof(*pathconf));
3348                 return 0;
3349         }
3350
3351         nfs_fattr_init(pathconf->fattr);
3352         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3353 }
3354
3355 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3356                 struct nfs_pathconf *pathconf)
3357 {
3358         struct nfs4_exception exception = { };
3359         int err;
3360
3361         do {
3362                 err = nfs4_handle_exception(server,
3363                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3364                                 &exception);
3365         } while (exception.retry);
3366         return err;
3367 }
3368
3369 void __nfs4_read_done_cb(struct nfs_read_data *data)
3370 {
3371         nfs_invalidate_atime(data->header->inode);
3372 }
3373
3374 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3375 {
3376         struct nfs_server *server = NFS_SERVER(data->header->inode);
3377
3378         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3379                 rpc_restart_call_prepare(task);
3380                 return -EAGAIN;
3381         }
3382
3383         __nfs4_read_done_cb(data);
3384         if (task->tk_status > 0)
3385                 renew_lease(server, data->timestamp);
3386         return 0;
3387 }
3388
3389 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3390 {
3391
3392         dprintk("--> %s\n", __func__);
3393
3394         if (!nfs4_sequence_done(task, &data->res.seq_res))
3395                 return -EAGAIN;
3396
3397         return data->read_done_cb ? data->read_done_cb(task, data) :
3398                                     nfs4_read_done_cb(task, data);
3399 }
3400
3401 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3402 {
3403         data->timestamp   = jiffies;
3404         data->read_done_cb = nfs4_read_done_cb;
3405         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3406         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3407 }
3408
3409 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3410 {
3411         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3412                                 &data->args.seq_args,
3413                                 &data->res.seq_res,
3414                                 task))
3415                 return;
3416         rpc_call_start(task);
3417 }
3418
3419 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3420 {
3421         struct inode *inode = data->header->inode;
3422
3423         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3424                 rpc_restart_call_prepare(task);
3425                 return -EAGAIN;
3426         }
3427         if (task->tk_status >= 0) {
3428                 renew_lease(NFS_SERVER(inode), data->timestamp);
3429                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3430         }
3431         return 0;
3432 }
3433
3434 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3435 {
3436         if (!nfs4_sequence_done(task, &data->res.seq_res))
3437                 return -EAGAIN;
3438         return data->write_done_cb ? data->write_done_cb(task, data) :
3439                 nfs4_write_done_cb(task, data);
3440 }
3441
3442 static
3443 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3444 {
3445         const struct nfs_pgio_header *hdr = data->header;
3446
3447         /* Don't request attributes for pNFS or O_DIRECT writes */
3448         if (data->ds_clp != NULL || hdr->dreq != NULL)
3449                 return false;
3450         /* Otherwise, request attributes if and only if we don't hold
3451          * a delegation
3452          */
3453         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3454 }
3455
3456 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3457 {
3458         struct nfs_server *server = NFS_SERVER(data->header->inode);
3459
3460         if (!nfs4_write_need_cache_consistency_data(data)) {
3461                 data->args.bitmask = NULL;
3462                 data->res.fattr = NULL;
3463         } else
3464                 data->args.bitmask = server->cache_consistency_bitmask;
3465
3466         if (!data->write_done_cb)
3467                 data->write_done_cb = nfs4_write_done_cb;
3468         data->res.server = server;
3469         data->timestamp   = jiffies;
3470
3471         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3472         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3473 }
3474
3475 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3476 {
3477         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3478                                 &data->args.seq_args,
3479                                 &data->res.seq_res,
3480                                 task))
3481                 return;
3482         rpc_call_start(task);
3483 }
3484
3485 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3486 {
3487         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3488                                 &data->args.seq_args,
3489                                 &data->res.seq_res,
3490                                 task))
3491                 return;
3492         rpc_call_start(task);
3493 }
3494
3495 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3496 {
3497         struct inode *inode = data->inode;
3498
3499         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3500                 rpc_restart_call_prepare(task);
3501                 return -EAGAIN;
3502         }
3503         return 0;
3504 }
3505
3506 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3507 {
3508         if (!nfs4_sequence_done(task, &data->res.seq_res))
3509                 return -EAGAIN;
3510         return data->commit_done_cb(task, data);
3511 }
3512
3513 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3514 {
3515         struct nfs_server *server = NFS_SERVER(data->inode);
3516
3517         if (data->commit_done_cb == NULL)
3518                 data->commit_done_cb = nfs4_commit_done_cb;
3519         data->res.server = server;
3520         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3521         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3522 }
3523
3524 struct nfs4_renewdata {
3525         struct nfs_client       *client;
3526         unsigned long           timestamp;
3527 };
3528
3529 /*
3530  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3531  * standalone procedure for queueing an asynchronous RENEW.
3532  */
3533 static void nfs4_renew_release(void *calldata)
3534 {
3535         struct nfs4_renewdata *data = calldata;
3536         struct nfs_client *clp = data->client;
3537
3538         if (atomic_read(&clp->cl_count) > 1)
3539                 nfs4_schedule_state_renewal(clp);
3540         nfs_put_client(clp);
3541         kfree(data);
3542 }
3543
3544 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3545 {
3546         struct nfs4_renewdata *data = calldata;
3547         struct nfs_client *clp = data->client;
3548         unsigned long timestamp = data->timestamp;
3549
3550         if (task->tk_status < 0) {
3551                 /* Unless we're shutting down, schedule state recovery! */
3552                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3553                         return;
3554                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3555                         nfs4_schedule_lease_recovery(clp);
3556                         return;
3557                 }
3558                 nfs4_schedule_path_down_recovery(clp);
3559         }
3560         do_renew_lease(clp, timestamp);
3561 }
3562
3563 static const struct rpc_call_ops nfs4_renew_ops = {
3564         .rpc_call_done = nfs4_renew_done,
3565         .rpc_release = nfs4_renew_release,
3566 };
3567
3568 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3569 {
3570         struct rpc_message msg = {
3571                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3572                 .rpc_argp       = clp,
3573                 .rpc_cred       = cred,
3574         };
3575         struct nfs4_renewdata *data;
3576
3577         if (renew_flags == 0)
3578                 return 0;
3579         if (!atomic_inc_not_zero(&clp->cl_count))
3580                 return -EIO;
3581         data = kmalloc(sizeof(*data), GFP_NOFS);
3582         if (data == NULL)
3583                 return -ENOMEM;
3584         data->client = clp;
3585         data->timestamp = jiffies;
3586         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3587                         &nfs4_renew_ops, data);
3588 }
3589
3590 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3591 {
3592         struct rpc_message msg = {
3593                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3594                 .rpc_argp       = clp,
3595                 .rpc_cred       = cred,
3596         };
3597         unsigned long now = jiffies;
3598         int status;
3599
3600         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3601         if (status < 0)
3602                 return status;
3603         do_renew_lease(clp, now);
3604         return 0;
3605 }
3606
3607 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3608 {
3609         return (server->caps & NFS_CAP_ACLS)
3610                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3611                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3612 }
3613
3614 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3615  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3616  * the stack.
3617  */
3618 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3619
3620 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3621                 struct page **pages, unsigned int *pgbase)
3622 {
3623         struct page *newpage, **spages;
3624         int rc = 0;
3625         size_t len;
3626         spages = pages;
3627
3628         do {
3629                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3630                 newpage = alloc_page(GFP_KERNEL);
3631
3632                 if (newpage == NULL)
3633                         goto unwind;
3634                 memcpy(page_address(newpage), buf, len);
3635                 buf += len;
3636                 buflen -= len;
3637                 *pages++ = newpage;
3638                 rc++;
3639         } while (buflen != 0);
3640
3641         return rc;
3642
3643 unwind:
3644         for(; rc > 0; rc--)
3645                 __free_page(spages[rc-1]);
3646         return -ENOMEM;
3647 }
3648
3649 struct nfs4_cached_acl {
3650         int cached;
3651         size_t len;
3652         char data[0];
3653 };
3654
3655 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3656 {
3657         struct nfs_inode *nfsi = NFS_I(inode);
3658
3659         spin_lock(&inode->i_lock);
3660         kfree(nfsi->nfs4_acl);
3661         nfsi->nfs4_acl = acl;
3662         spin_unlock(&inode->i_lock);
3663 }
3664
3665 static void nfs4_zap_acl_attr(struct inode *inode)
3666 {
3667         nfs4_set_cached_acl(inode, NULL);
3668 }
3669
3670 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3671 {
3672         struct nfs_inode *nfsi = NFS_I(inode);
3673         struct nfs4_cached_acl *acl;
3674         int ret = -ENOENT;
3675
3676         spin_lock(&inode->i_lock);
3677         acl = nfsi->nfs4_acl;
3678         if (acl == NULL)
3679                 goto out;
3680         if (buf == NULL) /* user is just asking for length */
3681                 goto out_len;
3682         if (acl->cached == 0)
3683                 goto out;
3684         ret = -ERANGE; /* see getxattr(2) man page */
3685         if (acl->len > buflen)
3686                 goto out;
3687         memcpy(buf, acl->data, acl->len);
3688 out_len:
3689         ret = acl->len;
3690 out:
3691         spin_unlock(&inode->i_lock);
3692         return ret;
3693 }
3694
3695 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3696 {
3697         struct nfs4_cached_acl *acl;
3698
3699         if (pages && acl_len <= PAGE_SIZE) {
3700                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3701                 if (acl == NULL)
3702                         goto out;
3703                 acl->cached = 1;
3704                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3705         } else {
3706                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3707                 if (acl == NULL)
3708                         goto out;
3709                 acl->cached = 0;
3710         }
3711         acl->len = acl_len;
3712 out:
3713         nfs4_set_cached_acl(inode, acl);
3714 }
3715
3716 /*
3717  * The getxattr API returns the required buffer length when called with a
3718  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3719  * the required buf.  On a NULL buf, we send a page of data to the server
3720  * guessing that the ACL request can be serviced by a page. If so, we cache
3721  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3722  * the cache. If not so, we throw away the page, and cache the required
3723  * length. The next getxattr call will then produce another round trip to
3724  * the server, this time with the input buf of the required size.
3725  */
3726 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3727 {
3728         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3729         struct nfs_getaclargs args = {
3730                 .fh = NFS_FH(inode),
3731                 .acl_pages = pages,
3732                 .acl_len = buflen,
3733         };
3734         struct nfs_getaclres res = {
3735                 .acl_len = buflen,
3736         };
3737         struct rpc_message msg = {
3738                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3739                 .rpc_argp = &args,
3740                 .rpc_resp = &res,
3741         };
3742         int ret = -ENOMEM, npages, i, acl_len = 0;
3743
3744         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3745         /* As long as we're doing a round trip to the server anyway,
3746          * let's be prepared for a page of acl data. */
3747         if (npages == 0)
3748                 npages = 1;
3749
3750         /* Add an extra page to handle the bitmap returned */
3751         npages++;
3752
3753         for (i = 0; i < npages; i++) {
3754                 pages[i] = alloc_page(GFP_KERNEL);
3755                 if (!pages[i])
3756                         goto out_free;
3757         }
3758
3759         /* for decoding across pages */
3760         res.acl_scratch = alloc_page(GFP_KERNEL);
3761         if (!res.acl_scratch)
3762                 goto out_free;
3763
3764         args.acl_len = npages * PAGE_SIZE;
3765         args.acl_pgbase = 0;
3766
3767         /* Let decode_getfacl know not to fail if the ACL data is larger than
3768          * the page we send as a guess */
3769         if (buf == NULL)
3770                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3771
3772         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3773                 __func__, buf, buflen, npages, args.acl_len);
3774         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3775                              &msg, &args.seq_args, &res.seq_res, 0);
3776         if (ret)
3777                 goto out_free;
3778
3779         acl_len = res.acl_len - res.acl_data_offset;
3780         if (acl_len > args.acl_len)
3781                 nfs4_write_cached_acl(inode, NULL, 0, acl_len);
3782         else
3783                 nfs4_write_cached_acl(inode, pages, res.acl_data_offset,
3784                                       acl_len);
3785         if (buf) {
3786                 ret = -ERANGE;
3787                 if (acl_len > buflen)
3788                         goto out_free;
3789                 _copy_from_pages(buf, pages, res.acl_data_offset,
3790                                 acl_len);
3791         }
3792         ret = acl_len;
3793 out_free:
3794         for (i = 0; i < npages; i++)
3795                 if (pages[i])
3796                         __free_page(pages[i]);
3797         if (res.acl_scratch)
3798                 __free_page(res.acl_scratch);
3799         return ret;
3800 }
3801
3802 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3803 {
3804         struct nfs4_exception exception = { };
3805         ssize_t ret;
3806         do {
3807                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3808                 if (ret >= 0)
3809                         break;
3810                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3811         } while (exception.retry);
3812         return ret;
3813 }
3814
3815 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3816 {
3817         struct nfs_server *server = NFS_SERVER(inode);
3818         int ret;
3819
3820         if (!nfs4_server_supports_acls(server))
3821                 return -EOPNOTSUPP;
3822         ret = nfs_revalidate_inode(server, inode);
3823         if (ret < 0)
3824                 return ret;
3825         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3826                 nfs_zap_acl_cache(inode);
3827         ret = nfs4_read_cached_acl(inode, buf, buflen);
3828         if (ret != -ENOENT)
3829                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3830                  * but no cached acl data, just the acl length */
3831                 return ret;
3832         return nfs4_get_acl_uncached(inode, buf, buflen);
3833 }
3834
3835 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3836 {
3837         struct nfs_server *server = NFS_SERVER(inode);
3838         struct page *pages[NFS4ACL_MAXPAGES];
3839         struct nfs_setaclargs arg = {
3840                 .fh             = NFS_FH(inode),
3841                 .acl_pages      = pages,
3842                 .acl_len        = buflen,
3843         };
3844         struct nfs_setaclres res;
3845         struct rpc_message msg = {
3846                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3847                 .rpc_argp       = &arg,
3848                 .rpc_resp       = &res,
3849         };
3850         int ret, i;
3851
3852         if (!nfs4_server_supports_acls(server))
3853                 return -EOPNOTSUPP;
3854         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3855         if (i < 0)
3856                 return i;
3857         nfs4_inode_return_delegation(inode);
3858         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3859
3860         /*
3861          * Free each page after tx, so the only ref left is
3862          * held by the network stack
3863          */
3864         for (; i > 0; i--)
3865                 put_page(pages[i-1]);
3866
3867         /*
3868          * Acl update can result in inode attribute update.
3869          * so mark the attribute cache invalid.
3870          */
3871         spin_lock(&inode->i_lock);
3872         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3873         spin_unlock(&inode->i_lock);
3874         nfs_access_zap_cache(inode);
3875         nfs_zap_acl_cache(inode);
3876         return ret;
3877 }
3878
3879 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3880 {
3881         struct nfs4_exception exception = { };
3882         int err;
3883         do {
3884                 err = nfs4_handle_exception(NFS_SERVER(inode),
3885                                 __nfs4_proc_set_acl(inode, buf, buflen),
3886                                 &exception);
3887         } while (exception.retry);
3888         return err;
3889 }
3890
3891 static int
3892 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3893 {
3894         struct nfs_client *clp = server->nfs_client;
3895
3896         if (task->tk_status >= 0)
3897                 return 0;
3898         switch(task->tk_status) {
3899                 case -NFS4ERR_DELEG_REVOKED:
3900                 case -NFS4ERR_ADMIN_REVOKED:
3901                 case -NFS4ERR_BAD_STATEID:
3902                         if (state == NULL)
3903                                 break;
3904                         nfs_remove_bad_delegation(state->inode);
3905                 case -NFS4ERR_OPENMODE:
3906                         if (state == NULL)
3907                                 break;
3908                         nfs4_schedule_stateid_recovery(server, state);
3909                         goto wait_on_recovery;
3910                 case -NFS4ERR_EXPIRED:
3911                         if (state != NULL)
3912                                 nfs4_schedule_stateid_recovery(server, state);
3913                 case -NFS4ERR_STALE_STATEID:
3914                 case -NFS4ERR_STALE_CLIENTID:
3915                         nfs4_schedule_lease_recovery(clp);
3916                         goto wait_on_recovery;
3917 #if defined(CONFIG_NFS_V4_1)
3918                 case -NFS4ERR_BADSESSION:
3919                 case -NFS4ERR_BADSLOT:
3920                 case -NFS4ERR_BAD_HIGH_SLOT:
3921                 case -NFS4ERR_DEADSESSION:
3922                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3923                 case -NFS4ERR_SEQ_FALSE_RETRY:
3924                 case -NFS4ERR_SEQ_MISORDERED:
3925                         dprintk("%s ERROR %d, Reset session\n", __func__,
3926                                 task->tk_status);
3927                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
3928                         task->tk_status = 0;
3929                         return -EAGAIN;
3930 #endif /* CONFIG_NFS_V4_1 */
3931                 case -NFS4ERR_DELAY:
3932                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3933                 case -NFS4ERR_GRACE:
3934                 case -EKEYEXPIRED:
3935                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3936                         task->tk_status = 0;
3937                         return -EAGAIN;
3938                 case -NFS4ERR_RETRY_UNCACHED_REP:
3939                 case -NFS4ERR_OLD_STATEID:
3940                         task->tk_status = 0;
3941                         return -EAGAIN;
3942         }
3943         task->tk_status = nfs4_map_errors(task->tk_status);
3944         return 0;
3945 wait_on_recovery:
3946         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3947         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3948                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3949         task->tk_status = 0;
3950         return -EAGAIN;
3951 }
3952
3953 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
3954                                     nfs4_verifier *bootverf)
3955 {
3956         __be32 verf[2];
3957
3958         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
3959                 /* An impossible timestamp guarantees this value
3960                  * will never match a generated boot time. */
3961                 verf[0] = 0;
3962                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
3963         } else {
3964                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
3965                 verf[0] = (__be32)nn->boot_time.tv_sec;
3966                 verf[1] = (__be32)nn->boot_time.tv_nsec;
3967         }
3968         memcpy(bootverf->data, verf, sizeof(bootverf->data));
3969 }
3970
3971 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3972                 unsigned short port, struct rpc_cred *cred,
3973                 struct nfs4_setclientid_res *res)
3974 {
3975         nfs4_verifier sc_verifier;
3976         struct nfs4_setclientid setclientid = {
3977                 .sc_verifier = &sc_verifier,
3978                 .sc_prog = program,
3979                 .sc_cb_ident = clp->cl_cb_ident,
3980         };
3981         struct rpc_message msg = {
3982                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3983                 .rpc_argp = &setclientid,
3984                 .rpc_resp = res,
3985                 .rpc_cred = cred,
3986         };
3987         int loop = 0;
3988         int status;
3989
3990         nfs4_init_boot_verifier(clp, &sc_verifier);
3991
3992         for(;;) {
3993                 rcu_read_lock();
3994                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3995                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3996                                 clp->cl_ipaddr,
3997                                 rpc_peeraddr2str(clp->cl_rpcclient,
3998                                                         RPC_DISPLAY_ADDR),
3999                                 rpc_peeraddr2str(clp->cl_rpcclient,
4000                                                         RPC_DISPLAY_PROTO),
4001                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4002                                 clp->cl_id_uniquifier);
4003                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4004                                 sizeof(setclientid.sc_netid),
4005                                 rpc_peeraddr2str(clp->cl_rpcclient,
4006                                                         RPC_DISPLAY_NETID));
4007                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4008                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4009                                 clp->cl_ipaddr, port >> 8, port & 255);
4010                 rcu_read_unlock();
4011
4012                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4013                 if (status != -NFS4ERR_CLID_INUSE)
4014                         break;
4015                 if (loop != 0) {
4016                         ++clp->cl_id_uniquifier;
4017                         break;
4018                 }
4019                 ++loop;
4020                 ssleep(clp->cl_lease_time / HZ + 1);
4021         }
4022         return status;
4023 }
4024
4025 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4026                 struct nfs4_setclientid_res *arg,
4027                 struct rpc_cred *cred)
4028 {
4029         struct nfs_fsinfo fsinfo;
4030         struct rpc_message msg = {
4031                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4032                 .rpc_argp = arg,
4033                 .rpc_resp = &fsinfo,
4034                 .rpc_cred = cred,
4035         };
4036         unsigned long now;
4037         int status;
4038
4039         now = jiffies;
4040         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4041         if (status == 0) {
4042                 spin_lock(&clp->cl_lock);
4043                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4044                 clp->cl_last_renewal = now;
4045                 spin_unlock(&clp->cl_lock);
4046         }
4047         return status;
4048 }
4049
4050 struct nfs4_delegreturndata {
4051         struct nfs4_delegreturnargs args;
4052         struct nfs4_delegreturnres res;
4053         struct nfs_fh fh;
4054         nfs4_stateid stateid;
4055         unsigned long timestamp;
4056         struct nfs_fattr fattr;
4057         int rpc_status;
4058 };
4059
4060 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4061 {
4062         struct nfs4_delegreturndata *data = calldata;
4063
4064         if (!nfs4_sequence_done(task, &data->res.seq_res))
4065                 return;
4066
4067         switch (task->tk_status) {
4068         case -NFS4ERR_STALE_STATEID:
4069         case -NFS4ERR_EXPIRED:
4070         case 0:
4071                 renew_lease(data->res.server, data->timestamp);
4072                 break;
4073         default:
4074                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4075                                 -EAGAIN) {
4076                         rpc_restart_call_prepare(task);
4077                         return;
4078                 }
4079         }
4080         data->rpc_status = task->tk_status;
4081 }
4082
4083 static void nfs4_delegreturn_release(void *calldata)
4084 {
4085         kfree(calldata);
4086 }
4087
4088 #if defined(CONFIG_NFS_V4_1)
4089 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4090 {
4091         struct nfs4_delegreturndata *d_data;
4092
4093         d_data = (struct nfs4_delegreturndata *)data;
4094
4095         if (nfs4_setup_sequence(d_data->res.server,
4096                                 &d_data->args.seq_args,
4097                                 &d_data->res.seq_res, task))
4098                 return;
4099         rpc_call_start(task);
4100 }
4101 #endif /* CONFIG_NFS_V4_1 */
4102
4103 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4104 #if defined(CONFIG_NFS_V4_1)
4105         .rpc_call_prepare = nfs4_delegreturn_prepare,
4106 #endif /* CONFIG_NFS_V4_1 */
4107         .rpc_call_done = nfs4_delegreturn_done,
4108         .rpc_release = nfs4_delegreturn_release,
4109 };
4110
4111 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4112 {
4113         struct nfs4_delegreturndata *data;
4114         struct nfs_server *server = NFS_SERVER(inode);
4115         struct rpc_task *task;
4116         struct rpc_message msg = {
4117                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4118                 .rpc_cred = cred,
4119         };
4120         struct rpc_task_setup task_setup_data = {
4121                 .rpc_client = server->client,
4122                 .rpc_message = &msg,
4123                 .callback_ops = &nfs4_delegreturn_ops,
4124                 .flags = RPC_TASK_ASYNC,
4125         };
4126         int status = 0;
4127
4128         data = kzalloc(sizeof(*data), GFP_NOFS);
4129         if (data == NULL)
4130                 return -ENOMEM;
4131         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4132         data->args.fhandle = &data->fh;
4133         data->args.stateid = &data->stateid;
4134         data->args.bitmask = server->cache_consistency_bitmask;
4135         nfs_copy_fh(&data->fh, NFS_FH(inode));
4136         nfs4_stateid_copy(&data->stateid, stateid);
4137         data->res.fattr = &data->fattr;
4138         data->res.server = server;
4139         nfs_fattr_init(data->res.fattr);
4140         data->timestamp = jiffies;
4141         data->rpc_status = 0;
4142
4143         task_setup_data.callback_data = data;
4144         msg.rpc_argp = &data->args;
4145         msg.rpc_resp = &data->res;
4146         task = rpc_run_task(&task_setup_data);
4147         if (IS_ERR(task))
4148                 return PTR_ERR(task);
4149         if (!issync)
4150                 goto out;
4151         status = nfs4_wait_for_completion_rpc_task(task);
4152         if (status != 0)
4153                 goto out;
4154         status = data->rpc_status;
4155         if (status == 0)
4156                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4157         else
4158                 nfs_refresh_inode(inode, &data->fattr);
4159 out:
4160         rpc_put_task(task);
4161         return status;
4162 }
4163
4164 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4165 {
4166         struct nfs_server *server = NFS_SERVER(inode);
4167         struct nfs4_exception exception = { };
4168         int err;
4169         do {
4170                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4171                 switch (err) {
4172                         case -NFS4ERR_STALE_STATEID:
4173                         case -NFS4ERR_EXPIRED:
4174                         case 0:
4175                                 return 0;
4176                 }
4177                 err = nfs4_handle_exception(server, err, &exception);
4178         } while (exception.retry);
4179         return err;
4180 }
4181
4182 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4183 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4184
4185 /*
4186  * sleep, with exponential backoff, and retry the LOCK operation.
4187  */
4188 static unsigned long
4189 nfs4_set_lock_task_retry(unsigned long timeout)
4190 {
4191         freezable_schedule_timeout_killable(timeout);
4192         timeout <<= 1;
4193         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4194                 return NFS4_LOCK_MAXTIMEOUT;
4195         return timeout;
4196 }
4197
4198 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4199 {
4200         struct inode *inode = state->inode;
4201         struct nfs_server *server = NFS_SERVER(inode);
4202         struct nfs_client *clp = server->nfs_client;
4203         struct nfs_lockt_args arg = {
4204                 .fh = NFS_FH(inode),
4205                 .fl = request,
4206         };
4207         struct nfs_lockt_res res = {
4208                 .denied = request,
4209         };
4210         struct rpc_message msg = {
4211                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4212                 .rpc_argp       = &arg,
4213                 .rpc_resp       = &res,
4214                 .rpc_cred       = state->owner->so_cred,
4215         };
4216         struct nfs4_lock_state *lsp;
4217         int status;
4218
4219         arg.lock_owner.clientid = clp->cl_clientid;
4220         status = nfs4_set_lock_state(state, request);
4221         if (status != 0)
4222                 goto out;
4223         lsp = request->fl_u.nfs4_fl.owner;
4224         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4225         arg.lock_owner.s_dev = server->s_dev;
4226         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4227         switch (status) {
4228                 case 0:
4229                         request->fl_type = F_UNLCK;
4230                         break;
4231                 case -NFS4ERR_DENIED:
4232                         status = 0;
4233         }
4234         request->fl_ops->fl_release_private(request);
4235 out:
4236         return status;
4237 }
4238
4239 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4240 {
4241         struct nfs4_exception exception = { };
4242         int err;
4243
4244         do {
4245                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4246                                 _nfs4_proc_getlk(state, cmd, request),
4247                                 &exception);
4248         } while (exception.retry);
4249         return err;
4250 }
4251
4252 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4253 {
4254         int res = 0;
4255         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4256                 case FL_POSIX:
4257                         res = posix_lock_file_wait(file, fl);
4258                         break;
4259                 case FL_FLOCK:
4260                         res = flock_lock_file_wait(file, fl);
4261                         break;
4262                 default:
4263                         BUG();
4264         }
4265         return res;
4266 }
4267
4268 struct nfs4_unlockdata {
4269         struct nfs_locku_args arg;
4270         struct nfs_locku_res res;
4271         struct nfs4_lock_state *lsp;
4272         struct nfs_open_context *ctx;
4273         struct file_lock fl;
4274         const struct nfs_server *server;
4275         unsigned long timestamp;
4276 };
4277
4278 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4279                 struct nfs_open_context *ctx,
4280                 struct nfs4_lock_state *lsp,
4281                 struct nfs_seqid *seqid)
4282 {
4283         struct nfs4_unlockdata *p;
4284         struct inode *inode = lsp->ls_state->inode;
4285
4286         p = kzalloc(sizeof(*p), GFP_NOFS);
4287         if (p == NULL)
4288                 return NULL;
4289         p->arg.fh = NFS_FH(inode);
4290         p->arg.fl = &p->fl;
4291         p->arg.seqid = seqid;
4292         p->res.seqid = seqid;
4293         p->arg.stateid = &lsp->ls_stateid;
4294         p->lsp = lsp;
4295         atomic_inc(&lsp->ls_count);
4296         /* Ensure we don't close file until we're done freeing locks! */
4297         p->ctx = get_nfs_open_context(ctx);
4298         memcpy(&p->fl, fl, sizeof(p->fl));
4299         p->server = NFS_SERVER(inode);
4300         return p;
4301 }
4302
4303 static void nfs4_locku_release_calldata(void *data)
4304 {
4305         struct nfs4_unlockdata *calldata = data;
4306         nfs_free_seqid(calldata->arg.seqid);
4307         nfs4_put_lock_state(calldata->lsp);
4308         put_nfs_open_context(calldata->ctx);
4309         kfree(calldata);
4310 }
4311
4312 static void nfs4_locku_done(struct rpc_task *task, void *data)
4313 {
4314         struct nfs4_unlockdata *calldata = data;
4315
4316         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4317                 return;
4318         switch (task->tk_status) {
4319                 case 0:
4320                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4321                                         &calldata->res.stateid);
4322                         renew_lease(calldata->server, calldata->timestamp);
4323                         break;
4324                 case -NFS4ERR_BAD_STATEID:
4325                 case -NFS4ERR_OLD_STATEID:
4326                 case -NFS4ERR_STALE_STATEID:
4327                 case -NFS4ERR_EXPIRED:
4328                         break;
4329                 default:
4330                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4331                                 rpc_restart_call_prepare(task);
4332         }
4333 }
4334
4335 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4336 {
4337         struct nfs4_unlockdata *calldata = data;
4338
4339         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4340                 return;
4341         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4342                 /* Note: exit _without_ running nfs4_locku_done */
4343                 task->tk_action = NULL;
4344                 return;
4345         }
4346         calldata->timestamp = jiffies;
4347         if (nfs4_setup_sequence(calldata->server,
4348                                 &calldata->arg.seq_args,
4349                                 &calldata->res.seq_res, task))
4350                 return;
4351         rpc_call_start(task);
4352 }
4353
4354 static const struct rpc_call_ops nfs4_locku_ops = {
4355         .rpc_call_prepare = nfs4_locku_prepare,
4356         .rpc_call_done = nfs4_locku_done,
4357         .rpc_release = nfs4_locku_release_calldata,
4358 };
4359
4360 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4361                 struct nfs_open_context *ctx,
4362                 struct nfs4_lock_state *lsp,
4363                 struct nfs_seqid *seqid)
4364 {
4365         struct nfs4_unlockdata *data;
4366         struct rpc_message msg = {
4367                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4368                 .rpc_cred = ctx->cred,
4369         };
4370         struct rpc_task_setup task_setup_data = {
4371                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4372                 .rpc_message = &msg,
4373                 .callback_ops = &nfs4_locku_ops,
4374                 .workqueue = nfsiod_workqueue,
4375                 .flags = RPC_TASK_ASYNC,
4376         };
4377
4378         /* Ensure this is an unlock - when canceling a lock, the
4379          * canceled lock is passed in, and it won't be an unlock.
4380          */
4381         fl->fl_type = F_UNLCK;
4382
4383         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4384         if (data == NULL) {
4385                 nfs_free_seqid(seqid);
4386                 return ERR_PTR(-ENOMEM);
4387         }
4388
4389         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4390         msg.rpc_argp = &data->arg;
4391         msg.rpc_resp = &data->res;
4392         task_setup_data.callback_data = data;
4393         return rpc_run_task(&task_setup_data);
4394 }
4395
4396 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4397 {
4398         struct nfs_inode *nfsi = NFS_I(state->inode);
4399         struct nfs_seqid *seqid;
4400         struct nfs4_lock_state *lsp;
4401         struct rpc_task *task;
4402         int status = 0;
4403         unsigned char fl_flags = request->fl_flags;
4404
4405         status = nfs4_set_lock_state(state, request);
4406         /* Unlock _before_ we do the RPC call */
4407         request->fl_flags |= FL_EXISTS;
4408         down_read(&nfsi->rwsem);
4409         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4410                 up_read(&nfsi->rwsem);
4411                 goto out;
4412         }
4413         up_read(&nfsi->rwsem);
4414         if (status != 0)
4415                 goto out;
4416         /* Is this a delegated lock? */
4417         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4418                 goto out;
4419         lsp = request->fl_u.nfs4_fl.owner;
4420         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4421         status = -ENOMEM;
4422         if (seqid == NULL)
4423                 goto out;
4424         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4425         status = PTR_ERR(task);
4426         if (IS_ERR(task))
4427                 goto out;
4428         status = nfs4_wait_for_completion_rpc_task(task);
4429         rpc_put_task(task);
4430 out:
4431         request->fl_flags = fl_flags;
4432         return status;
4433 }
4434
4435 struct nfs4_lockdata {
4436         struct nfs_lock_args arg;
4437         struct nfs_lock_res res;
4438         struct nfs4_lock_state *lsp;
4439         struct nfs_open_context *ctx;
4440         struct file_lock fl;
4441         unsigned long timestamp;
4442         int rpc_status;
4443         int cancelled;
4444         struct nfs_server *server;
4445 };
4446
4447 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4448                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4449                 gfp_t gfp_mask)
4450 {
4451         struct nfs4_lockdata *p;
4452         struct inode *inode = lsp->ls_state->inode;
4453         struct nfs_server *server = NFS_SERVER(inode);
4454
4455         p = kzalloc(sizeof(*p), gfp_mask);
4456         if (p == NULL)
4457                 return NULL;
4458
4459         p->arg.fh = NFS_FH(inode);
4460         p->arg.fl = &p->fl;
4461         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4462         if (p->arg.open_seqid == NULL)
4463                 goto out_free;
4464         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4465         if (p->arg.lock_seqid == NULL)
4466                 goto out_free_seqid;
4467         p->arg.lock_stateid = &lsp->ls_stateid;
4468         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4469         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4470         p->arg.lock_owner.s_dev = server->s_dev;
4471         p->res.lock_seqid = p->arg.lock_seqid;
4472         p->lsp = lsp;
4473         p->server = server;
4474         atomic_inc(&lsp->ls_count);
4475         p->ctx = get_nfs_open_context(ctx);
4476         memcpy(&p->fl, fl, sizeof(p->fl));
4477         return p;
4478 out_free_seqid:
4479         nfs_free_seqid(p->arg.open_seqid);
4480 out_free:
4481         kfree(p);
4482         return NULL;
4483 }
4484
4485 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4486 {
4487         struct nfs4_lockdata *data = calldata;
4488         struct nfs4_state *state = data->lsp->ls_state;
4489
4490         dprintk("%s: begin!\n", __func__);
4491         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4492                 return;
4493         /* Do we need to do an open_to_lock_owner? */
4494         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4495                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4496                         return;
4497                 data->arg.open_stateid = &state->stateid;
4498                 data->arg.new_lock_owner = 1;
4499                 data->res.open_seqid = data->arg.open_seqid;
4500         } else
4501                 data->arg.new_lock_owner = 0;
4502         data->timestamp = jiffies;
4503         if (nfs4_setup_sequence(data->server,
4504                                 &data->arg.seq_args,
4505                                 &data->res.seq_res, task))
4506                 return;
4507         rpc_call_start(task);
4508         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4509 }
4510
4511 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4512 {
4513         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4514         nfs4_lock_prepare(task, calldata);
4515 }
4516
4517 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4518 {
4519         struct nfs4_lockdata *data = calldata;
4520
4521         dprintk("%s: begin!\n", __func__);
4522
4523         if (!nfs4_sequence_done(task, &data->res.seq_res))
4524                 return;
4525
4526         data->rpc_status = task->tk_status;
4527         if (data->arg.new_lock_owner != 0) {
4528                 if (data->rpc_status == 0)
4529                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4530                 else
4531                         goto out;
4532         }
4533         if (data->rpc_status == 0) {
4534                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4535                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4536                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4537         }
4538 out:
4539         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4540 }
4541
4542 static void nfs4_lock_release(void *calldata)
4543 {
4544         struct nfs4_lockdata *data = calldata;
4545
4546         dprintk("%s: begin!\n", __func__);
4547         nfs_free_seqid(data->arg.open_seqid);
4548         if (data->cancelled != 0) {
4549                 struct rpc_task *task;
4550                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4551                                 data->arg.lock_seqid);
4552                 if (!IS_ERR(task))
4553                         rpc_put_task_async(task);
4554                 dprintk("%s: cancelling lock!\n", __func__);
4555         } else
4556                 nfs_free_seqid(data->arg.lock_seqid);
4557         nfs4_put_lock_state(data->lsp);
4558         put_nfs_open_context(data->ctx);
4559         kfree(data);
4560         dprintk("%s: done!\n", __func__);
4561 }
4562
4563 static const struct rpc_call_ops nfs4_lock_ops = {
4564         .rpc_call_prepare = nfs4_lock_prepare,
4565         .rpc_call_done = nfs4_lock_done,
4566         .rpc_release = nfs4_lock_release,
4567 };
4568
4569 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4570         .rpc_call_prepare = nfs4_recover_lock_prepare,
4571         .rpc_call_done = nfs4_lock_done,
4572         .rpc_release = nfs4_lock_release,
4573 };
4574
4575 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4576 {
4577         switch (error) {
4578         case -NFS4ERR_ADMIN_REVOKED:
4579         case -NFS4ERR_BAD_STATEID:
4580                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4581                 if (new_lock_owner != 0 ||
4582                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4583                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4584                 break;
4585         case -NFS4ERR_STALE_STATEID:
4586                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4587         case -NFS4ERR_EXPIRED:
4588                 nfs4_schedule_lease_recovery(server->nfs_client);
4589         };
4590 }
4591
4592 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4593 {
4594         struct nfs4_lockdata *data;
4595         struct rpc_task *task;
4596         struct rpc_message msg = {
4597                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4598                 .rpc_cred = state->owner->so_cred,
4599         };
4600         struct rpc_task_setup task_setup_data = {
4601                 .rpc_client = NFS_CLIENT(state->inode),
4602                 .rpc_message = &msg,
4603                 .callback_ops = &nfs4_lock_ops,
4604                 .workqueue = nfsiod_workqueue,
4605                 .flags = RPC_TASK_ASYNC,
4606         };
4607         int ret;
4608
4609         dprintk("%s: begin!\n", __func__);
4610         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4611                         fl->fl_u.nfs4_fl.owner,
4612                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4613         if (data == NULL)
4614                 return -ENOMEM;
4615         if (IS_SETLKW(cmd))
4616                 data->arg.block = 1;
4617         if (recovery_type > NFS_LOCK_NEW) {
4618                 if (recovery_type == NFS_LOCK_RECLAIM)
4619                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4620                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4621         }
4622         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4623         msg.rpc_argp = &data->arg;
4624         msg.rpc_resp = &data->res;
4625         task_setup_data.callback_data = data;
4626         task = rpc_run_task(&task_setup_data);
4627         if (IS_ERR(task))
4628                 return PTR_ERR(task);
4629         ret = nfs4_wait_for_completion_rpc_task(task);
4630         if (ret == 0) {
4631                 ret = data->rpc_status;
4632                 if (ret)
4633                         nfs4_handle_setlk_error(data->server, data->lsp,
4634                                         data->arg.new_lock_owner, ret);
4635         } else
4636                 data->cancelled = 1;
4637         rpc_put_task(task);
4638         dprintk("%s: done, ret = %d!\n", __func__, ret);
4639         return ret;
4640 }
4641
4642 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4643 {
4644         struct nfs_server *server = NFS_SERVER(state->inode);
4645         struct nfs4_exception exception = {
4646                 .inode = state->inode,
4647         };
4648         int err;
4649
4650         do {
4651                 /* Cache the lock if possible... */
4652                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4653                         return 0;
4654                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4655                 if (err != -NFS4ERR_DELAY)
4656                         break;
4657                 nfs4_handle_exception(server, err, &exception);
4658         } while (exception.retry);
4659         return err;
4660 }
4661
4662 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4663 {
4664         struct nfs_server *server = NFS_SERVER(state->inode);
4665         struct nfs4_exception exception = {
4666                 .inode = state->inode,
4667         };
4668         int err;
4669
4670         err = nfs4_set_lock_state(state, request);
4671         if (err != 0)
4672                 return err;
4673         do {
4674                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4675                         return 0;
4676                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4677                 switch (err) {
4678                 default:
4679                         goto out;
4680                 case -NFS4ERR_GRACE:
4681                 case -NFS4ERR_DELAY:
4682                         nfs4_handle_exception(server, err, &exception);
4683                         err = 0;
4684                 }
4685         } while (exception.retry);
4686 out:
4687         return err;
4688 }
4689
4690 #if defined(CONFIG_NFS_V4_1)
4691 static int nfs41_check_expired_locks(struct nfs4_state *state)
4692 {
4693         int status, ret = NFS_OK;
4694         struct nfs4_lock_state *lsp;
4695         struct nfs_server *server = NFS_SERVER(state->inode);
4696
4697         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4698                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4699                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4700                         if (status != NFS_OK) {
4701                                 if (status != -NFS4ERR_BAD_STATEID)
4702                                         nfs41_free_stateid(server,
4703                                                         &lsp->ls_stateid);
4704                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4705                                 ret = status;
4706                         }
4707                 }
4708         };
4709
4710         return ret;
4711 }
4712
4713 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4714 {
4715         int status = NFS_OK;
4716
4717         if (test_bit(LK_STATE_IN_USE, &state->flags))
4718                 status = nfs41_check_expired_locks(state);
4719         if (status == NFS_OK)
4720                 return status;
4721         return nfs4_lock_expired(state, request);
4722 }
4723 #endif
4724
4725 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4726 {
4727         struct nfs_inode *nfsi = NFS_I(state->inode);
4728         unsigned char fl_flags = request->fl_flags;
4729         int status = -ENOLCK;
4730
4731         if ((fl_flags & FL_POSIX) &&
4732                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4733                 goto out;
4734         /* Is this a delegated open? */
4735         status = nfs4_set_lock_state(state, request);
4736         if (status != 0)
4737                 goto out;
4738         request->fl_flags |= FL_ACCESS;
4739         status = do_vfs_lock(request->fl_file, request);
4740         if (status < 0)
4741                 goto out;
4742         down_read(&nfsi->rwsem);
4743         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4744                 /* Yes: cache locks! */
4745                 /* ...but avoid races with delegation recall... */
4746                 request->fl_flags = fl_flags & ~FL_SLEEP;
4747                 status = do_vfs_lock(request->fl_file, request);
4748                 goto out_unlock;
4749         }
4750         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4751         if (status != 0)
4752                 goto out_unlock;
4753         /* Note: we always want to sleep here! */
4754         request->fl_flags = fl_flags | FL_SLEEP;
4755         if (do_vfs_lock(request->fl_file, request) < 0)
4756                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4757                         "manager!\n", __func__);
4758 out_unlock:
4759         up_read(&nfsi->rwsem);
4760 out:
4761         request->fl_flags = fl_flags;
4762         return status;
4763 }
4764
4765 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4766 {
4767         struct nfs4_exception exception = {
4768                 .state = state,
4769                 .inode = state->inode,
4770         };
4771         int err;
4772
4773         do {
4774                 err = _nfs4_proc_setlk(state, cmd, request);
4775                 if (err == -NFS4ERR_DENIED)
4776                         err = -EAGAIN;
4777                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4778                                 err, &exception);
4779         } while (exception.retry);
4780         return err;
4781 }
4782
4783 static int
4784 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4785 {
4786         struct nfs_open_context *ctx;
4787         struct nfs4_state *state;
4788         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4789         int status;
4790
4791         /* verify open state */
4792         ctx = nfs_file_open_context(filp);
4793         state = ctx->state;
4794
4795         if (request->fl_start < 0 || request->fl_end < 0)
4796                 return -EINVAL;
4797
4798         if (IS_GETLK(cmd)) {
4799                 if (state != NULL)
4800                         return nfs4_proc_getlk(state, F_GETLK, request);
4801                 return 0;
4802         }
4803
4804         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4805                 return -EINVAL;
4806
4807         if (request->fl_type == F_UNLCK) {
4808                 if (state != NULL)
4809                         return nfs4_proc_unlck(state, cmd, request);
4810                 return 0;
4811         }
4812
4813         if (state == NULL)
4814                 return -ENOLCK;
4815         /*
4816          * Don't rely on the VFS having checked the file open mode,
4817          * since it won't do this for flock() locks.
4818          */
4819         switch (request->fl_type & (F_RDLCK|F_WRLCK|F_UNLCK)) {
4820         case F_RDLCK:
4821                 if (!(filp->f_mode & FMODE_READ))
4822                         return -EBADF;
4823                 break;
4824         case F_WRLCK:
4825                 if (!(filp->f_mode & FMODE_WRITE))
4826                         return -EBADF;
4827         }
4828
4829         do {
4830                 status = nfs4_proc_setlk(state, cmd, request);
4831                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4832                         break;
4833                 timeout = nfs4_set_lock_task_retry(timeout);
4834                 status = -ERESTARTSYS;
4835                 if (signalled())
4836                         break;
4837         } while(status < 0);
4838         return status;
4839 }
4840
4841 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4842 {
4843         struct nfs_server *server = NFS_SERVER(state->inode);
4844         struct nfs4_exception exception = { };
4845         int err;
4846
4847         err = nfs4_set_lock_state(state, fl);
4848         if (err != 0)
4849                 goto out;
4850         do {
4851                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4852                 switch (err) {
4853                         default:
4854                                 printk(KERN_ERR "NFS: %s: unhandled error "
4855                                         "%d.\n", __func__, err);
4856                         case 0:
4857                         case -ESTALE:
4858                                 goto out;
4859                         case -NFS4ERR_EXPIRED:
4860                                 nfs4_schedule_stateid_recovery(server, state);
4861                         case -NFS4ERR_STALE_CLIENTID:
4862                         case -NFS4ERR_STALE_STATEID:
4863                                 nfs4_schedule_lease_recovery(server->nfs_client);
4864                                 goto out;
4865                         case -NFS4ERR_BADSESSION:
4866                         case -NFS4ERR_BADSLOT:
4867                         case -NFS4ERR_BAD_HIGH_SLOT:
4868                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4869                         case -NFS4ERR_DEADSESSION:
4870                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
4871                                 goto out;
4872                         case -ERESTARTSYS:
4873                                 /*
4874                                  * The show must go on: exit, but mark the
4875                                  * stateid as needing recovery.
4876                                  */
4877                         case -NFS4ERR_DELEG_REVOKED:
4878                         case -NFS4ERR_ADMIN_REVOKED:
4879                         case -NFS4ERR_BAD_STATEID:
4880                         case -NFS4ERR_OPENMODE:
4881                                 nfs4_schedule_stateid_recovery(server, state);
4882                                 err = 0;
4883                                 goto out;
4884                         case -EKEYEXPIRED:
4885                                 /*
4886                                  * User RPCSEC_GSS context has expired.
4887                                  * We cannot recover this stateid now, so
4888                                  * skip it and allow recovery thread to
4889                                  * proceed.
4890                                  */
4891                                 err = 0;
4892                                 goto out;
4893                         case -ENOMEM:
4894                         case -NFS4ERR_DENIED:
4895                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4896                                 err = 0;
4897                                 goto out;
4898                         case -NFS4ERR_DELAY:
4899                                 break;
4900                 }
4901                 err = nfs4_handle_exception(server, err, &exception);
4902         } while (exception.retry);
4903 out:
4904         return err;
4905 }
4906
4907 struct nfs_release_lockowner_data {
4908         struct nfs4_lock_state *lsp;
4909         struct nfs_server *server;
4910         struct nfs_release_lockowner_args args;
4911 };
4912
4913 static void nfs4_release_lockowner_release(void *calldata)
4914 {
4915         struct nfs_release_lockowner_data *data = calldata;
4916         nfs4_free_lock_state(data->server, data->lsp);
4917         kfree(calldata);
4918 }
4919
4920 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
4921         .rpc_release = nfs4_release_lockowner_release,
4922 };
4923
4924 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
4925 {
4926         struct nfs_server *server = lsp->ls_state->owner->so_server;
4927         struct nfs_release_lockowner_data *data;
4928         struct rpc_message msg = {
4929                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4930         };
4931
4932         if (server->nfs_client->cl_mvops->minor_version != 0)
4933                 return -EINVAL;
4934         data = kmalloc(sizeof(*data), GFP_NOFS);
4935         if (!data)
4936                 return -ENOMEM;
4937         data->lsp = lsp;
4938         data->server = server;
4939         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
4940         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
4941         data->args.lock_owner.s_dev = server->s_dev;
4942         msg.rpc_argp = &data->args;
4943         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
4944         return 0;
4945 }
4946
4947 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4948
4949 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4950                                    const void *buf, size_t buflen,
4951                                    int flags, int type)
4952 {
4953         if (strcmp(key, "") != 0)
4954                 return -EINVAL;
4955
4956         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4957 }
4958
4959 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4960                                    void *buf, size_t buflen, int type)
4961 {
4962         if (strcmp(key, "") != 0)
4963                 return -EINVAL;
4964
4965         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4966 }
4967
4968 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4969                                        size_t list_len, const char *name,
4970                                        size_t name_len, int type)
4971 {
4972         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4973
4974         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4975                 return 0;
4976
4977         if (list && len <= list_len)
4978                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4979         return len;
4980 }
4981
4982 /*
4983  * nfs_fhget will use either the mounted_on_fileid or the fileid
4984  */
4985 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4986 {
4987         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4988                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4989               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4990               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
4991                 return;
4992
4993         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4994                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
4995         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4996         fattr->nlink = 2;
4997 }
4998
4999 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5000                                    const struct qstr *name,
5001                                    struct nfs4_fs_locations *fs_locations,
5002                                    struct page *page)
5003 {
5004         struct nfs_server *server = NFS_SERVER(dir);
5005         u32 bitmask[2] = {
5006                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5007         };
5008         struct nfs4_fs_locations_arg args = {
5009                 .dir_fh = NFS_FH(dir),
5010                 .name = name,
5011                 .page = page,
5012                 .bitmask = bitmask,
5013         };
5014         struct nfs4_fs_locations_res res = {
5015                 .fs_locations = fs_locations,
5016         };
5017         struct rpc_message msg = {
5018                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5019                 .rpc_argp = &args,
5020                 .rpc_resp = &res,
5021         };
5022         int status;
5023
5024         dprintk("%s: start\n", __func__);
5025
5026         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5027          * is not supported */
5028         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5029                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5030         else
5031                 bitmask[0] |= FATTR4_WORD0_FILEID;
5032
5033         nfs_fattr_init(&fs_locations->fattr);
5034         fs_locations->server = server;
5035         fs_locations->nlocations = 0;
5036         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5037         dprintk("%s: returned status = %d\n", __func__, status);
5038         return status;
5039 }
5040
5041 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5042                            const struct qstr *name,
5043                            struct nfs4_fs_locations *fs_locations,
5044                            struct page *page)
5045 {
5046         struct nfs4_exception exception = { };
5047         int err;
5048         do {
5049                 err = nfs4_handle_exception(NFS_SERVER(dir),
5050                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5051                                 &exception);
5052         } while (exception.retry);
5053         return err;
5054 }
5055
5056 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5057 {
5058         int status;
5059         struct nfs4_secinfo_arg args = {
5060                 .dir_fh = NFS_FH(dir),
5061                 .name   = name,
5062         };
5063         struct nfs4_secinfo_res res = {
5064                 .flavors     = flavors,
5065         };
5066         struct rpc_message msg = {
5067                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5068                 .rpc_argp = &args,
5069                 .rpc_resp = &res,
5070         };
5071
5072         dprintk("NFS call  secinfo %s\n", name->name);
5073         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5074         dprintk("NFS reply  secinfo: %d\n", status);
5075         return status;
5076 }
5077
5078 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5079                       struct nfs4_secinfo_flavors *flavors)
5080 {
5081         struct nfs4_exception exception = { };
5082         int err;
5083         do {
5084                 err = nfs4_handle_exception(NFS_SERVER(dir),
5085                                 _nfs4_proc_secinfo(dir, name, flavors),
5086                                 &exception);
5087         } while (exception.retry);
5088         return err;
5089 }
5090
5091 #ifdef CONFIG_NFS_V4_1
5092 /*
5093  * Check the exchange flags returned by the server for invalid flags, having
5094  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5095  * DS flags set.
5096  */
5097 static int nfs4_check_cl_exchange_flags(u32 flags)
5098 {
5099         if (flags & ~EXCHGID4_FLAG_MASK_R)
5100                 goto out_inval;
5101         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5102             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5103                 goto out_inval;
5104         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5105                 goto out_inval;
5106         return NFS_OK;
5107 out_inval:
5108         return -NFS4ERR_INVAL;
5109 }
5110
5111 static bool
5112 nfs41_same_server_scope(struct nfs41_server_scope *a,
5113                         struct nfs41_server_scope *b)
5114 {
5115         if (a->server_scope_sz == b->server_scope_sz &&
5116             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5117                 return true;
5118
5119         return false;
5120 }
5121
5122 /*
5123  * nfs4_proc_bind_conn_to_session()
5124  *
5125  * The 4.1 client currently uses the same TCP connection for the
5126  * fore and backchannel.
5127  */
5128 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5129 {
5130         int status;
5131         struct nfs41_bind_conn_to_session_res res;
5132         struct rpc_message msg = {
5133                 .rpc_proc =
5134                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5135                 .rpc_argp = clp,
5136                 .rpc_resp = &res,
5137                 .rpc_cred = cred,
5138         };
5139
5140         dprintk("--> %s\n", __func__);
5141         BUG_ON(clp == NULL);
5142
5143         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5144         if (unlikely(res.session == NULL)) {
5145                 status = -ENOMEM;
5146                 goto out;
5147         }
5148
5149         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5150         if (status == 0) {
5151                 if (memcmp(res.session->sess_id.data,
5152                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5153                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5154                         status = -EIO;
5155                         goto out_session;
5156                 }
5157                 if (res.dir != NFS4_CDFS4_BOTH) {
5158                         dprintk("NFS: %s: Unexpected direction from server\n",
5159                                 __func__);
5160                         status = -EIO;
5161                         goto out_session;
5162                 }
5163                 if (res.use_conn_in_rdma_mode) {
5164                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5165                                 __func__);
5166                         status = -EIO;
5167                         goto out_session;
5168                 }
5169         }
5170 out_session:
5171         kfree(res.session);
5172 out:
5173         dprintk("<-- %s status= %d\n", __func__, status);
5174         return status;
5175 }
5176
5177 /*
5178  * nfs4_proc_exchange_id()
5179  *
5180  * Since the clientid has expired, all compounds using sessions
5181  * associated with the stale clientid will be returning
5182  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5183  * be in some phase of session reset.
5184  */
5185 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5186 {
5187         nfs4_verifier verifier;
5188         struct nfs41_exchange_id_args args = {
5189                 .verifier = &verifier,
5190                 .client = clp,
5191                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5192         };
5193         struct nfs41_exchange_id_res res = {
5194                 0
5195         };
5196         int status;
5197         struct rpc_message msg = {
5198                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5199                 .rpc_argp = &args,
5200                 .rpc_resp = &res,
5201                 .rpc_cred = cred,
5202         };
5203
5204         dprintk("--> %s\n", __func__);
5205         BUG_ON(clp == NULL);
5206
5207         nfs4_init_boot_verifier(clp, &verifier);
5208
5209         args.id_len = scnprintf(args.id, sizeof(args.id),
5210                                 "%s/%s/%u",
5211                                 clp->cl_ipaddr,
5212                                 clp->cl_rpcclient->cl_nodename,
5213                                 clp->cl_rpcclient->cl_auth->au_flavor);
5214
5215         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5216                                         GFP_NOFS);
5217         if (unlikely(res.server_owner == NULL)) {
5218                 status = -ENOMEM;
5219                 goto out;
5220         }
5221
5222         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5223                                         GFP_NOFS);
5224         if (unlikely(res.server_scope == NULL)) {
5225                 status = -ENOMEM;
5226                 goto out_server_owner;
5227         }
5228
5229         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5230         if (unlikely(res.impl_id == NULL)) {
5231                 status = -ENOMEM;
5232                 goto out_server_scope;
5233         }
5234
5235         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5236         if (status == 0)
5237                 status = nfs4_check_cl_exchange_flags(res.flags);
5238
5239         if (status == 0) {
5240                 clp->cl_clientid = res.clientid;
5241                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5242                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5243                         clp->cl_seqid = res.seqid;
5244
5245                 kfree(clp->cl_serverowner);
5246                 clp->cl_serverowner = res.server_owner;
5247                 res.server_owner = NULL;
5248
5249                 /* use the most recent implementation id */
5250                 kfree(clp->cl_implid);
5251                 clp->cl_implid = res.impl_id;
5252
5253                 if (clp->cl_serverscope != NULL &&
5254                     !nfs41_same_server_scope(clp->cl_serverscope,
5255                                              res.server_scope)) {
5256                         dprintk("%s: server_scope mismatch detected\n",
5257                                 __func__);
5258                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5259                         kfree(clp->cl_serverscope);
5260                         clp->cl_serverscope = NULL;
5261                 }
5262
5263                 if (clp->cl_serverscope == NULL) {
5264                         clp->cl_serverscope = res.server_scope;
5265                         goto out;
5266                 }
5267         } else
5268                 kfree(res.impl_id);
5269
5270 out_server_owner:
5271         kfree(res.server_owner);
5272 out_server_scope:
5273         kfree(res.server_scope);
5274 out:
5275         if (clp->cl_implid != NULL)
5276                 dprintk("%s: Server Implementation ID: "
5277                         "domain: %s, name: %s, date: %llu,%u\n",
5278                         __func__, clp->cl_implid->domain, clp->cl_implid->name,
5279                         clp->cl_implid->date.seconds,
5280                         clp->cl_implid->date.nseconds);
5281         dprintk("<-- %s status= %d\n", __func__, status);
5282         return status;
5283 }
5284
5285 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5286                 struct rpc_cred *cred)
5287 {
5288         struct rpc_message msg = {
5289                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5290                 .rpc_argp = clp,
5291                 .rpc_cred = cred,
5292         };
5293         int status;
5294
5295         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5296         if (status)
5297                 dprintk("NFS: Got error %d from the server %s on "
5298                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5299         return status;
5300 }
5301
5302 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5303                 struct rpc_cred *cred)
5304 {
5305         unsigned int loop;
5306         int ret;
5307
5308         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5309                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5310                 switch (ret) {
5311                 case -NFS4ERR_DELAY:
5312                 case -NFS4ERR_CLIENTID_BUSY:
5313                         ssleep(1);
5314                         break;
5315                 default:
5316                         return ret;
5317                 }
5318         }
5319         return 0;
5320 }
5321
5322 int nfs4_destroy_clientid(struct nfs_client *clp)
5323 {
5324         struct rpc_cred *cred;
5325         int ret = 0;
5326
5327         if (clp->cl_mvops->minor_version < 1)
5328                 goto out;
5329         if (clp->cl_exchange_flags == 0)
5330                 goto out;
5331         cred = nfs4_get_exchange_id_cred(clp);
5332         ret = nfs4_proc_destroy_clientid(clp, cred);
5333         if (cred)
5334                 put_rpccred(cred);
5335         switch (ret) {
5336         case 0:
5337         case -NFS4ERR_STALE_CLIENTID:
5338                 clp->cl_exchange_flags = 0;
5339         }
5340 out:
5341         return ret;
5342 }
5343
5344 struct nfs4_get_lease_time_data {
5345         struct nfs4_get_lease_time_args *args;
5346         struct nfs4_get_lease_time_res *res;
5347         struct nfs_client *clp;
5348 };
5349
5350 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5351                                         void *calldata)
5352 {
5353         int ret;
5354         struct nfs4_get_lease_time_data *data =
5355                         (struct nfs4_get_lease_time_data *)calldata;
5356
5357         dprintk("--> %s\n", __func__);
5358         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5359         /* just setup sequence, do not trigger session recovery
5360            since we're invoked within one */
5361         ret = nfs41_setup_sequence(data->clp->cl_session,
5362                                    &data->args->la_seq_args,
5363                                    &data->res->lr_seq_res, task);
5364
5365         BUG_ON(ret == -EAGAIN);
5366         rpc_call_start(task);
5367         dprintk("<-- %s\n", __func__);
5368 }
5369
5370 /*
5371  * Called from nfs4_state_manager thread for session setup, so don't recover
5372  * from sequence operation or clientid errors.
5373  */
5374 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5375 {
5376         struct nfs4_get_lease_time_data *data =
5377                         (struct nfs4_get_lease_time_data *)calldata;
5378
5379         dprintk("--> %s\n", __func__);
5380         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5381                 return;
5382         switch (task->tk_status) {
5383         case -NFS4ERR_DELAY:
5384         case -NFS4ERR_GRACE:
5385                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5386                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5387                 task->tk_status = 0;
5388                 /* fall through */
5389         case -NFS4ERR_RETRY_UNCACHED_REP:
5390                 rpc_restart_call_prepare(task);
5391                 return;
5392         }
5393         dprintk("<-- %s\n", __func__);
5394 }
5395
5396 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5397         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5398         .rpc_call_done = nfs4_get_lease_time_done,
5399 };
5400
5401 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5402 {
5403         struct rpc_task *task;
5404         struct nfs4_get_lease_time_args args;
5405         struct nfs4_get_lease_time_res res = {
5406                 .lr_fsinfo = fsinfo,
5407         };
5408         struct nfs4_get_lease_time_data data = {
5409                 .args = &args,
5410                 .res = &res,
5411                 .clp = clp,
5412         };
5413         struct rpc_message msg = {
5414                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5415                 .rpc_argp = &args,
5416                 .rpc_resp = &res,
5417         };
5418         struct rpc_task_setup task_setup = {
5419                 .rpc_client = clp->cl_rpcclient,
5420                 .rpc_message = &msg,
5421                 .callback_ops = &nfs4_get_lease_time_ops,
5422                 .callback_data = &data,
5423                 .flags = RPC_TASK_TIMEOUT,
5424         };
5425         int status;
5426
5427         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5428         dprintk("--> %s\n", __func__);
5429         task = rpc_run_task(&task_setup);
5430
5431         if (IS_ERR(task))
5432                 status = PTR_ERR(task);
5433         else {
5434                 status = task->tk_status;
5435                 rpc_put_task(task);
5436         }
5437         dprintk("<-- %s return %d\n", __func__, status);
5438
5439         return status;
5440 }
5441
5442 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5443 {
5444         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5445 }
5446
5447 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5448                 struct nfs4_slot *new,
5449                 u32 max_slots,
5450                 u32 ivalue)
5451 {
5452         struct nfs4_slot *old = NULL;
5453         u32 i;
5454
5455         spin_lock(&tbl->slot_tbl_lock);
5456         if (new) {
5457                 old = tbl->slots;
5458                 tbl->slots = new;
5459                 tbl->max_slots = max_slots;
5460         }
5461         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5462         for (i = 0; i < tbl->max_slots; i++)
5463                 tbl->slots[i].seq_nr = ivalue;
5464         spin_unlock(&tbl->slot_tbl_lock);
5465         kfree(old);
5466 }
5467
5468 /*
5469  * (re)Initialise a slot table
5470  */
5471 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5472                                  u32 ivalue)
5473 {
5474         struct nfs4_slot *new = NULL;
5475         int ret = -ENOMEM;
5476
5477         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5478                 max_reqs, tbl->max_slots);
5479
5480         /* Does the newly negotiated max_reqs match the existing slot table? */
5481         if (max_reqs != tbl->max_slots) {
5482                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5483                 if (!new)
5484                         goto out;
5485         }
5486         ret = 0;
5487
5488         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5489         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5490                 tbl, tbl->slots, tbl->max_slots);
5491 out:
5492         dprintk("<-- %s: return %d\n", __func__, ret);
5493         return ret;
5494 }
5495
5496 /* Destroy the slot table */
5497 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5498 {
5499         if (session->fc_slot_table.slots != NULL) {
5500                 kfree(session->fc_slot_table.slots);
5501                 session->fc_slot_table.slots = NULL;
5502         }
5503         if (session->bc_slot_table.slots != NULL) {
5504                 kfree(session->bc_slot_table.slots);
5505                 session->bc_slot_table.slots = NULL;
5506         }
5507         return;
5508 }
5509
5510 /*
5511  * Initialize or reset the forechannel and backchannel tables
5512  */
5513 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5514 {
5515         struct nfs4_slot_table *tbl;
5516         int status;
5517
5518         dprintk("--> %s\n", __func__);
5519         /* Fore channel */
5520         tbl = &ses->fc_slot_table;
5521         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5522         if (status) /* -ENOMEM */
5523                 return status;
5524         /* Back channel */
5525         tbl = &ses->bc_slot_table;
5526         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5527         if (status && tbl->slots == NULL)
5528                 /* Fore and back channel share a connection so get
5529                  * both slot tables or neither */
5530                 nfs4_destroy_slot_tables(ses);
5531         return status;
5532 }
5533
5534 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5535 {
5536         struct nfs4_session *session;
5537         struct nfs4_slot_table *tbl;
5538
5539         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5540         if (!session)
5541                 return NULL;
5542
5543         tbl = &session->fc_slot_table;
5544         tbl->highest_used_slotid = NFS4_NO_SLOT;
5545         spin_lock_init(&tbl->slot_tbl_lock);
5546         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5547         init_completion(&tbl->complete);
5548
5549         tbl = &session->bc_slot_table;
5550         tbl->highest_used_slotid = NFS4_NO_SLOT;
5551         spin_lock_init(&tbl->slot_tbl_lock);
5552         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5553         init_completion(&tbl->complete);
5554
5555         session->session_state = 1<<NFS4_SESSION_INITING;
5556
5557         session->clp = clp;
5558         return session;
5559 }
5560
5561 void nfs4_destroy_session(struct nfs4_session *session)
5562 {
5563         struct rpc_xprt *xprt;
5564         struct rpc_cred *cred;
5565
5566         cred = nfs4_get_exchange_id_cred(session->clp);
5567         nfs4_proc_destroy_session(session, cred);
5568         if (cred)
5569                 put_rpccred(cred);
5570
5571         rcu_read_lock();
5572         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5573         rcu_read_unlock();
5574         dprintk("%s Destroy backchannel for xprt %p\n",
5575                 __func__, xprt);
5576         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5577         nfs4_destroy_slot_tables(session);
5578         kfree(session);
5579 }
5580
5581 /*
5582  * Initialize the values to be used by the client in CREATE_SESSION
5583  * If nfs4_init_session set the fore channel request and response sizes,
5584  * use them.
5585  *
5586  * Set the back channel max_resp_sz_cached to zero to force the client to
5587  * always set csa_cachethis to FALSE because the current implementation
5588  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5589  */
5590 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5591 {
5592         struct nfs4_session *session = args->client->cl_session;
5593         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5594                      mxresp_sz = session->fc_attrs.max_resp_sz;
5595
5596         if (mxrqst_sz == 0)
5597                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5598         if (mxresp_sz == 0)
5599                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5600         /* Fore channel attributes */
5601         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5602         args->fc_attrs.max_resp_sz = mxresp_sz;
5603         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5604         args->fc_attrs.max_reqs = max_session_slots;
5605
5606         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5607                 "max_ops=%u max_reqs=%u\n",
5608                 __func__,
5609                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5610                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5611
5612         /* Back channel attributes */
5613         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5614         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5615         args->bc_attrs.max_resp_sz_cached = 0;
5616         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5617         args->bc_attrs.max_reqs = 1;
5618
5619         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5620                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5621                 __func__,
5622                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5623                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5624                 args->bc_attrs.max_reqs);
5625 }
5626
5627 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5628 {
5629         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5630         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5631
5632         if (rcvd->max_resp_sz > sent->max_resp_sz)
5633                 return -EINVAL;
5634         /*
5635          * Our requested max_ops is the minimum we need; we're not
5636          * prepared to break up compounds into smaller pieces than that.
5637          * So, no point even trying to continue if the server won't
5638          * cooperate:
5639          */
5640         if (rcvd->max_ops < sent->max_ops)
5641                 return -EINVAL;
5642         if (rcvd->max_reqs == 0)
5643                 return -EINVAL;
5644         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5645                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5646         return 0;
5647 }
5648
5649 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5650 {
5651         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5652         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5653
5654         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5655                 return -EINVAL;
5656         if (rcvd->max_resp_sz < sent->max_resp_sz)
5657                 return -EINVAL;
5658         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5659                 return -EINVAL;
5660         /* These would render the backchannel useless: */
5661         if (rcvd->max_ops != sent->max_ops)
5662                 return -EINVAL;
5663         if (rcvd->max_reqs != sent->max_reqs)
5664                 return -EINVAL;
5665         return 0;
5666 }
5667
5668 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5669                                      struct nfs4_session *session)
5670 {
5671         int ret;
5672
5673         ret = nfs4_verify_fore_channel_attrs(args, session);
5674         if (ret)
5675                 return ret;
5676         return nfs4_verify_back_channel_attrs(args, session);
5677 }
5678
5679 static int _nfs4_proc_create_session(struct nfs_client *clp,
5680                 struct rpc_cred *cred)
5681 {
5682         struct nfs4_session *session = clp->cl_session;
5683         struct nfs41_create_session_args args = {
5684                 .client = clp,
5685                 .cb_program = NFS4_CALLBACK,
5686         };
5687         struct nfs41_create_session_res res = {
5688                 .client = clp,
5689         };
5690         struct rpc_message msg = {
5691                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5692                 .rpc_argp = &args,
5693                 .rpc_resp = &res,
5694                 .rpc_cred = cred,
5695         };
5696         int status;
5697
5698         nfs4_init_channel_attrs(&args);
5699         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5700
5701         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5702
5703         if (!status)
5704                 /* Verify the session's negotiated channel_attrs values */
5705                 status = nfs4_verify_channel_attrs(&args, session);
5706         if (!status) {
5707                 /* Increment the clientid slot sequence id */
5708                 clp->cl_seqid++;
5709         }
5710
5711         return status;
5712 }
5713
5714 /*
5715  * Issues a CREATE_SESSION operation to the server.
5716  * It is the responsibility of the caller to verify the session is
5717  * expired before calling this routine.
5718  */
5719 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5720 {
5721         int status;
5722         unsigned *ptr;
5723         struct nfs4_session *session = clp->cl_session;
5724
5725         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5726
5727         status = _nfs4_proc_create_session(clp, cred);
5728         if (status)
5729                 goto out;
5730
5731         /* Init or reset the session slot tables */
5732         status = nfs4_setup_session_slot_tables(session);
5733         dprintk("slot table setup returned %d\n", status);
5734         if (status)
5735                 goto out;
5736
5737         ptr = (unsigned *)&session->sess_id.data[0];
5738         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5739                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5740 out:
5741         dprintk("<-- %s\n", __func__);
5742         return status;
5743 }
5744
5745 /*
5746  * Issue the over-the-wire RPC DESTROY_SESSION.
5747  * The caller must serialize access to this routine.
5748  */
5749 int nfs4_proc_destroy_session(struct nfs4_session *session,
5750                 struct rpc_cred *cred)
5751 {
5752         struct rpc_message msg = {
5753                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5754                 .rpc_argp = session,
5755                 .rpc_cred = cred,
5756         };
5757         int status = 0;
5758
5759         dprintk("--> nfs4_proc_destroy_session\n");
5760
5761         /* session is still being setup */
5762         if (session->clp->cl_cons_state != NFS_CS_READY)
5763                 return status;
5764
5765         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5766
5767         if (status)
5768                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5769                         "Session has been destroyed regardless...\n", status);
5770
5771         dprintk("<-- nfs4_proc_destroy_session\n");
5772         return status;
5773 }
5774
5775 /*
5776  * With sessions, the client is not marked ready until after a
5777  * successful EXCHANGE_ID and CREATE_SESSION.
5778  *
5779  * Map errors cl_cons_state errors to EPROTONOSUPPORT to indicate
5780  * other versions of NFS can be tried.
5781  */
5782 static int nfs41_check_session_ready(struct nfs_client *clp)
5783 {
5784         int ret;
5785
5786         if (clp->cl_cons_state == NFS_CS_SESSION_INITING) {
5787                 ret = nfs4_client_recover_expired_lease(clp);
5788                 if (ret)
5789                         return ret;
5790         }
5791         if (clp->cl_cons_state < NFS_CS_READY)
5792                 return -EPROTONOSUPPORT;
5793         smp_rmb();
5794         return 0;
5795 }
5796
5797 int nfs4_init_session(struct nfs_server *server)
5798 {
5799         struct nfs_client *clp = server->nfs_client;
5800         struct nfs4_session *session;
5801         unsigned int rsize, wsize;
5802
5803         if (!nfs4_has_session(clp))
5804                 return 0;
5805
5806         session = clp->cl_session;
5807         spin_lock(&clp->cl_lock);
5808         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
5809
5810                 rsize = server->rsize;
5811                 if (rsize == 0)
5812                         rsize = NFS_MAX_FILE_IO_SIZE;
5813                 wsize = server->wsize;
5814                 if (wsize == 0)
5815                         wsize = NFS_MAX_FILE_IO_SIZE;
5816
5817                 session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5818                 session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5819         }
5820         spin_unlock(&clp->cl_lock);
5821
5822         return nfs41_check_session_ready(clp);
5823 }
5824
5825 int nfs4_init_ds_session(struct nfs_client *clp, unsigned long lease_time)
5826 {
5827         struct nfs4_session *session = clp->cl_session;
5828         int ret;
5829
5830         spin_lock(&clp->cl_lock);
5831         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
5832                 /*
5833                  * Do not set NFS_CS_CHECK_LEASE_TIME instead set the
5834                  * DS lease to be equal to the MDS lease.
5835                  */
5836                 clp->cl_lease_time = lease_time;
5837                 clp->cl_last_renewal = jiffies;
5838         }
5839         spin_unlock(&clp->cl_lock);
5840
5841         ret = nfs41_check_session_ready(clp);
5842         if (ret)
5843                 return ret;
5844         /* Test for the DS role */
5845         if (!is_ds_client(clp))
5846                 return -ENODEV;
5847         return 0;
5848 }
5849 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5850
5851
5852 /*
5853  * Renew the cl_session lease.
5854  */
5855 struct nfs4_sequence_data {
5856         struct nfs_client *clp;
5857         struct nfs4_sequence_args args;
5858         struct nfs4_sequence_res res;
5859 };
5860
5861 static void nfs41_sequence_release(void *data)
5862 {
5863         struct nfs4_sequence_data *calldata = data;
5864         struct nfs_client *clp = calldata->clp;
5865
5866         if (atomic_read(&clp->cl_count) > 1)
5867                 nfs4_schedule_state_renewal(clp);
5868         nfs_put_client(clp);
5869         kfree(calldata);
5870 }
5871
5872 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5873 {
5874         switch(task->tk_status) {
5875         case -NFS4ERR_DELAY:
5876                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5877                 return -EAGAIN;
5878         default:
5879                 nfs4_schedule_lease_recovery(clp);
5880         }
5881         return 0;
5882 }
5883
5884 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5885 {
5886         struct nfs4_sequence_data *calldata = data;
5887         struct nfs_client *clp = calldata->clp;
5888
5889         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5890                 return;
5891
5892         if (task->tk_status < 0) {
5893                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5894                 if (atomic_read(&clp->cl_count) == 1)
5895                         goto out;
5896
5897                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5898                         rpc_restart_call_prepare(task);
5899                         return;
5900                 }
5901         }
5902         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5903 out:
5904         dprintk("<-- %s\n", __func__);
5905 }
5906
5907 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5908 {
5909         struct nfs4_sequence_data *calldata = data;
5910         struct nfs_client *clp = calldata->clp;
5911         struct nfs4_sequence_args *args;
5912         struct nfs4_sequence_res *res;
5913
5914         args = task->tk_msg.rpc_argp;
5915         res = task->tk_msg.rpc_resp;
5916
5917         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5918                 return;
5919         rpc_call_start(task);
5920 }
5921
5922 static const struct rpc_call_ops nfs41_sequence_ops = {
5923         .rpc_call_done = nfs41_sequence_call_done,
5924         .rpc_call_prepare = nfs41_sequence_prepare,
5925         .rpc_release = nfs41_sequence_release,
5926 };
5927
5928 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5929 {
5930         struct nfs4_sequence_data *calldata;
5931         struct rpc_message msg = {
5932                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5933                 .rpc_cred = cred,
5934         };
5935         struct rpc_task_setup task_setup_data = {
5936                 .rpc_client = clp->cl_rpcclient,
5937                 .rpc_message = &msg,
5938                 .callback_ops = &nfs41_sequence_ops,
5939                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5940         };
5941
5942         if (!atomic_inc_not_zero(&clp->cl_count))
5943                 return ERR_PTR(-EIO);
5944         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5945         if (calldata == NULL) {
5946                 nfs_put_client(clp);
5947                 return ERR_PTR(-ENOMEM);
5948         }
5949         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5950         msg.rpc_argp = &calldata->args;
5951         msg.rpc_resp = &calldata->res;
5952         calldata->clp = clp;
5953         task_setup_data.callback_data = calldata;
5954
5955         return rpc_run_task(&task_setup_data);
5956 }
5957
5958 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5959 {
5960         struct rpc_task *task;
5961         int ret = 0;
5962
5963         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5964                 return 0;
5965         task = _nfs41_proc_sequence(clp, cred);
5966         if (IS_ERR(task))
5967                 ret = PTR_ERR(task);
5968         else
5969                 rpc_put_task_async(task);
5970         dprintk("<-- %s status=%d\n", __func__, ret);
5971         return ret;
5972 }
5973
5974 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5975 {
5976         struct rpc_task *task;
5977         int ret;
5978
5979         task = _nfs41_proc_sequence(clp, cred);
5980         if (IS_ERR(task)) {
5981                 ret = PTR_ERR(task);
5982                 goto out;
5983         }
5984         ret = rpc_wait_for_completion_task(task);
5985         if (!ret) {
5986                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5987
5988                 if (task->tk_status == 0)
5989                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5990                 ret = task->tk_status;
5991         }
5992         rpc_put_task(task);
5993 out:
5994         dprintk("<-- %s status=%d\n", __func__, ret);
5995         return ret;
5996 }
5997
5998 struct nfs4_reclaim_complete_data {
5999         struct nfs_client *clp;
6000         struct nfs41_reclaim_complete_args arg;
6001         struct nfs41_reclaim_complete_res res;
6002 };
6003
6004 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6005 {
6006         struct nfs4_reclaim_complete_data *calldata = data;
6007
6008         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
6009         if (nfs41_setup_sequence(calldata->clp->cl_session,
6010                                 &calldata->arg.seq_args,
6011                                 &calldata->res.seq_res, task))
6012                 return;
6013
6014         rpc_call_start(task);
6015 }
6016
6017 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6018 {
6019         switch(task->tk_status) {
6020         case 0:
6021         case -NFS4ERR_COMPLETE_ALREADY:
6022         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6023                 break;
6024         case -NFS4ERR_DELAY:
6025                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6026                 /* fall through */
6027         case -NFS4ERR_RETRY_UNCACHED_REP:
6028                 return -EAGAIN;
6029         default:
6030                 nfs4_schedule_lease_recovery(clp);
6031         }
6032         return 0;
6033 }
6034
6035 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6036 {
6037         struct nfs4_reclaim_complete_data *calldata = data;
6038         struct nfs_client *clp = calldata->clp;
6039         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6040
6041         dprintk("--> %s\n", __func__);
6042         if (!nfs41_sequence_done(task, res))
6043                 return;
6044
6045         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6046                 rpc_restart_call_prepare(task);
6047                 return;
6048         }
6049         dprintk("<-- %s\n", __func__);
6050 }
6051
6052 static void nfs4_free_reclaim_complete_data(void *data)
6053 {
6054         struct nfs4_reclaim_complete_data *calldata = data;
6055
6056         kfree(calldata);
6057 }
6058
6059 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6060         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6061         .rpc_call_done = nfs4_reclaim_complete_done,
6062         .rpc_release = nfs4_free_reclaim_complete_data,
6063 };
6064
6065 /*
6066  * Issue a global reclaim complete.
6067  */
6068 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6069 {
6070         struct nfs4_reclaim_complete_data *calldata;
6071         struct rpc_task *task;
6072         struct rpc_message msg = {
6073                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6074         };
6075         struct rpc_task_setup task_setup_data = {
6076                 .rpc_client = clp->cl_rpcclient,
6077                 .rpc_message = &msg,
6078                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6079                 .flags = RPC_TASK_ASYNC,
6080         };
6081         int status = -ENOMEM;
6082
6083         dprintk("--> %s\n", __func__);
6084         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6085         if (calldata == NULL)
6086                 goto out;
6087         calldata->clp = clp;
6088         calldata->arg.one_fs = 0;
6089
6090         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6091         msg.rpc_argp = &calldata->arg;
6092         msg.rpc_resp = &calldata->res;
6093         task_setup_data.callback_data = calldata;
6094         task = rpc_run_task(&task_setup_data);
6095         if (IS_ERR(task)) {
6096                 status = PTR_ERR(task);
6097                 goto out;
6098         }
6099         status = nfs4_wait_for_completion_rpc_task(task);
6100         if (status == 0)
6101                 status = task->tk_status;
6102         rpc_put_task(task);
6103         return 0;
6104 out:
6105         dprintk("<-- %s status=%d\n", __func__, status);
6106         return status;
6107 }
6108
6109 static void
6110 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6111 {
6112         struct nfs4_layoutget *lgp = calldata;
6113         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6114
6115         dprintk("--> %s\n", __func__);
6116         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6117          * right now covering the LAYOUTGET we are about to send.
6118          * However, that is not so catastrophic, and there seems
6119          * to be no way to prevent it completely.
6120          */
6121         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
6122                                 &lgp->res.seq_res, task))
6123                 return;
6124         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6125                                           NFS_I(lgp->args.inode)->layout,
6126                                           lgp->args.ctx->state)) {
6127                 rpc_exit(task, NFS4_OK);
6128                 return;
6129         }
6130         rpc_call_start(task);
6131 }
6132
6133 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6134 {
6135         struct nfs4_layoutget *lgp = calldata;
6136         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6137
6138         dprintk("--> %s\n", __func__);
6139
6140         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
6141                 return;
6142
6143         switch (task->tk_status) {
6144         case 0:
6145                 break;
6146         case -NFS4ERR_LAYOUTTRYLATER:
6147         case -NFS4ERR_RECALLCONFLICT:
6148                 task->tk_status = -NFS4ERR_DELAY;
6149                 /* Fall through */
6150         default:
6151                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6152                         rpc_restart_call_prepare(task);
6153                         return;
6154                 }
6155         }
6156         dprintk("<-- %s\n", __func__);
6157 }
6158
6159 static void nfs4_layoutget_release(void *calldata)
6160 {
6161         struct nfs4_layoutget *lgp = calldata;
6162
6163         dprintk("--> %s\n", __func__);
6164         put_nfs_open_context(lgp->args.ctx);
6165         kfree(calldata);
6166         dprintk("<-- %s\n", __func__);
6167 }
6168
6169 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6170         .rpc_call_prepare = nfs4_layoutget_prepare,
6171         .rpc_call_done = nfs4_layoutget_done,
6172         .rpc_release = nfs4_layoutget_release,
6173 };
6174
6175 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
6176 {
6177         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6178         struct rpc_task *task;
6179         struct rpc_message msg = {
6180                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6181                 .rpc_argp = &lgp->args,
6182                 .rpc_resp = &lgp->res,
6183         };
6184         struct rpc_task_setup task_setup_data = {
6185                 .rpc_client = server->client,
6186                 .rpc_message = &msg,
6187                 .callback_ops = &nfs4_layoutget_call_ops,
6188                 .callback_data = lgp,
6189                 .flags = RPC_TASK_ASYNC,
6190         };
6191         int status = 0;
6192
6193         dprintk("--> %s\n", __func__);
6194
6195         lgp->res.layoutp = &lgp->args.layout;
6196         lgp->res.seq_res.sr_slot = NULL;
6197         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6198         task = rpc_run_task(&task_setup_data);
6199         if (IS_ERR(task))
6200                 return PTR_ERR(task);
6201         status = nfs4_wait_for_completion_rpc_task(task);
6202         if (status == 0)
6203                 status = task->tk_status;
6204         if (status == 0)
6205                 status = pnfs_layout_process(lgp);
6206         rpc_put_task(task);
6207         dprintk("<-- %s status=%d\n", __func__, status);
6208         return status;
6209 }
6210
6211 static void
6212 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6213 {
6214         struct nfs4_layoutreturn *lrp = calldata;
6215
6216         dprintk("--> %s\n", __func__);
6217         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
6218                                 &lrp->res.seq_res, task))
6219                 return;
6220         rpc_call_start(task);
6221 }
6222
6223 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6224 {
6225         struct nfs4_layoutreturn *lrp = calldata;
6226         struct nfs_server *server;
6227         struct pnfs_layout_hdr *lo = lrp->args.layout;
6228
6229         dprintk("--> %s\n", __func__);
6230
6231         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
6232                 return;
6233
6234         server = NFS_SERVER(lrp->args.inode);
6235         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6236                 rpc_restart_call_prepare(task);
6237                 return;
6238         }
6239         spin_lock(&lo->plh_inode->i_lock);
6240         if (task->tk_status == 0) {
6241                 if (lrp->res.lrs_present) {
6242                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6243                 } else
6244                         BUG_ON(!list_empty(&lo->plh_segs));
6245         }
6246         lo->plh_block_lgets--;
6247         spin_unlock(&lo->plh_inode->i_lock);
6248         dprintk("<-- %s\n", __func__);
6249 }
6250
6251 static void nfs4_layoutreturn_release(void *calldata)
6252 {
6253         struct nfs4_layoutreturn *lrp = calldata;
6254
6255         dprintk("--> %s\n", __func__);
6256         put_layout_hdr(lrp->args.layout);
6257         kfree(calldata);
6258         dprintk("<-- %s\n", __func__);
6259 }
6260
6261 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6262         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6263         .rpc_call_done = nfs4_layoutreturn_done,
6264         .rpc_release = nfs4_layoutreturn_release,
6265 };
6266
6267 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6268 {
6269         struct rpc_task *task;
6270         struct rpc_message msg = {
6271                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6272                 .rpc_argp = &lrp->args,
6273                 .rpc_resp = &lrp->res,
6274         };
6275         struct rpc_task_setup task_setup_data = {
6276                 .rpc_client = lrp->clp->cl_rpcclient,
6277                 .rpc_message = &msg,
6278                 .callback_ops = &nfs4_layoutreturn_call_ops,
6279                 .callback_data = lrp,
6280         };
6281         int status;
6282
6283         dprintk("--> %s\n", __func__);
6284         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6285         task = rpc_run_task(&task_setup_data);
6286         if (IS_ERR(task))
6287                 return PTR_ERR(task);
6288         status = task->tk_status;
6289         dprintk("<-- %s status=%d\n", __func__, status);
6290         rpc_put_task(task);
6291         return status;
6292 }
6293
6294 /*
6295  * Retrieve the list of Data Server devices from the MDS.
6296  */
6297 static int _nfs4_getdevicelist(struct nfs_server *server,
6298                                     const struct nfs_fh *fh,
6299                                     struct pnfs_devicelist *devlist)
6300 {
6301         struct nfs4_getdevicelist_args args = {
6302                 .fh = fh,
6303                 .layoutclass = server->pnfs_curr_ld->id,
6304         };
6305         struct nfs4_getdevicelist_res res = {
6306                 .devlist = devlist,
6307         };
6308         struct rpc_message msg = {
6309                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6310                 .rpc_argp = &args,
6311                 .rpc_resp = &res,
6312         };
6313         int status;
6314
6315         dprintk("--> %s\n", __func__);
6316         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6317                                 &res.seq_res, 0);
6318         dprintk("<-- %s status=%d\n", __func__, status);
6319         return status;
6320 }
6321
6322 int nfs4_proc_getdevicelist(struct nfs_server *server,
6323                             const struct nfs_fh *fh,
6324                             struct pnfs_devicelist *devlist)
6325 {
6326         struct nfs4_exception exception = { };
6327         int err;
6328
6329         do {
6330                 err = nfs4_handle_exception(server,
6331                                 _nfs4_getdevicelist(server, fh, devlist),
6332                                 &exception);
6333         } while (exception.retry);
6334
6335         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6336                 err, devlist->num_devs);
6337
6338         return err;
6339 }
6340 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6341
6342 static int
6343 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6344 {
6345         struct nfs4_getdeviceinfo_args args = {
6346                 .pdev = pdev,
6347         };
6348         struct nfs4_getdeviceinfo_res res = {
6349                 .pdev = pdev,
6350         };
6351         struct rpc_message msg = {
6352                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6353                 .rpc_argp = &args,
6354                 .rpc_resp = &res,
6355         };
6356         int status;
6357
6358         dprintk("--> %s\n", __func__);
6359         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6360         dprintk("<-- %s status=%d\n", __func__, status);
6361
6362         return status;
6363 }
6364
6365 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6366 {
6367         struct nfs4_exception exception = { };
6368         int err;
6369
6370         do {
6371                 err = nfs4_handle_exception(server,
6372                                         _nfs4_proc_getdeviceinfo(server, pdev),
6373                                         &exception);
6374         } while (exception.retry);
6375         return err;
6376 }
6377 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6378
6379 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6380 {
6381         struct nfs4_layoutcommit_data *data = calldata;
6382         struct nfs_server *server = NFS_SERVER(data->args.inode);
6383
6384         if (nfs4_setup_sequence(server, &data->args.seq_args,
6385                                 &data->res.seq_res, task))
6386                 return;
6387         rpc_call_start(task);
6388 }
6389
6390 static void
6391 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6392 {
6393         struct nfs4_layoutcommit_data *data = calldata;
6394         struct nfs_server *server = NFS_SERVER(data->args.inode);
6395
6396         if (!nfs4_sequence_done(task, &data->res.seq_res))
6397                 return;
6398
6399         switch (task->tk_status) { /* Just ignore these failures */
6400         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6401         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6402         case -NFS4ERR_BADLAYOUT:     /* no layout */
6403         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6404                 task->tk_status = 0;
6405                 break;
6406         case 0:
6407                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6408                                                    data->res.fattr);
6409                 break;
6410         default:
6411                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6412                         rpc_restart_call_prepare(task);
6413                         return;
6414                 }
6415         }
6416 }
6417
6418 static void nfs4_layoutcommit_release(void *calldata)
6419 {
6420         struct nfs4_layoutcommit_data *data = calldata;
6421         struct pnfs_layout_segment *lseg, *tmp;
6422         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6423
6424         pnfs_cleanup_layoutcommit(data);
6425         /* Matched by references in pnfs_set_layoutcommit */
6426         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6427                 list_del_init(&lseg->pls_lc_list);
6428                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6429                                        &lseg->pls_flags))
6430                         put_lseg(lseg);
6431         }
6432
6433         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6434         smp_mb__after_clear_bit();
6435         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6436
6437         put_rpccred(data->cred);
6438         kfree(data);
6439 }
6440
6441 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6442         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6443         .rpc_call_done = nfs4_layoutcommit_done,
6444         .rpc_release = nfs4_layoutcommit_release,
6445 };
6446
6447 int
6448 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6449 {
6450         struct rpc_message msg = {
6451                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6452                 .rpc_argp = &data->args,
6453                 .rpc_resp = &data->res,
6454                 .rpc_cred = data->cred,
6455         };
6456         struct rpc_task_setup task_setup_data = {
6457                 .task = &data->task,
6458                 .rpc_client = NFS_CLIENT(data->args.inode),
6459                 .rpc_message = &msg,
6460                 .callback_ops = &nfs4_layoutcommit_ops,
6461                 .callback_data = data,
6462                 .flags = RPC_TASK_ASYNC,
6463         };
6464         struct rpc_task *task;
6465         int status = 0;
6466
6467         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6468                 "lbw: %llu inode %lu\n",
6469                 data->task.tk_pid, sync,
6470                 data->args.lastbytewritten,
6471                 data->args.inode->i_ino);
6472
6473         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6474         task = rpc_run_task(&task_setup_data);
6475         if (IS_ERR(task))
6476                 return PTR_ERR(task);
6477         if (sync == false)
6478                 goto out;
6479         status = nfs4_wait_for_completion_rpc_task(task);
6480         if (status != 0)
6481                 goto out;
6482         status = task->tk_status;
6483 out:
6484         dprintk("%s: status %d\n", __func__, status);
6485         rpc_put_task(task);
6486         return status;
6487 }
6488
6489 static int
6490 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6491                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6492 {
6493         struct nfs41_secinfo_no_name_args args = {
6494                 .style = SECINFO_STYLE_CURRENT_FH,
6495         };
6496         struct nfs4_secinfo_res res = {
6497                 .flavors = flavors,
6498         };
6499         struct rpc_message msg = {
6500                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6501                 .rpc_argp = &args,
6502                 .rpc_resp = &res,
6503         };
6504         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6505 }
6506
6507 static int
6508 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6509                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6510 {
6511         struct nfs4_exception exception = { };
6512         int err;
6513         do {
6514                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6515                 switch (err) {
6516                 case 0:
6517                 case -NFS4ERR_WRONGSEC:
6518                 case -NFS4ERR_NOTSUPP:
6519                         goto out;
6520                 default:
6521                         err = nfs4_handle_exception(server, err, &exception);
6522                 }
6523         } while (exception.retry);
6524 out:
6525         return err;
6526 }
6527
6528 static int
6529 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6530                     struct nfs_fsinfo *info)
6531 {
6532         int err;
6533         struct page *page;
6534         rpc_authflavor_t flavor;
6535         struct nfs4_secinfo_flavors *flavors;
6536
6537         page = alloc_page(GFP_KERNEL);
6538         if (!page) {
6539                 err = -ENOMEM;
6540                 goto out;
6541         }
6542
6543         flavors = page_address(page);
6544         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6545
6546         /*
6547          * Fall back on "guess and check" method if
6548          * the server doesn't support SECINFO_NO_NAME
6549          */
6550         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6551                 err = nfs4_find_root_sec(server, fhandle, info);
6552                 goto out_freepage;
6553         }
6554         if (err)
6555                 goto out_freepage;
6556
6557         flavor = nfs_find_best_sec(flavors);
6558         if (err == 0)
6559                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6560
6561 out_freepage:
6562         put_page(page);
6563         if (err == -EACCES)
6564                 return -EPERM;
6565 out:
6566         return err;
6567 }
6568
6569 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6570 {
6571         int status;
6572         struct nfs41_test_stateid_args args = {
6573                 .stateid = stateid,
6574         };
6575         struct nfs41_test_stateid_res res;
6576         struct rpc_message msg = {
6577                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6578                 .rpc_argp = &args,
6579                 .rpc_resp = &res,
6580         };
6581
6582         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6583         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6584         if (status != NFS_OK)
6585                 return status;
6586         return -res.status;
6587 }
6588
6589 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6590 {
6591         struct nfs4_exception exception = { };
6592         int err;
6593         do {
6594                 err = _nfs41_test_stateid(server, stateid);
6595                 if (err != -NFS4ERR_DELAY)
6596                         break;
6597                 nfs4_handle_exception(server, err, &exception);
6598         } while (exception.retry);
6599         return err;
6600 }
6601
6602 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6603 {
6604         struct nfs41_free_stateid_args args = {
6605                 .stateid = stateid,
6606         };
6607         struct nfs41_free_stateid_res res;
6608         struct rpc_message msg = {
6609                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6610                 .rpc_argp = &args,
6611                 .rpc_resp = &res,
6612         };
6613
6614         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6615         return nfs4_call_sync_sequence(server->client, server, &msg,
6616                                          &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_free_stateid(server, stateid);
6625                 if (err != -NFS4ERR_DELAY)
6626                         break;
6627                 nfs4_handle_exception(server, err, &exception);
6628         } while (exception.retry);
6629         return err;
6630 }
6631
6632 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6633                 const nfs4_stateid *s2)
6634 {
6635         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6636                 return false;
6637
6638         if (s1->seqid == s2->seqid)
6639                 return true;
6640         if (s1->seqid == 0 || s2->seqid == 0)
6641                 return true;
6642
6643         return false;
6644 }
6645
6646 #endif /* CONFIG_NFS_V4_1 */
6647
6648 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6649                 const nfs4_stateid *s2)
6650 {
6651         return nfs4_stateid_match(s1, s2);
6652 }
6653
6654
6655 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6656         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6657         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6658         .recover_open   = nfs4_open_reclaim,
6659         .recover_lock   = nfs4_lock_reclaim,
6660         .establish_clid = nfs4_init_clientid,
6661         .get_clid_cred  = nfs4_get_setclientid_cred,
6662 };
6663
6664 #if defined(CONFIG_NFS_V4_1)
6665 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6666         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6667         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6668         .recover_open   = nfs4_open_reclaim,
6669         .recover_lock   = nfs4_lock_reclaim,
6670         .establish_clid = nfs41_init_clientid,
6671         .get_clid_cred  = nfs4_get_exchange_id_cred,
6672         .reclaim_complete = nfs41_proc_reclaim_complete,
6673 };
6674 #endif /* CONFIG_NFS_V4_1 */
6675
6676 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6677         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6678         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6679         .recover_open   = nfs4_open_expired,
6680         .recover_lock   = nfs4_lock_expired,
6681         .establish_clid = nfs4_init_clientid,
6682         .get_clid_cred  = nfs4_get_setclientid_cred,
6683 };
6684
6685 #if defined(CONFIG_NFS_V4_1)
6686 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6687         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6688         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6689         .recover_open   = nfs41_open_expired,
6690         .recover_lock   = nfs41_lock_expired,
6691         .establish_clid = nfs41_init_clientid,
6692         .get_clid_cred  = nfs4_get_exchange_id_cred,
6693 };
6694 #endif /* CONFIG_NFS_V4_1 */
6695
6696 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6697         .sched_state_renewal = nfs4_proc_async_renew,
6698         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6699         .renew_lease = nfs4_proc_renew,
6700 };
6701
6702 #if defined(CONFIG_NFS_V4_1)
6703 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6704         .sched_state_renewal = nfs41_proc_async_sequence,
6705         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6706         .renew_lease = nfs4_proc_sequence,
6707 };
6708 #endif
6709
6710 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6711         .minor_version = 0,
6712         .call_sync = _nfs4_call_sync,
6713         .match_stateid = nfs4_match_stateid,
6714         .find_root_sec = nfs4_find_root_sec,
6715         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6716         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6717         .state_renewal_ops = &nfs40_state_renewal_ops,
6718 };
6719
6720 #if defined(CONFIG_NFS_V4_1)
6721 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6722         .minor_version = 1,
6723         .call_sync = _nfs4_call_sync_session,
6724         .match_stateid = nfs41_match_stateid,
6725         .find_root_sec = nfs41_find_root_sec,
6726         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6727         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6728         .state_renewal_ops = &nfs41_state_renewal_ops,
6729 };
6730 #endif
6731
6732 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6733         [0] = &nfs_v4_0_minor_ops,
6734 #if defined(CONFIG_NFS_V4_1)
6735         [1] = &nfs_v4_1_minor_ops,
6736 #endif
6737 };
6738
6739 static const struct inode_operations nfs4_file_inode_operations = {
6740         .permission     = nfs_permission,
6741         .getattr        = nfs_getattr,
6742         .setattr        = nfs_setattr,
6743         .getxattr       = generic_getxattr,
6744         .setxattr       = generic_setxattr,
6745         .listxattr      = generic_listxattr,
6746         .removexattr    = generic_removexattr,
6747 };
6748
6749 const struct nfs_rpc_ops nfs_v4_clientops = {
6750         .version        = 4,                    /* protocol version */
6751         .dentry_ops     = &nfs4_dentry_operations,
6752         .dir_inode_ops  = &nfs4_dir_inode_operations,
6753         .file_inode_ops = &nfs4_file_inode_operations,
6754         .file_ops       = &nfs4_file_operations,
6755         .getroot        = nfs4_proc_get_root,
6756         .submount       = nfs4_submount,
6757         .getattr        = nfs4_proc_getattr,
6758         .setattr        = nfs4_proc_setattr,
6759         .lookup         = nfs4_proc_lookup,
6760         .access         = nfs4_proc_access,
6761         .readlink       = nfs4_proc_readlink,
6762         .create         = nfs4_proc_create,
6763         .remove         = nfs4_proc_remove,
6764         .unlink_setup   = nfs4_proc_unlink_setup,
6765         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
6766         .unlink_done    = nfs4_proc_unlink_done,
6767         .rename         = nfs4_proc_rename,
6768         .rename_setup   = nfs4_proc_rename_setup,
6769         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
6770         .rename_done    = nfs4_proc_rename_done,
6771         .link           = nfs4_proc_link,
6772         .symlink        = nfs4_proc_symlink,
6773         .mkdir          = nfs4_proc_mkdir,
6774         .rmdir          = nfs4_proc_remove,
6775         .readdir        = nfs4_proc_readdir,
6776         .mknod          = nfs4_proc_mknod,
6777         .statfs         = nfs4_proc_statfs,
6778         .fsinfo         = nfs4_proc_fsinfo,
6779         .pathconf       = nfs4_proc_pathconf,
6780         .set_capabilities = nfs4_server_capabilities,
6781         .decode_dirent  = nfs4_decode_dirent,
6782         .read_setup     = nfs4_proc_read_setup,
6783         .read_pageio_init = pnfs_pageio_init_read,
6784         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
6785         .read_done      = nfs4_read_done,
6786         .write_setup    = nfs4_proc_write_setup,
6787         .write_pageio_init = pnfs_pageio_init_write,
6788         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
6789         .write_done     = nfs4_write_done,
6790         .commit_setup   = nfs4_proc_commit_setup,
6791         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
6792         .commit_done    = nfs4_commit_done,
6793         .lock           = nfs4_proc_lock,
6794         .clear_acl_cache = nfs4_zap_acl_attr,
6795         .close_context  = nfs4_close_context,
6796         .open_context   = nfs4_atomic_open,
6797         .have_delegation = nfs4_have_delegation,
6798         .return_delegation = nfs4_inode_return_delegation,
6799         .alloc_client   = nfs4_alloc_client,
6800         .init_client    = nfs4_init_client,
6801         .free_client    = nfs4_free_client,
6802 };
6803
6804 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6805         .prefix = XATTR_NAME_NFSV4_ACL,
6806         .list   = nfs4_xattr_list_nfs4_acl,
6807         .get    = nfs4_xattr_get_nfs4_acl,
6808         .set    = nfs4_xattr_set_nfs4_acl,
6809 };
6810
6811 const struct xattr_handler *nfs4_xattr_handlers[] = {
6812         &nfs4_xattr_nfs4_acl_handler,
6813         NULL
6814 };
6815
6816 module_param(max_session_slots, ushort, 0644);
6817 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
6818                 "requests the client will negotiate");
6819
6820 /*
6821  * Local variables:
6822  *  c-basic-offset: 8
6823  * End:
6824  */