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/slab.h>
  43 #include <linux/sunrpc/clnt.h>
  44 #include <linux/sunrpc/gss_api.h>
  45 #include <linux/nfs.h>
  46 #include <linux/nfs4.h>
  47 #include <linux/nfs_fs.h>
  48 #include <linux/nfs_page.h>
  49 #include <linux/namei.h>
  50 #include <linux/mount.h>
  51 #include <linux/module.h>
  52 #include <linux/sunrpc/bc_xprt.h>
  53 #include <linux/xattr.h>
  54 #include <linux/utsname.h>
  55
  56 #include "nfs4_fs.h"
  57 #include "delegation.h"
  58 #include "internal.h"
  59 #include "iostat.h"
  60 #include "callback.h"
  61 #include "pnfs.h"
  62
  63 #define NFSDBG_FACILITY         NFSDBG_PROC
  64
  65 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  66 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  67
  68 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  69
  70 struct nfs4_opendata;
  71 static int _nfs4_proc_open(struct nfs4_opendata *data);
  72 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  73 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  74 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  75 static int _nfs4_proc_lookup(struct rpc_clnt *client, struct inode *dir,
  76                              const struct qstr *name, struct nfs_fh *fhandle,
  77                              struct nfs_fattr *fattr);
  78 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  79 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  80                             struct nfs_fattr *fattr, struct iattr *sattr,
  81                             struct nfs4_state *state);
  82
  83 /* Prevent leaks of NFSv4 errors into userland */
  84 static int nfs4_map_errors(int err)
  85 {
  86         if (err >= -1000)
  87                 return err;
  88         switch (err) {
  89         case -NFS4ERR_RESOURCE:
  90                 return -EREMOTEIO;
  91         case -NFS4ERR_WRONGSEC:
  92                 return -EPERM;
  93         case -NFS4ERR_BADOWNER:
  94         case -NFS4ERR_BADNAME:
  95                 return -EINVAL;
  96         default:
  97                 dprintk("%s could not handle NFSv4 error %d\n",
  98                                 __func__, -err);
  99                 break;
 100         }
 101         return -EIO;
 102 }
 103
 104 /*
 105  * This is our standard bitmap for GETATTR requests.
 106  */
 107 const u32 nfs4_fattr_bitmap[2] = {
 108         FATTR4_WORD0_TYPE
 109         | FATTR4_WORD0_CHANGE
 110         | FATTR4_WORD0_SIZE
 111         | FATTR4_WORD0_FSID
 112         | FATTR4_WORD0_FILEID,
 113         FATTR4_WORD1_MODE
 114         | FATTR4_WORD1_NUMLINKS
 115         | FATTR4_WORD1_OWNER
 116         | FATTR4_WORD1_OWNER_GROUP
 117         | FATTR4_WORD1_RAWDEV
 118         | FATTR4_WORD1_SPACE_USED
 119         | FATTR4_WORD1_TIME_ACCESS
 120         | FATTR4_WORD1_TIME_METADATA
 121         | FATTR4_WORD1_TIME_MODIFY
 122 };
 123
 124 const u32 nfs4_statfs_bitmap[2] = {
 125         FATTR4_WORD0_FILES_AVAIL
 126         | FATTR4_WORD0_FILES_FREE
 127         | FATTR4_WORD0_FILES_TOTAL,
 128         FATTR4_WORD1_SPACE_AVAIL
 129         | FATTR4_WORD1_SPACE_FREE
 130         | FATTR4_WORD1_SPACE_TOTAL
 131 };
 132
 133 const u32 nfs4_pathconf_bitmap[2] = {
 134         FATTR4_WORD0_MAXLINK
 135         | FATTR4_WORD0_MAXNAME,
 136         0
 137 };
 138
 139 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
 140                         | FATTR4_WORD0_MAXREAD
 141                         | FATTR4_WORD0_MAXWRITE
 142                         | FATTR4_WORD0_LEASE_TIME,
 143                         FATTR4_WORD1_TIME_DELTA
 144                         | FATTR4_WORD1_FS_LAYOUT_TYPES
 145 };
 146
 147 const u32 nfs4_fs_locations_bitmap[2] = {
 148         FATTR4_WORD0_TYPE
 149         | FATTR4_WORD0_CHANGE
 150         | FATTR4_WORD0_SIZE
 151         | FATTR4_WORD0_FSID
 152         | FATTR4_WORD0_FILEID
 153         | FATTR4_WORD0_FS_LOCATIONS,
 154         FATTR4_WORD1_MODE
 155         | FATTR4_WORD1_NUMLINKS
 156         | FATTR4_WORD1_OWNER
 157         | FATTR4_WORD1_OWNER_GROUP
 158         | FATTR4_WORD1_RAWDEV
 159         | FATTR4_WORD1_SPACE_USED
 160         | FATTR4_WORD1_TIME_ACCESS
 161         | FATTR4_WORD1_TIME_METADATA
 162         | FATTR4_WORD1_TIME_MODIFY
 163         | FATTR4_WORD1_MOUNTED_ON_FILEID
 164 };
 165
 166 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 167                 struct nfs4_readdir_arg *readdir)
 168 {
 169         __be32 *start, *p;
 170
 171         BUG_ON(readdir->count < 80);
 172         if (cookie > 2) {
 173                 readdir->cookie = cookie;
 174                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 175                 return;
 176         }
 177
 178         readdir->cookie = 0;
 179         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 180         if (cookie == 2)
 181                 return;
 182
 183         /*
 184          * NFSv4 servers do not return entries for '.' and '..'
 185          * Therefore, we fake these entries here.  We let '.'
 186          * have cookie 0 and '..' have cookie 1.  Note that
 187          * when talking to the server, we always send cookie 0
 188          * instead of 1 or 2.
 189          */
 190         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 191
 192         if (cookie == 0) {
 193                 *p++ = xdr_one;                                  /* next */
 194                 *p++ = xdr_zero;                   /* cookie, first word */
 195                 *p++ = xdr_one;                   /* cookie, second word */
 196                 *p++ = xdr_one;                             /* entry len */
 197                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 198                 p++;
 199                 *p++ = xdr_one;                         /* bitmap length */
 200                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 201                 *p++ = htonl(8);              /* attribute buffer length */
 202                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 203         }
 204
 205         *p++ = xdr_one;                                  /* next */
 206         *p++ = xdr_zero;                   /* cookie, first word */
 207         *p++ = xdr_two;                   /* cookie, second word */
 208         *p++ = xdr_two;                             /* entry len */
 209         memcpy(p, "..\0\0", 4);                         /* entry */
 210         p++;
 211         *p++ = xdr_one;                         /* bitmap length */
 212         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 213         *p++ = htonl(8);              /* attribute buffer length */
 214         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 215
 216         readdir->pgbase = (char *)p - (char *)start;
 217         readdir->count -= readdir->pgbase;
 218         kunmap_atomic(start, KM_USER0);
 219 }
 220
 221 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 222 {
 223         int res;
 224
 225         might_sleep();
 226
 227         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 228                         nfs_wait_bit_killable, TASK_KILLABLE);
 229         return res;
 230 }
 231
 232 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 233 {
 234         int res = 0;
 235
 236         might_sleep();
 237
 238         if (*timeout <= 0)
 239                 *timeout = NFS4_POLL_RETRY_MIN;
 240         if (*timeout > NFS4_POLL_RETRY_MAX)
 241                 *timeout = NFS4_POLL_RETRY_MAX;
 242         schedule_timeout_killable(*timeout);
 243         if (fatal_signal_pending(current))
 244                 res = -ERESTARTSYS;
 245         *timeout <<= 1;
 246         return res;
 247 }
 248
 249 /* This is the error handling routine for processes that are allowed
 250  * to sleep.
 251  */
 252 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 253 {
 254         struct nfs_client *clp = server->nfs_client;
 255         struct nfs4_state *state = exception->state;
 256         int ret = errorcode;
 257
 258         exception->retry = 0;
 259         switch(errorcode) {
 260                 case 0:
 261                         return 0;
 262                 case -NFS4ERR_ADMIN_REVOKED:
 263                 case -NFS4ERR_BAD_STATEID:
 264                 case -NFS4ERR_OPENMODE:
 265                         if (state == NULL)
 266                                 break;
 267                         nfs4_schedule_stateid_recovery(server, state);
 268                         goto wait_on_recovery;
 269                 case -NFS4ERR_STALE_STATEID:
 270                 case -NFS4ERR_STALE_CLIENTID:
 271                 case -NFS4ERR_EXPIRED:
 272                         nfs4_schedule_lease_recovery(clp);
 273                         goto wait_on_recovery;
 274 #if defined(CONFIG_NFS_V4_1)
 275                 case -NFS4ERR_BADSESSION:
 276                 case -NFS4ERR_BADSLOT:
 277                 case -NFS4ERR_BAD_HIGH_SLOT:
 278                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 279                 case -NFS4ERR_DEADSESSION:
 280                 case -NFS4ERR_SEQ_FALSE_RETRY:
 281                 case -NFS4ERR_SEQ_MISORDERED:
 282                         dprintk("%s ERROR: %d Reset session\n", __func__,
 283                                 errorcode);
 284                         nfs4_schedule_session_recovery(clp->cl_session);
 285                         exception->retry = 1;
 286                         break;
 287 #endif /* defined(CONFIG_NFS_V4_1) */
 288                 case -NFS4ERR_FILE_OPEN:
 289                         if (exception->timeout > HZ) {
 290                                 /* We have retried a decent amount, time to
 291                                  * fail
 292                                  */
 293                                 ret = -EBUSY;
 294                                 break;
 295                         }
 296                 case -NFS4ERR_GRACE:
 297                 case -NFS4ERR_DELAY:
 298                 case -EKEYEXPIRED:
 299                         ret = nfs4_delay(server->client, &exception->timeout);
 300                         if (ret != 0)
 301                                 break;
 302                 case -NFS4ERR_OLD_STATEID:
 303                         exception->retry = 1;
 304                         break;
 305                 case -NFS4ERR_BADOWNER:
 306                         /* The following works around a Linux server bug! */
 307                 case -NFS4ERR_BADNAME:
 308                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 309                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 310                                 exception->retry = 1;
 311                                 printk(KERN_WARNING "NFS: v4 server %s "
 312                                                 "does not accept raw "
 313                                                 "uid/gids. "
 314                                                 "Reenabling the idmapper.\n",
 315                                                 server->nfs_client->cl_hostname);
 316                         }
 317         }
 318         /* We failed to handle the error */
 319         return nfs4_map_errors(ret);
 320 wait_on_recovery:
 321         ret = nfs4_wait_clnt_recover(clp);
 322         if (ret == 0)
 323                 exception->retry = 1;
 324         return ret;
 325 }
 326
 327
 328 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 329 {
 330         spin_lock(&clp->cl_lock);
 331         if (time_before(clp->cl_last_renewal,timestamp))
 332                 clp->cl_last_renewal = timestamp;
 333         spin_unlock(&clp->cl_lock);
 334 }
 335
 336 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 337 {
 338         do_renew_lease(server->nfs_client, timestamp);
 339 }
 340
 341 #if defined(CONFIG_NFS_V4_1)
 342
 343 /*
 344  * nfs4_free_slot - free a slot and efficiently update slot table.
 345  *
 346  * freeing a slot is trivially done by clearing its respective bit
 347  * in the bitmap.
 348  * If the freed slotid equals highest_used_slotid we want to update it
 349  * so that the server would be able to size down the slot table if needed,
 350  * otherwise we know that the highest_used_slotid is still in use.
 351  * When updating highest_used_slotid there may be "holes" in the bitmap
 352  * so we need to scan down from highest_used_slotid to 0 looking for the now
 353  * highest slotid in use.
 354  * If none found, highest_used_slotid is set to -1.
 355  *
 356  * Must be called while holding tbl->slot_tbl_lock
 357  */
 358 static void
 359 nfs4_free_slot(struct nfs4_slot_table *tbl, struct nfs4_slot *free_slot)
 360 {
 361         int free_slotid = free_slot - tbl->slots;
 362         int slotid = free_slotid;
 363
 364         BUG_ON(slotid < 0 || slotid >= NFS4_MAX_SLOT_TABLE);
 365         /* clear used bit in bitmap */
 366         __clear_bit(slotid, tbl->used_slots);
 367
 368         /* update highest_used_slotid when it is freed */
 369         if (slotid == tbl->highest_used_slotid) {
 370                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 371                 if (slotid < tbl->max_slots)
 372                         tbl->highest_used_slotid = slotid;
 373                 else
 374                         tbl->highest_used_slotid = -1;
 375         }
 376         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
 377                 free_slotid, tbl->highest_used_slotid);
 378 }
 379
 380 /*
 381  * Signal state manager thread if session fore channel is drained
 382  */
 383 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 384 {
 385         struct rpc_task *task;
 386
 387         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 388                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
 389                 if (task)
 390                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 391                 return;
 392         }
 393
 394         if (ses->fc_slot_table.highest_used_slotid != -1)
 395                 return;
 396
 397         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 398         complete(&ses->fc_slot_table.complete);
 399 }
 400
 401 /*
 402  * Signal state manager thread if session back channel is drained
 403  */
 404 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 405 {
 406         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 407             ses->bc_slot_table.highest_used_slotid != -1)
 408                 return;
 409         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 410         complete(&ses->bc_slot_table.complete);
 411 }
 412
 413 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 414 {
 415         struct nfs4_slot_table *tbl;
 416
 417         tbl = &res->sr_session->fc_slot_table;
 418         if (!res->sr_slot) {
 419                 /* just wake up the next guy waiting since
 420                  * we may have not consumed a slot after all */
 421                 dprintk("%s: No slot\n", __func__);
 422                 return;
 423         }
 424
 425         spin_lock(&tbl->slot_tbl_lock);
 426         nfs4_free_slot(tbl, res->sr_slot);
 427         nfs4_check_drain_fc_complete(res->sr_session);
 428         spin_unlock(&tbl->slot_tbl_lock);
 429         res->sr_slot = NULL;
 430 }
 431
 432 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 433 {
 434         unsigned long timestamp;
 435         struct nfs_client *clp;
 436
 437         /*
 438          * sr_status remains 1 if an RPC level error occurred. The server
 439          * may or may not have processed the sequence operation..
 440          * Proceed as if the server received and processed the sequence
 441          * operation.
 442          */
 443         if (res->sr_status == 1)
 444                 res->sr_status = NFS_OK;
 445
 446         /* -ERESTARTSYS can result in skipping nfs41_sequence_setup */
 447         if (!res->sr_slot)
 448                 goto out;
 449
 450         /* Check the SEQUENCE operation status */
 451         switch (res->sr_status) {
 452         case 0:
 453                 /* Update the slot's sequence and clientid lease timer */
 454                 ++res->sr_slot->seq_nr;
 455                 timestamp = res->sr_renewal_time;
 456                 clp = res->sr_session->clp;
 457                 do_renew_lease(clp, timestamp);
 458                 /* Check sequence flags */
 459                 if (res->sr_status_flags != 0)
 460                         nfs4_schedule_lease_recovery(clp);
 461                 break;
 462         case -NFS4ERR_DELAY:
 463                 /* The server detected a resend of the RPC call and
 464                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 465                  * of RFC5661.
 466                  */
 467                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 468                         __func__,
 469                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 470                         res->sr_slot->seq_nr);
 471                 goto out_retry;
 472         default:
 473                 /* Just update the slot sequence no. */
 474                 ++res->sr_slot->seq_nr;
 475         }
 476 out:
 477         /* The session may be reset by one of the error handlers. */
 478         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 479         nfs41_sequence_free_slot(res);
 480         return 1;
 481 out_retry:
 482         if (!rpc_restart_call(task))
 483                 goto out;
 484         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 485         return 0;
 486 }
 487
 488 static int nfs4_sequence_done(struct rpc_task *task,
 489                                struct nfs4_sequence_res *res)
 490 {
 491         if (res->sr_session == NULL)
 492                 return 1;
 493         return nfs41_sequence_done(task, res);
 494 }
 495
 496 /*
 497  * nfs4_find_slot - efficiently look for a free slot
 498  *
 499  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 500  * If found, we mark the slot as used, update the highest_used_slotid,
 501  * and respectively set up the sequence operation args.
 502  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
 503  *
 504  * Note: must be called with under the slot_tbl_lock.
 505  */
 506 static u8
 507 nfs4_find_slot(struct nfs4_slot_table *tbl)
 508 {
 509         int slotid;
 510         u8 ret_id = NFS4_MAX_SLOT_TABLE;
 511         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
 512
 513         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
 514                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 515                 tbl->max_slots);
 516         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 517         if (slotid >= tbl->max_slots)
 518                 goto out;
 519         __set_bit(slotid, tbl->used_slots);
 520         if (slotid > tbl->highest_used_slotid)
 521                 tbl->highest_used_slotid = slotid;
 522         ret_id = slotid;
 523 out:
 524         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 525                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 526         return ret_id;
 527 }
 528
 529 int nfs41_setup_sequence(struct nfs4_session *session,
 530                                 struct nfs4_sequence_args *args,
 531                                 struct nfs4_sequence_res *res,
 532                                 int cache_reply,
 533                                 struct rpc_task *task)
 534 {
 535         struct nfs4_slot *slot;
 536         struct nfs4_slot_table *tbl;
 537         u8 slotid;
 538
 539         dprintk("--> %s\n", __func__);
 540         /* slot already allocated? */
 541         if (res->sr_slot != NULL)
 542                 return 0;
 543
 544         tbl = &session->fc_slot_table;
 545
 546         spin_lock(&tbl->slot_tbl_lock);
 547         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 548             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 549                 /*
 550                  * The state manager will wait until the slot table is empty.
 551                  * Schedule the reset thread
 552                  */
 553                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 554                 spin_unlock(&tbl->slot_tbl_lock);
 555                 dprintk("%s Schedule Session Reset\n", __func__);
 556                 return -EAGAIN;
 557         }
 558
 559         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 560             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 561                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 562                 spin_unlock(&tbl->slot_tbl_lock);
 563                 dprintk("%s enforce FIFO order\n", __func__);
 564                 return -EAGAIN;
 565         }
 566
 567         slotid = nfs4_find_slot(tbl);
 568         if (slotid == NFS4_MAX_SLOT_TABLE) {
 569                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 570                 spin_unlock(&tbl->slot_tbl_lock);
 571                 dprintk("<-- %s: no free slots\n", __func__);
 572                 return -EAGAIN;
 573         }
 574         spin_unlock(&tbl->slot_tbl_lock);
 575
 576         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 577         slot = tbl->slots + slotid;
 578         args->sa_session = session;
 579         args->sa_slotid = slotid;
 580         args->sa_cache_this = cache_reply;
 581
 582         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 583
 584         res->sr_session = session;
 585         res->sr_slot = slot;
 586         res->sr_renewal_time = jiffies;
 587         res->sr_status_flags = 0;
 588         /*
 589          * sr_status is only set in decode_sequence, and so will remain
 590          * set to 1 if an rpc level failure occurs.
 591          */
 592         res->sr_status = 1;
 593         return 0;
 594 }
 595 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 596
 597 int nfs4_setup_sequence(const struct nfs_server *server,
 598                         struct nfs4_sequence_args *args,
 599                         struct nfs4_sequence_res *res,
 600                         int cache_reply,
 601                         struct rpc_task *task)
 602 {
 603         struct nfs4_session *session = nfs4_get_session(server);
 604         int ret = 0;
 605
 606         if (session == NULL) {
 607                 args->sa_session = NULL;
 608                 res->sr_session = NULL;
 609                 goto out;
 610         }
 611
 612         dprintk("--> %s clp %p session %p sr_slot %td\n",
 613                 __func__, session->clp, session, res->sr_slot ?
 614                         res->sr_slot - session->fc_slot_table.slots : -1);
 615
 616         ret = nfs41_setup_sequence(session, args, res, cache_reply,
 617                                    task);
 618 out:
 619         dprintk("<-- %s status=%d\n", __func__, ret);
 620         return ret;
 621 }
 622
 623 struct nfs41_call_sync_data {
 624         const struct nfs_server *seq_server;
 625         struct nfs4_sequence_args *seq_args;
 626         struct nfs4_sequence_res *seq_res;
 627         int cache_reply;
 628 };
 629
 630 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 631 {
 632         struct nfs41_call_sync_data *data = calldata;
 633
 634         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 635
 636         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 637                                 data->seq_res, data->cache_reply, task))
 638                 return;
 639         rpc_call_start(task);
 640 }
 641
 642 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 643 {
 644         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 645         nfs41_call_sync_prepare(task, calldata);
 646 }
 647
 648 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 649 {
 650         struct nfs41_call_sync_data *data = calldata;
 651
 652         nfs41_sequence_done(task, data->seq_res);
 653 }
 654
 655 struct rpc_call_ops nfs41_call_sync_ops = {
 656         .rpc_call_prepare = nfs41_call_sync_prepare,
 657         .rpc_call_done = nfs41_call_sync_done,
 658 };
 659
 660 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 661         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 662         .rpc_call_done = nfs41_call_sync_done,
 663 };
 664
 665 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 666                                    struct nfs_server *server,
 667                                    struct rpc_message *msg,
 668                                    struct nfs4_sequence_args *args,
 669                                    struct nfs4_sequence_res *res,
 670                                    int cache_reply,
 671                                    int privileged)
 672 {
 673         int ret;
 674         struct rpc_task *task;
 675         struct nfs41_call_sync_data data = {
 676                 .seq_server = server,
 677                 .seq_args = args,
 678                 .seq_res = res,
 679                 .cache_reply = cache_reply,
 680         };
 681         struct rpc_task_setup task_setup = {
 682                 .rpc_client = clnt,
 683                 .rpc_message = msg,
 684                 .callback_ops = &nfs41_call_sync_ops,
 685                 .callback_data = &data
 686         };
 687
 688         res->sr_slot = NULL;
 689         if (privileged)
 690                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 691         task = rpc_run_task(&task_setup);
 692         if (IS_ERR(task))
 693                 ret = PTR_ERR(task);
 694         else {
 695                 ret = task->tk_status;
 696                 rpc_put_task(task);
 697         }
 698         return ret;
 699 }
 700
 701 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 702                             struct nfs_server *server,
 703                             struct rpc_message *msg,
 704                             struct nfs4_sequence_args *args,
 705                             struct nfs4_sequence_res *res,
 706                             int cache_reply)
 707 {
 708         return nfs4_call_sync_sequence(clnt, server, msg, args, res, cache_reply, 0);
 709 }
 710
 711 #else
 712 static int nfs4_sequence_done(struct rpc_task *task,
 713                                struct nfs4_sequence_res *res)
 714 {
 715         return 1;
 716 }
 717 #endif /* CONFIG_NFS_V4_1 */
 718
 719 int _nfs4_call_sync(struct rpc_clnt *clnt,
 720                     struct nfs_server *server,
 721                     struct rpc_message *msg,
 722                     struct nfs4_sequence_args *args,
 723                     struct nfs4_sequence_res *res,
 724                     int cache_reply)
 725 {
 726         args->sa_session = res->sr_session = NULL;
 727         return rpc_call_sync(clnt, msg, 0);
 728 }
 729
 730 static inline
 731 int nfs4_call_sync(struct rpc_clnt *clnt,
 732                    struct nfs_server *server,
 733                    struct rpc_message *msg,
 734                    struct nfs4_sequence_args *args,
 735                    struct nfs4_sequence_res *res,
 736                    int cache_reply)
 737 {
 738         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 739                                                 args, res, cache_reply);
 740 }
 741
 742 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 743 {
 744         struct nfs_inode *nfsi = NFS_I(dir);
 745
 746         spin_lock(&dir->i_lock);
 747         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 748         if (!cinfo->atomic || cinfo->before != nfsi->change_attr)
 749                 nfs_force_lookup_revalidate(dir);
 750         nfsi->change_attr = cinfo->after;
 751         spin_unlock(&dir->i_lock);
 752 }
 753
 754 struct nfs4_opendata {
 755         struct kref kref;
 756         struct nfs_openargs o_arg;
 757         struct nfs_openres o_res;
 758         struct nfs_open_confirmargs c_arg;
 759         struct nfs_open_confirmres c_res;
 760         struct nfs_fattr f_attr;
 761         struct nfs_fattr dir_attr;
 762         struct path path;
 763         struct dentry *dir;
 764         struct nfs4_state_owner *owner;
 765         struct nfs4_state *state;
 766         struct iattr attrs;
 767         unsigned long timestamp;
 768         unsigned int rpc_done : 1;
 769         int rpc_status;
 770         int cancelled;
 771 };
 772
 773
 774 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 775 {
 776         p->o_res.f_attr = &p->f_attr;
 777         p->o_res.dir_attr = &p->dir_attr;
 778         p->o_res.seqid = p->o_arg.seqid;
 779         p->c_res.seqid = p->c_arg.seqid;
 780         p->o_res.server = p->o_arg.server;
 781         nfs_fattr_init(&p->f_attr);
 782         nfs_fattr_init(&p->dir_attr);
 783 }
 784
 785 static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
 786                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 787                 const struct iattr *attrs,
 788                 gfp_t gfp_mask)
 789 {
 790         struct dentry *parent = dget_parent(path->dentry);
 791         struct inode *dir = parent->d_inode;
 792         struct nfs_server *server = NFS_SERVER(dir);
 793         struct nfs4_opendata *p;
 794
 795         p = kzalloc(sizeof(*p), gfp_mask);
 796         if (p == NULL)
 797                 goto err;
 798         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 799         if (p->o_arg.seqid == NULL)
 800                 goto err_free;
 801         path_get(path);
 802         p->path = *path;
 803         p->dir = parent;
 804         p->owner = sp;
 805         atomic_inc(&sp->so_count);
 806         p->o_arg.fh = NFS_FH(dir);
 807         p->o_arg.open_flags = flags;
 808         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 809         p->o_arg.clientid = server->nfs_client->cl_clientid;
 810         p->o_arg.id = sp->so_owner_id.id;
 811         p->o_arg.name = &p->path.dentry->d_name;
 812         p->o_arg.server = server;
 813         p->o_arg.bitmask = server->attr_bitmask;
 814         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 815         if (flags & O_CREAT) {
 816                 u32 *s;
 817
 818                 p->o_arg.u.attrs = &p->attrs;
 819                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 820                 s = (u32 *) p->o_arg.u.verifier.data;
 821                 s[0] = jiffies;
 822                 s[1] = current->pid;
 823         }
 824         p->c_arg.fh = &p->o_res.fh;
 825         p->c_arg.stateid = &p->o_res.stateid;
 826         p->c_arg.seqid = p->o_arg.seqid;
 827         nfs4_init_opendata_res(p);
 828         kref_init(&p->kref);
 829         return p;
 830 err_free:
 831         kfree(p);
 832 err:
 833         dput(parent);
 834         return NULL;
 835 }
 836
 837 static void nfs4_opendata_free(struct kref *kref)
 838 {
 839         struct nfs4_opendata *p = container_of(kref,
 840                         struct nfs4_opendata, kref);
 841
 842         nfs_free_seqid(p->o_arg.seqid);
 843         if (p->state != NULL)
 844                 nfs4_put_open_state(p->state);
 845         nfs4_put_state_owner(p->owner);
 846         dput(p->dir);
 847         path_put(&p->path);
 848         kfree(p);
 849 }
 850
 851 static void nfs4_opendata_put(struct nfs4_opendata *p)
 852 {
 853         if (p != NULL)
 854                 kref_put(&p->kref, nfs4_opendata_free);
 855 }
 856
 857 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 858 {
 859         int ret;
 860
 861         ret = rpc_wait_for_completion_task(task);
 862         return ret;
 863 }
 864
 865 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 866 {
 867         int ret = 0;
 868
 869         if (open_mode & O_EXCL)
 870                 goto out;
 871         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 872                 case FMODE_READ:
 873                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 874                                 && state->n_rdonly != 0;
 875                         break;
 876                 case FMODE_WRITE:
 877                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 878                                 && state->n_wronly != 0;
 879                         break;
 880                 case FMODE_READ|FMODE_WRITE:
 881                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 882                                 && state->n_rdwr != 0;
 883         }
 884 out:
 885         return ret;
 886 }
 887
 888 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 889 {
 890         if ((delegation->type & fmode) != fmode)
 891                 return 0;
 892         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 893                 return 0;
 894         nfs_mark_delegation_referenced(delegation);
 895         return 1;
 896 }
 897
 898 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 899 {
 900         switch (fmode) {
 901                 case FMODE_WRITE:
 902                         state->n_wronly++;
 903                         break;
 904                 case FMODE_READ:
 905                         state->n_rdonly++;
 906                         break;
 907                 case FMODE_READ|FMODE_WRITE:
 908                         state->n_rdwr++;
 909         }
 910         nfs4_state_set_mode_locked(state, state->state | fmode);
 911 }
 912
 913 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 914 {
 915         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 916                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 917         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 918         switch (fmode) {
 919                 case FMODE_READ:
 920                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 921                         break;
 922                 case FMODE_WRITE:
 923                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 924                         break;
 925                 case FMODE_READ|FMODE_WRITE:
 926                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 927         }
 928 }
 929
 930 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 931 {
 932         write_seqlock(&state->seqlock);
 933         nfs_set_open_stateid_locked(state, stateid, fmode);
 934         write_sequnlock(&state->seqlock);
 935 }
 936
 937 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 938 {
 939         /*
 940          * Protect the call to nfs4_state_set_mode_locked and
 941          * serialise the stateid update
 942          */
 943         write_seqlock(&state->seqlock);
 944         if (deleg_stateid != NULL) {
 945                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 946                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 947         }
 948         if (open_stateid != NULL)
 949                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 950         write_sequnlock(&state->seqlock);
 951         spin_lock(&state->owner->so_lock);
 952         update_open_stateflags(state, fmode);
 953         spin_unlock(&state->owner->so_lock);
 954 }
 955
 956 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 957 {
 958         struct nfs_inode *nfsi = NFS_I(state->inode);
 959         struct nfs_delegation *deleg_cur;
 960         int ret = 0;
 961
 962         fmode &= (FMODE_READ|FMODE_WRITE);
 963
 964         rcu_read_lock();
 965         deleg_cur = rcu_dereference(nfsi->delegation);
 966         if (deleg_cur == NULL)
 967                 goto no_delegation;
 968
 969         spin_lock(&deleg_cur->lock);
 970         if (nfsi->delegation != deleg_cur ||
 971             (deleg_cur->type & fmode) != fmode)
 972                 goto no_delegation_unlock;
 973
 974         if (delegation == NULL)
 975                 delegation = &deleg_cur->stateid;
 976         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
 977                 goto no_delegation_unlock;
 978
 979         nfs_mark_delegation_referenced(deleg_cur);
 980         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
 981         ret = 1;
 982 no_delegation_unlock:
 983         spin_unlock(&deleg_cur->lock);
 984 no_delegation:
 985         rcu_read_unlock();
 986
 987         if (!ret && open_stateid != NULL) {
 988                 __update_open_stateid(state, open_stateid, NULL, fmode);
 989                 ret = 1;
 990         }
 991
 992         return ret;
 993 }
 994
 995
 996 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
 997 {
 998         struct nfs_delegation *delegation;
 999
1000         rcu_read_lock();
1001         delegation = rcu_dereference(NFS_I(inode)->delegation);
1002         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1003                 rcu_read_unlock();
1004                 return;
1005         }
1006         rcu_read_unlock();
1007         nfs_inode_return_delegation(inode);
1008 }
1009
1010 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1011 {
1012         struct nfs4_state *state = opendata->state;
1013         struct nfs_inode *nfsi = NFS_I(state->inode);
1014         struct nfs_delegation *delegation;
1015         int open_mode = opendata->o_arg.open_flags & O_EXCL;
1016         fmode_t fmode = opendata->o_arg.fmode;
1017         nfs4_stateid stateid;
1018         int ret = -EAGAIN;
1019
1020         for (;;) {
1021                 if (can_open_cached(state, fmode, open_mode)) {
1022                         spin_lock(&state->owner->so_lock);
1023                         if (can_open_cached(state, fmode, open_mode)) {
1024                                 update_open_stateflags(state, fmode);
1025                                 spin_unlock(&state->owner->so_lock);
1026                                 goto out_return_state;
1027                         }
1028                         spin_unlock(&state->owner->so_lock);
1029                 }
1030                 rcu_read_lock();
1031                 delegation = rcu_dereference(nfsi->delegation);
1032                 if (delegation == NULL ||
1033                     !can_open_delegated(delegation, fmode)) {
1034                         rcu_read_unlock();
1035                         break;
1036                 }
1037                 /* Save the delegation */
1038                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1039                 rcu_read_unlock();
1040                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1041                 if (ret != 0)
1042                         goto out;
1043                 ret = -EAGAIN;
1044
1045                 /* Try to update the stateid using the delegation */
1046                 if (update_open_stateid(state, NULL, &stateid, fmode))
1047                         goto out_return_state;
1048         }
1049 out:
1050         return ERR_PTR(ret);
1051 out_return_state:
1052         atomic_inc(&state->count);
1053         return state;
1054 }
1055
1056 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1057 {
1058         struct inode *inode;
1059         struct nfs4_state *state = NULL;
1060         struct nfs_delegation *delegation;
1061         int ret;
1062
1063         if (!data->rpc_done) {
1064                 state = nfs4_try_open_cached(data);
1065                 goto out;
1066         }
1067
1068         ret = -EAGAIN;
1069         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1070                 goto err;
1071         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1072         ret = PTR_ERR(inode);
1073         if (IS_ERR(inode))
1074                 goto err;
1075         ret = -ENOMEM;
1076         state = nfs4_get_open_state(inode, data->owner);
1077         if (state == NULL)
1078                 goto err_put_inode;
1079         if (data->o_res.delegation_type != 0) {
1080                 int delegation_flags = 0;
1081
1082                 rcu_read_lock();
1083                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1084                 if (delegation)
1085                         delegation_flags = delegation->flags;
1086                 rcu_read_unlock();
1087                 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1088                         nfs_inode_set_delegation(state->inode,
1089                                         data->owner->so_cred,
1090                                         &data->o_res);
1091                 else
1092                         nfs_inode_reclaim_delegation(state->inode,
1093                                         data->owner->so_cred,
1094                                         &data->o_res);
1095         }
1096
1097         update_open_stateid(state, &data->o_res.stateid, NULL,
1098                         data->o_arg.fmode);
1099         iput(inode);
1100 out:
1101         return state;
1102 err_put_inode:
1103         iput(inode);
1104 err:
1105         return ERR_PTR(ret);
1106 }
1107
1108 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1109 {
1110         struct nfs_inode *nfsi = NFS_I(state->inode);
1111         struct nfs_open_context *ctx;
1112
1113         spin_lock(&state->inode->i_lock);
1114         list_for_each_entry(ctx, &nfsi->open_files, list) {
1115                 if (ctx->state != state)
1116                         continue;
1117                 get_nfs_open_context(ctx);
1118                 spin_unlock(&state->inode->i_lock);
1119                 return ctx;
1120         }
1121         spin_unlock(&state->inode->i_lock);
1122         return ERR_PTR(-ENOENT);
1123 }
1124
1125 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1126 {
1127         struct nfs4_opendata *opendata;
1128
1129         opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, 0, NULL, GFP_NOFS);
1130         if (opendata == NULL)
1131                 return ERR_PTR(-ENOMEM);
1132         opendata->state = state;
1133         atomic_inc(&state->count);
1134         return opendata;
1135 }
1136
1137 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1138 {
1139         struct nfs4_state *newstate;
1140         int ret;
1141
1142         opendata->o_arg.open_flags = 0;
1143         opendata->o_arg.fmode = fmode;
1144         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1145         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1146         nfs4_init_opendata_res(opendata);
1147         ret = _nfs4_recover_proc_open(opendata);
1148         if (ret != 0)
1149                 return ret;
1150         newstate = nfs4_opendata_to_nfs4_state(opendata);
1151         if (IS_ERR(newstate))
1152                 return PTR_ERR(newstate);
1153         nfs4_close_state(&opendata->path, newstate, fmode);
1154         *res = newstate;
1155         return 0;
1156 }
1157
1158 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1159 {
1160         struct nfs4_state *newstate;
1161         int ret;
1162
1163         /* memory barrier prior to reading state->n_* */
1164         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1165         smp_rmb();
1166         if (state->n_rdwr != 0) {
1167                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1168                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1169                 if (ret != 0)
1170                         return ret;
1171                 if (newstate != state)
1172                         return -ESTALE;
1173         }
1174         if (state->n_wronly != 0) {
1175                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1176                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1177                 if (ret != 0)
1178                         return ret;
1179                 if (newstate != state)
1180                         return -ESTALE;
1181         }
1182         if (state->n_rdonly != 0) {
1183                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1184                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1185                 if (ret != 0)
1186                         return ret;
1187                 if (newstate != state)
1188                         return -ESTALE;
1189         }
1190         /*
1191          * We may have performed cached opens for all three recoveries.
1192          * Check if we need to update the current stateid.
1193          */
1194         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1195             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1196                 write_seqlock(&state->seqlock);
1197                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1198                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1199                 write_sequnlock(&state->seqlock);
1200         }
1201         return 0;
1202 }
1203
1204 /*
1205  * OPEN_RECLAIM:
1206  *      reclaim state on the server after a reboot.
1207  */
1208 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1209 {
1210         struct nfs_delegation *delegation;
1211         struct nfs4_opendata *opendata;
1212         fmode_t delegation_type = 0;
1213         int status;
1214
1215         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1216         if (IS_ERR(opendata))
1217                 return PTR_ERR(opendata);
1218         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1219         opendata->o_arg.fh = NFS_FH(state->inode);
1220         rcu_read_lock();
1221         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1222         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1223                 delegation_type = delegation->type;
1224         rcu_read_unlock();
1225         opendata->o_arg.u.delegation_type = delegation_type;
1226         status = nfs4_open_recover(opendata, state);
1227         nfs4_opendata_put(opendata);
1228         return status;
1229 }
1230
1231 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1232 {
1233         struct nfs_server *server = NFS_SERVER(state->inode);
1234         struct nfs4_exception exception = { };
1235         int err;
1236         do {
1237                 err = _nfs4_do_open_reclaim(ctx, state);
1238                 if (err != -NFS4ERR_DELAY)
1239                         break;
1240                 nfs4_handle_exception(server, err, &exception);
1241         } while (exception.retry);
1242         return err;
1243 }
1244
1245 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1246 {
1247         struct nfs_open_context *ctx;
1248         int ret;
1249
1250         ctx = nfs4_state_find_open_context(state);
1251         if (IS_ERR(ctx))
1252                 return PTR_ERR(ctx);
1253         ret = nfs4_do_open_reclaim(ctx, state);
1254         put_nfs_open_context(ctx);
1255         return ret;
1256 }
1257
1258 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1259 {
1260         struct nfs4_opendata *opendata;
1261         int ret;
1262
1263         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1264         if (IS_ERR(opendata))
1265                 return PTR_ERR(opendata);
1266         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1267         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1268                         sizeof(opendata->o_arg.u.delegation.data));
1269         ret = nfs4_open_recover(opendata, state);
1270         nfs4_opendata_put(opendata);
1271         return ret;
1272 }
1273
1274 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1275 {
1276         struct nfs4_exception exception = { };
1277         struct nfs_server *server = NFS_SERVER(state->inode);
1278         int err;
1279         do {
1280                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1281                 switch (err) {
1282                         case 0:
1283                         case -ENOENT:
1284                         case -ESTALE:
1285                                 goto out;
1286                         case -NFS4ERR_BADSESSION:
1287                         case -NFS4ERR_BADSLOT:
1288                         case -NFS4ERR_BAD_HIGH_SLOT:
1289                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1290                         case -NFS4ERR_DEADSESSION:
1291                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1292                                 goto out;
1293                         case -NFS4ERR_STALE_CLIENTID:
1294                         case -NFS4ERR_STALE_STATEID:
1295                         case -NFS4ERR_EXPIRED:
1296                                 /* Don't recall a delegation if it was lost */
1297                                 nfs4_schedule_lease_recovery(server->nfs_client);
1298                                 goto out;
1299                         case -ERESTARTSYS:
1300                                 /*
1301                                  * The show must go on: exit, but mark the
1302                                  * stateid as needing recovery.
1303                                  */
1304                         case -NFS4ERR_ADMIN_REVOKED:
1305                         case -NFS4ERR_BAD_STATEID:
1306                                 nfs4_schedule_stateid_recovery(server, state);
1307                         case -EKEYEXPIRED:
1308                                 /*
1309                                  * User RPCSEC_GSS context has expired.
1310                                  * We cannot recover this stateid now, so
1311                                  * skip it and allow recovery thread to
1312                                  * proceed.
1313                                  */
1314                         case -ENOMEM:
1315                                 err = 0;
1316                                 goto out;
1317                 }
1318                 err = nfs4_handle_exception(server, err, &exception);
1319         } while (exception.retry);
1320 out:
1321         return err;
1322 }
1323
1324 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1325 {
1326         struct nfs4_opendata *data = calldata;
1327
1328         data->rpc_status = task->tk_status;
1329         if (data->rpc_status == 0) {
1330                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1331                                 sizeof(data->o_res.stateid.data));
1332                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1333                 renew_lease(data->o_res.server, data->timestamp);
1334                 data->rpc_done = 1;
1335         }
1336 }
1337
1338 static void nfs4_open_confirm_release(void *calldata)
1339 {
1340         struct nfs4_opendata *data = calldata;
1341         struct nfs4_state *state = NULL;
1342
1343         /* If this request hasn't been cancelled, do nothing */
1344         if (data->cancelled == 0)
1345                 goto out_free;
1346         /* In case of error, no cleanup! */
1347         if (!data->rpc_done)
1348                 goto out_free;
1349         state = nfs4_opendata_to_nfs4_state(data);
1350         if (!IS_ERR(state))
1351                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1352 out_free:
1353         nfs4_opendata_put(data);
1354 }
1355
1356 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1357         .rpc_call_done = nfs4_open_confirm_done,
1358         .rpc_release = nfs4_open_confirm_release,
1359 };
1360
1361 /*
1362  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1363  */
1364 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1365 {
1366         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1367         struct rpc_task *task;
1368         struct  rpc_message msg = {
1369                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1370                 .rpc_argp = &data->c_arg,
1371                 .rpc_resp = &data->c_res,
1372                 .rpc_cred = data->owner->so_cred,
1373         };
1374         struct rpc_task_setup task_setup_data = {
1375                 .rpc_client = server->client,
1376                 .rpc_message = &msg,
1377                 .callback_ops = &nfs4_open_confirm_ops,
1378                 .callback_data = data,
1379                 .workqueue = nfsiod_workqueue,
1380                 .flags = RPC_TASK_ASYNC,
1381         };
1382         int status;
1383
1384         kref_get(&data->kref);
1385         data->rpc_done = 0;
1386         data->rpc_status = 0;
1387         data->timestamp = jiffies;
1388         task = rpc_run_task(&task_setup_data);
1389         if (IS_ERR(task))
1390                 return PTR_ERR(task);
1391         status = nfs4_wait_for_completion_rpc_task(task);
1392         if (status != 0) {
1393                 data->cancelled = 1;
1394                 smp_wmb();
1395         } else
1396                 status = data->rpc_status;
1397         rpc_put_task(task);
1398         return status;
1399 }
1400
1401 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1402 {
1403         struct nfs4_opendata *data = calldata;
1404         struct nfs4_state_owner *sp = data->owner;
1405
1406         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1407                 return;
1408         /*
1409          * Check if we still need to send an OPEN call, or if we can use
1410          * a delegation instead.
1411          */
1412         if (data->state != NULL) {
1413                 struct nfs_delegation *delegation;
1414
1415                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1416                         goto out_no_action;
1417                 rcu_read_lock();
1418                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1419                 if (delegation != NULL &&
1420                     test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) == 0) {
1421                         rcu_read_unlock();
1422                         goto out_no_action;
1423                 }
1424                 rcu_read_unlock();
1425         }
1426         /* Update sequence id. */
1427         data->o_arg.id = sp->so_owner_id.id;
1428         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1429         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1430                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1431                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1432         }
1433         data->timestamp = jiffies;
1434         if (nfs4_setup_sequence(data->o_arg.server,
1435                                 &data->o_arg.seq_args,
1436                                 &data->o_res.seq_res, 1, task))
1437                 return;
1438         rpc_call_start(task);
1439         return;
1440 out_no_action:
1441         task->tk_action = NULL;
1442
1443 }
1444
1445 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1446 {
1447         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1448         nfs4_open_prepare(task, calldata);
1449 }
1450
1451 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1452 {
1453         struct nfs4_opendata *data = calldata;
1454
1455         data->rpc_status = task->tk_status;
1456
1457         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1458                 return;
1459
1460         if (task->tk_status == 0) {
1461                 switch (data->o_res.f_attr->mode & S_IFMT) {
1462                         case S_IFREG:
1463                                 break;
1464                         case S_IFLNK:
1465                                 data->rpc_status = -ELOOP;
1466                                 break;
1467                         case S_IFDIR:
1468                                 data->rpc_status = -EISDIR;
1469                                 break;
1470                         default:
1471                                 data->rpc_status = -ENOTDIR;
1472                 }
1473                 renew_lease(data->o_res.server, data->timestamp);
1474                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1475                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1476         }
1477         data->rpc_done = 1;
1478 }
1479
1480 static void nfs4_open_release(void *calldata)
1481 {
1482         struct nfs4_opendata *data = calldata;
1483         struct nfs4_state *state = NULL;
1484
1485         /* If this request hasn't been cancelled, do nothing */
1486         if (data->cancelled == 0)
1487                 goto out_free;
1488         /* In case of error, no cleanup! */
1489         if (data->rpc_status != 0 || !data->rpc_done)
1490                 goto out_free;
1491         /* In case we need an open_confirm, no cleanup! */
1492         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1493                 goto out_free;
1494         state = nfs4_opendata_to_nfs4_state(data);
1495         if (!IS_ERR(state))
1496                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1497 out_free:
1498         nfs4_opendata_put(data);
1499 }
1500
1501 static const struct rpc_call_ops nfs4_open_ops = {
1502         .rpc_call_prepare = nfs4_open_prepare,
1503         .rpc_call_done = nfs4_open_done,
1504         .rpc_release = nfs4_open_release,
1505 };
1506
1507 static const struct rpc_call_ops nfs4_recover_open_ops = {
1508         .rpc_call_prepare = nfs4_recover_open_prepare,
1509         .rpc_call_done = nfs4_open_done,
1510         .rpc_release = nfs4_open_release,
1511 };
1512
1513 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1514 {
1515         struct inode *dir = data->dir->d_inode;
1516         struct nfs_server *server = NFS_SERVER(dir);
1517         struct nfs_openargs *o_arg = &data->o_arg;
1518         struct nfs_openres *o_res = &data->o_res;
1519         struct rpc_task *task;
1520         struct rpc_message msg = {
1521                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1522                 .rpc_argp = o_arg,
1523                 .rpc_resp = o_res,
1524                 .rpc_cred = data->owner->so_cred,
1525         };
1526         struct rpc_task_setup task_setup_data = {
1527                 .rpc_client = server->client,
1528                 .rpc_message = &msg,
1529                 .callback_ops = &nfs4_open_ops,
1530                 .callback_data = data,
1531                 .workqueue = nfsiod_workqueue,
1532                 .flags = RPC_TASK_ASYNC,
1533         };
1534         int status;
1535
1536         kref_get(&data->kref);
1537         data->rpc_done = 0;
1538         data->rpc_status = 0;
1539         data->cancelled = 0;
1540         if (isrecover)
1541                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1542         task = rpc_run_task(&task_setup_data);
1543         if (IS_ERR(task))
1544                 return PTR_ERR(task);
1545         status = nfs4_wait_for_completion_rpc_task(task);
1546         if (status != 0) {
1547                 data->cancelled = 1;
1548                 smp_wmb();
1549         } else
1550                 status = data->rpc_status;
1551         rpc_put_task(task);
1552
1553         return status;
1554 }
1555
1556 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1557 {
1558         struct inode *dir = data->dir->d_inode;
1559         struct nfs_openres *o_res = &data->o_res;
1560         int status;
1561
1562         status = nfs4_run_open_task(data, 1);
1563         if (status != 0 || !data->rpc_done)
1564                 return status;
1565
1566         nfs_refresh_inode(dir, o_res->dir_attr);
1567
1568         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1569                 status = _nfs4_proc_open_confirm(data);
1570                 if (status != 0)
1571                         return status;
1572         }
1573
1574         return status;
1575 }
1576
1577 /*
1578  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1579  */
1580 static int _nfs4_proc_open(struct nfs4_opendata *data)
1581 {
1582         struct inode *dir = data->dir->d_inode;
1583         struct nfs_server *server = NFS_SERVER(dir);
1584         struct nfs_openargs *o_arg = &data->o_arg;
1585         struct nfs_openres *o_res = &data->o_res;
1586         int status;
1587
1588         status = nfs4_run_open_task(data, 0);
1589         if (status != 0 || !data->rpc_done)
1590                 return status;
1591
1592         if (o_arg->open_flags & O_CREAT) {
1593                 update_changeattr(dir, &o_res->cinfo);
1594                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1595         } else
1596                 nfs_refresh_inode(dir, o_res->dir_attr);
1597         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1598                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1599         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1600                 status = _nfs4_proc_open_confirm(data);
1601                 if (status != 0)
1602                         return status;
1603         }
1604         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1605                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1606         return 0;
1607 }
1608
1609 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1610 {
1611         unsigned int loop;
1612         int ret;
1613
1614         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1615                 ret = nfs4_wait_clnt_recover(clp);
1616                 if (ret != 0)
1617                         break;
1618                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1619                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1620                         break;
1621                 nfs4_schedule_state_manager(clp);
1622                 ret = -EIO;
1623         }
1624         return ret;
1625 }
1626
1627 static int nfs4_recover_expired_lease(struct nfs_server *server)
1628 {
1629         return nfs4_client_recover_expired_lease(server->nfs_client);
1630 }
1631
1632 /*
1633  * OPEN_EXPIRED:
1634  *      reclaim state on the server after a network partition.
1635  *      Assumes caller holds the appropriate lock
1636  */
1637 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1638 {
1639         struct nfs4_opendata *opendata;
1640         int ret;
1641
1642         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1643         if (IS_ERR(opendata))
1644                 return PTR_ERR(opendata);
1645         ret = nfs4_open_recover(opendata, state);
1646         if (ret == -ESTALE)
1647                 d_drop(ctx->path.dentry);
1648         nfs4_opendata_put(opendata);
1649         return ret;
1650 }
1651
1652 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1653 {
1654         struct nfs_server *server = NFS_SERVER(state->inode);
1655         struct nfs4_exception exception = { };
1656         int err;
1657
1658         do {
1659                 err = _nfs4_open_expired(ctx, state);
1660                 switch (err) {
1661                 default:
1662                         goto out;
1663                 case -NFS4ERR_GRACE:
1664                 case -NFS4ERR_DELAY:
1665                         nfs4_handle_exception(server, err, &exception);
1666                         err = 0;
1667                 }
1668         } while (exception.retry);
1669 out:
1670         return err;
1671 }
1672
1673 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1674 {
1675         struct nfs_open_context *ctx;
1676         int ret;
1677
1678         ctx = nfs4_state_find_open_context(state);
1679         if (IS_ERR(ctx))
1680                 return PTR_ERR(ctx);
1681         ret = nfs4_do_open_expired(ctx, state);
1682         put_nfs_open_context(ctx);
1683         return ret;
1684 }
1685
1686 /*
1687  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1688  * fields corresponding to attributes that were used to store the verifier.
1689  * Make sure we clobber those fields in the later setattr call
1690  */
1691 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1692 {
1693         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1694             !(sattr->ia_valid & ATTR_ATIME_SET))
1695                 sattr->ia_valid |= ATTR_ATIME;
1696
1697         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1698             !(sattr->ia_valid & ATTR_MTIME_SET))
1699                 sattr->ia_valid |= ATTR_MTIME;
1700 }
1701
1702 /*
1703  * Returns a referenced nfs4_state
1704  */
1705 static int _nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1706 {
1707         struct nfs4_state_owner  *sp;
1708         struct nfs4_state     *state = NULL;
1709         struct nfs_server       *server = NFS_SERVER(dir);
1710         struct nfs4_opendata *opendata;
1711         int status;
1712
1713         /* Protect against reboot recovery conflicts */
1714         status = -ENOMEM;
1715         if (!(sp = nfs4_get_state_owner(server, cred))) {
1716                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1717                 goto out_err;
1718         }
1719         status = nfs4_recover_expired_lease(server);
1720         if (status != 0)
1721                 goto err_put_state_owner;
1722         if (path->dentry->d_inode != NULL)
1723                 nfs4_return_incompatible_delegation(path->dentry->d_inode, fmode);
1724         status = -ENOMEM;
1725         opendata = nfs4_opendata_alloc(path, sp, fmode, flags, sattr, GFP_KERNEL);
1726         if (opendata == NULL)
1727                 goto err_put_state_owner;
1728
1729         if (path->dentry->d_inode != NULL)
1730                 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1731
1732         status = _nfs4_proc_open(opendata);
1733         if (status != 0)
1734                 goto err_opendata_put;
1735
1736         state = nfs4_opendata_to_nfs4_state(opendata);
1737         status = PTR_ERR(state);
1738         if (IS_ERR(state))
1739                 goto err_opendata_put;
1740         if (server->caps & NFS_CAP_POSIX_LOCK)
1741                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1742
1743         if (opendata->o_arg.open_flags & O_EXCL) {
1744                 nfs4_exclusive_attrset(opendata, sattr);
1745
1746                 nfs_fattr_init(opendata->o_res.f_attr);
1747                 status = nfs4_do_setattr(state->inode, cred,
1748                                 opendata->o_res.f_attr, sattr,
1749                                 state);
1750                 if (status == 0)
1751                         nfs_setattr_update_inode(state->inode, sattr);
1752                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1753         }
1754         nfs4_opendata_put(opendata);
1755         nfs4_put_state_owner(sp);
1756         *res = state;
1757         return 0;
1758 err_opendata_put:
1759         nfs4_opendata_put(opendata);
1760 err_put_state_owner:
1761         nfs4_put_state_owner(sp);
1762 out_err:
1763         *res = NULL;
1764         return status;
1765 }
1766
1767
1768 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1769 {
1770         struct nfs4_exception exception = { };
1771         struct nfs4_state *res;
1772         int status;
1773
1774         do {
1775                 status = _nfs4_do_open(dir, path, fmode, flags, sattr, cred, &res);
1776                 if (status == 0)
1777                         break;
1778                 /* NOTE: BAD_SEQID means the server and client disagree about the
1779                  * book-keeping w.r.t. state-changing operations
1780                  * (OPEN/CLOSE/LOCK/LOCKU...)
1781                  * It is actually a sign of a bug on the client or on the server.
1782                  *
1783                  * If we receive a BAD_SEQID error in the particular case of
1784                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1785                  * have unhashed the old state_owner for us, and that we can
1786                  * therefore safely retry using a new one. We should still warn
1787                  * the user though...
1788                  */
1789                 if (status == -NFS4ERR_BAD_SEQID) {
1790                         printk(KERN_WARNING "NFS: v4 server %s "
1791                                         " returned a bad sequence-id error!\n",
1792                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1793                         exception.retry = 1;
1794                         continue;
1795                 }
1796                 /*
1797                  * BAD_STATEID on OPEN means that the server cancelled our
1798                  * state before it received the OPEN_CONFIRM.
1799                  * Recover by retrying the request as per the discussion
1800                  * on Page 181 of RFC3530.
1801                  */
1802                 if (status == -NFS4ERR_BAD_STATEID) {
1803                         exception.retry = 1;
1804                         continue;
1805                 }
1806                 if (status == -EAGAIN) {
1807                         /* We must have found a delegation */
1808                         exception.retry = 1;
1809                         continue;
1810                 }
1811                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1812                                         status, &exception));
1813         } while (exception.retry);
1814         return res;
1815 }
1816
1817 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1818                             struct nfs_fattr *fattr, struct iattr *sattr,
1819                             struct nfs4_state *state)
1820 {
1821         struct nfs_server *server = NFS_SERVER(inode);
1822         struct nfs_setattrargs  arg = {
1823                 .fh             = NFS_FH(inode),
1824                 .iap            = sattr,
1825                 .server         = server,
1826                 .bitmask = server->attr_bitmask,
1827         };
1828         struct nfs_setattrres  res = {
1829                 .fattr          = fattr,
1830                 .server         = server,
1831         };
1832         struct rpc_message msg = {
1833                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1834                 .rpc_argp       = &arg,
1835                 .rpc_resp       = &res,
1836                 .rpc_cred       = cred,
1837         };
1838         unsigned long timestamp = jiffies;
1839         int status;
1840
1841         nfs_fattr_init(fattr);
1842
1843         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1844                 /* Use that stateid */
1845         } else if (state != NULL) {
1846                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1847         } else
1848                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1849
1850         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1851         if (status == 0 && state != NULL)
1852                 renew_lease(server, timestamp);
1853         return status;
1854 }
1855
1856 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1857                            struct nfs_fattr *fattr, struct iattr *sattr,
1858                            struct nfs4_state *state)
1859 {
1860         struct nfs_server *server = NFS_SERVER(inode);
1861         struct nfs4_exception exception = { };
1862         int err;
1863         do {
1864                 err = nfs4_handle_exception(server,
1865                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1866                                 &exception);
1867         } while (exception.retry);
1868         return err;
1869 }
1870
1871 struct nfs4_closedata {
1872         struct path path;
1873         struct inode *inode;
1874         struct nfs4_state *state;
1875         struct nfs_closeargs arg;
1876         struct nfs_closeres res;
1877         struct nfs_fattr fattr;
1878         unsigned long timestamp;
1879         bool roc;
1880         u32 roc_barrier;
1881 };
1882
1883 static void nfs4_free_closedata(void *data)
1884 {
1885         struct nfs4_closedata *calldata = data;
1886         struct nfs4_state_owner *sp = calldata->state->owner;
1887
1888         if (calldata->roc)
1889                 pnfs_roc_release(calldata->state->inode);
1890         nfs4_put_open_state(calldata->state);
1891         nfs_free_seqid(calldata->arg.seqid);
1892         nfs4_put_state_owner(sp);
1893         path_put(&calldata->path);
1894         kfree(calldata);
1895 }
1896
1897 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1898                 fmode_t fmode)
1899 {
1900         spin_lock(&state->owner->so_lock);
1901         if (!(fmode & FMODE_READ))
1902                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1903         if (!(fmode & FMODE_WRITE))
1904                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1905         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1906         spin_unlock(&state->owner->so_lock);
1907 }
1908
1909 static void nfs4_close_done(struct rpc_task *task, void *data)
1910 {
1911         struct nfs4_closedata *calldata = data;
1912         struct nfs4_state *state = calldata->state;
1913         struct nfs_server *server = NFS_SERVER(calldata->inode);
1914
1915         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1916                 return;
1917         /* hmm. we are done with the inode, and in the process of freeing
1918          * the state_owner. we keep this around to process errors
1919          */
1920         switch (task->tk_status) {
1921                 case 0:
1922                         if (calldata->roc)
1923                                 pnfs_roc_set_barrier(state->inode,
1924                                                      calldata->roc_barrier);
1925                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1926                         renew_lease(server, calldata->timestamp);
1927                         nfs4_close_clear_stateid_flags(state,
1928                                         calldata->arg.fmode);
1929                         break;
1930                 case -NFS4ERR_STALE_STATEID:
1931                 case -NFS4ERR_OLD_STATEID:
1932                 case -NFS4ERR_BAD_STATEID:
1933                 case -NFS4ERR_EXPIRED:
1934                         if (calldata->arg.fmode == 0)
1935                                 break;
1936                 default:
1937                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1938                                 rpc_restart_call_prepare(task);
1939         }
1940         nfs_release_seqid(calldata->arg.seqid);
1941         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1942 }
1943
1944 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1945 {
1946         struct nfs4_closedata *calldata = data;
1947         struct nfs4_state *state = calldata->state;
1948         int call_close = 0;
1949
1950         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1951                 return;
1952
1953         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1954         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1955         spin_lock(&state->owner->so_lock);
1956         /* Calculate the change in open mode */
1957         if (state->n_rdwr == 0) {
1958                 if (state->n_rdonly == 0) {
1959                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
1960                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1961                         calldata->arg.fmode &= ~FMODE_READ;
1962                 }
1963                 if (state->n_wronly == 0) {
1964                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
1965                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1966                         calldata->arg.fmode &= ~FMODE_WRITE;
1967                 }
1968         }
1969         spin_unlock(&state->owner->so_lock);
1970
1971         if (!call_close) {
1972                 /* Note: exit _without_ calling nfs4_close_done */
1973                 task->tk_action = NULL;
1974                 return;
1975         }
1976
1977         if (calldata->arg.fmode == 0) {
1978                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
1979                 if (calldata->roc &&
1980                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
1981                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
1982                                      task, NULL);
1983                         return;
1984                 }
1985         }
1986
1987         nfs_fattr_init(calldata->res.fattr);
1988         calldata->timestamp = jiffies;
1989         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
1990                                 &calldata->arg.seq_args, &calldata->res.seq_res,
1991                                 1, task))
1992                 return;
1993         rpc_call_start(task);
1994 }
1995
1996 static const struct rpc_call_ops nfs4_close_ops = {
1997         .rpc_call_prepare = nfs4_close_prepare,
1998         .rpc_call_done = nfs4_close_done,
1999         .rpc_release = nfs4_free_closedata,
2000 };
2001
2002 /*
2003  * It is possible for data to be read/written from a mem-mapped file
2004  * after the sys_close call (which hits the vfs layer as a flush).
2005  * This means that we can't safely call nfsv4 close on a file until
2006  * the inode is cleared. This in turn means that we are not good
2007  * NFSv4 citizens - we do not indicate to the server to update the file's
2008  * share state even when we are done with one of the three share
2009  * stateid's in the inode.
2010  *
2011  * NOTE: Caller must be holding the sp->so_owner semaphore!
2012  */
2013 int nfs4_do_close(struct path *path, struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2014 {
2015         struct nfs_server *server = NFS_SERVER(state->inode);
2016         struct nfs4_closedata *calldata;
2017         struct nfs4_state_owner *sp = state->owner;
2018         struct rpc_task *task;
2019         struct rpc_message msg = {
2020                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2021                 .rpc_cred = state->owner->so_cred,
2022         };
2023         struct rpc_task_setup task_setup_data = {
2024                 .rpc_client = server->client,
2025                 .rpc_message = &msg,
2026                 .callback_ops = &nfs4_close_ops,
2027                 .workqueue = nfsiod_workqueue,
2028                 .flags = RPC_TASK_ASYNC,
2029         };
2030         int status = -ENOMEM;
2031
2032         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2033         if (calldata == NULL)
2034                 goto out;
2035         calldata->inode = state->inode;
2036         calldata->state = state;
2037         calldata->arg.fh = NFS_FH(state->inode);
2038         calldata->arg.stateid = &state->open_stateid;
2039         /* Serialization for the sequence id */
2040         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2041         if (calldata->arg.seqid == NULL)
2042                 goto out_free_calldata;
2043         calldata->arg.fmode = 0;
2044         calldata->arg.bitmask = server->cache_consistency_bitmask;
2045         calldata->res.fattr = &calldata->fattr;
2046         calldata->res.seqid = calldata->arg.seqid;
2047         calldata->res.server = server;
2048         calldata->roc = roc;
2049         path_get(path);
2050         calldata->path = *path;
2051
2052         msg.rpc_argp = &calldata->arg;
2053         msg.rpc_resp = &calldata->res;
2054         task_setup_data.callback_data = calldata;
2055         task = rpc_run_task(&task_setup_data);
2056         if (IS_ERR(task))
2057                 return PTR_ERR(task);
2058         status = 0;
2059         if (wait)
2060                 status = rpc_wait_for_completion_task(task);
2061         rpc_put_task(task);
2062         return status;
2063 out_free_calldata:
2064         kfree(calldata);
2065 out:
2066         if (roc)
2067                 pnfs_roc_release(state->inode);
2068         nfs4_put_open_state(state);
2069         nfs4_put_state_owner(sp);
2070         return status;
2071 }
2072
2073 static struct inode *
2074 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2075 {
2076         struct nfs4_state *state;
2077
2078         /* Protect against concurrent sillydeletes */
2079         state = nfs4_do_open(dir, &ctx->path, ctx->mode, open_flags, attr, ctx->cred);
2080         if (IS_ERR(state))
2081                 return ERR_CAST(state);
2082         ctx->state = state;
2083         return igrab(state->inode);
2084 }
2085
2086 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2087 {
2088         if (ctx->state == NULL)
2089                 return;
2090         if (is_sync)
2091                 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
2092         else
2093                 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
2094 }
2095
2096 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2097 {
2098         struct nfs4_server_caps_arg args = {
2099                 .fhandle = fhandle,
2100         };
2101         struct nfs4_server_caps_res res = {};
2102         struct rpc_message msg = {
2103                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2104                 .rpc_argp = &args,
2105                 .rpc_resp = &res,
2106         };
2107         int status;
2108
2109         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2110         if (status == 0) {
2111                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2112                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2113                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2114                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2115                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2116                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2117                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2118                         server->caps |= NFS_CAP_ACLS;
2119                 if (res.has_links != 0)
2120                         server->caps |= NFS_CAP_HARDLINKS;
2121                 if (res.has_symlinks != 0)
2122                         server->caps |= NFS_CAP_SYMLINKS;
2123                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2124                         server->caps |= NFS_CAP_FILEID;
2125                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2126                         server->caps |= NFS_CAP_MODE;
2127                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2128                         server->caps |= NFS_CAP_NLINK;
2129                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2130                         server->caps |= NFS_CAP_OWNER;
2131                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2132                         server->caps |= NFS_CAP_OWNER_GROUP;
2133                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2134                         server->caps |= NFS_CAP_ATIME;
2135                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2136                         server->caps |= NFS_CAP_CTIME;
2137                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2138                         server->caps |= NFS_CAP_MTIME;
2139
2140                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2141                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2142                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2143                 server->acl_bitmask = res.acl_bitmask;
2144         }
2145
2146         return status;
2147 }
2148
2149 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2150 {
2151         struct nfs4_exception exception = { };
2152         int err;
2153         do {
2154                 err = nfs4_handle_exception(server,
2155                                 _nfs4_server_capabilities(server, fhandle),
2156                                 &exception);
2157         } while (exception.retry);
2158         return err;
2159 }
2160
2161 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2162                 struct nfs_fsinfo *info)
2163 {
2164         struct nfs4_lookup_root_arg args = {
2165                 .bitmask = nfs4_fattr_bitmap,
2166         };
2167         struct nfs4_lookup_res res = {
2168                 .server = server,
2169                 .fattr = info->fattr,
2170                 .fh = fhandle,
2171         };
2172         struct rpc_message msg = {
2173                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2174                 .rpc_argp = &args,
2175                 .rpc_resp = &res,
2176         };
2177
2178         nfs_fattr_init(info->fattr);
2179         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2180 }
2181
2182 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2183                 struct nfs_fsinfo *info)
2184 {
2185         struct nfs4_exception exception = { };
2186         int err;
2187         do {
2188                 err = nfs4_handle_exception(server,
2189                                 _nfs4_lookup_root(server, fhandle, info),
2190                                 &exception);
2191         } while (exception.retry);
2192         return err;
2193 }
2194
2195 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2196                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2197 {
2198         struct rpc_auth *auth;
2199         int ret;
2200
2201         auth = rpcauth_create(flavor, server->client);
2202         if (!auth) {
2203                 ret = -EIO;
2204                 goto out;
2205         }
2206         ret = nfs4_lookup_root(server, fhandle, info);
2207 out:
2208         return ret;
2209 }
2210
2211 /*
2212  * get the file handle for the "/" directory on the server
2213  */
2214 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2215                               struct nfs_fsinfo *info)
2216 {
2217         int i, len, status = 0;
2218         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS + 2];
2219
2220         flav_array[0] = RPC_AUTH_UNIX;
2221         len = gss_mech_list_pseudoflavors(&flav_array[1]);
2222         flav_array[1+len] = RPC_AUTH_NULL;
2223         len += 2;
2224
2225         for (i = 0; i < len; i++) {
2226                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2227                 if (status == -EPERM || status == -EACCES)
2228                         continue;
2229                 break;
2230         }
2231         if (status == 0)
2232                 status = nfs4_server_capabilities(server, fhandle);
2233         if (status == 0)
2234                 status = nfs4_do_fsinfo(server, fhandle, info);
2235         return nfs4_map_errors(status);
2236 }
2237
2238 /*
2239  * Get locations and (maybe) other attributes of a referral.
2240  * Note that we'll actually follow the referral later when
2241  * we detect fsid mismatch in inode revalidation
2242  */
2243 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2244 {
2245         int status = -ENOMEM;
2246         struct page *page = NULL;
2247         struct nfs4_fs_locations *locations = NULL;
2248
2249         page = alloc_page(GFP_KERNEL);
2250         if (page == NULL)
2251                 goto out;
2252         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2253         if (locations == NULL)
2254                 goto out;
2255
2256         status = nfs4_proc_fs_locations(dir, name, locations, page);
2257         if (status != 0)
2258                 goto out;
2259         /* Make sure server returned a different fsid for the referral */
2260         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2261                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2262                 status = -EIO;
2263                 goto out;
2264         }
2265
2266         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2267         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2268         if (!fattr->mode)
2269                 fattr->mode = S_IFDIR;
2270         memset(fhandle, 0, sizeof(struct nfs_fh));
2271 out:
2272         if (page)
2273                 __free_page(page);
2274         kfree(locations);
2275         return status;
2276 }
2277
2278 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2279 {
2280         struct nfs4_getattr_arg args = {
2281                 .fh = fhandle,
2282                 .bitmask = server->attr_bitmask,
2283         };
2284         struct nfs4_getattr_res res = {
2285                 .fattr = fattr,
2286                 .server = server,
2287         };
2288         struct rpc_message msg = {
2289                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2290                 .rpc_argp = &args,
2291                 .rpc_resp = &res,
2292         };
2293
2294         nfs_fattr_init(fattr);
2295         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2296 }
2297
2298 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2299 {
2300         struct nfs4_exception exception = { };
2301         int err;
2302         do {
2303                 err = nfs4_handle_exception(server,
2304                                 _nfs4_proc_getattr(server, fhandle, fattr),
2305                                 &exception);
2306         } while (exception.retry);
2307         return err;
2308 }
2309
2310 /*
2311  * The file is not closed if it is opened due to the a request to change
2312  * the size of the file. The open call will not be needed once the
2313  * VFS layer lookup-intents are implemented.
2314  *
2315  * Close is called when the inode is destroyed.
2316  * If we haven't opened the file for O_WRONLY, we
2317  * need to in the size_change case to obtain a stateid.
2318  *
2319  * Got race?
2320  * Because OPEN is always done by name in nfsv4, it is
2321  * possible that we opened a different file by the same
2322  * name.  We can recognize this race condition, but we
2323  * can't do anything about it besides returning an error.
2324  *
2325  * This will be fixed with VFS changes (lookup-intent).
2326  */
2327 static int
2328 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2329                   struct iattr *sattr)
2330 {
2331         struct inode *inode = dentry->d_inode;
2332         struct rpc_cred *cred = NULL;
2333         struct nfs4_state *state = NULL;
2334         int status;
2335
2336         nfs_fattr_init(fattr);
2337
2338         /* Search for an existing open(O_WRITE) file */
2339         if (sattr->ia_valid & ATTR_FILE) {
2340                 struct nfs_open_context *ctx;
2341
2342                 ctx = nfs_file_open_context(sattr->ia_file);
2343                 if (ctx) {
2344                         cred = ctx->cred;
2345                         state = ctx->state;
2346                 }
2347         }
2348
2349         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2350         if (status == 0)
2351                 nfs_setattr_update_inode(inode, sattr);
2352         return status;
2353 }
2354
2355 static int _nfs4_proc_lookupfh(struct rpc_clnt *clnt, struct nfs_server *server,
2356                 const struct nfs_fh *dirfh, const struct qstr *name,
2357                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2358 {
2359         int                    status;
2360         struct nfs4_lookup_arg args = {
2361                 .bitmask = server->attr_bitmask,
2362                 .dir_fh = dirfh,
2363                 .name = name,
2364         };
2365         struct nfs4_lookup_res res = {
2366                 .server = server,
2367                 .fattr = fattr,
2368                 .fh = fhandle,
2369         };
2370         struct rpc_message msg = {
2371                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2372                 .rpc_argp = &args,
2373                 .rpc_resp = &res,
2374         };
2375
2376         nfs_fattr_init(fattr);
2377
2378         dprintk("NFS call  lookupfh %s\n", name->name);
2379         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2380         dprintk("NFS reply lookupfh: %d\n", status);
2381         return status;
2382 }
2383
2384 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2385                               struct qstr *name, struct nfs_fh *fhandle,
2386                               struct nfs_fattr *fattr)
2387 {
2388         struct nfs4_exception exception = { };
2389         int err;
2390         do {
2391                 err = _nfs4_proc_lookupfh(server->client, server, dirfh, name, fhandle, fattr);
2392                 /* FIXME: !!!! */
2393                 if (err == -NFS4ERR_MOVED) {
2394                         err = -EREMOTE;
2395                         break;
2396                 }
2397                 err = nfs4_handle_exception(server, err, &exception);
2398         } while (exception.retry);
2399         return err;
2400 }
2401
2402 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2403                 const struct qstr *name, struct nfs_fh *fhandle,
2404                 struct nfs_fattr *fattr)
2405 {
2406         int status;
2407
2408         dprintk("NFS call  lookup %s\n", name->name);
2409         status = _nfs4_proc_lookupfh(clnt, NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2410         if (status == -NFS4ERR_MOVED)
2411                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2412         dprintk("NFS reply lookup: %d\n", status);
2413         return status;
2414 }
2415
2416 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2417 {
2418         memset(fh, 0, sizeof(struct nfs_fh));
2419         fattr->fsid.major = 1;
2420         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2421                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2422         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2423         fattr->nlink = 2;
2424 }
2425
2426 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2427                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2428 {
2429         struct nfs4_exception exception = { };
2430         int err;
2431         do {
2432                 err = nfs4_handle_exception(NFS_SERVER(dir),
2433                                 _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr),
2434                                 &exception);
2435                 if (err == -EPERM)
2436                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2437         } while (exception.retry);
2438         return err;
2439 }
2440
2441 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2442 {
2443         struct nfs_server *server = NFS_SERVER(inode);
2444         struct nfs4_accessargs args = {
2445                 .fh = NFS_FH(inode),
2446                 .bitmask = server->attr_bitmask,
2447         };
2448         struct nfs4_accessres res = {
2449                 .server = server,
2450         };
2451         struct rpc_message msg = {
2452                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2453                 .rpc_argp = &args,
2454                 .rpc_resp = &res,
2455                 .rpc_cred = entry->cred,
2456         };
2457         int mode = entry->mask;
2458         int status;
2459
2460         /*
2461          * Determine which access bits we want to ask for...
2462          */
2463         if (mode & MAY_READ)
2464                 args.access |= NFS4_ACCESS_READ;
2465         if (S_ISDIR(inode->i_mode)) {
2466                 if (mode & MAY_WRITE)
2467                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2468                 if (mode & MAY_EXEC)
2469                         args.access |= NFS4_ACCESS_LOOKUP;
2470         } else {
2471                 if (mode & MAY_WRITE)
2472                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2473                 if (mode & MAY_EXEC)
2474                         args.access |= NFS4_ACCESS_EXECUTE;
2475         }
2476
2477         res.fattr = nfs_alloc_fattr();
2478         if (res.fattr == NULL)
2479                 return -ENOMEM;
2480
2481         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2482         if (!status) {
2483                 entry->mask = 0;
2484                 if (res.access & NFS4_ACCESS_READ)
2485                         entry->mask |= MAY_READ;
2486                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2487                         entry->mask |= MAY_WRITE;
2488                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2489                         entry->mask |= MAY_EXEC;
2490                 nfs_refresh_inode(inode, res.fattr);
2491         }
2492         nfs_free_fattr(res.fattr);
2493         return status;
2494 }
2495
2496 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2497 {
2498         struct nfs4_exception exception = { };
2499         int err;
2500         do {
2501                 err = nfs4_handle_exception(NFS_SERVER(inode),
2502                                 _nfs4_proc_access(inode, entry),
2503                                 &exception);
2504         } while (exception.retry);
2505         return err;
2506 }
2507
2508 /*
2509  * TODO: For the time being, we don't try to get any attributes
2510  * along with any of the zero-copy operations READ, READDIR,
2511  * READLINK, WRITE.
2512  *
2513  * In the case of the first three, we want to put the GETATTR
2514  * after the read-type operation -- this is because it is hard
2515  * to predict the length of a GETATTR response in v4, and thus
2516  * align the READ data correctly.  This means that the GETATTR
2517  * may end up partially falling into the page cache, and we should
2518  * shift it into the 'tail' of the xdr_buf before processing.
2519  * To do this efficiently, we need to know the total length
2520  * of data received, which doesn't seem to be available outside
2521  * of the RPC layer.
2522  *
2523  * In the case of WRITE, we also want to put the GETATTR after
2524  * the operation -- in this case because we want to make sure
2525  * we get the post-operation mtime and size.  This means that
2526  * we can't use xdr_encode_pages() as written: we need a variant
2527  * of it which would leave room in the 'tail' iovec.
2528  *
2529  * Both of these changes to the XDR layer would in fact be quite
2530  * minor, but I decided to leave them for a subsequent patch.
2531  */
2532 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2533                 unsigned int pgbase, unsigned int pglen)
2534 {
2535         struct nfs4_readlink args = {
2536                 .fh       = NFS_FH(inode),
2537                 .pgbase   = pgbase,
2538                 .pglen    = pglen,
2539                 .pages    = &page,
2540         };
2541         struct nfs4_readlink_res res;
2542         struct rpc_message msg = {
2543                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2544                 .rpc_argp = &args,
2545                 .rpc_resp = &res,
2546         };
2547
2548         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2549 }
2550
2551 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2552                 unsigned int pgbase, unsigned int pglen)
2553 {
2554         struct nfs4_exception exception = { };
2555         int err;
2556         do {
2557                 err = nfs4_handle_exception(NFS_SERVER(inode),
2558                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2559                                 &exception);
2560         } while (exception.retry);
2561         return err;
2562 }
2563
2564 /*
2565  * Got race?
2566  * We will need to arrange for the VFS layer to provide an atomic open.
2567  * Until then, this create/open method is prone to inefficiency and race
2568  * conditions due to the lookup, create, and open VFS calls from sys_open()
2569  * placed on the wire.
2570  *
2571  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2572  * The file will be opened again in the subsequent VFS open call
2573  * (nfs4_proc_file_open).
2574  *
2575  * The open for read will just hang around to be used by any process that
2576  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2577  */
2578
2579 static int
2580 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2581                  int flags, struct nfs_open_context *ctx)
2582 {
2583         struct path my_path = {
2584                 .dentry = dentry,
2585         };
2586         struct path *path = &my_path;
2587         struct nfs4_state *state;
2588         struct rpc_cred *cred = NULL;
2589         fmode_t fmode = 0;
2590         int status = 0;
2591
2592         if (ctx != NULL) {
2593                 cred = ctx->cred;
2594                 path = &ctx->path;
2595                 fmode = ctx->mode;
2596         }
2597         sattr->ia_mode &= ~current_umask();
2598         state = nfs4_do_open(dir, path, fmode, flags, sattr, cred);
2599         d_drop(dentry);
2600         if (IS_ERR(state)) {
2601                 status = PTR_ERR(state);
2602                 goto out;
2603         }
2604         d_add(dentry, igrab(state->inode));
2605         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2606         if (ctx != NULL)
2607                 ctx->state = state;
2608         else
2609                 nfs4_close_sync(path, state, fmode);
2610 out:
2611         return status;
2612 }
2613
2614 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2615 {
2616         struct nfs_server *server = NFS_SERVER(dir);
2617         struct nfs_removeargs args = {
2618                 .fh = NFS_FH(dir),
2619                 .name.len = name->len,
2620                 .name.name = name->name,
2621                 .bitmask = server->attr_bitmask,
2622         };
2623         struct nfs_removeres res = {
2624                 .server = server,
2625         };
2626         struct rpc_message msg = {
2627                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2628                 .rpc_argp = &args,
2629                 .rpc_resp = &res,
2630         };
2631         int status = -ENOMEM;
2632
2633         res.dir_attr = nfs_alloc_fattr();
2634         if (res.dir_attr == NULL)
2635                 goto out;
2636
2637         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2638         if (status == 0) {
2639                 update_changeattr(dir, &res.cinfo);
2640                 nfs_post_op_update_inode(dir, res.dir_attr);
2641         }
2642         nfs_free_fattr(res.dir_attr);
2643 out:
2644         return status;
2645 }
2646
2647 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2648 {
2649         struct nfs4_exception exception = { };
2650         int err;
2651         do {
2652                 err = nfs4_handle_exception(NFS_SERVER(dir),
2653                                 _nfs4_proc_remove(dir, name),
2654                                 &exception);
2655         } while (exception.retry);
2656         return err;
2657 }
2658
2659 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2660 {
2661         struct nfs_server *server = NFS_SERVER(dir);
2662         struct nfs_removeargs *args = msg->rpc_argp;
2663         struct nfs_removeres *res = msg->rpc_resp;
2664
2665         args->bitmask = server->cache_consistency_bitmask;
2666         res->server = server;
2667         res->seq_res.sr_slot = NULL;
2668         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2669 }
2670
2671 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2672 {
2673         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2674
2675         if (!nfs4_sequence_done(task, &res->seq_res))
2676                 return 0;
2677         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2678                 return 0;
2679         update_changeattr(dir, &res->cinfo);
2680         nfs_post_op_update_inode(dir, res->dir_attr);
2681         return 1;
2682 }
2683
2684 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2685 {
2686         struct nfs_server *server = NFS_SERVER(dir);
2687         struct nfs_renameargs *arg = msg->rpc_argp;
2688         struct nfs_renameres *res = msg->rpc_resp;
2689
2690         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2691         arg->bitmask = server->attr_bitmask;
2692         res->server = server;
2693 }
2694
2695 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2696                                  struct inode *new_dir)
2697 {
2698         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2699
2700         if (!nfs4_sequence_done(task, &res->seq_res))
2701                 return 0;
2702         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2703                 return 0;
2704
2705         update_changeattr(old_dir, &res->old_cinfo);
2706         nfs_post_op_update_inode(old_dir, res->old_fattr);
2707         update_changeattr(new_dir, &res->new_cinfo);
2708         nfs_post_op_update_inode(new_dir, res->new_fattr);
2709         return 1;
2710 }
2711
2712 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2713                 struct inode *new_dir, struct qstr *new_name)
2714 {
2715         struct nfs_server *server = NFS_SERVER(old_dir);
2716         struct nfs_renameargs arg = {
2717                 .old_dir = NFS_FH(old_dir),
2718                 .new_dir = NFS_FH(new_dir),
2719                 .old_name = old_name,
2720                 .new_name = new_name,
2721                 .bitmask = server->attr_bitmask,
2722         };
2723         struct nfs_renameres res = {
2724                 .server = server,
2725         };
2726         struct rpc_message msg = {
2727                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2728                 .rpc_argp = &arg,
2729                 .rpc_resp = &res,
2730         };
2731         int status = -ENOMEM;
2732
2733         res.old_fattr = nfs_alloc_fattr();
2734         res.new_fattr = nfs_alloc_fattr();
2735         if (res.old_fattr == NULL || res.new_fattr == NULL)
2736                 goto out;
2737
2738         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2739         if (!status) {
2740                 update_changeattr(old_dir, &res.old_cinfo);
2741                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2742                 update_changeattr(new_dir, &res.new_cinfo);
2743                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2744         }
2745 out:
2746         nfs_free_fattr(res.new_fattr);
2747         nfs_free_fattr(res.old_fattr);
2748         return status;
2749 }
2750
2751 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2752                 struct inode *new_dir, struct qstr *new_name)
2753 {
2754         struct nfs4_exception exception = { };
2755         int err;
2756         do {
2757                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2758                                 _nfs4_proc_rename(old_dir, old_name,
2759                                         new_dir, new_name),
2760                                 &exception);
2761         } while (exception.retry);
2762         return err;
2763 }
2764
2765 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2766 {
2767         struct nfs_server *server = NFS_SERVER(inode);
2768         struct nfs4_link_arg arg = {
2769                 .fh     = NFS_FH(inode),
2770                 .dir_fh = NFS_FH(dir),
2771                 .name   = name,
2772                 .bitmask = server->attr_bitmask,
2773         };
2774         struct nfs4_link_res res = {
2775                 .server = server,
2776         };
2777         struct rpc_message msg = {
2778                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2779                 .rpc_argp = &arg,
2780                 .rpc_resp = &res,
2781         };
2782         int status = -ENOMEM;
2783
2784         res.fattr = nfs_alloc_fattr();
2785         res.dir_attr = nfs_alloc_fattr();
2786         if (res.fattr == NULL || res.dir_attr == NULL)
2787                 goto out;
2788
2789         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2790         if (!status) {
2791                 update_changeattr(dir, &res.cinfo);
2792                 nfs_post_op_update_inode(dir, res.dir_attr);
2793                 nfs_post_op_update_inode(inode, res.fattr);
2794         }
2795 out:
2796         nfs_free_fattr(res.dir_attr);
2797         nfs_free_fattr(res.fattr);
2798         return status;
2799 }
2800
2801 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2802 {
2803         struct nfs4_exception exception = { };
2804         int err;
2805         do {
2806                 err = nfs4_handle_exception(NFS_SERVER(inode),
2807                                 _nfs4_proc_link(inode, dir, name),
2808                                 &exception);
2809         } while (exception.retry);
2810         return err;
2811 }
2812
2813 struct nfs4_createdata {
2814         struct rpc_message msg;
2815         struct nfs4_create_arg arg;
2816         struct nfs4_create_res res;
2817         struct nfs_fh fh;
2818         struct nfs_fattr fattr;
2819         struct nfs_fattr dir_fattr;
2820 };
2821
2822 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2823                 struct qstr *name, struct iattr *sattr, u32 ftype)
2824 {
2825         struct nfs4_createdata *data;
2826
2827         data = kzalloc(sizeof(*data), GFP_KERNEL);
2828         if (data != NULL) {
2829                 struct nfs_server *server = NFS_SERVER(dir);
2830
2831                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2832                 data->msg.rpc_argp = &data->arg;
2833                 data->msg.rpc_resp = &data->res;
2834                 data->arg.dir_fh = NFS_FH(dir);
2835                 data->arg.server = server;
2836                 data->arg.name = name;
2837                 data->arg.attrs = sattr;
2838                 data->arg.ftype = ftype;
2839                 data->arg.bitmask = server->attr_bitmask;
2840                 data->res.server = server;
2841                 data->res.fh = &data->fh;
2842                 data->res.fattr = &data->fattr;
2843                 data->res.dir_fattr = &data->dir_fattr;
2844                 nfs_fattr_init(data->res.fattr);
2845                 nfs_fattr_init(data->res.dir_fattr);
2846         }
2847         return data;
2848 }
2849
2850 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2851 {
2852         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2853                                     &data->arg.seq_args, &data->res.seq_res, 1);
2854         if (status == 0) {
2855                 update_changeattr(dir, &data->res.dir_cinfo);
2856                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2857                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2858         }
2859         return status;
2860 }
2861
2862 static void nfs4_free_createdata(struct nfs4_createdata *data)
2863 {
2864         kfree(data);
2865 }
2866
2867 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2868                 struct page *page, unsigned int len, struct iattr *sattr)
2869 {
2870         struct nfs4_createdata *data;
2871         int status = -ENAMETOOLONG;
2872
2873         if (len > NFS4_MAXPATHLEN)
2874                 goto out;
2875
2876         status = -ENOMEM;
2877         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2878         if (data == NULL)
2879                 goto out;
2880
2881         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2882         data->arg.u.symlink.pages = &page;
2883         data->arg.u.symlink.len = len;
2884
2885         status = nfs4_do_create(dir, dentry, data);
2886
2887         nfs4_free_createdata(data);
2888 out:
2889         return status;
2890 }
2891
2892 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2893                 struct page *page, unsigned int len, struct iattr *sattr)
2894 {
2895         struct nfs4_exception exception = { };
2896         int err;
2897         do {
2898                 err = nfs4_handle_exception(NFS_SERVER(dir),
2899                                 _nfs4_proc_symlink(dir, dentry, page,
2900                                                         len, sattr),
2901                                 &exception);
2902         } while (exception.retry);
2903         return err;
2904 }
2905
2906 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2907                 struct iattr *sattr)
2908 {
2909         struct nfs4_createdata *data;
2910         int status = -ENOMEM;
2911
2912         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2913         if (data == NULL)
2914                 goto out;
2915
2916         status = nfs4_do_create(dir, dentry, data);
2917
2918         nfs4_free_createdata(data);
2919 out:
2920         return status;
2921 }
2922
2923 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2924                 struct iattr *sattr)
2925 {
2926         struct nfs4_exception exception = { };
2927         int err;
2928
2929         sattr->ia_mode &= ~current_umask();
2930         do {
2931                 err = nfs4_handle_exception(NFS_SERVER(dir),
2932                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2933                                 &exception);
2934         } while (exception.retry);
2935         return err;
2936 }
2937
2938 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2939                 u64 cookie, struct page **pages, unsigned int count, int plus)
2940 {
2941         struct inode            *dir = dentry->d_inode;
2942         struct nfs4_readdir_arg args = {
2943                 .fh = NFS_FH(dir),
2944                 .pages = pages,
2945                 .pgbase = 0,
2946                 .count = count,
2947                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2948                 .plus = plus,
2949         };
2950         struct nfs4_readdir_res res;
2951         struct rpc_message msg = {
2952                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2953                 .rpc_argp = &args,
2954                 .rpc_resp = &res,
2955                 .rpc_cred = cred,
2956         };
2957         int                     status;
2958
2959         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2960                         dentry->d_parent->d_name.name,
2961                         dentry->d_name.name,
2962                         (unsigned long long)cookie);
2963         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2964         res.pgbase = args.pgbase;
2965         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
2966         if (status >= 0) {
2967                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2968                 status += args.pgbase;
2969         }
2970
2971         nfs_invalidate_atime(dir);
2972
2973         dprintk("%s: returns %d\n", __func__, status);
2974         return status;
2975 }
2976
2977 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2978                 u64 cookie, struct page **pages, unsigned int count, int plus)
2979 {
2980         struct nfs4_exception exception = { };
2981         int err;
2982         do {
2983                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
2984                                 _nfs4_proc_readdir(dentry, cred, cookie,
2985                                         pages, count, plus),
2986                                 &exception);
2987         } while (exception.retry);
2988         return err;
2989 }
2990
2991 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2992                 struct iattr *sattr, dev_t rdev)
2993 {
2994         struct nfs4_createdata *data;
2995         int mode = sattr->ia_mode;
2996         int status = -ENOMEM;
2997
2998         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
2999         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3000
3001         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3002         if (data == NULL)
3003                 goto out;
3004
3005         if (S_ISFIFO(mode))
3006                 data->arg.ftype = NF4FIFO;
3007         else if (S_ISBLK(mode)) {
3008                 data->arg.ftype = NF4BLK;
3009                 data->arg.u.device.specdata1 = MAJOR(rdev);
3010                 data->arg.u.device.specdata2 = MINOR(rdev);
3011         }
3012         else if (S_ISCHR(mode)) {
3013                 data->arg.ftype = NF4CHR;
3014                 data->arg.u.device.specdata1 = MAJOR(rdev);
3015                 data->arg.u.device.specdata2 = MINOR(rdev);
3016         }
3017
3018         status = nfs4_do_create(dir, dentry, data);
3019
3020         nfs4_free_createdata(data);
3021 out:
3022         return status;
3023 }
3024
3025 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3026                 struct iattr *sattr, dev_t rdev)
3027 {
3028         struct nfs4_exception exception = { };
3029         int err;
3030
3031         sattr->ia_mode &= ~current_umask();
3032         do {
3033                 err = nfs4_handle_exception(NFS_SERVER(dir),
3034                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3035                                 &exception);
3036         } while (exception.retry);
3037         return err;
3038 }
3039
3040 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3041                  struct nfs_fsstat *fsstat)
3042 {
3043         struct nfs4_statfs_arg args = {
3044                 .fh = fhandle,
3045                 .bitmask = server->attr_bitmask,
3046         };
3047         struct nfs4_statfs_res res = {
3048                 .fsstat = fsstat,
3049         };
3050         struct rpc_message msg = {
3051                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3052                 .rpc_argp = &args,
3053                 .rpc_resp = &res,
3054         };
3055
3056         nfs_fattr_init(fsstat->fattr);
3057         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3058 }
3059
3060 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3061 {
3062         struct nfs4_exception exception = { };
3063         int err;
3064         do {
3065                 err = nfs4_handle_exception(server,
3066                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3067                                 &exception);
3068         } while (exception.retry);
3069         return err;
3070 }
3071
3072 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3073                 struct nfs_fsinfo *fsinfo)
3074 {
3075         struct nfs4_fsinfo_arg args = {
3076                 .fh = fhandle,
3077                 .bitmask = server->attr_bitmask,
3078         };
3079         struct nfs4_fsinfo_res res = {
3080                 .fsinfo = fsinfo,
3081         };
3082         struct rpc_message msg = {
3083                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3084                 .rpc_argp = &args,
3085                 .rpc_resp = &res,
3086         };
3087
3088         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3089 }
3090
3091 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3092 {
3093         struct nfs4_exception exception = { };
3094         int err;
3095
3096         do {
3097                 err = nfs4_handle_exception(server,
3098                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3099                                 &exception);
3100         } while (exception.retry);
3101         return err;
3102 }
3103
3104 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3105 {
3106         nfs_fattr_init(fsinfo->fattr);
3107         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3108 }
3109
3110 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3111                 struct nfs_pathconf *pathconf)
3112 {
3113         struct nfs4_pathconf_arg args = {
3114                 .fh = fhandle,
3115                 .bitmask = server->attr_bitmask,
3116         };
3117         struct nfs4_pathconf_res res = {
3118                 .pathconf = pathconf,
3119         };
3120         struct rpc_message msg = {
3121                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3122                 .rpc_argp = &args,
3123                 .rpc_resp = &res,
3124         };
3125
3126         /* None of the pathconf attributes are mandatory to implement */
3127         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3128                 memset(pathconf, 0, sizeof(*pathconf));
3129                 return 0;
3130         }
3131
3132         nfs_fattr_init(pathconf->fattr);
3133         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3134 }
3135
3136 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3137                 struct nfs_pathconf *pathconf)
3138 {
3139         struct nfs4_exception exception = { };
3140         int err;
3141
3142         do {
3143                 err = nfs4_handle_exception(server,
3144                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3145                                 &exception);
3146         } while (exception.retry);
3147         return err;
3148 }
3149
3150 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3151 {
3152         struct nfs_server *server = NFS_SERVER(data->inode);
3153
3154         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3155                 nfs_restart_rpc(task, server->nfs_client);
3156                 return -EAGAIN;
3157         }
3158
3159         nfs_invalidate_atime(data->inode);
3160         if (task->tk_status > 0)
3161                 renew_lease(server, data->timestamp);
3162         return 0;
3163 }
3164
3165 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3166 {
3167
3168         dprintk("--> %s\n", __func__);
3169
3170         if (!nfs4_sequence_done(task, &data->res.seq_res))
3171                 return -EAGAIN;
3172
3173         return data->read_done_cb(task, data);
3174 }
3175
3176 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3177 {
3178         data->timestamp   = jiffies;
3179         data->read_done_cb = nfs4_read_done_cb;
3180         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3181 }
3182
3183 /* Reset the the nfs_read_data to send the read to the MDS. */
3184 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3185 {
3186         dprintk("%s Reset task for i/o through\n", __func__);
3187         put_lseg(data->lseg);
3188         data->lseg = NULL;
3189         /* offsets will differ in the dense stripe case */
3190         data->args.offset = data->mds_offset;
3191         data->ds_clp = NULL;
3192         data->args.fh     = NFS_FH(data->inode);
3193         data->read_done_cb = nfs4_read_done_cb;
3194         task->tk_ops = data->mds_ops;
3195         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3196 }
3197 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3198
3199 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3200 {
3201         struct inode *inode = data->inode;
3202
3203         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3204                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3205                 return -EAGAIN;
3206         }
3207         if (task->tk_status >= 0) {
3208                 renew_lease(NFS_SERVER(inode), data->timestamp);
3209                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3210         }
3211         return 0;
3212 }
3213
3214 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3215 {
3216         if (!nfs4_sequence_done(task, &data->res.seq_res))
3217                 return -EAGAIN;
3218         return data->write_done_cb(task, data);
3219 }
3220
3221 /* Reset the the nfs_write_data to send the write to the MDS. */
3222 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3223 {
3224         dprintk("%s Reset task for i/o through\n", __func__);
3225         put_lseg(data->lseg);
3226         data->lseg          = NULL;
3227         data->ds_clp        = NULL;
3228         data->write_done_cb = nfs4_write_done_cb;
3229         data->args.fh       = NFS_FH(data->inode);
3230         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3231         data->args.offset   = data->mds_offset;
3232         data->res.fattr     = &data->fattr;
3233         task->tk_ops        = data->mds_ops;
3234         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3235 }
3236 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3237
3238 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3239 {
3240         struct nfs_server *server = NFS_SERVER(data->inode);
3241
3242         if (data->lseg) {
3243                 data->args.bitmask = NULL;
3244                 data->res.fattr = NULL;
3245         } else
3246                 data->args.bitmask = server->cache_consistency_bitmask;
3247         if (!data->write_done_cb)
3248                 data->write_done_cb = nfs4_write_done_cb;
3249         data->res.server = server;
3250         data->timestamp   = jiffies;
3251
3252         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3253 }
3254
3255 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3256 {
3257         struct inode *inode = data->inode;
3258
3259         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3260                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3261                 return -EAGAIN;
3262         }
3263         nfs_refresh_inode(inode, data->res.fattr);
3264         return 0;
3265 }
3266
3267 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3268 {
3269         if (!nfs4_sequence_done(task, &data->res.seq_res))
3270                 return -EAGAIN;
3271         return data->write_done_cb(task, data);
3272 }
3273
3274 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3275 {
3276         struct nfs_server *server = NFS_SERVER(data->inode);
3277
3278         if (data->lseg) {
3279                 data->args.bitmask = NULL;
3280                 data->res.fattr = NULL;
3281         } else
3282                 data->args.bitmask = server->cache_consistency_bitmask;
3283         if (!data->write_done_cb)
3284                 data->write_done_cb = nfs4_commit_done_cb;
3285         data->res.server = server;
3286         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3287 }
3288
3289 struct nfs4_renewdata {
3290         struct nfs_client       *client;
3291         unsigned long           timestamp;
3292 };
3293
3294 /*
3295  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3296  * standalone procedure for queueing an asynchronous RENEW.
3297  */
3298 static void nfs4_renew_release(void *calldata)
3299 {
3300         struct nfs4_renewdata *data = calldata;
3301         struct nfs_client *clp = data->client;
3302
3303         if (atomic_read(&clp->cl_count) > 1)
3304                 nfs4_schedule_state_renewal(clp);
3305         nfs_put_client(clp);
3306         kfree(data);
3307 }
3308
3309 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3310 {
3311         struct nfs4_renewdata *data = calldata;
3312         struct nfs_client *clp = data->client;
3313         unsigned long timestamp = data->timestamp;
3314
3315         if (task->tk_status < 0) {
3316                 /* Unless we're shutting down, schedule state recovery! */
3317                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
3318                         nfs4_schedule_lease_recovery(clp);
3319                 return;
3320         }
3321         do_renew_lease(clp, timestamp);
3322 }
3323
3324 static const struct rpc_call_ops nfs4_renew_ops = {
3325         .rpc_call_done = nfs4_renew_done,
3326         .rpc_release = nfs4_renew_release,
3327 };
3328
3329 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
3330 {
3331         struct rpc_message msg = {
3332                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3333                 .rpc_argp       = clp,
3334                 .rpc_cred       = cred,
3335         };
3336         struct nfs4_renewdata *data;
3337
3338         if (!atomic_inc_not_zero(&clp->cl_count))
3339                 return -EIO;
3340         data = kmalloc(sizeof(*data), GFP_KERNEL);
3341         if (data == NULL)
3342                 return -ENOMEM;
3343         data->client = clp;
3344         data->timestamp = jiffies;
3345         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3346                         &nfs4_renew_ops, data);
3347 }
3348
3349 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3350 {
3351         struct rpc_message msg = {
3352                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3353                 .rpc_argp       = clp,
3354                 .rpc_cred       = cred,
3355         };
3356         unsigned long now = jiffies;
3357         int status;
3358
3359         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3360         if (status < 0)
3361                 return status;
3362         do_renew_lease(clp, now);
3363         return 0;
3364 }
3365
3366 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3367 {
3368         return (server->caps & NFS_CAP_ACLS)
3369                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3370                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3371 }
3372
3373 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3374  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3375  * the stack.
3376  */
3377 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3378
3379 static void buf_to_pages(const void *buf, size_t buflen,
3380                 struct page **pages, unsigned int *pgbase)
3381 {
3382         const void *p = buf;
3383
3384         *pgbase = offset_in_page(buf);
3385         p -= *pgbase;
3386         while (p < buf + buflen) {
3387                 *(pages++) = virt_to_page(p);
3388                 p += PAGE_CACHE_SIZE;
3389         }
3390 }
3391
3392 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3393                 struct page **pages, unsigned int *pgbase)
3394 {
3395         struct page *newpage, **spages;
3396         int rc = 0;
3397         size_t len;
3398         spages = pages;
3399
3400         do {
3401                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3402                 newpage = alloc_page(GFP_KERNEL);
3403
3404                 if (newpage == NULL)
3405                         goto unwind;
3406                 memcpy(page_address(newpage), buf, len);
3407                 buf += len;
3408                 buflen -= len;
3409                 *pages++ = newpage;
3410                 rc++;
3411         } while (buflen != 0);
3412
3413         return rc;
3414
3415 unwind:
3416         for(; rc > 0; rc--)
3417                 __free_page(spages[rc-1]);
3418         return -ENOMEM;
3419 }
3420
3421 struct nfs4_cached_acl {
3422         int cached;
3423         size_t len;
3424         char data[0];
3425 };
3426
3427 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3428 {
3429         struct nfs_inode *nfsi = NFS_I(inode);
3430
3431         spin_lock(&inode->i_lock);
3432         kfree(nfsi->nfs4_acl);
3433         nfsi->nfs4_acl = acl;
3434         spin_unlock(&inode->i_lock);
3435 }
3436
3437 static void nfs4_zap_acl_attr(struct inode *inode)
3438 {
3439         nfs4_set_cached_acl(inode, NULL);
3440 }
3441
3442 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3443 {
3444         struct nfs_inode *nfsi = NFS_I(inode);
3445         struct nfs4_cached_acl *acl;
3446         int ret = -ENOENT;
3447
3448         spin_lock(&inode->i_lock);
3449         acl = nfsi->nfs4_acl;
3450         if (acl == NULL)
3451                 goto out;
3452         if (buf == NULL) /* user is just asking for length */
3453                 goto out_len;
3454         if (acl->cached == 0)
3455                 goto out;
3456         ret = -ERANGE; /* see getxattr(2) man page */
3457         if (acl->len > buflen)
3458                 goto out;
3459         memcpy(buf, acl->data, acl->len);
3460 out_len:
3461         ret = acl->len;
3462 out:
3463         spin_unlock(&inode->i_lock);
3464         return ret;
3465 }
3466
3467 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3468 {
3469         struct nfs4_cached_acl *acl;
3470
3471         if (buf && acl_len <= PAGE_SIZE) {
3472                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3473                 if (acl == NULL)
3474                         goto out;
3475                 acl->cached = 1;
3476                 memcpy(acl->data, buf, acl_len);
3477         } else {
3478                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3479                 if (acl == NULL)
3480                         goto out;
3481                 acl->cached = 0;
3482         }
3483         acl->len = acl_len;
3484 out:
3485         nfs4_set_cached_acl(inode, acl);
3486 }
3487
3488 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3489 {
3490         struct page *pages[NFS4ACL_MAXPAGES];
3491         struct nfs_getaclargs args = {
3492                 .fh = NFS_FH(inode),
3493                 .acl_pages = pages,
3494                 .acl_len = buflen,
3495         };
3496         struct nfs_getaclres res = {
3497                 .acl_len = buflen,
3498         };
3499         void *resp_buf;
3500         struct rpc_message msg = {
3501                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3502                 .rpc_argp = &args,
3503                 .rpc_resp = &res,
3504         };
3505         struct page *localpage = NULL;
3506         int ret;
3507
3508         if (buflen < PAGE_SIZE) {
3509                 /* As long as we're doing a round trip to the server anyway,
3510                  * let's be prepared for a page of acl data. */
3511                 localpage = alloc_page(GFP_KERNEL);
3512                 resp_buf = page_address(localpage);
3513                 if (localpage == NULL)
3514                         return -ENOMEM;
3515                 args.acl_pages[0] = localpage;
3516                 args.acl_pgbase = 0;
3517                 args.acl_len = PAGE_SIZE;
3518         } else {
3519                 resp_buf = buf;
3520                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3521         }
3522         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3523         if (ret)
3524                 goto out_free;
3525         if (res.acl_len > args.acl_len)
3526                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3527         else
3528                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3529         if (buf) {
3530                 ret = -ERANGE;
3531                 if (res.acl_len > buflen)
3532                         goto out_free;
3533                 if (localpage)
3534                         memcpy(buf, resp_buf, res.acl_len);
3535         }
3536         ret = res.acl_len;
3537 out_free:
3538         if (localpage)
3539                 __free_page(localpage);
3540         return ret;
3541 }
3542
3543 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3544 {
3545         struct nfs4_exception exception = { };
3546         ssize_t ret;
3547         do {
3548                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3549                 if (ret >= 0)
3550                         break;
3551                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3552         } while (exception.retry);
3553         return ret;
3554 }
3555
3556 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3557 {
3558         struct nfs_server *server = NFS_SERVER(inode);
3559         int ret;
3560
3561         if (!nfs4_server_supports_acls(server))
3562                 return -EOPNOTSUPP;
3563         ret = nfs_revalidate_inode(server, inode);
3564         if (ret < 0)
3565                 return ret;
3566         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3567                 nfs_zap_acl_cache(inode);
3568         ret = nfs4_read_cached_acl(inode, buf, buflen);
3569         if (ret != -ENOENT)
3570                 return ret;
3571         return nfs4_get_acl_uncached(inode, buf, buflen);
3572 }
3573
3574 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3575 {
3576         struct nfs_server *server = NFS_SERVER(inode);
3577         struct page *pages[NFS4ACL_MAXPAGES];
3578         struct nfs_setaclargs arg = {
3579                 .fh             = NFS_FH(inode),
3580                 .acl_pages      = pages,
3581                 .acl_len        = buflen,
3582         };
3583         struct nfs_setaclres res;
3584         struct rpc_message msg = {
3585                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3586                 .rpc_argp       = &arg,
3587                 .rpc_resp       = &res,
3588         };
3589         int ret, i;
3590
3591         if (!nfs4_server_supports_acls(server))
3592                 return -EOPNOTSUPP;
3593         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3594         if (i < 0)
3595                 return i;
3596         nfs_inode_return_delegation(inode);
3597         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3598
3599         /*
3600          * Free each page after tx, so the only ref left is
3601          * held by the network stack
3602          */
3603         for (; i > 0; i--)
3604                 put_page(pages[i-1]);
3605
3606         /*
3607          * Acl update can result in inode attribute update.
3608          * so mark the attribute cache invalid.
3609          */
3610         spin_lock(&inode->i_lock);
3611         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3612         spin_unlock(&inode->i_lock);
3613         nfs_access_zap_cache(inode);
3614         nfs_zap_acl_cache(inode);
3615         return ret;
3616 }
3617
3618 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3619 {
3620         struct nfs4_exception exception = { };
3621         int err;
3622         do {
3623                 err = nfs4_handle_exception(NFS_SERVER(inode),
3624                                 __nfs4_proc_set_acl(inode, buf, buflen),
3625                                 &exception);
3626         } while (exception.retry);
3627         return err;
3628 }
3629
3630 static int
3631 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3632 {
3633         struct nfs_client *clp = server->nfs_client;
3634
3635         if (task->tk_status >= 0)
3636                 return 0;
3637         switch(task->tk_status) {
3638                 case -NFS4ERR_ADMIN_REVOKED:
3639                 case -NFS4ERR_BAD_STATEID:
3640                 case -NFS4ERR_OPENMODE:
3641                         if (state == NULL)
3642                                 break;
3643                         nfs4_schedule_stateid_recovery(server, state);
3644                         goto wait_on_recovery;
3645                 case -NFS4ERR_STALE_STATEID:
3646                 case -NFS4ERR_STALE_CLIENTID:
3647                 case -NFS4ERR_EXPIRED:
3648                         nfs4_schedule_lease_recovery(clp);
3649                         goto wait_on_recovery;
3650 #if defined(CONFIG_NFS_V4_1)
3651                 case -NFS4ERR_BADSESSION:
3652                 case -NFS4ERR_BADSLOT:
3653                 case -NFS4ERR_BAD_HIGH_SLOT:
3654                 case -NFS4ERR_DEADSESSION:
3655                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3656                 case -NFS4ERR_SEQ_FALSE_RETRY:
3657                 case -NFS4ERR_SEQ_MISORDERED:
3658                         dprintk("%s ERROR %d, Reset session\n", __func__,
3659                                 task->tk_status);
3660                         nfs4_schedule_session_recovery(clp->cl_session);
3661                         task->tk_status = 0;
3662                         return -EAGAIN;
3663 #endif /* CONFIG_NFS_V4_1 */
3664                 case -NFS4ERR_DELAY:
3665                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3666                 case -NFS4ERR_GRACE:
3667                 case -EKEYEXPIRED:
3668                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3669                         task->tk_status = 0;
3670                         return -EAGAIN;
3671                 case -NFS4ERR_OLD_STATEID:
3672                         task->tk_status = 0;
3673                         return -EAGAIN;
3674         }
3675         task->tk_status = nfs4_map_errors(task->tk_status);
3676         return 0;
3677 wait_on_recovery:
3678         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3679         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3680                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3681         task->tk_status = 0;
3682         return -EAGAIN;
3683 }
3684
3685 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3686                 unsigned short port, struct rpc_cred *cred,
3687                 struct nfs4_setclientid_res *res)
3688 {
3689         nfs4_verifier sc_verifier;
3690         struct nfs4_setclientid setclientid = {
3691                 .sc_verifier = &sc_verifier,
3692                 .sc_prog = program,
3693                 .sc_cb_ident = clp->cl_cb_ident,
3694         };
3695         struct rpc_message msg = {
3696                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3697                 .rpc_argp = &setclientid,
3698                 .rpc_resp = res,
3699                 .rpc_cred = cred,
3700         };
3701         __be32 *p;
3702         int loop = 0;
3703         int status;
3704
3705         p = (__be32*)sc_verifier.data;
3706         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3707         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3708
3709         for(;;) {
3710                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3711                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3712                                 clp->cl_ipaddr,
3713                                 rpc_peeraddr2str(clp->cl_rpcclient,
3714                                                         RPC_DISPLAY_ADDR),
3715                                 rpc_peeraddr2str(clp->cl_rpcclient,
3716                                                         RPC_DISPLAY_PROTO),
3717                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3718                                 clp->cl_id_uniquifier);
3719                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3720                                 sizeof(setclientid.sc_netid),
3721                                 rpc_peeraddr2str(clp->cl_rpcclient,
3722                                                         RPC_DISPLAY_NETID));
3723                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3724                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3725                                 clp->cl_ipaddr, port >> 8, port & 255);
3726
3727                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3728                 if (status != -NFS4ERR_CLID_INUSE)
3729                         break;
3730                 if (signalled())
3731                         break;
3732                 if (loop++ & 1)
3733                         ssleep(clp->cl_lease_time / HZ + 1);
3734                 else
3735                         if (++clp->cl_id_uniquifier == 0)
3736                                 break;
3737         }
3738         return status;
3739 }
3740
3741 static int _nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3742                 struct nfs4_setclientid_res *arg,
3743                 struct rpc_cred *cred)
3744 {
3745         struct nfs_fsinfo fsinfo;
3746         struct rpc_message msg = {
3747                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3748                 .rpc_argp = arg,
3749                 .rpc_resp = &fsinfo,
3750                 .rpc_cred = cred,
3751         };
3752         unsigned long now;
3753         int status;
3754
3755         now = jiffies;
3756         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3757         if (status == 0) {
3758                 spin_lock(&clp->cl_lock);
3759                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3760                 clp->cl_last_renewal = now;
3761                 spin_unlock(&clp->cl_lock);
3762         }
3763         return status;
3764 }
3765
3766 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3767                 struct nfs4_setclientid_res *arg,
3768                 struct rpc_cred *cred)
3769 {
3770         long timeout = 0;
3771         int err;
3772         do {
3773                 err = _nfs4_proc_setclientid_confirm(clp, arg, cred);
3774                 switch (err) {
3775                         case 0:
3776                                 return err;
3777                         case -NFS4ERR_RESOURCE:
3778                                 /* The IBM lawyers misread another document! */
3779                         case -NFS4ERR_DELAY:
3780                                 err = nfs4_delay(clp->cl_rpcclient, &timeout);
3781                 }
3782         } while (err == 0);
3783         return err;
3784 }
3785
3786 struct nfs4_delegreturndata {
3787         struct nfs4_delegreturnargs args;
3788         struct nfs4_delegreturnres res;
3789         struct nfs_fh fh;
3790         nfs4_stateid stateid;
3791         unsigned long timestamp;
3792         struct nfs_fattr fattr;
3793         int rpc_status;
3794 };
3795
3796 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3797 {
3798         struct nfs4_delegreturndata *data = calldata;
3799
3800         if (!nfs4_sequence_done(task, &data->res.seq_res))
3801                 return;
3802
3803         switch (task->tk_status) {
3804         case -NFS4ERR_STALE_STATEID:
3805         case -NFS4ERR_EXPIRED:
3806         case 0:
3807                 renew_lease(data->res.server, data->timestamp);
3808                 break;
3809         default:
3810                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3811                                 -EAGAIN) {
3812                         nfs_restart_rpc(task, data->res.server->nfs_client);
3813                         return;
3814                 }
3815         }
3816         data->rpc_status = task->tk_status;
3817 }
3818
3819 static void nfs4_delegreturn_release(void *calldata)
3820 {
3821         kfree(calldata);
3822 }
3823
3824 #if defined(CONFIG_NFS_V4_1)
3825 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3826 {
3827         struct nfs4_delegreturndata *d_data;
3828
3829         d_data = (struct nfs4_delegreturndata *)data;
3830
3831         if (nfs4_setup_sequence(d_data->res.server,
3832                                 &d_data->args.seq_args,
3833                                 &d_data->res.seq_res, 1, task))
3834                 return;
3835         rpc_call_start(task);
3836 }
3837 #endif /* CONFIG_NFS_V4_1 */
3838
3839 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3840 #if defined(CONFIG_NFS_V4_1)
3841         .rpc_call_prepare = nfs4_delegreturn_prepare,
3842 #endif /* CONFIG_NFS_V4_1 */
3843         .rpc_call_done = nfs4_delegreturn_done,
3844         .rpc_release = nfs4_delegreturn_release,
3845 };
3846
3847 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3848 {
3849         struct nfs4_delegreturndata *data;
3850         struct nfs_server *server = NFS_SERVER(inode);
3851         struct rpc_task *task;
3852         struct rpc_message msg = {
3853                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3854                 .rpc_cred = cred,
3855         };
3856         struct rpc_task_setup task_setup_data = {
3857                 .rpc_client = server->client,
3858                 .rpc_message = &msg,
3859                 .callback_ops = &nfs4_delegreturn_ops,
3860                 .flags = RPC_TASK_ASYNC,
3861         };
3862         int status = 0;
3863
3864         data = kzalloc(sizeof(*data), GFP_NOFS);
3865         if (data == NULL)
3866                 return -ENOMEM;
3867         data->args.fhandle = &data->fh;
3868         data->args.stateid = &data->stateid;
3869         data->args.bitmask = server->attr_bitmask;
3870         nfs_copy_fh(&data->fh, NFS_FH(inode));
3871         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3872         data->res.fattr = &data->fattr;
3873         data->res.server = server;
3874         nfs_fattr_init(data->res.fattr);
3875         data->timestamp = jiffies;
3876         data->rpc_status = 0;
3877
3878         task_setup_data.callback_data = data;
3879         msg.rpc_argp = &data->args;
3880         msg.rpc_resp = &data->res;
3881         task = rpc_run_task(&task_setup_data);
3882         if (IS_ERR(task))
3883                 return PTR_ERR(task);
3884         if (!issync)
3885                 goto out;
3886         status = nfs4_wait_for_completion_rpc_task(task);
3887         if (status != 0)
3888                 goto out;
3889         status = data->rpc_status;
3890         if (status != 0)
3891                 goto out;
3892         nfs_refresh_inode(inode, &data->fattr);
3893 out:
3894         rpc_put_task(task);
3895         return status;
3896 }
3897
3898 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3899 {
3900         struct nfs_server *server = NFS_SERVER(inode);
3901         struct nfs4_exception exception = { };
3902         int err;
3903         do {
3904                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3905                 switch (err) {
3906                         case -NFS4ERR_STALE_STATEID:
3907                         case -NFS4ERR_EXPIRED:
3908                         case 0:
3909                                 return 0;
3910                 }
3911                 err = nfs4_handle_exception(server, err, &exception);
3912         } while (exception.retry);
3913         return err;
3914 }
3915
3916 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3917 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3918
3919 /*
3920  * sleep, with exponential backoff, and retry the LOCK operation.
3921  */
3922 static unsigned long
3923 nfs4_set_lock_task_retry(unsigned long timeout)
3924 {
3925         schedule_timeout_killable(timeout);
3926         timeout <<= 1;
3927         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3928                 return NFS4_LOCK_MAXTIMEOUT;
3929         return timeout;
3930 }
3931
3932 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3933 {
3934         struct inode *inode = state->inode;
3935         struct nfs_server *server = NFS_SERVER(inode);
3936         struct nfs_client *clp = server->nfs_client;
3937         struct nfs_lockt_args arg = {
3938                 .fh = NFS_FH(inode),
3939                 .fl = request,
3940         };
3941         struct nfs_lockt_res res = {
3942                 .denied = request,
3943         };
3944         struct rpc_message msg = {
3945                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3946                 .rpc_argp       = &arg,
3947                 .rpc_resp       = &res,
3948                 .rpc_cred       = state->owner->so_cred,
3949         };
3950         struct nfs4_lock_state *lsp;
3951         int status;
3952
3953         arg.lock_owner.clientid = clp->cl_clientid;
3954         status = nfs4_set_lock_state(state, request);
3955         if (status != 0)
3956                 goto out;
3957         lsp = request->fl_u.nfs4_fl.owner;
3958         arg.lock_owner.id = lsp->ls_id.id;
3959         arg.lock_owner.s_dev = server->s_dev;
3960         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3961         switch (status) {
3962                 case 0:
3963                         request->fl_type = F_UNLCK;
3964                         break;
3965                 case -NFS4ERR_DENIED:
3966                         status = 0;
3967         }
3968         request->fl_ops->fl_release_private(request);
3969 out:
3970         return status;
3971 }
3972
3973 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3974 {
3975         struct nfs4_exception exception = { };
3976         int err;
3977
3978         do {
3979                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3980                                 _nfs4_proc_getlk(state, cmd, request),
3981                                 &exception);
3982         } while (exception.retry);
3983         return err;
3984 }
3985
3986 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3987 {
3988         int res = 0;
3989         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3990                 case FL_POSIX:
3991                         res = posix_lock_file_wait(file, fl);
3992                         break;
3993                 case FL_FLOCK:
3994                         res = flock_lock_file_wait(file, fl);
3995                         break;
3996                 default:
3997                         BUG();
3998         }
3999         return res;
4000 }
4001
4002 struct nfs4_unlockdata {
4003         struct nfs_locku_args arg;
4004         struct nfs_locku_res res;
4005         struct nfs4_lock_state *lsp;
4006         struct nfs_open_context *ctx;
4007         struct file_lock fl;
4008         const struct nfs_server *server;
4009         unsigned long timestamp;
4010 };
4011
4012 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4013                 struct nfs_open_context *ctx,
4014                 struct nfs4_lock_state *lsp,
4015                 struct nfs_seqid *seqid)
4016 {
4017         struct nfs4_unlockdata *p;
4018         struct inode *inode = lsp->ls_state->inode;
4019
4020         p = kzalloc(sizeof(*p), GFP_NOFS);
4021         if (p == NULL)
4022                 return NULL;
4023         p->arg.fh = NFS_FH(inode);
4024         p->arg.fl = &p->fl;
4025         p->arg.seqid = seqid;
4026         p->res.seqid = seqid;
4027         p->arg.stateid = &lsp->ls_stateid;
4028         p->lsp = lsp;
4029         atomic_inc(&lsp->ls_count);
4030         /* Ensure we don't close file until we're done freeing locks! */
4031         p->ctx = get_nfs_open_context(ctx);
4032         memcpy(&p->fl, fl, sizeof(p->fl));
4033         p->server = NFS_SERVER(inode);
4034         return p;
4035 }
4036
4037 static void nfs4_locku_release_calldata(void *data)
4038 {
4039         struct nfs4_unlockdata *calldata = data;
4040         nfs_free_seqid(calldata->arg.seqid);
4041         nfs4_put_lock_state(calldata->lsp);
4042         put_nfs_open_context(calldata->ctx);
4043         kfree(calldata);
4044 }
4045
4046 static void nfs4_locku_done(struct rpc_task *task, void *data)
4047 {
4048         struct nfs4_unlockdata *calldata = data;
4049
4050         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4051                 return;
4052         switch (task->tk_status) {
4053                 case 0:
4054                         memcpy(calldata->lsp->ls_stateid.data,
4055                                         calldata->res.stateid.data,
4056                                         sizeof(calldata->lsp->ls_stateid.data));
4057                         renew_lease(calldata->server, calldata->timestamp);
4058                         break;
4059                 case -NFS4ERR_BAD_STATEID:
4060                 case -NFS4ERR_OLD_STATEID:
4061                 case -NFS4ERR_STALE_STATEID:
4062                 case -NFS4ERR_EXPIRED:
4063                         break;
4064                 default:
4065                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4066                                 nfs_restart_rpc(task,
4067                                                  calldata->server->nfs_client);
4068         }
4069 }
4070
4071 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4072 {
4073         struct nfs4_unlockdata *calldata = data;
4074
4075         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4076                 return;
4077         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4078                 /* Note: exit _without_ running nfs4_locku_done */
4079                 task->tk_action = NULL;
4080                 return;
4081         }
4082         calldata->timestamp = jiffies;
4083         if (nfs4_setup_sequence(calldata->server,
4084                                 &calldata->arg.seq_args,
4085                                 &calldata->res.seq_res, 1, task))
4086                 return;
4087         rpc_call_start(task);
4088 }
4089
4090 static const struct rpc_call_ops nfs4_locku_ops = {
4091         .rpc_call_prepare = nfs4_locku_prepare,
4092         .rpc_call_done = nfs4_locku_done,
4093         .rpc_release = nfs4_locku_release_calldata,
4094 };
4095
4096 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4097                 struct nfs_open_context *ctx,
4098                 struct nfs4_lock_state *lsp,
4099                 struct nfs_seqid *seqid)
4100 {
4101         struct nfs4_unlockdata *data;
4102         struct rpc_message msg = {
4103                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4104                 .rpc_cred = ctx->cred,
4105         };
4106         struct rpc_task_setup task_setup_data = {
4107                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4108                 .rpc_message = &msg,
4109                 .callback_ops = &nfs4_locku_ops,
4110                 .workqueue = nfsiod_workqueue,
4111                 .flags = RPC_TASK_ASYNC,
4112         };
4113
4114         /* Ensure this is an unlock - when canceling a lock, the
4115          * canceled lock is passed in, and it won't be an unlock.
4116          */
4117         fl->fl_type = F_UNLCK;
4118
4119         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4120         if (data == NULL) {
4121                 nfs_free_seqid(seqid);
4122                 return ERR_PTR(-ENOMEM);
4123         }
4124
4125         msg.rpc_argp = &data->arg;
4126         msg.rpc_resp = &data->res;
4127         task_setup_data.callback_data = data;
4128         return rpc_run_task(&task_setup_data);
4129 }
4130
4131 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4132 {
4133         struct nfs_inode *nfsi = NFS_I(state->inode);
4134         struct nfs_seqid *seqid;
4135         struct nfs4_lock_state *lsp;
4136         struct rpc_task *task;
4137         int status = 0;
4138         unsigned char fl_flags = request->fl_flags;
4139
4140         status = nfs4_set_lock_state(state, request);
4141         /* Unlock _before_ we do the RPC call */
4142         request->fl_flags |= FL_EXISTS;
4143         down_read(&nfsi->rwsem);
4144         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4145                 up_read(&nfsi->rwsem);
4146                 goto out;
4147         }
4148         up_read(&nfsi->rwsem);
4149         if (status != 0)
4150                 goto out;
4151         /* Is this a delegated lock? */
4152         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4153                 goto out;
4154         lsp = request->fl_u.nfs4_fl.owner;
4155         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4156         status = -ENOMEM;
4157         if (seqid == NULL)
4158                 goto out;
4159         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4160         status = PTR_ERR(task);
4161         if (IS_ERR(task))
4162                 goto out;
4163         status = nfs4_wait_for_completion_rpc_task(task);
4164         rpc_put_task(task);
4165 out:
4166         request->fl_flags = fl_flags;
4167         return status;
4168 }
4169
4170 struct nfs4_lockdata {
4171         struct nfs_lock_args arg;
4172         struct nfs_lock_res res;
4173         struct nfs4_lock_state *lsp;
4174         struct nfs_open_context *ctx;
4175         struct file_lock fl;
4176         unsigned long timestamp;
4177         int rpc_status;
4178         int cancelled;
4179         struct nfs_server *server;
4180 };
4181
4182 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4183                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4184                 gfp_t gfp_mask)
4185 {
4186         struct nfs4_lockdata *p;
4187         struct inode *inode = lsp->ls_state->inode;
4188         struct nfs_server *server = NFS_SERVER(inode);
4189
4190         p = kzalloc(sizeof(*p), gfp_mask);
4191         if (p == NULL)
4192                 return NULL;
4193
4194         p->arg.fh = NFS_FH(inode);
4195         p->arg.fl = &p->fl;
4196         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4197         if (p->arg.open_seqid == NULL)
4198                 goto out_free;
4199         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4200         if (p->arg.lock_seqid == NULL)
4201                 goto out_free_seqid;
4202         p->arg.lock_stateid = &lsp->ls_stateid;
4203         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4204         p->arg.lock_owner.id = lsp->ls_id.id;
4205         p->arg.lock_owner.s_dev = server->s_dev;
4206         p->res.lock_seqid = p->arg.lock_seqid;
4207         p->lsp = lsp;
4208         p->server = server;
4209         atomic_inc(&lsp->ls_count);
4210         p->ctx = get_nfs_open_context(ctx);
4211         memcpy(&p->fl, fl, sizeof(p->fl));
4212         return p;
4213 out_free_seqid:
4214         nfs_free_seqid(p->arg.open_seqid);
4215 out_free:
4216         kfree(p);
4217         return NULL;
4218 }
4219
4220 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4221 {
4222         struct nfs4_lockdata *data = calldata;
4223         struct nfs4_state *state = data->lsp->ls_state;
4224
4225         dprintk("%s: begin!\n", __func__);
4226         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4227                 return;
4228         /* Do we need to do an open_to_lock_owner? */
4229         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4230                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4231                         return;
4232                 data->arg.open_stateid = &state->stateid;
4233                 data->arg.new_lock_owner = 1;
4234                 data->res.open_seqid = data->arg.open_seqid;
4235         } else
4236                 data->arg.new_lock_owner = 0;
4237         data->timestamp = jiffies;
4238         if (nfs4_setup_sequence(data->server,
4239                                 &data->arg.seq_args,
4240                                 &data->res.seq_res, 1, task))
4241                 return;
4242         rpc_call_start(task);
4243         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4244 }
4245
4246 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4247 {
4248         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4249         nfs4_lock_prepare(task, calldata);
4250 }
4251
4252 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4253 {
4254         struct nfs4_lockdata *data = calldata;
4255
4256         dprintk("%s: begin!\n", __func__);
4257
4258         if (!nfs4_sequence_done(task, &data->res.seq_res))
4259                 return;
4260
4261         data->rpc_status = task->tk_status;
4262         if (data->arg.new_lock_owner != 0) {
4263                 if (data->rpc_status == 0)
4264                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4265                 else
4266                         goto out;
4267         }
4268         if (data->rpc_status == 0) {
4269                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4270                                         sizeof(data->lsp->ls_stateid.data));
4271                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4272                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
4273         }
4274 out:
4275         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4276 }
4277
4278 static void nfs4_lock_release(void *calldata)
4279 {
4280         struct nfs4_lockdata *data = calldata;
4281
4282         dprintk("%s: begin!\n", __func__);
4283         nfs_free_seqid(data->arg.open_seqid);
4284         if (data->cancelled != 0) {
4285                 struct rpc_task *task;
4286                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4287                                 data->arg.lock_seqid);
4288                 if (!IS_ERR(task))
4289                         rpc_put_task_async(task);
4290                 dprintk("%s: cancelling lock!\n", __func__);
4291         } else
4292                 nfs_free_seqid(data->arg.lock_seqid);
4293         nfs4_put_lock_state(data->lsp);
4294         put_nfs_open_context(data->ctx);
4295         kfree(data);
4296         dprintk("%s: done!\n", __func__);
4297 }
4298
4299 static const struct rpc_call_ops nfs4_lock_ops = {
4300         .rpc_call_prepare = nfs4_lock_prepare,
4301         .rpc_call_done = nfs4_lock_done,
4302         .rpc_release = nfs4_lock_release,
4303 };
4304
4305 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4306         .rpc_call_prepare = nfs4_recover_lock_prepare,
4307         .rpc_call_done = nfs4_lock_done,
4308         .rpc_release = nfs4_lock_release,
4309 };
4310
4311 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4312 {
4313         switch (error) {
4314         case -NFS4ERR_ADMIN_REVOKED:
4315         case -NFS4ERR_BAD_STATEID:
4316                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4317                 if (new_lock_owner != 0 ||
4318                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4319                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4320                 break;
4321         case -NFS4ERR_STALE_STATEID:
4322                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4323         case -NFS4ERR_EXPIRED:
4324                 nfs4_schedule_lease_recovery(server->nfs_client);
4325         };
4326 }
4327
4328 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4329 {
4330         struct nfs4_lockdata *data;
4331         struct rpc_task *task;
4332         struct rpc_message msg = {
4333                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4334                 .rpc_cred = state->owner->so_cred,
4335         };
4336         struct rpc_task_setup task_setup_data = {
4337                 .rpc_client = NFS_CLIENT(state->inode),
4338                 .rpc_message = &msg,
4339                 .callback_ops = &nfs4_lock_ops,
4340                 .workqueue = nfsiod_workqueue,
4341                 .flags = RPC_TASK_ASYNC,
4342         };
4343         int ret;
4344
4345         dprintk("%s: begin!\n", __func__);
4346         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4347                         fl->fl_u.nfs4_fl.owner,
4348                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4349         if (data == NULL)
4350                 return -ENOMEM;
4351         if (IS_SETLKW(cmd))
4352                 data->arg.block = 1;
4353         if (recovery_type > NFS_LOCK_NEW) {
4354                 if (recovery_type == NFS_LOCK_RECLAIM)
4355                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4356                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4357         }
4358         msg.rpc_argp = &data->arg;
4359         msg.rpc_resp = &data->res;
4360         task_setup_data.callback_data = data;
4361         task = rpc_run_task(&task_setup_data);
4362         if (IS_ERR(task))
4363                 return PTR_ERR(task);
4364         ret = nfs4_wait_for_completion_rpc_task(task);
4365         if (ret == 0) {
4366                 ret = data->rpc_status;
4367                 if (ret)
4368                         nfs4_handle_setlk_error(data->server, data->lsp,
4369                                         data->arg.new_lock_owner, ret);
4370         } else
4371                 data->cancelled = 1;
4372         rpc_put_task(task);
4373         dprintk("%s: done, ret = %d!\n", __func__, ret);
4374         return ret;
4375 }
4376
4377 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4378 {
4379         struct nfs_server *server = NFS_SERVER(state->inode);
4380         struct nfs4_exception exception = { };
4381         int err;
4382
4383         do {
4384                 /* Cache the lock if possible... */
4385                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4386                         return 0;
4387                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4388                 if (err != -NFS4ERR_DELAY)
4389                         break;
4390                 nfs4_handle_exception(server, err, &exception);
4391         } while (exception.retry);
4392         return err;
4393 }
4394
4395 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4396 {
4397         struct nfs_server *server = NFS_SERVER(state->inode);
4398         struct nfs4_exception exception = { };
4399         int err;
4400
4401         err = nfs4_set_lock_state(state, request);
4402         if (err != 0)
4403                 return err;
4404         do {
4405                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4406                         return 0;
4407                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4408                 switch (err) {
4409                 default:
4410                         goto out;
4411                 case -NFS4ERR_GRACE:
4412                 case -NFS4ERR_DELAY:
4413                         nfs4_handle_exception(server, err, &exception);
4414                         err = 0;
4415                 }
4416         } while (exception.retry);
4417 out:
4418         return err;
4419 }
4420
4421 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4422 {
4423         struct nfs_inode *nfsi = NFS_I(state->inode);
4424         unsigned char fl_flags = request->fl_flags;
4425         int status = -ENOLCK;
4426
4427         if ((fl_flags & FL_POSIX) &&
4428                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4429                 goto out;
4430         /* Is this a delegated open? */
4431         status = nfs4_set_lock_state(state, request);
4432         if (status != 0)
4433                 goto out;
4434         request->fl_flags |= FL_ACCESS;
4435         status = do_vfs_lock(request->fl_file, request);
4436         if (status < 0)
4437                 goto out;
4438         down_read(&nfsi->rwsem);
4439         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4440                 /* Yes: cache locks! */
4441                 /* ...but avoid races with delegation recall... */
4442                 request->fl_flags = fl_flags & ~FL_SLEEP;
4443                 status = do_vfs_lock(request->fl_file, request);
4444                 goto out_unlock;
4445         }
4446         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4447         if (status != 0)
4448                 goto out_unlock;
4449         /* Note: we always want to sleep here! */
4450         request->fl_flags = fl_flags | FL_SLEEP;
4451         if (do_vfs_lock(request->fl_file, request) < 0)
4452                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4453 out_unlock:
4454         up_read(&nfsi->rwsem);
4455 out:
4456         request->fl_flags = fl_flags;
4457         return status;
4458 }
4459
4460 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4461 {
4462         struct nfs4_exception exception = { };
4463         int err;
4464
4465         do {
4466                 err = _nfs4_proc_setlk(state, cmd, request);
4467                 if (err == -NFS4ERR_DENIED)
4468                         err = -EAGAIN;
4469                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4470                                 err, &exception);
4471         } while (exception.retry);
4472         return err;
4473 }
4474
4475 static int
4476 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4477 {
4478         struct nfs_open_context *ctx;
4479         struct nfs4_state *state;
4480         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4481         int status;
4482
4483         /* verify open state */
4484         ctx = nfs_file_open_context(filp);
4485         state = ctx->state;
4486
4487         if (request->fl_start < 0 || request->fl_end < 0)
4488                 return -EINVAL;
4489
4490         if (IS_GETLK(cmd)) {
4491                 if (state != NULL)
4492                         return nfs4_proc_getlk(state, F_GETLK, request);
4493                 return 0;
4494         }
4495
4496         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4497                 return -EINVAL;
4498
4499         if (request->fl_type == F_UNLCK) {
4500                 if (state != NULL)
4501                         return nfs4_proc_unlck(state, cmd, request);
4502                 return 0;
4503         }
4504
4505         if (state == NULL)
4506                 return -ENOLCK;
4507         do {
4508                 status = nfs4_proc_setlk(state, cmd, request);
4509                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4510                         break;
4511                 timeout = nfs4_set_lock_task_retry(timeout);
4512                 status = -ERESTARTSYS;
4513                 if (signalled())
4514                         break;
4515         } while(status < 0);
4516         return status;
4517 }
4518
4519 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4520 {
4521         struct nfs_server *server = NFS_SERVER(state->inode);
4522         struct nfs4_exception exception = { };
4523         int err;
4524
4525         err = nfs4_set_lock_state(state, fl);
4526         if (err != 0)
4527                 goto out;
4528         do {
4529                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4530                 switch (err) {
4531                         default:
4532                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4533                                                 __func__, err);
4534                         case 0:
4535                         case -ESTALE:
4536                                 goto out;
4537                         case -NFS4ERR_EXPIRED:
4538                         case -NFS4ERR_STALE_CLIENTID:
4539                         case -NFS4ERR_STALE_STATEID:
4540                                 nfs4_schedule_lease_recovery(server->nfs_client);
4541                                 goto out;
4542                         case -NFS4ERR_BADSESSION:
4543                         case -NFS4ERR_BADSLOT:
4544                         case -NFS4ERR_BAD_HIGH_SLOT:
4545                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4546                         case -NFS4ERR_DEADSESSION:
4547                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4548                                 goto out;
4549                         case -ERESTARTSYS:
4550                                 /*
4551                                  * The show must go on: exit, but mark the
4552                                  * stateid as needing recovery.
4553                                  */
4554                         case -NFS4ERR_ADMIN_REVOKED:
4555                         case -NFS4ERR_BAD_STATEID:
4556                         case -NFS4ERR_OPENMODE:
4557                                 nfs4_schedule_stateid_recovery(server, state);
4558                                 err = 0;
4559                                 goto out;
4560                         case -EKEYEXPIRED:
4561                                 /*
4562                                  * User RPCSEC_GSS context has expired.
4563                                  * We cannot recover this stateid now, so
4564                                  * skip it and allow recovery thread to
4565                                  * proceed.
4566                                  */
4567                                 err = 0;
4568                                 goto out;
4569                         case -ENOMEM:
4570                         case -NFS4ERR_DENIED:
4571                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4572                                 err = 0;
4573                                 goto out;
4574                         case -NFS4ERR_DELAY:
4575                                 break;
4576                 }
4577                 err = nfs4_handle_exception(server, err, &exception);
4578         } while (exception.retry);
4579 out:
4580         return err;
4581 }
4582
4583 static void nfs4_release_lockowner_release(void *calldata)
4584 {
4585         kfree(calldata);
4586 }
4587
4588 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4589         .rpc_release = nfs4_release_lockowner_release,
4590 };
4591
4592 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4593 {
4594         struct nfs_server *server = lsp->ls_state->owner->so_server;
4595         struct nfs_release_lockowner_args *args;
4596         struct rpc_message msg = {
4597                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4598         };
4599
4600         if (server->nfs_client->cl_mvops->minor_version != 0)
4601                 return;
4602         args = kmalloc(sizeof(*args), GFP_NOFS);
4603         if (!args)
4604                 return;
4605         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4606         args->lock_owner.id = lsp->ls_id.id;
4607         args->lock_owner.s_dev = server->s_dev;
4608         msg.rpc_argp = args;
4609         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4610 }
4611
4612 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4613
4614 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4615                                    const void *buf, size_t buflen,
4616                                    int flags, int type)
4617 {
4618         if (strcmp(key, "") != 0)
4619                 return -EINVAL;
4620
4621         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4622 }
4623
4624 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4625                                    void *buf, size_t buflen, int type)
4626 {
4627         if (strcmp(key, "") != 0)
4628                 return -EINVAL;
4629
4630         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4631 }
4632
4633 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4634                                        size_t list_len, const char *name,
4635                                        size_t name_len, int type)
4636 {
4637         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4638
4639         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4640                 return 0;
4641
4642         if (list && len <= list_len)
4643                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4644         return len;
4645 }
4646
4647 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4648 {
4649         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4650                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4651                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4652                 return;
4653
4654         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4655                 NFS_ATTR_FATTR_NLINK;
4656         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4657         fattr->nlink = 2;
4658 }
4659
4660 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4661                 struct nfs4_fs_locations *fs_locations, struct page *page)
4662 {
4663         struct nfs_server *server = NFS_SERVER(dir);
4664         u32 bitmask[2] = {
4665                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4666                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4667         };
4668         struct nfs4_fs_locations_arg args = {
4669                 .dir_fh = NFS_FH(dir),
4670                 .name = name,
4671                 .page = page,
4672                 .bitmask = bitmask,
4673         };
4674         struct nfs4_fs_locations_res res = {
4675                 .fs_locations = fs_locations,
4676         };
4677         struct rpc_message msg = {
4678                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4679                 .rpc_argp = &args,
4680                 .rpc_resp = &res,
4681         };
4682         int status;
4683
4684         dprintk("%s: start\n", __func__);
4685         nfs_fattr_init(&fs_locations->fattr);
4686         fs_locations->server = server;
4687         fs_locations->nlocations = 0;
4688         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4689         nfs_fixup_referral_attributes(&fs_locations->fattr);
4690         dprintk("%s: returned status = %d\n", __func__, status);
4691         return status;
4692 }
4693
4694 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4695 {
4696         int status;
4697         struct nfs4_secinfo_arg args = {
4698                 .dir_fh = NFS_FH(dir),
4699                 .name   = name,
4700         };
4701         struct nfs4_secinfo_res res = {
4702                 .flavors     = flavors,
4703         };
4704         struct rpc_message msg = {
4705                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4706                 .rpc_argp = &args,
4707                 .rpc_resp = &res,
4708         };
4709
4710         dprintk("NFS call  secinfo %s\n", name->name);
4711         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4712         dprintk("NFS reply  secinfo: %d\n", status);
4713         return status;
4714 }
4715
4716 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4717 {
4718         struct nfs4_exception exception = { };
4719         int err;
4720         do {
4721                 err = nfs4_handle_exception(NFS_SERVER(dir),
4722                                 _nfs4_proc_secinfo(dir, name, flavors),
4723                                 &exception);
4724         } while (exception.retry);
4725         return err;
4726 }
4727
4728 #ifdef CONFIG_NFS_V4_1
4729 /*
4730  * Check the exchange flags returned by the server for invalid flags, having
4731  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4732  * DS flags set.
4733  */
4734 static int nfs4_check_cl_exchange_flags(u32 flags)
4735 {
4736         if (flags & ~EXCHGID4_FLAG_MASK_R)
4737                 goto out_inval;
4738         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4739             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4740                 goto out_inval;
4741         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4742                 goto out_inval;
4743         return NFS_OK;
4744 out_inval:
4745         return -NFS4ERR_INVAL;
4746 }
4747
4748 /*
4749  * nfs4_proc_exchange_id()
4750  *
4751  * Since the clientid has expired, all compounds using sessions
4752  * associated with the stale clientid will be returning
4753  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4754  * be in some phase of session reset.
4755  */
4756 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4757 {
4758         nfs4_verifier verifier;
4759         struct nfs41_exchange_id_args args = {
4760                 .client = clp,
4761                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4762         };
4763         struct nfs41_exchange_id_res res = {
4764                 .client = clp,
4765         };
4766         int status;
4767         struct rpc_message msg = {
4768                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4769                 .rpc_argp = &args,
4770                 .rpc_resp = &res,
4771                 .rpc_cred = cred,
4772         };
4773         __be32 *p;
4774
4775         dprintk("--> %s\n", __func__);
4776         BUG_ON(clp == NULL);
4777
4778         p = (u32 *)verifier.data;
4779         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4780         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4781         args.verifier = &verifier;
4782
4783         args.id_len = scnprintf(args.id, sizeof(args.id),
4784                                 "%s/%s.%s/%u",
4785                                 clp->cl_ipaddr,
4786                                 init_utsname()->nodename,
4787                                 init_utsname()->domainname,
4788                                 clp->cl_rpcclient->cl_auth->au_flavor);
4789
4790         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
4791         if (!status)
4792                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4793         dprintk("<-- %s status= %d\n", __func__, status);
4794         return status;
4795 }
4796
4797 struct nfs4_get_lease_time_data {
4798         struct nfs4_get_lease_time_args *args;
4799         struct nfs4_get_lease_time_res *res;
4800         struct nfs_client *clp;
4801 };
4802
4803 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4804                                         void *calldata)
4805 {
4806         int ret;
4807         struct nfs4_get_lease_time_data *data =
4808                         (struct nfs4_get_lease_time_data *)calldata;
4809
4810         dprintk("--> %s\n", __func__);
4811         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4812         /* just setup sequence, do not trigger session recovery
4813            since we're invoked within one */
4814         ret = nfs41_setup_sequence(data->clp->cl_session,
4815                                    &data->args->la_seq_args,
4816                                    &data->res->lr_seq_res, 0, task);
4817
4818         BUG_ON(ret == -EAGAIN);
4819         rpc_call_start(task);
4820         dprintk("<-- %s\n", __func__);
4821 }
4822
4823 /*
4824  * Called from nfs4_state_manager thread for session setup, so don't recover
4825  * from sequence operation or clientid errors.
4826  */
4827 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4828 {
4829         struct nfs4_get_lease_time_data *data =
4830                         (struct nfs4_get_lease_time_data *)calldata;
4831
4832         dprintk("--> %s\n", __func__);
4833         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
4834                 return;
4835         switch (task->tk_status) {
4836         case -NFS4ERR_DELAY:
4837         case -NFS4ERR_GRACE:
4838                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4839                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4840                 task->tk_status = 0;
4841                 nfs_restart_rpc(task, data->clp);
4842                 return;
4843         }
4844         dprintk("<-- %s\n", __func__);
4845 }
4846
4847 struct rpc_call_ops nfs4_get_lease_time_ops = {
4848         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4849         .rpc_call_done = nfs4_get_lease_time_done,
4850 };
4851
4852 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4853 {
4854         struct rpc_task *task;
4855         struct nfs4_get_lease_time_args args;
4856         struct nfs4_get_lease_time_res res = {
4857                 .lr_fsinfo = fsinfo,
4858         };
4859         struct nfs4_get_lease_time_data data = {
4860                 .args = &args,
4861                 .res = &res,
4862                 .clp = clp,
4863         };
4864         struct rpc_message msg = {
4865                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4866                 .rpc_argp = &args,
4867                 .rpc_resp = &res,
4868         };
4869         struct rpc_task_setup task_setup = {
4870                 .rpc_client = clp->cl_rpcclient,
4871                 .rpc_message = &msg,
4872                 .callback_ops = &nfs4_get_lease_time_ops,
4873                 .callback_data = &data
4874         };
4875         int status;
4876
4877         dprintk("--> %s\n", __func__);
4878         task = rpc_run_task(&task_setup);
4879
4880         if (IS_ERR(task))
4881                 status = PTR_ERR(task);
4882         else {
4883                 status = task->tk_status;
4884                 rpc_put_task(task);
4885         }
4886         dprintk("<-- %s return %d\n", __func__, status);
4887
4888         return status;
4889 }
4890
4891 /*
4892  * Reset a slot table
4893  */
4894 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
4895                                  int ivalue)
4896 {
4897         struct nfs4_slot *new = NULL;
4898         int i;
4899         int ret = 0;
4900
4901         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
4902                 max_reqs, tbl->max_slots);
4903
4904         /* Does the newly negotiated max_reqs match the existing slot table? */
4905         if (max_reqs != tbl->max_slots) {
4906                 ret = -ENOMEM;
4907                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
4908                               GFP_NOFS);
4909                 if (!new)
4910                         goto out;
4911                 ret = 0;
4912                 kfree(tbl->slots);
4913         }
4914         spin_lock(&tbl->slot_tbl_lock);
4915         if (new) {
4916                 tbl->slots = new;
4917                 tbl->max_slots = max_reqs;
4918         }
4919         for (i = 0; i < tbl->max_slots; ++i)
4920                 tbl->slots[i].seq_nr = ivalue;
4921         spin_unlock(&tbl->slot_tbl_lock);
4922         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4923                 tbl, tbl->slots, tbl->max_slots);
4924 out:
4925         dprintk("<-- %s: return %d\n", __func__, ret);
4926         return ret;
4927 }
4928
4929 /*
4930  * Reset the forechannel and backchannel slot tables
4931  */
4932 static int nfs4_reset_slot_tables(struct nfs4_session *session)
4933 {
4934         int status;
4935
4936         status = nfs4_reset_slot_table(&session->fc_slot_table,
4937                         session->fc_attrs.max_reqs, 1);
4938         if (status)
4939                 return status;
4940
4941         status = nfs4_reset_slot_table(&session->bc_slot_table,
4942                         session->bc_attrs.max_reqs, 0);
4943         return status;
4944 }
4945
4946 /* Destroy the slot table */
4947 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
4948 {
4949         if (session->fc_slot_table.slots != NULL) {
4950                 kfree(session->fc_slot_table.slots);
4951                 session->fc_slot_table.slots = NULL;
4952         }
4953         if (session->bc_slot_table.slots != NULL) {
4954                 kfree(session->bc_slot_table.slots);
4955                 session->bc_slot_table.slots = NULL;
4956         }
4957         return;
4958 }
4959
4960 /*
4961  * Initialize slot table
4962  */
4963 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
4964                 int max_slots, int ivalue)
4965 {
4966         struct nfs4_slot *slot;
4967         int ret = -ENOMEM;
4968
4969         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
4970
4971         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
4972
4973         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
4974         if (!slot)
4975                 goto out;
4976         ret = 0;
4977
4978         spin_lock(&tbl->slot_tbl_lock);
4979         tbl->max_slots = max_slots;
4980         tbl->slots = slot;
4981         tbl->highest_used_slotid = -1;  /* no slot is currently used */
4982         spin_unlock(&tbl->slot_tbl_lock);
4983         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4984                 tbl, tbl->slots, tbl->max_slots);
4985 out:
4986         dprintk("<-- %s: return %d\n", __func__, ret);
4987         return ret;
4988 }
4989
4990 /*
4991  * Initialize the forechannel and backchannel tables
4992  */
4993 static int nfs4_init_slot_tables(struct nfs4_session *session)
4994 {
4995         struct nfs4_slot_table *tbl;
4996         int status = 0;
4997
4998         tbl = &session->fc_slot_table;
4999         if (tbl->slots == NULL) {
5000                 status = nfs4_init_slot_table(tbl,
5001                                 session->fc_attrs.max_reqs, 1);
5002                 if (status)
5003                         return status;
5004         }
5005
5006         tbl = &session->bc_slot_table;
5007         if (tbl->slots == NULL) {
5008                 status = nfs4_init_slot_table(tbl,
5009                                 session->bc_attrs.max_reqs, 0);
5010                 if (status)
5011                         nfs4_destroy_slot_tables(session);
5012         }
5013
5014         return status;
5015 }
5016
5017 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5018 {
5019         struct nfs4_session *session;
5020         struct nfs4_slot_table *tbl;
5021
5022         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5023         if (!session)
5024                 return NULL;
5025
5026         tbl = &session->fc_slot_table;
5027         tbl->highest_used_slotid = -1;
5028         spin_lock_init(&tbl->slot_tbl_lock);
5029         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5030         init_completion(&tbl->complete);
5031
5032         tbl = &session->bc_slot_table;
5033         tbl->highest_used_slotid = -1;
5034         spin_lock_init(&tbl->slot_tbl_lock);
5035         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5036         init_completion(&tbl->complete);
5037
5038         session->session_state = 1<<NFS4_SESSION_INITING;
5039
5040         session->clp = clp;
5041         return session;
5042 }
5043
5044 void nfs4_destroy_session(struct nfs4_session *session)
5045 {
5046         nfs4_proc_destroy_session(session);
5047         dprintk("%s Destroy backchannel for xprt %p\n",
5048                 __func__, session->clp->cl_rpcclient->cl_xprt);
5049         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
5050                                 NFS41_BC_MIN_CALLBACKS);
5051         nfs4_destroy_slot_tables(session);
5052         kfree(session);
5053 }
5054
5055 /*
5056  * Initialize the values to be used by the client in CREATE_SESSION
5057  * If nfs4_init_session set the fore channel request and response sizes,
5058  * use them.
5059  *
5060  * Set the back channel max_resp_sz_cached to zero to force the client to
5061  * always set csa_cachethis to FALSE because the current implementation
5062  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5063  */
5064 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5065 {
5066         struct nfs4_session *session = args->client->cl_session;
5067         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5068                      mxresp_sz = session->fc_attrs.max_resp_sz;
5069
5070         if (mxrqst_sz == 0)
5071                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5072         if (mxresp_sz == 0)
5073                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5074         /* Fore channel attributes */
5075         args->fc_attrs.headerpadsz = 0;
5076         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5077         args->fc_attrs.max_resp_sz = mxresp_sz;
5078         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5079         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
5080
5081         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5082                 "max_ops=%u max_reqs=%u\n",
5083                 __func__,
5084                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5085                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5086
5087         /* Back channel attributes */
5088         args->bc_attrs.headerpadsz = 0;
5089         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5090         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5091         args->bc_attrs.max_resp_sz_cached = 0;
5092         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5093         args->bc_attrs.max_reqs = 1;
5094
5095         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5096                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5097                 __func__,
5098                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5099                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5100                 args->bc_attrs.max_reqs);
5101 }
5102
5103 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5104 {
5105         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5106         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5107
5108         if (rcvd->headerpadsz > sent->headerpadsz)
5109                 return -EINVAL;
5110         if (rcvd->max_resp_sz > sent->max_resp_sz)
5111                 return -EINVAL;
5112         /*
5113          * Our requested max_ops is the minimum we need; we're not
5114          * prepared to break up compounds into smaller pieces than that.
5115          * So, no point even trying to continue if the server won't
5116          * cooperate:
5117          */
5118         if (rcvd->max_ops < sent->max_ops)
5119                 return -EINVAL;
5120         if (rcvd->max_reqs == 0)
5121                 return -EINVAL;
5122         return 0;
5123 }
5124
5125 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5126 {
5127         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5128         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5129
5130         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5131                 return -EINVAL;
5132         if (rcvd->max_resp_sz < sent->max_resp_sz)
5133                 return -EINVAL;
5134         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5135                 return -EINVAL;
5136         /* These would render the backchannel useless: */
5137         if (rcvd->max_ops  == 0)
5138                 return -EINVAL;
5139         if (rcvd->max_reqs == 0)
5140                 return -EINVAL;
5141         return 0;
5142 }
5143
5144 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5145                                      struct nfs4_session *session)
5146 {
5147         int ret;
5148
5149         ret = nfs4_verify_fore_channel_attrs(args, session);
5150         if (ret)
5151                 return ret;
5152         return nfs4_verify_back_channel_attrs(args, session);
5153 }
5154
5155 static int _nfs4_proc_create_session(struct nfs_client *clp)
5156 {
5157         struct nfs4_session *session = clp->cl_session;
5158         struct nfs41_create_session_args args = {
5159                 .client = clp,
5160                 .cb_program = NFS4_CALLBACK,
5161         };
5162         struct nfs41_create_session_res res = {
5163                 .client = clp,
5164         };
5165         struct rpc_message msg = {
5166                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5167                 .rpc_argp = &args,
5168                 .rpc_resp = &res,
5169         };
5170         int status;
5171
5172         nfs4_init_channel_attrs(&args);
5173         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5174
5175         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5176
5177         if (!status)
5178                 /* Verify the session's negotiated channel_attrs values */
5179                 status = nfs4_verify_channel_attrs(&args, session);
5180         if (!status) {
5181                 /* Increment the clientid slot sequence id */
5182                 clp->cl_seqid++;
5183         }
5184
5185         return status;
5186 }
5187
5188 /*
5189  * Issues a CREATE_SESSION operation to the server.
5190  * It is the responsibility of the caller to verify the session is
5191  * expired before calling this routine.
5192  */
5193 int nfs4_proc_create_session(struct nfs_client *clp)
5194 {
5195         int status;
5196         unsigned *ptr;
5197         struct nfs4_session *session = clp->cl_session;
5198         long timeout = 0;
5199         int err;
5200
5201         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5202
5203         do {
5204                 status = _nfs4_proc_create_session(clp);
5205                 if (status == -NFS4ERR_DELAY) {
5206                         err = nfs4_delay(clp->cl_rpcclient, &timeout);
5207                         if (err)
5208                                 status = err;
5209                 }
5210         } while (status == -NFS4ERR_DELAY);
5211
5212         if (status)
5213                 goto out;
5214
5215         /* Init and reset the fore channel */
5216         status = nfs4_init_slot_tables(session);
5217         dprintk("slot table initialization returned %d\n", status);
5218         if (status)
5219                 goto out;
5220         status = nfs4_reset_slot_tables(session);
5221         dprintk("slot table reset returned %d\n", status);
5222         if (status)
5223                 goto out;
5224
5225         ptr = (unsigned *)&session->sess_id.data[0];
5226         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5227                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5228 out:
5229         dprintk("<-- %s\n", __func__);
5230         return status;
5231 }
5232
5233 /*
5234  * Issue the over-the-wire RPC DESTROY_SESSION.
5235  * The caller must serialize access to this routine.
5236  */
5237 int nfs4_proc_destroy_session(struct nfs4_session *session)
5238 {
5239         int status = 0;
5240         struct rpc_message msg;
5241
5242         dprintk("--> nfs4_proc_destroy_session\n");
5243
5244         /* session is still being setup */
5245         if (session->clp->cl_cons_state != NFS_CS_READY)
5246                 return status;
5247
5248         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5249         msg.rpc_argp = session;
5250         msg.rpc_resp = NULL;
5251         msg.rpc_cred = NULL;
5252         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5253
5254         if (status)
5255                 printk(KERN_WARNING
5256                         "Got error %d from the server on DESTROY_SESSION. "
5257                         "Session has been destroyed regardless...\n", status);
5258
5259         dprintk("<-- nfs4_proc_destroy_session\n");
5260         return status;
5261 }
5262
5263 int nfs4_init_session(struct nfs_server *server)
5264 {
5265         struct nfs_client *clp = server->nfs_client;
5266         struct nfs4_session *session;
5267         unsigned int rsize, wsize;
5268         int ret;
5269
5270         if (!nfs4_has_session(clp))
5271                 return 0;
5272
5273         session = clp->cl_session;
5274         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5275                 return 0;
5276
5277         rsize = server->rsize;
5278         if (rsize == 0)
5279                 rsize = NFS_MAX_FILE_IO_SIZE;
5280         wsize = server->wsize;
5281         if (wsize == 0)
5282                 wsize = NFS_MAX_FILE_IO_SIZE;
5283
5284         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5285         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5286
5287         ret = nfs4_recover_expired_lease(server);
5288         if (!ret)
5289                 ret = nfs4_check_client_ready(clp);
5290         return ret;
5291 }
5292
5293 int nfs4_init_ds_session(struct nfs_client *clp)
5294 {
5295         struct nfs4_session *session = clp->cl_session;
5296         int ret;
5297
5298         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5299                 return 0;
5300
5301         ret = nfs4_client_recover_expired_lease(clp);
5302         if (!ret)
5303                 /* Test for the DS role */
5304                 if (!is_ds_client(clp))
5305                         ret = -ENODEV;
5306         if (!ret)
5307                 ret = nfs4_check_client_ready(clp);
5308         return ret;
5309
5310 }
5311 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5312
5313
5314 /*
5315  * Renew the cl_session lease.
5316  */
5317 struct nfs4_sequence_data {
5318         struct nfs_client *clp;
5319         struct nfs4_sequence_args args;
5320         struct nfs4_sequence_res res;
5321 };
5322
5323 static void nfs41_sequence_release(void *data)
5324 {
5325         struct nfs4_sequence_data *calldata = data;
5326         struct nfs_client *clp = calldata->clp;
5327
5328         if (atomic_read(&clp->cl_count) > 1)
5329                 nfs4_schedule_state_renewal(clp);
5330         nfs_put_client(clp);
5331         kfree(calldata);
5332 }
5333
5334 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5335 {
5336         switch(task->tk_status) {
5337         case -NFS4ERR_DELAY:
5338                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5339                 return -EAGAIN;
5340         default:
5341                 nfs4_schedule_lease_recovery(clp);
5342         }
5343         return 0;
5344 }
5345
5346 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5347 {
5348         struct nfs4_sequence_data *calldata = data;
5349         struct nfs_client *clp = calldata->clp;
5350
5351         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5352                 return;
5353
5354         if (task->tk_status < 0) {
5355                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5356                 if (atomic_read(&clp->cl_count) == 1)
5357                         goto out;
5358
5359                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5360                         rpc_restart_call_prepare(task);
5361                         return;
5362                 }
5363         }
5364         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5365 out:
5366         dprintk("<-- %s\n", __func__);
5367 }
5368
5369 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5370 {
5371         struct nfs4_sequence_data *calldata = data;
5372         struct nfs_client *clp = calldata->clp;
5373         struct nfs4_sequence_args *args;
5374         struct nfs4_sequence_res *res;
5375
5376         args = task->tk_msg.rpc_argp;
5377         res = task->tk_msg.rpc_resp;
5378
5379         if (nfs41_setup_sequence(clp->cl_session, args, res, 0, task))
5380                 return;
5381         rpc_call_start(task);
5382 }
5383
5384 static const struct rpc_call_ops nfs41_sequence_ops = {
5385         .rpc_call_done = nfs41_sequence_call_done,
5386         .rpc_call_prepare = nfs41_sequence_prepare,
5387         .rpc_release = nfs41_sequence_release,
5388 };
5389
5390 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5391 {
5392         struct nfs4_sequence_data *calldata;
5393         struct rpc_message msg = {
5394                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5395                 .rpc_cred = cred,
5396         };
5397         struct rpc_task_setup task_setup_data = {
5398                 .rpc_client = clp->cl_rpcclient,
5399                 .rpc_message = &msg,
5400                 .callback_ops = &nfs41_sequence_ops,
5401                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5402         };
5403
5404         if (!atomic_inc_not_zero(&clp->cl_count))
5405                 return ERR_PTR(-EIO);
5406         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5407         if (calldata == NULL) {
5408                 nfs_put_client(clp);
5409                 return ERR_PTR(-ENOMEM);
5410         }
5411         msg.rpc_argp = &calldata->args;
5412         msg.rpc_resp = &calldata->res;
5413         calldata->clp = clp;
5414         task_setup_data.callback_data = calldata;
5415
5416         return rpc_run_task(&task_setup_data);
5417 }
5418
5419 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5420 {
5421         struct rpc_task *task;
5422         int ret = 0;
5423
5424         task = _nfs41_proc_sequence(clp, cred);
5425         if (IS_ERR(task))
5426                 ret = PTR_ERR(task);
5427         else
5428                 rpc_put_task_async(task);
5429         dprintk("<-- %s status=%d\n", __func__, ret);
5430         return ret;
5431 }
5432
5433 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5434 {
5435         struct rpc_task *task;
5436         int ret;
5437
5438         task = _nfs41_proc_sequence(clp, cred);
5439         if (IS_ERR(task)) {
5440                 ret = PTR_ERR(task);
5441                 goto out;
5442         }
5443         ret = rpc_wait_for_completion_task(task);
5444         if (!ret) {
5445                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5446
5447                 if (task->tk_status == 0)
5448                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5449                 ret = task->tk_status;
5450         }
5451         rpc_put_task(task);
5452 out:
5453         dprintk("<-- %s status=%d\n", __func__, ret);
5454         return ret;
5455 }
5456
5457 struct nfs4_reclaim_complete_data {
5458         struct nfs_client *clp;
5459         struct nfs41_reclaim_complete_args arg;
5460         struct nfs41_reclaim_complete_res res;
5461 };
5462
5463 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5464 {
5465         struct nfs4_reclaim_complete_data *calldata = data;
5466
5467         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5468         if (nfs41_setup_sequence(calldata->clp->cl_session,
5469                                 &calldata->arg.seq_args,
5470                                 &calldata->res.seq_res, 0, task))
5471                 return;
5472
5473         rpc_call_start(task);
5474 }
5475
5476 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5477 {
5478         switch(task->tk_status) {
5479         case 0:
5480         case -NFS4ERR_COMPLETE_ALREADY:
5481         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5482                 break;
5483         case -NFS4ERR_DELAY:
5484                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5485                 return -EAGAIN;
5486         default:
5487                 nfs4_schedule_lease_recovery(clp);
5488         }
5489         return 0;
5490 }
5491
5492 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5493 {
5494         struct nfs4_reclaim_complete_data *calldata = data;
5495         struct nfs_client *clp = calldata->clp;
5496         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5497
5498         dprintk("--> %s\n", __func__);
5499         if (!nfs41_sequence_done(task, res))
5500                 return;
5501
5502         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5503                 rpc_restart_call_prepare(task);
5504                 return;
5505         }
5506         dprintk("<-- %s\n", __func__);
5507 }
5508
5509 static void nfs4_free_reclaim_complete_data(void *data)
5510 {
5511         struct nfs4_reclaim_complete_data *calldata = data;
5512
5513         kfree(calldata);
5514 }
5515
5516 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5517         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5518         .rpc_call_done = nfs4_reclaim_complete_done,
5519         .rpc_release = nfs4_free_reclaim_complete_data,
5520 };
5521
5522 /*
5523  * Issue a global reclaim complete.
5524  */
5525 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5526 {
5527         struct nfs4_reclaim_complete_data *calldata;
5528         struct rpc_task *task;
5529         struct rpc_message msg = {
5530                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5531         };
5532         struct rpc_task_setup task_setup_data = {
5533                 .rpc_client = clp->cl_rpcclient,
5534                 .rpc_message = &msg,
5535                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5536                 .flags = RPC_TASK_ASYNC,
5537         };
5538         int status = -ENOMEM;
5539
5540         dprintk("--> %s\n", __func__);
5541         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5542         if (calldata == NULL)
5543                 goto out;
5544         calldata->clp = clp;
5545         calldata->arg.one_fs = 0;
5546
5547         msg.rpc_argp = &calldata->arg;
5548         msg.rpc_resp = &calldata->res;
5549         task_setup_data.callback_data = calldata;
5550         task = rpc_run_task(&task_setup_data);
5551         if (IS_ERR(task)) {
5552                 status = PTR_ERR(task);
5553                 goto out;
5554         }
5555         status = nfs4_wait_for_completion_rpc_task(task);
5556         if (status == 0)
5557                 status = task->tk_status;
5558         rpc_put_task(task);
5559         return 0;
5560 out:
5561         dprintk("<-- %s status=%d\n", __func__, status);
5562         return status;
5563 }
5564
5565 static void
5566 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5567 {
5568         struct nfs4_layoutget *lgp = calldata;
5569         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5570
5571         dprintk("--> %s\n", __func__);
5572         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5573          * right now covering the LAYOUTGET we are about to send.
5574          * However, that is not so catastrophic, and there seems
5575          * to be no way to prevent it completely.
5576          */
5577         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5578                                 &lgp->res.seq_res, 0, task))
5579                 return;
5580         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5581                                           NFS_I(lgp->args.inode)->layout,
5582                                           lgp->args.ctx->state)) {
5583                 rpc_exit(task, NFS4_OK);
5584                 return;
5585         }
5586         rpc_call_start(task);
5587 }
5588
5589 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5590 {
5591         struct nfs4_layoutget *lgp = calldata;
5592         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5593
5594         dprintk("--> %s\n", __func__);
5595
5596         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5597                 return;
5598
5599         switch (task->tk_status) {
5600         case 0:
5601                 break;
5602         case -NFS4ERR_LAYOUTTRYLATER:
5603         case -NFS4ERR_RECALLCONFLICT:
5604                 task->tk_status = -NFS4ERR_DELAY;
5605                 /* Fall through */
5606         default:
5607                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5608                         rpc_restart_call_prepare(task);
5609                         return;
5610                 }
5611         }
5612         dprintk("<-- %s\n", __func__);
5613 }
5614
5615 static void nfs4_layoutget_release(void *calldata)
5616 {
5617         struct nfs4_layoutget *lgp = calldata;
5618
5619         dprintk("--> %s\n", __func__);
5620         put_nfs_open_context(lgp->args.ctx);
5621         kfree(calldata);
5622         dprintk("<-- %s\n", __func__);
5623 }
5624
5625 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5626         .rpc_call_prepare = nfs4_layoutget_prepare,
5627         .rpc_call_done = nfs4_layoutget_done,
5628         .rpc_release = nfs4_layoutget_release,
5629 };
5630
5631 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5632 {
5633         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5634         struct rpc_task *task;
5635         struct rpc_message msg = {
5636                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5637                 .rpc_argp = &lgp->args,
5638                 .rpc_resp = &lgp->res,
5639         };
5640         struct rpc_task_setup task_setup_data = {
5641                 .rpc_client = server->client,
5642                 .rpc_message = &msg,
5643                 .callback_ops = &nfs4_layoutget_call_ops,
5644                 .callback_data = lgp,
5645                 .flags = RPC_TASK_ASYNC,
5646         };
5647         int status = 0;
5648
5649         dprintk("--> %s\n", __func__);
5650
5651         lgp->res.layoutp = &lgp->args.layout;
5652         lgp->res.seq_res.sr_slot = NULL;
5653         task = rpc_run_task(&task_setup_data);
5654         if (IS_ERR(task))
5655                 return PTR_ERR(task);
5656         status = nfs4_wait_for_completion_rpc_task(task);
5657         if (status == 0)
5658                 status = task->tk_status;
5659         if (status == 0)
5660                 status = pnfs_layout_process(lgp);
5661         rpc_put_task(task);
5662         dprintk("<-- %s status=%d\n", __func__, status);
5663         return status;
5664 }
5665
5666 static int
5667 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5668 {
5669         struct nfs4_getdeviceinfo_args args = {
5670                 .pdev = pdev,
5671         };
5672         struct nfs4_getdeviceinfo_res res = {
5673                 .pdev = pdev,
5674         };
5675         struct rpc_message msg = {
5676                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5677                 .rpc_argp = &args,
5678                 .rpc_resp = &res,
5679         };
5680         int status;
5681
5682         dprintk("--> %s\n", __func__);
5683         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5684         dprintk("<-- %s status=%d\n", __func__, status);
5685
5686         return status;
5687 }
5688
5689 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5690 {
5691         struct nfs4_exception exception = { };
5692         int err;
5693
5694         do {
5695                 err = nfs4_handle_exception(server,
5696                                         _nfs4_proc_getdeviceinfo(server, pdev),
5697                                         &exception);
5698         } while (exception.retry);
5699         return err;
5700 }
5701 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5702
5703 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
5704 {
5705         struct nfs4_layoutcommit_data *data = calldata;
5706         struct nfs_server *server = NFS_SERVER(data->args.inode);
5707
5708         if (nfs4_setup_sequence(server, &data->args.seq_args,
5709                                 &data->res.seq_res, 1, task))
5710                 return;
5711         rpc_call_start(task);
5712 }
5713
5714 static void
5715 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
5716 {
5717         struct nfs4_layoutcommit_data *data = calldata;
5718         struct nfs_server *server = NFS_SERVER(data->args.inode);
5719
5720         if (!nfs4_sequence_done(task, &data->res.seq_res))
5721                 return;
5722
5723         switch (task->tk_status) { /* Just ignore these failures */
5724         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
5725         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
5726         case NFS4ERR_BADLAYOUT:     /* no layout */
5727         case NFS4ERR_GRACE:         /* loca_recalim always false */
5728                 task->tk_status = 0;
5729         }
5730
5731         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5732                 nfs_restart_rpc(task, server->nfs_client);
5733                 return;
5734         }
5735
5736         if (task->tk_status == 0)
5737                 nfs_post_op_update_inode_force_wcc(data->args.inode,
5738                                                    data->res.fattr);
5739 }
5740
5741 static void nfs4_layoutcommit_release(void *calldata)
5742 {
5743         struct nfs4_layoutcommit_data *data = calldata;
5744
5745         /* Matched by references in pnfs_set_layoutcommit */
5746         put_lseg(data->lseg);
5747         put_rpccred(data->cred);
5748         kfree(data);
5749 }
5750
5751 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
5752         .rpc_call_prepare = nfs4_layoutcommit_prepare,
5753         .rpc_call_done = nfs4_layoutcommit_done,
5754         .rpc_release = nfs4_layoutcommit_release,
5755 };
5756
5757 int
5758 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
5759 {
5760         struct rpc_message msg = {
5761                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
5762                 .rpc_argp = &data->args,
5763                 .rpc_resp = &data->res,
5764                 .rpc_cred = data->cred,
5765         };
5766         struct rpc_task_setup task_setup_data = {
5767                 .task = &data->task,
5768                 .rpc_client = NFS_CLIENT(data->args.inode),
5769                 .rpc_message = &msg,
5770                 .callback_ops = &nfs4_layoutcommit_ops,
5771                 .callback_data = data,
5772                 .flags = RPC_TASK_ASYNC,
5773         };
5774         struct rpc_task *task;
5775         int status = 0;
5776
5777         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
5778                 "lbw: %llu inode %lu\n",
5779                 data->task.tk_pid, sync,
5780                 data->args.lastbytewritten,
5781                 data->args.inode->i_ino);
5782
5783         task = rpc_run_task(&task_setup_data);
5784         if (IS_ERR(task))
5785                 return PTR_ERR(task);
5786         if (sync == false)
5787                 goto out;
5788         status = nfs4_wait_for_completion_rpc_task(task);
5789         if (status != 0)
5790                 goto out;
5791         status = task->tk_status;
5792 out:
5793         dprintk("%s: status %d\n", __func__, status);
5794         rpc_put_task(task);
5795         return status;
5796 }
5797 #endif /* CONFIG_NFS_V4_1 */
5798
5799 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
5800         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5801         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5802         .recover_open   = nfs4_open_reclaim,
5803         .recover_lock   = nfs4_lock_reclaim,
5804         .establish_clid = nfs4_init_clientid,
5805         .get_clid_cred  = nfs4_get_setclientid_cred,
5806 };
5807
5808 #if defined(CONFIG_NFS_V4_1)
5809 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
5810         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5811         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5812         .recover_open   = nfs4_open_reclaim,
5813         .recover_lock   = nfs4_lock_reclaim,
5814         .establish_clid = nfs41_init_clientid,
5815         .get_clid_cred  = nfs4_get_exchange_id_cred,
5816         .reclaim_complete = nfs41_proc_reclaim_complete,
5817 };
5818 #endif /* CONFIG_NFS_V4_1 */
5819
5820 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
5821         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5822         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5823         .recover_open   = nfs4_open_expired,
5824         .recover_lock   = nfs4_lock_expired,
5825         .establish_clid = nfs4_init_clientid,
5826         .get_clid_cred  = nfs4_get_setclientid_cred,
5827 };
5828
5829 #if defined(CONFIG_NFS_V4_1)
5830 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
5831         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5832         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5833         .recover_open   = nfs4_open_expired,
5834         .recover_lock   = nfs4_lock_expired,
5835         .establish_clid = nfs41_init_clientid,
5836         .get_clid_cred  = nfs4_get_exchange_id_cred,
5837 };
5838 #endif /* CONFIG_NFS_V4_1 */
5839
5840 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
5841         .sched_state_renewal = nfs4_proc_async_renew,
5842         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
5843         .renew_lease = nfs4_proc_renew,
5844 };
5845
5846 #if defined(CONFIG_NFS_V4_1)
5847 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
5848         .sched_state_renewal = nfs41_proc_async_sequence,
5849         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
5850         .renew_lease = nfs4_proc_sequence,
5851 };
5852 #endif
5853
5854 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
5855         .minor_version = 0,
5856         .call_sync = _nfs4_call_sync,
5857         .validate_stateid = nfs4_validate_delegation_stateid,
5858         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
5859         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
5860         .state_renewal_ops = &nfs40_state_renewal_ops,
5861 };
5862
5863 #if defined(CONFIG_NFS_V4_1)
5864 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
5865         .minor_version = 1,
5866         .call_sync = _nfs4_call_sync_session,
5867         .validate_stateid = nfs41_validate_delegation_stateid,
5868         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
5869         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
5870         .state_renewal_ops = &nfs41_state_renewal_ops,
5871 };
5872 #endif
5873
5874 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
5875         [0] = &nfs_v4_0_minor_ops,
5876 #if defined(CONFIG_NFS_V4_1)
5877         [1] = &nfs_v4_1_minor_ops,
5878 #endif
5879 };
5880
5881 static const struct inode_operations nfs4_file_inode_operations = {
5882         .permission     = nfs_permission,
5883         .getattr        = nfs_getattr,
5884         .setattr        = nfs_setattr,
5885         .getxattr       = generic_getxattr,
5886         .setxattr       = generic_setxattr,
5887         .listxattr      = generic_listxattr,
5888         .removexattr    = generic_removexattr,
5889 };
5890
5891 const struct nfs_rpc_ops nfs_v4_clientops = {
5892         .version        = 4,                    /* protocol version */
5893         .dentry_ops     = &nfs4_dentry_operations,
5894         .dir_inode_ops  = &nfs4_dir_inode_operations,
5895         .file_inode_ops = &nfs4_file_inode_operations,
5896         .getroot        = nfs4_proc_get_root,
5897         .getattr        = nfs4_proc_getattr,
5898         .setattr        = nfs4_proc_setattr,
5899         .lookupfh       = nfs4_proc_lookupfh,
5900         .lookup         = nfs4_proc_lookup,
5901         .access         = nfs4_proc_access,
5902         .readlink       = nfs4_proc_readlink,
5903         .create         = nfs4_proc_create,
5904         .remove         = nfs4_proc_remove,
5905         .unlink_setup   = nfs4_proc_unlink_setup,
5906         .unlink_done    = nfs4_proc_unlink_done,
5907         .rename         = nfs4_proc_rename,
5908         .rename_setup   = nfs4_proc_rename_setup,
5909         .rename_done    = nfs4_proc_rename_done,
5910         .link           = nfs4_proc_link,
5911         .symlink        = nfs4_proc_symlink,
5912         .mkdir          = nfs4_proc_mkdir,
5913         .rmdir          = nfs4_proc_remove,
5914         .readdir        = nfs4_proc_readdir,
5915         .mknod          = nfs4_proc_mknod,
5916         .statfs         = nfs4_proc_statfs,
5917         .fsinfo         = nfs4_proc_fsinfo,
5918         .pathconf       = nfs4_proc_pathconf,
5919         .set_capabilities = nfs4_server_capabilities,
5920         .decode_dirent  = nfs4_decode_dirent,
5921         .read_setup     = nfs4_proc_read_setup,
5922         .read_done      = nfs4_read_done,
5923         .write_setup    = nfs4_proc_write_setup,
5924         .write_done     = nfs4_write_done,
5925         .commit_setup   = nfs4_proc_commit_setup,
5926         .commit_done    = nfs4_commit_done,
5927         .lock           = nfs4_proc_lock,
5928         .clear_acl_cache = nfs4_zap_acl_attr,
5929         .close_context  = nfs4_close_context,
5930         .open_context   = nfs4_atomic_open,
5931         .init_client    = nfs4_init_client,
5932         .secinfo        = nfs4_proc_secinfo,
5933 };
5934
5935 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
5936         .prefix = XATTR_NAME_NFSV4_ACL,
5937         .list   = nfs4_xattr_list_nfs4_acl,
5938         .get    = nfs4_xattr_get_nfs4_acl,
5939         .set    = nfs4_xattr_set_nfs4_acl,
5940 };
5941
5942 const struct xattr_handler *nfs4_xattr_handlers[] = {
5943         &nfs4_xattr_nfs4_acl_handler,
5944         NULL
5945 };
5946
5947 /*
5948  * Local variables:
5949  *  c-basic-offset: 8
5950  * End:
5951  */