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         if (ret < 0)
2208                 ret = -EAGAIN;
2209 out:
2210         return ret;
2211 }
2212
2213 /*
2214  * get the file handle for the "/" directory on the server
2215  */
2216 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2217                               struct nfs_fsinfo *info)
2218 {
2219         int i, len, status = 0;
2220         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS + 2];
2221
2222         flav_array[0] = RPC_AUTH_UNIX;
2223         len = gss_mech_list_pseudoflavors(&flav_array[1]);
2224         flav_array[1+len] = RPC_AUTH_NULL;
2225         len += 2;
2226
2227         for (i = 0; i < len; i++) {
2228                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2229                 if (status == 0)
2230                         break;
2231         }
2232         if (status == 0)
2233                 status = nfs4_server_capabilities(server, fhandle);
2234         if (status == 0)
2235                 status = nfs4_do_fsinfo(server, fhandle, info);
2236         return nfs4_map_errors(status);
2237 }
2238
2239 /*
2240  * Get locations and (maybe) other attributes of a referral.
2241  * Note that we'll actually follow the referral later when
2242  * we detect fsid mismatch in inode revalidation
2243  */
2244 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2245 {
2246         int status = -ENOMEM;
2247         struct page *page = NULL;
2248         struct nfs4_fs_locations *locations = NULL;
2249
2250         page = alloc_page(GFP_KERNEL);
2251         if (page == NULL)
2252                 goto out;
2253         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2254         if (locations == NULL)
2255                 goto out;
2256
2257         status = nfs4_proc_fs_locations(dir, name, locations, page);
2258         if (status != 0)
2259                 goto out;
2260         /* Make sure server returned a different fsid for the referral */
2261         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2262                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2263                 status = -EIO;
2264                 goto out;
2265         }
2266
2267         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2268         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2269         if (!fattr->mode)
2270                 fattr->mode = S_IFDIR;
2271         memset(fhandle, 0, sizeof(struct nfs_fh));
2272 out:
2273         if (page)
2274                 __free_page(page);
2275         kfree(locations);
2276         return status;
2277 }
2278
2279 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2280 {
2281         struct nfs4_getattr_arg args = {
2282                 .fh = fhandle,
2283                 .bitmask = server->attr_bitmask,
2284         };
2285         struct nfs4_getattr_res res = {
2286                 .fattr = fattr,
2287                 .server = server,
2288         };
2289         struct rpc_message msg = {
2290                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2291                 .rpc_argp = &args,
2292                 .rpc_resp = &res,
2293         };
2294
2295         nfs_fattr_init(fattr);
2296         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2297 }
2298
2299 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2300 {
2301         struct nfs4_exception exception = { };
2302         int err;
2303         do {
2304                 err = nfs4_handle_exception(server,
2305                                 _nfs4_proc_getattr(server, fhandle, fattr),
2306                                 &exception);
2307         } while (exception.retry);
2308         return err;
2309 }
2310
2311 /*
2312  * The file is not closed if it is opened due to the a request to change
2313  * the size of the file. The open call will not be needed once the
2314  * VFS layer lookup-intents are implemented.
2315  *
2316  * Close is called when the inode is destroyed.
2317  * If we haven't opened the file for O_WRONLY, we
2318  * need to in the size_change case to obtain a stateid.
2319  *
2320  * Got race?
2321  * Because OPEN is always done by name in nfsv4, it is
2322  * possible that we opened a different file by the same
2323  * name.  We can recognize this race condition, but we
2324  * can't do anything about it besides returning an error.
2325  *
2326  * This will be fixed with VFS changes (lookup-intent).
2327  */
2328 static int
2329 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2330                   struct iattr *sattr)
2331 {
2332         struct inode *inode = dentry->d_inode;
2333         struct rpc_cred *cred = NULL;
2334         struct nfs4_state *state = NULL;
2335         int status;
2336
2337         nfs_fattr_init(fattr);
2338
2339         /* Search for an existing open(O_WRITE) file */
2340         if (sattr->ia_valid & ATTR_FILE) {
2341                 struct nfs_open_context *ctx;
2342
2343                 ctx = nfs_file_open_context(sattr->ia_file);
2344                 if (ctx) {
2345                         cred = ctx->cred;
2346                         state = ctx->state;
2347                 }
2348         }
2349
2350         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2351         if (status == 0)
2352                 nfs_setattr_update_inode(inode, sattr);
2353         return status;
2354 }
2355
2356 static int _nfs4_proc_lookupfh(struct rpc_clnt *clnt, struct nfs_server *server,
2357                 const struct nfs_fh *dirfh, const struct qstr *name,
2358                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2359 {
2360         int                    status;
2361         struct nfs4_lookup_arg args = {
2362                 .bitmask = server->attr_bitmask,
2363                 .dir_fh = dirfh,
2364                 .name = name,
2365         };
2366         struct nfs4_lookup_res res = {
2367                 .server = server,
2368                 .fattr = fattr,
2369                 .fh = fhandle,
2370         };
2371         struct rpc_message msg = {
2372                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2373                 .rpc_argp = &args,
2374                 .rpc_resp = &res,
2375         };
2376
2377         nfs_fattr_init(fattr);
2378
2379         dprintk("NFS call  lookupfh %s\n", name->name);
2380         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2381         dprintk("NFS reply lookupfh: %d\n", status);
2382         return status;
2383 }
2384
2385 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2386                               struct qstr *name, struct nfs_fh *fhandle,
2387                               struct nfs_fattr *fattr)
2388 {
2389         struct nfs4_exception exception = { };
2390         int err;
2391         do {
2392                 err = _nfs4_proc_lookupfh(server->client, server, dirfh, name, fhandle, fattr);
2393                 /* FIXME: !!!! */
2394                 if (err == -NFS4ERR_MOVED) {
2395                         err = -EREMOTE;
2396                         break;
2397                 }
2398                 err = nfs4_handle_exception(server, err, &exception);
2399         } while (exception.retry);
2400         return err;
2401 }
2402
2403 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2404                 const struct qstr *name, struct nfs_fh *fhandle,
2405                 struct nfs_fattr *fattr)
2406 {
2407         int status;
2408
2409         dprintk("NFS call  lookup %s\n", name->name);
2410         status = _nfs4_proc_lookupfh(clnt, NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2411         if (status == -NFS4ERR_MOVED)
2412                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2413         dprintk("NFS reply lookup: %d\n", status);
2414         return status;
2415 }
2416
2417 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2418 {
2419         memset(fh, 0, sizeof(struct nfs_fh));
2420         fattr->fsid.major = 1;
2421         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2422                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2423         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2424         fattr->nlink = 2;
2425 }
2426
2427 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2428                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2429 {
2430         struct nfs4_exception exception = { };
2431         int err;
2432         do {
2433                 err = nfs4_handle_exception(NFS_SERVER(dir),
2434                                 _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr),
2435                                 &exception);
2436                 if (err == -EPERM)
2437                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2438         } while (exception.retry);
2439         return err;
2440 }
2441
2442 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2443 {
2444         struct nfs_server *server = NFS_SERVER(inode);
2445         struct nfs4_accessargs args = {
2446                 .fh = NFS_FH(inode),
2447                 .bitmask = server->attr_bitmask,
2448         };
2449         struct nfs4_accessres res = {
2450                 .server = server,
2451         };
2452         struct rpc_message msg = {
2453                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2454                 .rpc_argp = &args,
2455                 .rpc_resp = &res,
2456                 .rpc_cred = entry->cred,
2457         };
2458         int mode = entry->mask;
2459         int status;
2460
2461         /*
2462          * Determine which access bits we want to ask for...
2463          */
2464         if (mode & MAY_READ)
2465                 args.access |= NFS4_ACCESS_READ;
2466         if (S_ISDIR(inode->i_mode)) {
2467                 if (mode & MAY_WRITE)
2468                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2469                 if (mode & MAY_EXEC)
2470                         args.access |= NFS4_ACCESS_LOOKUP;
2471         } else {
2472                 if (mode & MAY_WRITE)
2473                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2474                 if (mode & MAY_EXEC)
2475                         args.access |= NFS4_ACCESS_EXECUTE;
2476         }
2477
2478         res.fattr = nfs_alloc_fattr();
2479         if (res.fattr == NULL)
2480                 return -ENOMEM;
2481
2482         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2483         if (!status) {
2484                 entry->mask = 0;
2485                 if (res.access & NFS4_ACCESS_READ)
2486                         entry->mask |= MAY_READ;
2487                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2488                         entry->mask |= MAY_WRITE;
2489                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2490                         entry->mask |= MAY_EXEC;
2491                 nfs_refresh_inode(inode, res.fattr);
2492         }
2493         nfs_free_fattr(res.fattr);
2494         return status;
2495 }
2496
2497 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2498 {
2499         struct nfs4_exception exception = { };
2500         int err;
2501         do {
2502                 err = nfs4_handle_exception(NFS_SERVER(inode),
2503                                 _nfs4_proc_access(inode, entry),
2504                                 &exception);
2505         } while (exception.retry);
2506         return err;
2507 }
2508
2509 /*
2510  * TODO: For the time being, we don't try to get any attributes
2511  * along with any of the zero-copy operations READ, READDIR,
2512  * READLINK, WRITE.
2513  *
2514  * In the case of the first three, we want to put the GETATTR
2515  * after the read-type operation -- this is because it is hard
2516  * to predict the length of a GETATTR response in v4, and thus
2517  * align the READ data correctly.  This means that the GETATTR
2518  * may end up partially falling into the page cache, and we should
2519  * shift it into the 'tail' of the xdr_buf before processing.
2520  * To do this efficiently, we need to know the total length
2521  * of data received, which doesn't seem to be available outside
2522  * of the RPC layer.
2523  *
2524  * In the case of WRITE, we also want to put the GETATTR after
2525  * the operation -- in this case because we want to make sure
2526  * we get the post-operation mtime and size.  This means that
2527  * we can't use xdr_encode_pages() as written: we need a variant
2528  * of it which would leave room in the 'tail' iovec.
2529  *
2530  * Both of these changes to the XDR layer would in fact be quite
2531  * minor, but I decided to leave them for a subsequent patch.
2532  */
2533 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2534                 unsigned int pgbase, unsigned int pglen)
2535 {
2536         struct nfs4_readlink args = {
2537                 .fh       = NFS_FH(inode),
2538                 .pgbase   = pgbase,
2539                 .pglen    = pglen,
2540                 .pages    = &page,
2541         };
2542         struct nfs4_readlink_res res;
2543         struct rpc_message msg = {
2544                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2545                 .rpc_argp = &args,
2546                 .rpc_resp = &res,
2547         };
2548
2549         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2550 }
2551
2552 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2553                 unsigned int pgbase, unsigned int pglen)
2554 {
2555         struct nfs4_exception exception = { };
2556         int err;
2557         do {
2558                 err = nfs4_handle_exception(NFS_SERVER(inode),
2559                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2560                                 &exception);
2561         } while (exception.retry);
2562         return err;
2563 }
2564
2565 /*
2566  * Got race?
2567  * We will need to arrange for the VFS layer to provide an atomic open.
2568  * Until then, this create/open method is prone to inefficiency and race
2569  * conditions due to the lookup, create, and open VFS calls from sys_open()
2570  * placed on the wire.
2571  *
2572  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2573  * The file will be opened again in the subsequent VFS open call
2574  * (nfs4_proc_file_open).
2575  *
2576  * The open for read will just hang around to be used by any process that
2577  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2578  */
2579
2580 static int
2581 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2582                  int flags, struct nfs_open_context *ctx)
2583 {
2584         struct path my_path = {
2585                 .dentry = dentry,
2586         };
2587         struct path *path = &my_path;
2588         struct nfs4_state *state;
2589         struct rpc_cred *cred = NULL;
2590         fmode_t fmode = 0;
2591         int status = 0;
2592
2593         if (ctx != NULL) {
2594                 cred = ctx->cred;
2595                 path = &ctx->path;
2596                 fmode = ctx->mode;
2597         }
2598         sattr->ia_mode &= ~current_umask();
2599         state = nfs4_do_open(dir, path, fmode, flags, sattr, cred);
2600         d_drop(dentry);
2601         if (IS_ERR(state)) {
2602                 status = PTR_ERR(state);
2603                 goto out;
2604         }
2605         d_add(dentry, igrab(state->inode));
2606         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2607         if (ctx != NULL)
2608                 ctx->state = state;
2609         else
2610                 nfs4_close_sync(path, state, fmode);
2611 out:
2612         return status;
2613 }
2614
2615 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2616 {
2617         struct nfs_server *server = NFS_SERVER(dir);
2618         struct nfs_removeargs args = {
2619                 .fh = NFS_FH(dir),
2620                 .name.len = name->len,
2621                 .name.name = name->name,
2622                 .bitmask = server->attr_bitmask,
2623         };
2624         struct nfs_removeres res = {
2625                 .server = server,
2626         };
2627         struct rpc_message msg = {
2628                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2629                 .rpc_argp = &args,
2630                 .rpc_resp = &res,
2631         };
2632         int status = -ENOMEM;
2633
2634         res.dir_attr = nfs_alloc_fattr();
2635         if (res.dir_attr == NULL)
2636                 goto out;
2637
2638         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2639         if (status == 0) {
2640                 update_changeattr(dir, &res.cinfo);
2641                 nfs_post_op_update_inode(dir, res.dir_attr);
2642         }
2643         nfs_free_fattr(res.dir_attr);
2644 out:
2645         return status;
2646 }
2647
2648 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2649 {
2650         struct nfs4_exception exception = { };
2651         int err;
2652         do {
2653                 err = nfs4_handle_exception(NFS_SERVER(dir),
2654                                 _nfs4_proc_remove(dir, name),
2655                                 &exception);
2656         } while (exception.retry);
2657         return err;
2658 }
2659
2660 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2661 {
2662         struct nfs_server *server = NFS_SERVER(dir);
2663         struct nfs_removeargs *args = msg->rpc_argp;
2664         struct nfs_removeres *res = msg->rpc_resp;
2665
2666         args->bitmask = server->cache_consistency_bitmask;
2667         res->server = server;
2668         res->seq_res.sr_slot = NULL;
2669         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2670 }
2671
2672 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2673 {
2674         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2675
2676         if (!nfs4_sequence_done(task, &res->seq_res))
2677                 return 0;
2678         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2679                 return 0;
2680         update_changeattr(dir, &res->cinfo);
2681         nfs_post_op_update_inode(dir, res->dir_attr);
2682         return 1;
2683 }
2684
2685 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2686 {
2687         struct nfs_server *server = NFS_SERVER(dir);
2688         struct nfs_renameargs *arg = msg->rpc_argp;
2689         struct nfs_renameres *res = msg->rpc_resp;
2690
2691         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2692         arg->bitmask = server->attr_bitmask;
2693         res->server = server;
2694 }
2695
2696 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2697                                  struct inode *new_dir)
2698 {
2699         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2700
2701         if (!nfs4_sequence_done(task, &res->seq_res))
2702                 return 0;
2703         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2704                 return 0;
2705
2706         update_changeattr(old_dir, &res->old_cinfo);
2707         nfs_post_op_update_inode(old_dir, res->old_fattr);
2708         update_changeattr(new_dir, &res->new_cinfo);
2709         nfs_post_op_update_inode(new_dir, res->new_fattr);
2710         return 1;
2711 }
2712
2713 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2714                 struct inode *new_dir, struct qstr *new_name)
2715 {
2716         struct nfs_server *server = NFS_SERVER(old_dir);
2717         struct nfs_renameargs arg = {
2718                 .old_dir = NFS_FH(old_dir),
2719                 .new_dir = NFS_FH(new_dir),
2720                 .old_name = old_name,
2721                 .new_name = new_name,
2722                 .bitmask = server->attr_bitmask,
2723         };
2724         struct nfs_renameres res = {
2725                 .server = server,
2726         };
2727         struct rpc_message msg = {
2728                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2729                 .rpc_argp = &arg,
2730                 .rpc_resp = &res,
2731         };
2732         int status = -ENOMEM;
2733
2734         res.old_fattr = nfs_alloc_fattr();
2735         res.new_fattr = nfs_alloc_fattr();
2736         if (res.old_fattr == NULL || res.new_fattr == NULL)
2737                 goto out;
2738
2739         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2740         if (!status) {
2741                 update_changeattr(old_dir, &res.old_cinfo);
2742                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2743                 update_changeattr(new_dir, &res.new_cinfo);
2744                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2745         }
2746 out:
2747         nfs_free_fattr(res.new_fattr);
2748         nfs_free_fattr(res.old_fattr);
2749         return status;
2750 }
2751
2752 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2753                 struct inode *new_dir, struct qstr *new_name)
2754 {
2755         struct nfs4_exception exception = { };
2756         int err;
2757         do {
2758                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2759                                 _nfs4_proc_rename(old_dir, old_name,
2760                                         new_dir, new_name),
2761                                 &exception);
2762         } while (exception.retry);
2763         return err;
2764 }
2765
2766 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2767 {
2768         struct nfs_server *server = NFS_SERVER(inode);
2769         struct nfs4_link_arg arg = {
2770                 .fh     = NFS_FH(inode),
2771                 .dir_fh = NFS_FH(dir),
2772                 .name   = name,
2773                 .bitmask = server->attr_bitmask,
2774         };
2775         struct nfs4_link_res res = {
2776                 .server = server,
2777         };
2778         struct rpc_message msg = {
2779                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2780                 .rpc_argp = &arg,
2781                 .rpc_resp = &res,
2782         };
2783         int status = -ENOMEM;
2784
2785         res.fattr = nfs_alloc_fattr();
2786         res.dir_attr = nfs_alloc_fattr();
2787         if (res.fattr == NULL || res.dir_attr == NULL)
2788                 goto out;
2789
2790         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2791         if (!status) {
2792                 update_changeattr(dir, &res.cinfo);
2793                 nfs_post_op_update_inode(dir, res.dir_attr);
2794                 nfs_post_op_update_inode(inode, res.fattr);
2795         }
2796 out:
2797         nfs_free_fattr(res.dir_attr);
2798         nfs_free_fattr(res.fattr);
2799         return status;
2800 }
2801
2802 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2803 {
2804         struct nfs4_exception exception = { };
2805         int err;
2806         do {
2807                 err = nfs4_handle_exception(NFS_SERVER(inode),
2808                                 _nfs4_proc_link(inode, dir, name),
2809                                 &exception);
2810         } while (exception.retry);
2811         return err;
2812 }
2813
2814 struct nfs4_createdata {
2815         struct rpc_message msg;
2816         struct nfs4_create_arg arg;
2817         struct nfs4_create_res res;
2818         struct nfs_fh fh;
2819         struct nfs_fattr fattr;
2820         struct nfs_fattr dir_fattr;
2821 };
2822
2823 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2824                 struct qstr *name, struct iattr *sattr, u32 ftype)
2825 {
2826         struct nfs4_createdata *data;
2827
2828         data = kzalloc(sizeof(*data), GFP_KERNEL);
2829         if (data != NULL) {
2830                 struct nfs_server *server = NFS_SERVER(dir);
2831
2832                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2833                 data->msg.rpc_argp = &data->arg;
2834                 data->msg.rpc_resp = &data->res;
2835                 data->arg.dir_fh = NFS_FH(dir);
2836                 data->arg.server = server;
2837                 data->arg.name = name;
2838                 data->arg.attrs = sattr;
2839                 data->arg.ftype = ftype;
2840                 data->arg.bitmask = server->attr_bitmask;
2841                 data->res.server = server;
2842                 data->res.fh = &data->fh;
2843                 data->res.fattr = &data->fattr;
2844                 data->res.dir_fattr = &data->dir_fattr;
2845                 nfs_fattr_init(data->res.fattr);
2846                 nfs_fattr_init(data->res.dir_fattr);
2847         }
2848         return data;
2849 }
2850
2851 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2852 {
2853         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2854                                     &data->arg.seq_args, &data->res.seq_res, 1);
2855         if (status == 0) {
2856                 update_changeattr(dir, &data->res.dir_cinfo);
2857                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2858                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2859         }
2860         return status;
2861 }
2862
2863 static void nfs4_free_createdata(struct nfs4_createdata *data)
2864 {
2865         kfree(data);
2866 }
2867
2868 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2869                 struct page *page, unsigned int len, struct iattr *sattr)
2870 {
2871         struct nfs4_createdata *data;
2872         int status = -ENAMETOOLONG;
2873
2874         if (len > NFS4_MAXPATHLEN)
2875                 goto out;
2876
2877         status = -ENOMEM;
2878         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2879         if (data == NULL)
2880                 goto out;
2881
2882         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2883         data->arg.u.symlink.pages = &page;
2884         data->arg.u.symlink.len = len;
2885
2886         status = nfs4_do_create(dir, dentry, data);
2887
2888         nfs4_free_createdata(data);
2889 out:
2890         return status;
2891 }
2892
2893 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2894                 struct page *page, unsigned int len, struct iattr *sattr)
2895 {
2896         struct nfs4_exception exception = { };
2897         int err;
2898         do {
2899                 err = nfs4_handle_exception(NFS_SERVER(dir),
2900                                 _nfs4_proc_symlink(dir, dentry, page,
2901                                                         len, sattr),
2902                                 &exception);
2903         } while (exception.retry);
2904         return err;
2905 }
2906
2907 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2908                 struct iattr *sattr)
2909 {
2910         struct nfs4_createdata *data;
2911         int status = -ENOMEM;
2912
2913         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2914         if (data == NULL)
2915                 goto out;
2916
2917         status = nfs4_do_create(dir, dentry, data);
2918
2919         nfs4_free_createdata(data);
2920 out:
2921         return status;
2922 }
2923
2924 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2925                 struct iattr *sattr)
2926 {
2927         struct nfs4_exception exception = { };
2928         int err;
2929
2930         sattr->ia_mode &= ~current_umask();
2931         do {
2932                 err = nfs4_handle_exception(NFS_SERVER(dir),
2933                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2934                                 &exception);
2935         } while (exception.retry);
2936         return err;
2937 }
2938
2939 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2940                 u64 cookie, struct page **pages, unsigned int count, int plus)
2941 {
2942         struct inode            *dir = dentry->d_inode;
2943         struct nfs4_readdir_arg args = {
2944                 .fh = NFS_FH(dir),
2945                 .pages = pages,
2946                 .pgbase = 0,
2947                 .count = count,
2948                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2949                 .plus = plus,
2950         };
2951         struct nfs4_readdir_res res;
2952         struct rpc_message msg = {
2953                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2954                 .rpc_argp = &args,
2955                 .rpc_resp = &res,
2956                 .rpc_cred = cred,
2957         };
2958         int                     status;
2959
2960         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2961                         dentry->d_parent->d_name.name,
2962                         dentry->d_name.name,
2963                         (unsigned long long)cookie);
2964         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2965         res.pgbase = args.pgbase;
2966         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
2967         if (status >= 0) {
2968                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2969                 status += args.pgbase;
2970         }
2971
2972         nfs_invalidate_atime(dir);
2973
2974         dprintk("%s: returns %d\n", __func__, status);
2975         return status;
2976 }
2977
2978 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2979                 u64 cookie, struct page **pages, unsigned int count, int plus)
2980 {
2981         struct nfs4_exception exception = { };
2982         int err;
2983         do {
2984                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
2985                                 _nfs4_proc_readdir(dentry, cred, cookie,
2986                                         pages, count, plus),
2987                                 &exception);
2988         } while (exception.retry);
2989         return err;
2990 }
2991
2992 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2993                 struct iattr *sattr, dev_t rdev)
2994 {
2995         struct nfs4_createdata *data;
2996         int mode = sattr->ia_mode;
2997         int status = -ENOMEM;
2998
2999         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3000         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3001
3002         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3003         if (data == NULL)
3004                 goto out;
3005
3006         if (S_ISFIFO(mode))
3007                 data->arg.ftype = NF4FIFO;
3008         else if (S_ISBLK(mode)) {
3009                 data->arg.ftype = NF4BLK;
3010                 data->arg.u.device.specdata1 = MAJOR(rdev);
3011                 data->arg.u.device.specdata2 = MINOR(rdev);
3012         }
3013         else if (S_ISCHR(mode)) {
3014                 data->arg.ftype = NF4CHR;
3015                 data->arg.u.device.specdata1 = MAJOR(rdev);
3016                 data->arg.u.device.specdata2 = MINOR(rdev);
3017         }
3018
3019         status = nfs4_do_create(dir, dentry, data);
3020
3021         nfs4_free_createdata(data);
3022 out:
3023         return status;
3024 }
3025
3026 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3027                 struct iattr *sattr, dev_t rdev)
3028 {
3029         struct nfs4_exception exception = { };
3030         int err;
3031
3032         sattr->ia_mode &= ~current_umask();
3033         do {
3034                 err = nfs4_handle_exception(NFS_SERVER(dir),
3035                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3036                                 &exception);
3037         } while (exception.retry);
3038         return err;
3039 }
3040
3041 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3042                  struct nfs_fsstat *fsstat)
3043 {
3044         struct nfs4_statfs_arg args = {
3045                 .fh = fhandle,
3046                 .bitmask = server->attr_bitmask,
3047         };
3048         struct nfs4_statfs_res res = {
3049                 .fsstat = fsstat,
3050         };
3051         struct rpc_message msg = {
3052                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3053                 .rpc_argp = &args,
3054                 .rpc_resp = &res,
3055         };
3056
3057         nfs_fattr_init(fsstat->fattr);
3058         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3059 }
3060
3061 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3062 {
3063         struct nfs4_exception exception = { };
3064         int err;
3065         do {
3066                 err = nfs4_handle_exception(server,
3067                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3068                                 &exception);
3069         } while (exception.retry);
3070         return err;
3071 }
3072
3073 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3074                 struct nfs_fsinfo *fsinfo)
3075 {
3076         struct nfs4_fsinfo_arg args = {
3077                 .fh = fhandle,
3078                 .bitmask = server->attr_bitmask,
3079         };
3080         struct nfs4_fsinfo_res res = {
3081                 .fsinfo = fsinfo,
3082         };
3083         struct rpc_message msg = {
3084                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3085                 .rpc_argp = &args,
3086                 .rpc_resp = &res,
3087         };
3088
3089         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3090 }
3091
3092 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3093 {
3094         struct nfs4_exception exception = { };
3095         int err;
3096
3097         do {
3098                 err = nfs4_handle_exception(server,
3099                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3100                                 &exception);
3101         } while (exception.retry);
3102         return err;
3103 }
3104
3105 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3106 {
3107         nfs_fattr_init(fsinfo->fattr);
3108         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3109 }
3110
3111 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3112                 struct nfs_pathconf *pathconf)
3113 {
3114         struct nfs4_pathconf_arg args = {
3115                 .fh = fhandle,
3116                 .bitmask = server->attr_bitmask,
3117         };
3118         struct nfs4_pathconf_res res = {
3119                 .pathconf = pathconf,
3120         };
3121         struct rpc_message msg = {
3122                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3123                 .rpc_argp = &args,
3124                 .rpc_resp = &res,
3125         };
3126
3127         /* None of the pathconf attributes are mandatory to implement */
3128         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3129                 memset(pathconf, 0, sizeof(*pathconf));
3130                 return 0;
3131         }
3132
3133         nfs_fattr_init(pathconf->fattr);
3134         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3135 }
3136
3137 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3138                 struct nfs_pathconf *pathconf)
3139 {
3140         struct nfs4_exception exception = { };
3141         int err;
3142
3143         do {
3144                 err = nfs4_handle_exception(server,
3145                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3146                                 &exception);
3147         } while (exception.retry);
3148         return err;
3149 }
3150
3151 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3152 {
3153         struct nfs_server *server = NFS_SERVER(data->inode);
3154
3155         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3156                 nfs_restart_rpc(task, server->nfs_client);
3157                 return -EAGAIN;
3158         }
3159
3160         nfs_invalidate_atime(data->inode);
3161         if (task->tk_status > 0)
3162                 renew_lease(server, data->timestamp);
3163         return 0;
3164 }
3165
3166 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3167 {
3168
3169         dprintk("--> %s\n", __func__);
3170
3171         if (!nfs4_sequence_done(task, &data->res.seq_res))
3172                 return -EAGAIN;
3173
3174         return data->read_done_cb(task, data);
3175 }
3176
3177 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3178 {
3179         data->timestamp   = jiffies;
3180         data->read_done_cb = nfs4_read_done_cb;
3181         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3182 }
3183
3184 /* Reset the the nfs_read_data to send the read to the MDS. */
3185 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3186 {
3187         dprintk("%s Reset task for i/o through\n", __func__);
3188         put_lseg(data->lseg);
3189         data->lseg = NULL;
3190         /* offsets will differ in the dense stripe case */
3191         data->args.offset = data->mds_offset;
3192         data->ds_clp = NULL;
3193         data->args.fh     = NFS_FH(data->inode);
3194         data->read_done_cb = nfs4_read_done_cb;
3195         task->tk_ops = data->mds_ops;
3196         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3197 }
3198 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3199
3200 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3201 {
3202         struct inode *inode = data->inode;
3203
3204         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3205                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3206                 return -EAGAIN;
3207         }
3208         if (task->tk_status >= 0) {
3209                 renew_lease(NFS_SERVER(inode), data->timestamp);
3210                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3211         }
3212         return 0;
3213 }
3214
3215 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3216 {
3217         if (!nfs4_sequence_done(task, &data->res.seq_res))
3218                 return -EAGAIN;
3219         return data->write_done_cb(task, data);
3220 }
3221
3222 /* Reset the the nfs_write_data to send the write to the MDS. */
3223 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3224 {
3225         dprintk("%s Reset task for i/o through\n", __func__);
3226         put_lseg(data->lseg);
3227         data->lseg          = NULL;
3228         data->ds_clp        = NULL;
3229         data->write_done_cb = nfs4_write_done_cb;
3230         data->args.fh       = NFS_FH(data->inode);
3231         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3232         data->args.offset   = data->mds_offset;
3233         data->res.fattr     = &data->fattr;
3234         task->tk_ops        = data->mds_ops;
3235         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3236 }
3237 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3238
3239 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3240 {
3241         struct nfs_server *server = NFS_SERVER(data->inode);
3242
3243         if (data->lseg) {
3244                 data->args.bitmask = NULL;
3245                 data->res.fattr = NULL;
3246         } else
3247                 data->args.bitmask = server->cache_consistency_bitmask;
3248         if (!data->write_done_cb)
3249                 data->write_done_cb = nfs4_write_done_cb;
3250         data->res.server = server;
3251         data->timestamp   = jiffies;
3252
3253         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3254 }
3255
3256 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3257 {
3258         struct inode *inode = data->inode;
3259
3260         if (!nfs4_sequence_done(task, &data->res.seq_res))
3261                 return -EAGAIN;
3262
3263         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3264                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3265                 return -EAGAIN;
3266         }
3267         nfs_refresh_inode(inode, data->res.fattr);
3268         return 0;
3269 }
3270
3271 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3272 {
3273         struct nfs_server *server = NFS_SERVER(data->inode);
3274
3275         data->args.bitmask = server->cache_consistency_bitmask;
3276         data->res.server = server;
3277         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3278 }
3279
3280 struct nfs4_renewdata {
3281         struct nfs_client       *client;
3282         unsigned long           timestamp;
3283 };
3284
3285 /*
3286  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3287  * standalone procedure for queueing an asynchronous RENEW.
3288  */
3289 static void nfs4_renew_release(void *calldata)
3290 {
3291         struct nfs4_renewdata *data = calldata;
3292         struct nfs_client *clp = data->client;
3293
3294         if (atomic_read(&clp->cl_count) > 1)
3295                 nfs4_schedule_state_renewal(clp);
3296         nfs_put_client(clp);
3297         kfree(data);
3298 }
3299
3300 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3301 {
3302         struct nfs4_renewdata *data = calldata;
3303         struct nfs_client *clp = data->client;
3304         unsigned long timestamp = data->timestamp;
3305
3306         if (task->tk_status < 0) {
3307                 /* Unless we're shutting down, schedule state recovery! */
3308                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
3309                         nfs4_schedule_lease_recovery(clp);
3310                 return;
3311         }
3312         do_renew_lease(clp, timestamp);
3313 }
3314
3315 static const struct rpc_call_ops nfs4_renew_ops = {
3316         .rpc_call_done = nfs4_renew_done,
3317         .rpc_release = nfs4_renew_release,
3318 };
3319
3320 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
3321 {
3322         struct rpc_message msg = {
3323                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3324                 .rpc_argp       = clp,
3325                 .rpc_cred       = cred,
3326         };
3327         struct nfs4_renewdata *data;
3328
3329         if (!atomic_inc_not_zero(&clp->cl_count))
3330                 return -EIO;
3331         data = kmalloc(sizeof(*data), GFP_KERNEL);
3332         if (data == NULL)
3333                 return -ENOMEM;
3334         data->client = clp;
3335         data->timestamp = jiffies;
3336         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3337                         &nfs4_renew_ops, data);
3338 }
3339
3340 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3341 {
3342         struct rpc_message msg = {
3343                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3344                 .rpc_argp       = clp,
3345                 .rpc_cred       = cred,
3346         };
3347         unsigned long now = jiffies;
3348         int status;
3349
3350         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3351         if (status < 0)
3352                 return status;
3353         do_renew_lease(clp, now);
3354         return 0;
3355 }
3356
3357 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3358 {
3359         return (server->caps & NFS_CAP_ACLS)
3360                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3361                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3362 }
3363
3364 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3365  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3366  * the stack.
3367  */
3368 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3369
3370 static void buf_to_pages(const void *buf, size_t buflen,
3371                 struct page **pages, unsigned int *pgbase)
3372 {
3373         const void *p = buf;
3374
3375         *pgbase = offset_in_page(buf);
3376         p -= *pgbase;
3377         while (p < buf + buflen) {
3378                 *(pages++) = virt_to_page(p);
3379                 p += PAGE_CACHE_SIZE;
3380         }
3381 }
3382
3383 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3384                 struct page **pages, unsigned int *pgbase)
3385 {
3386         struct page *newpage, **spages;
3387         int rc = 0;
3388         size_t len;
3389         spages = pages;
3390
3391         do {
3392                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3393                 newpage = alloc_page(GFP_KERNEL);
3394
3395                 if (newpage == NULL)
3396                         goto unwind;
3397                 memcpy(page_address(newpage), buf, len);
3398                 buf += len;
3399                 buflen -= len;
3400                 *pages++ = newpage;
3401                 rc++;
3402         } while (buflen != 0);
3403
3404         return rc;
3405
3406 unwind:
3407         for(; rc > 0; rc--)
3408                 __free_page(spages[rc-1]);
3409         return -ENOMEM;
3410 }
3411
3412 struct nfs4_cached_acl {
3413         int cached;
3414         size_t len;
3415         char data[0];
3416 };
3417
3418 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3419 {
3420         struct nfs_inode *nfsi = NFS_I(inode);
3421
3422         spin_lock(&inode->i_lock);
3423         kfree(nfsi->nfs4_acl);
3424         nfsi->nfs4_acl = acl;
3425         spin_unlock(&inode->i_lock);
3426 }
3427
3428 static void nfs4_zap_acl_attr(struct inode *inode)
3429 {
3430         nfs4_set_cached_acl(inode, NULL);
3431 }
3432
3433 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3434 {
3435         struct nfs_inode *nfsi = NFS_I(inode);
3436         struct nfs4_cached_acl *acl;
3437         int ret = -ENOENT;
3438
3439         spin_lock(&inode->i_lock);
3440         acl = nfsi->nfs4_acl;
3441         if (acl == NULL)
3442                 goto out;
3443         if (buf == NULL) /* user is just asking for length */
3444                 goto out_len;
3445         if (acl->cached == 0)
3446                 goto out;
3447         ret = -ERANGE; /* see getxattr(2) man page */
3448         if (acl->len > buflen)
3449                 goto out;
3450         memcpy(buf, acl->data, acl->len);
3451 out_len:
3452         ret = acl->len;
3453 out:
3454         spin_unlock(&inode->i_lock);
3455         return ret;
3456 }
3457
3458 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3459 {
3460         struct nfs4_cached_acl *acl;
3461
3462         if (buf && acl_len <= PAGE_SIZE) {
3463                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3464                 if (acl == NULL)
3465                         goto out;
3466                 acl->cached = 1;
3467                 memcpy(acl->data, buf, acl_len);
3468         } else {
3469                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3470                 if (acl == NULL)
3471                         goto out;
3472                 acl->cached = 0;
3473         }
3474         acl->len = acl_len;
3475 out:
3476         nfs4_set_cached_acl(inode, acl);
3477 }
3478
3479 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3480 {
3481         struct page *pages[NFS4ACL_MAXPAGES];
3482         struct nfs_getaclargs args = {
3483                 .fh = NFS_FH(inode),
3484                 .acl_pages = pages,
3485                 .acl_len = buflen,
3486         };
3487         struct nfs_getaclres res = {
3488                 .acl_len = buflen,
3489         };
3490         void *resp_buf;
3491         struct rpc_message msg = {
3492                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3493                 .rpc_argp = &args,
3494                 .rpc_resp = &res,
3495         };
3496         struct page *localpage = NULL;
3497         int ret;
3498
3499         if (buflen < PAGE_SIZE) {
3500                 /* As long as we're doing a round trip to the server anyway,
3501                  * let's be prepared for a page of acl data. */
3502                 localpage = alloc_page(GFP_KERNEL);
3503                 resp_buf = page_address(localpage);
3504                 if (localpage == NULL)
3505                         return -ENOMEM;
3506                 args.acl_pages[0] = localpage;
3507                 args.acl_pgbase = 0;
3508                 args.acl_len = PAGE_SIZE;
3509         } else {
3510                 resp_buf = buf;
3511                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3512         }
3513         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3514         if (ret)
3515                 goto out_free;
3516         if (res.acl_len > args.acl_len)
3517                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3518         else
3519                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3520         if (buf) {
3521                 ret = -ERANGE;
3522                 if (res.acl_len > buflen)
3523                         goto out_free;
3524                 if (localpage)
3525                         memcpy(buf, resp_buf, res.acl_len);
3526         }
3527         ret = res.acl_len;
3528 out_free:
3529         if (localpage)
3530                 __free_page(localpage);
3531         return ret;
3532 }
3533
3534 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3535 {
3536         struct nfs4_exception exception = { };
3537         ssize_t ret;
3538         do {
3539                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3540                 if (ret >= 0)
3541                         break;
3542                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3543         } while (exception.retry);
3544         return ret;
3545 }
3546
3547 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3548 {
3549         struct nfs_server *server = NFS_SERVER(inode);
3550         int ret;
3551
3552         if (!nfs4_server_supports_acls(server))
3553                 return -EOPNOTSUPP;
3554         ret = nfs_revalidate_inode(server, inode);
3555         if (ret < 0)
3556                 return ret;
3557         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3558                 nfs_zap_acl_cache(inode);
3559         ret = nfs4_read_cached_acl(inode, buf, buflen);
3560         if (ret != -ENOENT)
3561                 return ret;
3562         return nfs4_get_acl_uncached(inode, buf, buflen);
3563 }
3564
3565 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3566 {
3567         struct nfs_server *server = NFS_SERVER(inode);
3568         struct page *pages[NFS4ACL_MAXPAGES];
3569         struct nfs_setaclargs arg = {
3570                 .fh             = NFS_FH(inode),
3571                 .acl_pages      = pages,
3572                 .acl_len        = buflen,
3573         };
3574         struct nfs_setaclres res;
3575         struct rpc_message msg = {
3576                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3577                 .rpc_argp       = &arg,
3578                 .rpc_resp       = &res,
3579         };
3580         int ret, i;
3581
3582         if (!nfs4_server_supports_acls(server))
3583                 return -EOPNOTSUPP;
3584         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3585         if (i < 0)
3586                 return i;
3587         nfs_inode_return_delegation(inode);
3588         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3589
3590         /*
3591          * Free each page after tx, so the only ref left is
3592          * held by the network stack
3593          */
3594         for (; i > 0; i--)
3595                 put_page(pages[i-1]);
3596
3597         /*
3598          * Acl update can result in inode attribute update.
3599          * so mark the attribute cache invalid.
3600          */
3601         spin_lock(&inode->i_lock);
3602         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3603         spin_unlock(&inode->i_lock);
3604         nfs_access_zap_cache(inode);
3605         nfs_zap_acl_cache(inode);
3606         return ret;
3607 }
3608
3609 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3610 {
3611         struct nfs4_exception exception = { };
3612         int err;
3613         do {
3614                 err = nfs4_handle_exception(NFS_SERVER(inode),
3615                                 __nfs4_proc_set_acl(inode, buf, buflen),
3616                                 &exception);
3617         } while (exception.retry);
3618         return err;
3619 }
3620
3621 static int
3622 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3623 {
3624         struct nfs_client *clp = server->nfs_client;
3625
3626         if (task->tk_status >= 0)
3627                 return 0;
3628         switch(task->tk_status) {
3629                 case -NFS4ERR_ADMIN_REVOKED:
3630                 case -NFS4ERR_BAD_STATEID:
3631                 case -NFS4ERR_OPENMODE:
3632                         if (state == NULL)
3633                                 break;
3634                         nfs4_schedule_stateid_recovery(server, state);
3635                         goto wait_on_recovery;
3636                 case -NFS4ERR_STALE_STATEID:
3637                 case -NFS4ERR_STALE_CLIENTID:
3638                 case -NFS4ERR_EXPIRED:
3639                         nfs4_schedule_lease_recovery(clp);
3640                         goto wait_on_recovery;
3641 #if defined(CONFIG_NFS_V4_1)
3642                 case -NFS4ERR_BADSESSION:
3643                 case -NFS4ERR_BADSLOT:
3644                 case -NFS4ERR_BAD_HIGH_SLOT:
3645                 case -NFS4ERR_DEADSESSION:
3646                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3647                 case -NFS4ERR_SEQ_FALSE_RETRY:
3648                 case -NFS4ERR_SEQ_MISORDERED:
3649                         dprintk("%s ERROR %d, Reset session\n", __func__,
3650                                 task->tk_status);
3651                         nfs4_schedule_session_recovery(clp->cl_session);
3652                         task->tk_status = 0;
3653                         return -EAGAIN;
3654 #endif /* CONFIG_NFS_V4_1 */
3655                 case -NFS4ERR_DELAY:
3656                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3657                 case -NFS4ERR_GRACE:
3658                 case -EKEYEXPIRED:
3659                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3660                         task->tk_status = 0;
3661                         return -EAGAIN;
3662                 case -NFS4ERR_OLD_STATEID:
3663                         task->tk_status = 0;
3664                         return -EAGAIN;
3665         }
3666         task->tk_status = nfs4_map_errors(task->tk_status);
3667         return 0;
3668 wait_on_recovery:
3669         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3670         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3671                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3672         task->tk_status = 0;
3673         return -EAGAIN;
3674 }
3675
3676 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3677                 unsigned short port, struct rpc_cred *cred,
3678                 struct nfs4_setclientid_res *res)
3679 {
3680         nfs4_verifier sc_verifier;
3681         struct nfs4_setclientid setclientid = {
3682                 .sc_verifier = &sc_verifier,
3683                 .sc_prog = program,
3684                 .sc_cb_ident = clp->cl_cb_ident,
3685         };
3686         struct rpc_message msg = {
3687                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3688                 .rpc_argp = &setclientid,
3689                 .rpc_resp = res,
3690                 .rpc_cred = cred,
3691         };
3692         __be32 *p;
3693         int loop = 0;
3694         int status;
3695
3696         p = (__be32*)sc_verifier.data;
3697         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3698         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3699
3700         for(;;) {
3701                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3702                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3703                                 clp->cl_ipaddr,
3704                                 rpc_peeraddr2str(clp->cl_rpcclient,
3705                                                         RPC_DISPLAY_ADDR),
3706                                 rpc_peeraddr2str(clp->cl_rpcclient,
3707                                                         RPC_DISPLAY_PROTO),
3708                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3709                                 clp->cl_id_uniquifier);
3710                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3711                                 sizeof(setclientid.sc_netid),
3712                                 rpc_peeraddr2str(clp->cl_rpcclient,
3713                                                         RPC_DISPLAY_NETID));
3714                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3715                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3716                                 clp->cl_ipaddr, port >> 8, port & 255);
3717
3718                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3719                 if (status != -NFS4ERR_CLID_INUSE)
3720                         break;
3721                 if (signalled())
3722                         break;
3723                 if (loop++ & 1)
3724                         ssleep(clp->cl_lease_time / HZ + 1);
3725                 else
3726                         if (++clp->cl_id_uniquifier == 0)
3727                                 break;
3728         }
3729         return status;
3730 }
3731
3732 static int _nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3733                 struct nfs4_setclientid_res *arg,
3734                 struct rpc_cred *cred)
3735 {
3736         struct nfs_fsinfo fsinfo;
3737         struct rpc_message msg = {
3738                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3739                 .rpc_argp = arg,
3740                 .rpc_resp = &fsinfo,
3741                 .rpc_cred = cred,
3742         };
3743         unsigned long now;
3744         int status;
3745
3746         now = jiffies;
3747         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3748         if (status == 0) {
3749                 spin_lock(&clp->cl_lock);
3750                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3751                 clp->cl_last_renewal = now;
3752                 spin_unlock(&clp->cl_lock);
3753         }
3754         return status;
3755 }
3756
3757 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3758                 struct nfs4_setclientid_res *arg,
3759                 struct rpc_cred *cred)
3760 {
3761         long timeout = 0;
3762         int err;
3763         do {
3764                 err = _nfs4_proc_setclientid_confirm(clp, arg, cred);
3765                 switch (err) {
3766                         case 0:
3767                                 return err;
3768                         case -NFS4ERR_RESOURCE:
3769                                 /* The IBM lawyers misread another document! */
3770                         case -NFS4ERR_DELAY:
3771                                 err = nfs4_delay(clp->cl_rpcclient, &timeout);
3772                 }
3773         } while (err == 0);
3774         return err;
3775 }
3776
3777 struct nfs4_delegreturndata {
3778         struct nfs4_delegreturnargs args;
3779         struct nfs4_delegreturnres res;
3780         struct nfs_fh fh;
3781         nfs4_stateid stateid;
3782         unsigned long timestamp;
3783         struct nfs_fattr fattr;
3784         int rpc_status;
3785 };
3786
3787 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3788 {
3789         struct nfs4_delegreturndata *data = calldata;
3790
3791         if (!nfs4_sequence_done(task, &data->res.seq_res))
3792                 return;
3793
3794         switch (task->tk_status) {
3795         case -NFS4ERR_STALE_STATEID:
3796         case -NFS4ERR_EXPIRED:
3797         case 0:
3798                 renew_lease(data->res.server, data->timestamp);
3799                 break;
3800         default:
3801                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3802                                 -EAGAIN) {
3803                         nfs_restart_rpc(task, data->res.server->nfs_client);
3804                         return;
3805                 }
3806         }
3807         data->rpc_status = task->tk_status;
3808 }
3809
3810 static void nfs4_delegreturn_release(void *calldata)
3811 {
3812         kfree(calldata);
3813 }
3814
3815 #if defined(CONFIG_NFS_V4_1)
3816 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3817 {
3818         struct nfs4_delegreturndata *d_data;
3819
3820         d_data = (struct nfs4_delegreturndata *)data;
3821
3822         if (nfs4_setup_sequence(d_data->res.server,
3823                                 &d_data->args.seq_args,
3824                                 &d_data->res.seq_res, 1, task))
3825                 return;
3826         rpc_call_start(task);
3827 }
3828 #endif /* CONFIG_NFS_V4_1 */
3829
3830 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3831 #if defined(CONFIG_NFS_V4_1)
3832         .rpc_call_prepare = nfs4_delegreturn_prepare,
3833 #endif /* CONFIG_NFS_V4_1 */
3834         .rpc_call_done = nfs4_delegreturn_done,
3835         .rpc_release = nfs4_delegreturn_release,
3836 };
3837
3838 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3839 {
3840         struct nfs4_delegreturndata *data;
3841         struct nfs_server *server = NFS_SERVER(inode);
3842         struct rpc_task *task;
3843         struct rpc_message msg = {
3844                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3845                 .rpc_cred = cred,
3846         };
3847         struct rpc_task_setup task_setup_data = {
3848                 .rpc_client = server->client,
3849                 .rpc_message = &msg,
3850                 .callback_ops = &nfs4_delegreturn_ops,
3851                 .flags = RPC_TASK_ASYNC,
3852         };
3853         int status = 0;
3854
3855         data = kzalloc(sizeof(*data), GFP_NOFS);
3856         if (data == NULL)
3857                 return -ENOMEM;
3858         data->args.fhandle = &data->fh;
3859         data->args.stateid = &data->stateid;
3860         data->args.bitmask = server->attr_bitmask;
3861         nfs_copy_fh(&data->fh, NFS_FH(inode));
3862         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3863         data->res.fattr = &data->fattr;
3864         data->res.server = server;
3865         nfs_fattr_init(data->res.fattr);
3866         data->timestamp = jiffies;
3867         data->rpc_status = 0;
3868
3869         task_setup_data.callback_data = data;
3870         msg.rpc_argp = &data->args;
3871         msg.rpc_resp = &data->res;
3872         task = rpc_run_task(&task_setup_data);
3873         if (IS_ERR(task))
3874                 return PTR_ERR(task);
3875         if (!issync)
3876                 goto out;
3877         status = nfs4_wait_for_completion_rpc_task(task);
3878         if (status != 0)
3879                 goto out;
3880         status = data->rpc_status;
3881         if (status != 0)
3882                 goto out;
3883         nfs_refresh_inode(inode, &data->fattr);
3884 out:
3885         rpc_put_task(task);
3886         return status;
3887 }
3888
3889 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3890 {
3891         struct nfs_server *server = NFS_SERVER(inode);
3892         struct nfs4_exception exception = { };
3893         int err;
3894         do {
3895                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3896                 switch (err) {
3897                         case -NFS4ERR_STALE_STATEID:
3898                         case -NFS4ERR_EXPIRED:
3899                         case 0:
3900                                 return 0;
3901                 }
3902                 err = nfs4_handle_exception(server, err, &exception);
3903         } while (exception.retry);
3904         return err;
3905 }
3906
3907 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3908 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3909
3910 /*
3911  * sleep, with exponential backoff, and retry the LOCK operation.
3912  */
3913 static unsigned long
3914 nfs4_set_lock_task_retry(unsigned long timeout)
3915 {
3916         schedule_timeout_killable(timeout);
3917         timeout <<= 1;
3918         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3919                 return NFS4_LOCK_MAXTIMEOUT;
3920         return timeout;
3921 }
3922
3923 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3924 {
3925         struct inode *inode = state->inode;
3926         struct nfs_server *server = NFS_SERVER(inode);
3927         struct nfs_client *clp = server->nfs_client;
3928         struct nfs_lockt_args arg = {
3929                 .fh = NFS_FH(inode),
3930                 .fl = request,
3931         };
3932         struct nfs_lockt_res res = {
3933                 .denied = request,
3934         };
3935         struct rpc_message msg = {
3936                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3937                 .rpc_argp       = &arg,
3938                 .rpc_resp       = &res,
3939                 .rpc_cred       = state->owner->so_cred,
3940         };
3941         struct nfs4_lock_state *lsp;
3942         int status;
3943
3944         arg.lock_owner.clientid = clp->cl_clientid;
3945         status = nfs4_set_lock_state(state, request);
3946         if (status != 0)
3947                 goto out;
3948         lsp = request->fl_u.nfs4_fl.owner;
3949         arg.lock_owner.id = lsp->ls_id.id;
3950         arg.lock_owner.s_dev = server->s_dev;
3951         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3952         switch (status) {
3953                 case 0:
3954                         request->fl_type = F_UNLCK;
3955                         break;
3956                 case -NFS4ERR_DENIED:
3957                         status = 0;
3958         }
3959         request->fl_ops->fl_release_private(request);
3960 out:
3961         return status;
3962 }
3963
3964 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3965 {
3966         struct nfs4_exception exception = { };
3967         int err;
3968
3969         do {
3970                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3971                                 _nfs4_proc_getlk(state, cmd, request),
3972                                 &exception);
3973         } while (exception.retry);
3974         return err;
3975 }
3976
3977 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3978 {
3979         int res = 0;
3980         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3981                 case FL_POSIX:
3982                         res = posix_lock_file_wait(file, fl);
3983                         break;
3984                 case FL_FLOCK:
3985                         res = flock_lock_file_wait(file, fl);
3986                         break;
3987                 default:
3988                         BUG();
3989         }
3990         return res;
3991 }
3992
3993 struct nfs4_unlockdata {
3994         struct nfs_locku_args arg;
3995         struct nfs_locku_res res;
3996         struct nfs4_lock_state *lsp;
3997         struct nfs_open_context *ctx;
3998         struct file_lock fl;
3999         const struct nfs_server *server;
4000         unsigned long timestamp;
4001 };
4002
4003 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4004                 struct nfs_open_context *ctx,
4005                 struct nfs4_lock_state *lsp,
4006                 struct nfs_seqid *seqid)
4007 {
4008         struct nfs4_unlockdata *p;
4009         struct inode *inode = lsp->ls_state->inode;
4010
4011         p = kzalloc(sizeof(*p), GFP_NOFS);
4012         if (p == NULL)
4013                 return NULL;
4014         p->arg.fh = NFS_FH(inode);
4015         p->arg.fl = &p->fl;
4016         p->arg.seqid = seqid;
4017         p->res.seqid = seqid;
4018         p->arg.stateid = &lsp->ls_stateid;
4019         p->lsp = lsp;
4020         atomic_inc(&lsp->ls_count);
4021         /* Ensure we don't close file until we're done freeing locks! */
4022         p->ctx = get_nfs_open_context(ctx);
4023         memcpy(&p->fl, fl, sizeof(p->fl));
4024         p->server = NFS_SERVER(inode);
4025         return p;
4026 }
4027
4028 static void nfs4_locku_release_calldata(void *data)
4029 {
4030         struct nfs4_unlockdata *calldata = data;
4031         nfs_free_seqid(calldata->arg.seqid);
4032         nfs4_put_lock_state(calldata->lsp);
4033         put_nfs_open_context(calldata->ctx);
4034         kfree(calldata);
4035 }
4036
4037 static void nfs4_locku_done(struct rpc_task *task, void *data)
4038 {
4039         struct nfs4_unlockdata *calldata = data;
4040
4041         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4042                 return;
4043         switch (task->tk_status) {
4044                 case 0:
4045                         memcpy(calldata->lsp->ls_stateid.data,
4046                                         calldata->res.stateid.data,
4047                                         sizeof(calldata->lsp->ls_stateid.data));
4048                         renew_lease(calldata->server, calldata->timestamp);
4049                         break;
4050                 case -NFS4ERR_BAD_STATEID:
4051                 case -NFS4ERR_OLD_STATEID:
4052                 case -NFS4ERR_STALE_STATEID:
4053                 case -NFS4ERR_EXPIRED:
4054                         break;
4055                 default:
4056                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4057                                 nfs_restart_rpc(task,
4058                                                  calldata->server->nfs_client);
4059         }
4060 }
4061
4062 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4063 {
4064         struct nfs4_unlockdata *calldata = data;
4065
4066         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4067                 return;
4068         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4069                 /* Note: exit _without_ running nfs4_locku_done */
4070                 task->tk_action = NULL;
4071                 return;
4072         }
4073         calldata->timestamp = jiffies;
4074         if (nfs4_setup_sequence(calldata->server,
4075                                 &calldata->arg.seq_args,
4076                                 &calldata->res.seq_res, 1, task))
4077                 return;
4078         rpc_call_start(task);
4079 }
4080
4081 static const struct rpc_call_ops nfs4_locku_ops = {
4082         .rpc_call_prepare = nfs4_locku_prepare,
4083         .rpc_call_done = nfs4_locku_done,
4084         .rpc_release = nfs4_locku_release_calldata,
4085 };
4086
4087 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4088                 struct nfs_open_context *ctx,
4089                 struct nfs4_lock_state *lsp,
4090                 struct nfs_seqid *seqid)
4091 {
4092         struct nfs4_unlockdata *data;
4093         struct rpc_message msg = {
4094                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4095                 .rpc_cred = ctx->cred,
4096         };
4097         struct rpc_task_setup task_setup_data = {
4098                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4099                 .rpc_message = &msg,
4100                 .callback_ops = &nfs4_locku_ops,
4101                 .workqueue = nfsiod_workqueue,
4102                 .flags = RPC_TASK_ASYNC,
4103         };
4104
4105         /* Ensure this is an unlock - when canceling a lock, the
4106          * canceled lock is passed in, and it won't be an unlock.
4107          */
4108         fl->fl_type = F_UNLCK;
4109
4110         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4111         if (data == NULL) {
4112                 nfs_free_seqid(seqid);
4113                 return ERR_PTR(-ENOMEM);
4114         }
4115
4116         msg.rpc_argp = &data->arg;
4117         msg.rpc_resp = &data->res;
4118         task_setup_data.callback_data = data;
4119         return rpc_run_task(&task_setup_data);
4120 }
4121
4122 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4123 {
4124         struct nfs_inode *nfsi = NFS_I(state->inode);
4125         struct nfs_seqid *seqid;
4126         struct nfs4_lock_state *lsp;
4127         struct rpc_task *task;
4128         int status = 0;
4129         unsigned char fl_flags = request->fl_flags;
4130
4131         status = nfs4_set_lock_state(state, request);
4132         /* Unlock _before_ we do the RPC call */
4133         request->fl_flags |= FL_EXISTS;
4134         down_read(&nfsi->rwsem);
4135         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4136                 up_read(&nfsi->rwsem);
4137                 goto out;
4138         }
4139         up_read(&nfsi->rwsem);
4140         if (status != 0)
4141                 goto out;
4142         /* Is this a delegated lock? */
4143         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4144                 goto out;
4145         lsp = request->fl_u.nfs4_fl.owner;
4146         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4147         status = -ENOMEM;
4148         if (seqid == NULL)
4149                 goto out;
4150         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4151         status = PTR_ERR(task);
4152         if (IS_ERR(task))
4153                 goto out;
4154         status = nfs4_wait_for_completion_rpc_task(task);
4155         rpc_put_task(task);
4156 out:
4157         request->fl_flags = fl_flags;
4158         return status;
4159 }
4160
4161 struct nfs4_lockdata {
4162         struct nfs_lock_args arg;
4163         struct nfs_lock_res res;
4164         struct nfs4_lock_state *lsp;
4165         struct nfs_open_context *ctx;
4166         struct file_lock fl;
4167         unsigned long timestamp;
4168         int rpc_status;
4169         int cancelled;
4170         struct nfs_server *server;
4171 };
4172
4173 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4174                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4175                 gfp_t gfp_mask)
4176 {
4177         struct nfs4_lockdata *p;
4178         struct inode *inode = lsp->ls_state->inode;
4179         struct nfs_server *server = NFS_SERVER(inode);
4180
4181         p = kzalloc(sizeof(*p), gfp_mask);
4182         if (p == NULL)
4183                 return NULL;
4184
4185         p->arg.fh = NFS_FH(inode);
4186         p->arg.fl = &p->fl;
4187         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4188         if (p->arg.open_seqid == NULL)
4189                 goto out_free;
4190         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4191         if (p->arg.lock_seqid == NULL)
4192                 goto out_free_seqid;
4193         p->arg.lock_stateid = &lsp->ls_stateid;
4194         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4195         p->arg.lock_owner.id = lsp->ls_id.id;
4196         p->arg.lock_owner.s_dev = server->s_dev;
4197         p->res.lock_seqid = p->arg.lock_seqid;
4198         p->lsp = lsp;
4199         p->server = server;
4200         atomic_inc(&lsp->ls_count);
4201         p->ctx = get_nfs_open_context(ctx);
4202         memcpy(&p->fl, fl, sizeof(p->fl));
4203         return p;
4204 out_free_seqid:
4205         nfs_free_seqid(p->arg.open_seqid);
4206 out_free:
4207         kfree(p);
4208         return NULL;
4209 }
4210
4211 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4212 {
4213         struct nfs4_lockdata *data = calldata;
4214         struct nfs4_state *state = data->lsp->ls_state;
4215
4216         dprintk("%s: begin!\n", __func__);
4217         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4218                 return;
4219         /* Do we need to do an open_to_lock_owner? */
4220         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4221                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4222                         return;
4223                 data->arg.open_stateid = &state->stateid;
4224                 data->arg.new_lock_owner = 1;
4225                 data->res.open_seqid = data->arg.open_seqid;
4226         } else
4227                 data->arg.new_lock_owner = 0;
4228         data->timestamp = jiffies;
4229         if (nfs4_setup_sequence(data->server,
4230                                 &data->arg.seq_args,
4231                                 &data->res.seq_res, 1, task))
4232                 return;
4233         rpc_call_start(task);
4234         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4235 }
4236
4237 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4238 {
4239         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4240         nfs4_lock_prepare(task, calldata);
4241 }
4242
4243 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4244 {
4245         struct nfs4_lockdata *data = calldata;
4246
4247         dprintk("%s: begin!\n", __func__);
4248
4249         if (!nfs4_sequence_done(task, &data->res.seq_res))
4250                 return;
4251
4252         data->rpc_status = task->tk_status;
4253         if (data->arg.new_lock_owner != 0) {
4254                 if (data->rpc_status == 0)
4255                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4256                 else
4257                         goto out;
4258         }
4259         if (data->rpc_status == 0) {
4260                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4261                                         sizeof(data->lsp->ls_stateid.data));
4262                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4263                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
4264         }
4265 out:
4266         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4267 }
4268
4269 static void nfs4_lock_release(void *calldata)
4270 {
4271         struct nfs4_lockdata *data = calldata;
4272
4273         dprintk("%s: begin!\n", __func__);
4274         nfs_free_seqid(data->arg.open_seqid);
4275         if (data->cancelled != 0) {
4276                 struct rpc_task *task;
4277                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4278                                 data->arg.lock_seqid);
4279                 if (!IS_ERR(task))
4280                         rpc_put_task_async(task);
4281                 dprintk("%s: cancelling lock!\n", __func__);
4282         } else
4283                 nfs_free_seqid(data->arg.lock_seqid);
4284         nfs4_put_lock_state(data->lsp);
4285         put_nfs_open_context(data->ctx);
4286         kfree(data);
4287         dprintk("%s: done!\n", __func__);
4288 }
4289
4290 static const struct rpc_call_ops nfs4_lock_ops = {
4291         .rpc_call_prepare = nfs4_lock_prepare,
4292         .rpc_call_done = nfs4_lock_done,
4293         .rpc_release = nfs4_lock_release,
4294 };
4295
4296 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4297         .rpc_call_prepare = nfs4_recover_lock_prepare,
4298         .rpc_call_done = nfs4_lock_done,
4299         .rpc_release = nfs4_lock_release,
4300 };
4301
4302 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4303 {
4304         switch (error) {
4305         case -NFS4ERR_ADMIN_REVOKED:
4306         case -NFS4ERR_BAD_STATEID:
4307                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4308                 if (new_lock_owner != 0 ||
4309                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4310                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4311                 break;
4312         case -NFS4ERR_STALE_STATEID:
4313                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4314         case -NFS4ERR_EXPIRED:
4315                 nfs4_schedule_lease_recovery(server->nfs_client);
4316         };
4317 }
4318
4319 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4320 {
4321         struct nfs4_lockdata *data;
4322         struct rpc_task *task;
4323         struct rpc_message msg = {
4324                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4325                 .rpc_cred = state->owner->so_cred,
4326         };
4327         struct rpc_task_setup task_setup_data = {
4328                 .rpc_client = NFS_CLIENT(state->inode),
4329                 .rpc_message = &msg,
4330                 .callback_ops = &nfs4_lock_ops,
4331                 .workqueue = nfsiod_workqueue,
4332                 .flags = RPC_TASK_ASYNC,
4333         };
4334         int ret;
4335
4336         dprintk("%s: begin!\n", __func__);
4337         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4338                         fl->fl_u.nfs4_fl.owner,
4339                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4340         if (data == NULL)
4341                 return -ENOMEM;
4342         if (IS_SETLKW(cmd))
4343                 data->arg.block = 1;
4344         if (recovery_type > NFS_LOCK_NEW) {
4345                 if (recovery_type == NFS_LOCK_RECLAIM)
4346                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4347                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4348         }
4349         msg.rpc_argp = &data->arg;
4350         msg.rpc_resp = &data->res;
4351         task_setup_data.callback_data = data;
4352         task = rpc_run_task(&task_setup_data);
4353         if (IS_ERR(task))
4354                 return PTR_ERR(task);
4355         ret = nfs4_wait_for_completion_rpc_task(task);
4356         if (ret == 0) {
4357                 ret = data->rpc_status;
4358                 if (ret)
4359                         nfs4_handle_setlk_error(data->server, data->lsp,
4360                                         data->arg.new_lock_owner, ret);
4361         } else
4362                 data->cancelled = 1;
4363         rpc_put_task(task);
4364         dprintk("%s: done, ret = %d!\n", __func__, ret);
4365         return ret;
4366 }
4367
4368 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4369 {
4370         struct nfs_server *server = NFS_SERVER(state->inode);
4371         struct nfs4_exception exception = { };
4372         int err;
4373
4374         do {
4375                 /* Cache the lock if possible... */
4376                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4377                         return 0;
4378                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4379                 if (err != -NFS4ERR_DELAY)
4380                         break;
4381                 nfs4_handle_exception(server, err, &exception);
4382         } while (exception.retry);
4383         return err;
4384 }
4385
4386 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4387 {
4388         struct nfs_server *server = NFS_SERVER(state->inode);
4389         struct nfs4_exception exception = { };
4390         int err;
4391
4392         err = nfs4_set_lock_state(state, request);
4393         if (err != 0)
4394                 return err;
4395         do {
4396                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4397                         return 0;
4398                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4399                 switch (err) {
4400                 default:
4401                         goto out;
4402                 case -NFS4ERR_GRACE:
4403                 case -NFS4ERR_DELAY:
4404                         nfs4_handle_exception(server, err, &exception);
4405                         err = 0;
4406                 }
4407         } while (exception.retry);
4408 out:
4409         return err;
4410 }
4411
4412 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4413 {
4414         struct nfs_inode *nfsi = NFS_I(state->inode);
4415         unsigned char fl_flags = request->fl_flags;
4416         int status = -ENOLCK;
4417
4418         if ((fl_flags & FL_POSIX) &&
4419                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4420                 goto out;
4421         /* Is this a delegated open? */
4422         status = nfs4_set_lock_state(state, request);
4423         if (status != 0)
4424                 goto out;
4425         request->fl_flags |= FL_ACCESS;
4426         status = do_vfs_lock(request->fl_file, request);
4427         if (status < 0)
4428                 goto out;
4429         down_read(&nfsi->rwsem);
4430         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4431                 /* Yes: cache locks! */
4432                 /* ...but avoid races with delegation recall... */
4433                 request->fl_flags = fl_flags & ~FL_SLEEP;
4434                 status = do_vfs_lock(request->fl_file, request);
4435                 goto out_unlock;
4436         }
4437         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4438         if (status != 0)
4439                 goto out_unlock;
4440         /* Note: we always want to sleep here! */
4441         request->fl_flags = fl_flags | FL_SLEEP;
4442         if (do_vfs_lock(request->fl_file, request) < 0)
4443                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4444 out_unlock:
4445         up_read(&nfsi->rwsem);
4446 out:
4447         request->fl_flags = fl_flags;
4448         return status;
4449 }
4450
4451 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4452 {
4453         struct nfs4_exception exception = { };
4454         int err;
4455
4456         do {
4457                 err = _nfs4_proc_setlk(state, cmd, request);
4458                 if (err == -NFS4ERR_DENIED)
4459                         err = -EAGAIN;
4460                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4461                                 err, &exception);
4462         } while (exception.retry);
4463         return err;
4464 }
4465
4466 static int
4467 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4468 {
4469         struct nfs_open_context *ctx;
4470         struct nfs4_state *state;
4471         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4472         int status;
4473
4474         /* verify open state */
4475         ctx = nfs_file_open_context(filp);
4476         state = ctx->state;
4477
4478         if (request->fl_start < 0 || request->fl_end < 0)
4479                 return -EINVAL;
4480
4481         if (IS_GETLK(cmd)) {
4482                 if (state != NULL)
4483                         return nfs4_proc_getlk(state, F_GETLK, request);
4484                 return 0;
4485         }
4486
4487         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4488                 return -EINVAL;
4489
4490         if (request->fl_type == F_UNLCK) {
4491                 if (state != NULL)
4492                         return nfs4_proc_unlck(state, cmd, request);
4493                 return 0;
4494         }
4495
4496         if (state == NULL)
4497                 return -ENOLCK;
4498         do {
4499                 status = nfs4_proc_setlk(state, cmd, request);
4500                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4501                         break;
4502                 timeout = nfs4_set_lock_task_retry(timeout);
4503                 status = -ERESTARTSYS;
4504                 if (signalled())
4505                         break;
4506         } while(status < 0);
4507         return status;
4508 }
4509
4510 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4511 {
4512         struct nfs_server *server = NFS_SERVER(state->inode);
4513         struct nfs4_exception exception = { };
4514         int err;
4515
4516         err = nfs4_set_lock_state(state, fl);
4517         if (err != 0)
4518                 goto out;
4519         do {
4520                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4521                 switch (err) {
4522                         default:
4523                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4524                                                 __func__, err);
4525                         case 0:
4526                         case -ESTALE:
4527                                 goto out;
4528                         case -NFS4ERR_EXPIRED:
4529                         case -NFS4ERR_STALE_CLIENTID:
4530                         case -NFS4ERR_STALE_STATEID:
4531                                 nfs4_schedule_lease_recovery(server->nfs_client);
4532                                 goto out;
4533                         case -NFS4ERR_BADSESSION:
4534                         case -NFS4ERR_BADSLOT:
4535                         case -NFS4ERR_BAD_HIGH_SLOT:
4536                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4537                         case -NFS4ERR_DEADSESSION:
4538                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4539                                 goto out;
4540                         case -ERESTARTSYS:
4541                                 /*
4542                                  * The show must go on: exit, but mark the
4543                                  * stateid as needing recovery.
4544                                  */
4545                         case -NFS4ERR_ADMIN_REVOKED:
4546                         case -NFS4ERR_BAD_STATEID:
4547                         case -NFS4ERR_OPENMODE:
4548                                 nfs4_schedule_stateid_recovery(server, state);
4549                                 err = 0;
4550                                 goto out;
4551                         case -EKEYEXPIRED:
4552                                 /*
4553                                  * User RPCSEC_GSS context has expired.
4554                                  * We cannot recover this stateid now, so
4555                                  * skip it and allow recovery thread to
4556                                  * proceed.
4557                                  */
4558                                 err = 0;
4559                                 goto out;
4560                         case -ENOMEM:
4561                         case -NFS4ERR_DENIED:
4562                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4563                                 err = 0;
4564                                 goto out;
4565                         case -NFS4ERR_DELAY:
4566                                 break;
4567                 }
4568                 err = nfs4_handle_exception(server, err, &exception);
4569         } while (exception.retry);
4570 out:
4571         return err;
4572 }
4573
4574 static void nfs4_release_lockowner_release(void *calldata)
4575 {
4576         kfree(calldata);
4577 }
4578
4579 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4580         .rpc_release = nfs4_release_lockowner_release,
4581 };
4582
4583 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4584 {
4585         struct nfs_server *server = lsp->ls_state->owner->so_server;
4586         struct nfs_release_lockowner_args *args;
4587         struct rpc_message msg = {
4588                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4589         };
4590
4591         if (server->nfs_client->cl_mvops->minor_version != 0)
4592                 return;
4593         args = kmalloc(sizeof(*args), GFP_NOFS);
4594         if (!args)
4595                 return;
4596         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4597         args->lock_owner.id = lsp->ls_id.id;
4598         args->lock_owner.s_dev = server->s_dev;
4599         msg.rpc_argp = args;
4600         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4601 }
4602
4603 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4604
4605 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4606                                    const void *buf, size_t buflen,
4607                                    int flags, int type)
4608 {
4609         if (strcmp(key, "") != 0)
4610                 return -EINVAL;
4611
4612         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4613 }
4614
4615 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4616                                    void *buf, size_t buflen, int type)
4617 {
4618         if (strcmp(key, "") != 0)
4619                 return -EINVAL;
4620
4621         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4622 }
4623
4624 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4625                                        size_t list_len, const char *name,
4626                                        size_t name_len, int type)
4627 {
4628         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4629
4630         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4631                 return 0;
4632
4633         if (list && len <= list_len)
4634                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4635         return len;
4636 }
4637
4638 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4639 {
4640         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4641                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4642                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4643                 return;
4644
4645         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4646                 NFS_ATTR_FATTR_NLINK;
4647         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4648         fattr->nlink = 2;
4649 }
4650
4651 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4652                 struct nfs4_fs_locations *fs_locations, struct page *page)
4653 {
4654         struct nfs_server *server = NFS_SERVER(dir);
4655         u32 bitmask[2] = {
4656                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4657                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4658         };
4659         struct nfs4_fs_locations_arg args = {
4660                 .dir_fh = NFS_FH(dir),
4661                 .name = name,
4662                 .page = page,
4663                 .bitmask = bitmask,
4664         };
4665         struct nfs4_fs_locations_res res = {
4666                 .fs_locations = fs_locations,
4667         };
4668         struct rpc_message msg = {
4669                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4670                 .rpc_argp = &args,
4671                 .rpc_resp = &res,
4672         };
4673         int status;
4674
4675         dprintk("%s: start\n", __func__);
4676         nfs_fattr_init(&fs_locations->fattr);
4677         fs_locations->server = server;
4678         fs_locations->nlocations = 0;
4679         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4680         nfs_fixup_referral_attributes(&fs_locations->fattr);
4681         dprintk("%s: returned status = %d\n", __func__, status);
4682         return status;
4683 }
4684
4685 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4686 {
4687         int status;
4688         struct nfs4_secinfo_arg args = {
4689                 .dir_fh = NFS_FH(dir),
4690                 .name   = name,
4691         };
4692         struct nfs4_secinfo_res res = {
4693                 .flavors     = flavors,
4694         };
4695         struct rpc_message msg = {
4696                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4697                 .rpc_argp = &args,
4698                 .rpc_resp = &res,
4699         };
4700
4701         dprintk("NFS call  secinfo %s\n", name->name);
4702         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4703         dprintk("NFS reply  secinfo: %d\n", status);
4704         return status;
4705 }
4706
4707 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4708 {
4709         struct nfs4_exception exception = { };
4710         int err;
4711         do {
4712                 err = nfs4_handle_exception(NFS_SERVER(dir),
4713                                 _nfs4_proc_secinfo(dir, name, flavors),
4714                                 &exception);
4715         } while (exception.retry);
4716         return err;
4717 }
4718
4719 #ifdef CONFIG_NFS_V4_1
4720 /*
4721  * Check the exchange flags returned by the server for invalid flags, having
4722  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4723  * DS flags set.
4724  */
4725 static int nfs4_check_cl_exchange_flags(u32 flags)
4726 {
4727         if (flags & ~EXCHGID4_FLAG_MASK_R)
4728                 goto out_inval;
4729         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4730             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4731                 goto out_inval;
4732         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4733                 goto out_inval;
4734         return NFS_OK;
4735 out_inval:
4736         return -NFS4ERR_INVAL;
4737 }
4738
4739 /*
4740  * nfs4_proc_exchange_id()
4741  *
4742  * Since the clientid has expired, all compounds using sessions
4743  * associated with the stale clientid will be returning
4744  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4745  * be in some phase of session reset.
4746  */
4747 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4748 {
4749         nfs4_verifier verifier;
4750         struct nfs41_exchange_id_args args = {
4751                 .client = clp,
4752                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4753         };
4754         struct nfs41_exchange_id_res res = {
4755                 .client = clp,
4756         };
4757         int status;
4758         struct rpc_message msg = {
4759                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4760                 .rpc_argp = &args,
4761                 .rpc_resp = &res,
4762                 .rpc_cred = cred,
4763         };
4764         __be32 *p;
4765
4766         dprintk("--> %s\n", __func__);
4767         BUG_ON(clp == NULL);
4768
4769         p = (u32 *)verifier.data;
4770         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4771         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4772         args.verifier = &verifier;
4773
4774         args.id_len = scnprintf(args.id, sizeof(args.id),
4775                                 "%s/%s.%s/%u",
4776                                 clp->cl_ipaddr,
4777                                 init_utsname()->nodename,
4778                                 init_utsname()->domainname,
4779                                 clp->cl_rpcclient->cl_auth->au_flavor);
4780
4781         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
4782         if (!status)
4783                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4784         dprintk("<-- %s status= %d\n", __func__, status);
4785         return status;
4786 }
4787
4788 struct nfs4_get_lease_time_data {
4789         struct nfs4_get_lease_time_args *args;
4790         struct nfs4_get_lease_time_res *res;
4791         struct nfs_client *clp;
4792 };
4793
4794 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4795                                         void *calldata)
4796 {
4797         int ret;
4798         struct nfs4_get_lease_time_data *data =
4799                         (struct nfs4_get_lease_time_data *)calldata;
4800
4801         dprintk("--> %s\n", __func__);
4802         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4803         /* just setup sequence, do not trigger session recovery
4804            since we're invoked within one */
4805         ret = nfs41_setup_sequence(data->clp->cl_session,
4806                                    &data->args->la_seq_args,
4807                                    &data->res->lr_seq_res, 0, task);
4808
4809         BUG_ON(ret == -EAGAIN);
4810         rpc_call_start(task);
4811         dprintk("<-- %s\n", __func__);
4812 }
4813
4814 /*
4815  * Called from nfs4_state_manager thread for session setup, so don't recover
4816  * from sequence operation or clientid errors.
4817  */
4818 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4819 {
4820         struct nfs4_get_lease_time_data *data =
4821                         (struct nfs4_get_lease_time_data *)calldata;
4822
4823         dprintk("--> %s\n", __func__);
4824         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
4825                 return;
4826         switch (task->tk_status) {
4827         case -NFS4ERR_DELAY:
4828         case -NFS4ERR_GRACE:
4829                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4830                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4831                 task->tk_status = 0;
4832                 nfs_restart_rpc(task, data->clp);
4833                 return;
4834         }
4835         dprintk("<-- %s\n", __func__);
4836 }
4837
4838 struct rpc_call_ops nfs4_get_lease_time_ops = {
4839         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4840         .rpc_call_done = nfs4_get_lease_time_done,
4841 };
4842
4843 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4844 {
4845         struct rpc_task *task;
4846         struct nfs4_get_lease_time_args args;
4847         struct nfs4_get_lease_time_res res = {
4848                 .lr_fsinfo = fsinfo,
4849         };
4850         struct nfs4_get_lease_time_data data = {
4851                 .args = &args,
4852                 .res = &res,
4853                 .clp = clp,
4854         };
4855         struct rpc_message msg = {
4856                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4857                 .rpc_argp = &args,
4858                 .rpc_resp = &res,
4859         };
4860         struct rpc_task_setup task_setup = {
4861                 .rpc_client = clp->cl_rpcclient,
4862                 .rpc_message = &msg,
4863                 .callback_ops = &nfs4_get_lease_time_ops,
4864                 .callback_data = &data
4865         };
4866         int status;
4867
4868         dprintk("--> %s\n", __func__);
4869         task = rpc_run_task(&task_setup);
4870
4871         if (IS_ERR(task))
4872                 status = PTR_ERR(task);
4873         else {
4874                 status = task->tk_status;
4875                 rpc_put_task(task);
4876         }
4877         dprintk("<-- %s return %d\n", __func__, status);
4878
4879         return status;
4880 }
4881
4882 /*
4883  * Reset a slot table
4884  */
4885 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
4886                                  int ivalue)
4887 {
4888         struct nfs4_slot *new = NULL;
4889         int i;
4890         int ret = 0;
4891
4892         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
4893                 max_reqs, tbl->max_slots);
4894
4895         /* Does the newly negotiated max_reqs match the existing slot table? */
4896         if (max_reqs != tbl->max_slots) {
4897                 ret = -ENOMEM;
4898                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
4899                               GFP_NOFS);
4900                 if (!new)
4901                         goto out;
4902                 ret = 0;
4903                 kfree(tbl->slots);
4904         }
4905         spin_lock(&tbl->slot_tbl_lock);
4906         if (new) {
4907                 tbl->slots = new;
4908                 tbl->max_slots = max_reqs;
4909         }
4910         for (i = 0; i < tbl->max_slots; ++i)
4911                 tbl->slots[i].seq_nr = ivalue;
4912         spin_unlock(&tbl->slot_tbl_lock);
4913         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4914                 tbl, tbl->slots, tbl->max_slots);
4915 out:
4916         dprintk("<-- %s: return %d\n", __func__, ret);
4917         return ret;
4918 }
4919
4920 /*
4921  * Reset the forechannel and backchannel slot tables
4922  */
4923 static int nfs4_reset_slot_tables(struct nfs4_session *session)
4924 {
4925         int status;
4926
4927         status = nfs4_reset_slot_table(&session->fc_slot_table,
4928                         session->fc_attrs.max_reqs, 1);
4929         if (status)
4930                 return status;
4931
4932         status = nfs4_reset_slot_table(&session->bc_slot_table,
4933                         session->bc_attrs.max_reqs, 0);
4934         return status;
4935 }
4936
4937 /* Destroy the slot table */
4938 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
4939 {
4940         if (session->fc_slot_table.slots != NULL) {
4941                 kfree(session->fc_slot_table.slots);
4942                 session->fc_slot_table.slots = NULL;
4943         }
4944         if (session->bc_slot_table.slots != NULL) {
4945                 kfree(session->bc_slot_table.slots);
4946                 session->bc_slot_table.slots = NULL;
4947         }
4948         return;
4949 }
4950
4951 /*
4952  * Initialize slot table
4953  */
4954 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
4955                 int max_slots, int ivalue)
4956 {
4957         struct nfs4_slot *slot;
4958         int ret = -ENOMEM;
4959
4960         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
4961
4962         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
4963
4964         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
4965         if (!slot)
4966                 goto out;
4967         ret = 0;
4968
4969         spin_lock(&tbl->slot_tbl_lock);
4970         tbl->max_slots = max_slots;
4971         tbl->slots = slot;
4972         tbl->highest_used_slotid = -1;  /* no slot is currently used */
4973         spin_unlock(&tbl->slot_tbl_lock);
4974         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4975                 tbl, tbl->slots, tbl->max_slots);
4976 out:
4977         dprintk("<-- %s: return %d\n", __func__, ret);
4978         return ret;
4979 }
4980
4981 /*
4982  * Initialize the forechannel and backchannel tables
4983  */
4984 static int nfs4_init_slot_tables(struct nfs4_session *session)
4985 {
4986         struct nfs4_slot_table *tbl;
4987         int status = 0;
4988
4989         tbl = &session->fc_slot_table;
4990         if (tbl->slots == NULL) {
4991                 status = nfs4_init_slot_table(tbl,
4992                                 session->fc_attrs.max_reqs, 1);
4993                 if (status)
4994                         return status;
4995         }
4996
4997         tbl = &session->bc_slot_table;
4998         if (tbl->slots == NULL) {
4999                 status = nfs4_init_slot_table(tbl,
5000                                 session->bc_attrs.max_reqs, 0);
5001                 if (status)
5002                         nfs4_destroy_slot_tables(session);
5003         }
5004
5005         return status;
5006 }
5007
5008 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5009 {
5010         struct nfs4_session *session;
5011         struct nfs4_slot_table *tbl;
5012
5013         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5014         if (!session)
5015                 return NULL;
5016
5017         tbl = &session->fc_slot_table;
5018         tbl->highest_used_slotid = -1;
5019         spin_lock_init(&tbl->slot_tbl_lock);
5020         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5021         init_completion(&tbl->complete);
5022
5023         tbl = &session->bc_slot_table;
5024         tbl->highest_used_slotid = -1;
5025         spin_lock_init(&tbl->slot_tbl_lock);
5026         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5027         init_completion(&tbl->complete);
5028
5029         session->session_state = 1<<NFS4_SESSION_INITING;
5030
5031         session->clp = clp;
5032         return session;
5033 }
5034
5035 void nfs4_destroy_session(struct nfs4_session *session)
5036 {
5037         nfs4_proc_destroy_session(session);
5038         dprintk("%s Destroy backchannel for xprt %p\n",
5039                 __func__, session->clp->cl_rpcclient->cl_xprt);
5040         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
5041                                 NFS41_BC_MIN_CALLBACKS);
5042         nfs4_destroy_slot_tables(session);
5043         kfree(session);
5044 }
5045
5046 /*
5047  * Initialize the values to be used by the client in CREATE_SESSION
5048  * If nfs4_init_session set the fore channel request and response sizes,
5049  * use them.
5050  *
5051  * Set the back channel max_resp_sz_cached to zero to force the client to
5052  * always set csa_cachethis to FALSE because the current implementation
5053  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5054  */
5055 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5056 {
5057         struct nfs4_session *session = args->client->cl_session;
5058         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5059                      mxresp_sz = session->fc_attrs.max_resp_sz;
5060
5061         if (mxrqst_sz == 0)
5062                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5063         if (mxresp_sz == 0)
5064                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5065         /* Fore channel attributes */
5066         args->fc_attrs.headerpadsz = 0;
5067         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5068         args->fc_attrs.max_resp_sz = mxresp_sz;
5069         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5070         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
5071
5072         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5073                 "max_ops=%u max_reqs=%u\n",
5074                 __func__,
5075                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5076                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5077
5078         /* Back channel attributes */
5079         args->bc_attrs.headerpadsz = 0;
5080         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5081         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5082         args->bc_attrs.max_resp_sz_cached = 0;
5083         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5084         args->bc_attrs.max_reqs = 1;
5085
5086         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5087                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5088                 __func__,
5089                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5090                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5091                 args->bc_attrs.max_reqs);
5092 }
5093
5094 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5095 {
5096         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5097         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5098
5099         if (rcvd->headerpadsz > sent->headerpadsz)
5100                 return -EINVAL;
5101         if (rcvd->max_resp_sz > sent->max_resp_sz)
5102                 return -EINVAL;
5103         /*
5104          * Our requested max_ops is the minimum we need; we're not
5105          * prepared to break up compounds into smaller pieces than that.
5106          * So, no point even trying to continue if the server won't
5107          * cooperate:
5108          */
5109         if (rcvd->max_ops < sent->max_ops)
5110                 return -EINVAL;
5111         if (rcvd->max_reqs == 0)
5112                 return -EINVAL;
5113         return 0;
5114 }
5115
5116 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5117 {
5118         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5119         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5120
5121         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5122                 return -EINVAL;
5123         if (rcvd->max_resp_sz < sent->max_resp_sz)
5124                 return -EINVAL;
5125         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5126                 return -EINVAL;
5127         /* These would render the backchannel useless: */
5128         if (rcvd->max_ops  == 0)
5129                 return -EINVAL;
5130         if (rcvd->max_reqs == 0)
5131                 return -EINVAL;
5132         return 0;
5133 }
5134
5135 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5136                                      struct nfs4_session *session)
5137 {
5138         int ret;
5139
5140         ret = nfs4_verify_fore_channel_attrs(args, session);
5141         if (ret)
5142                 return ret;
5143         return nfs4_verify_back_channel_attrs(args, session);
5144 }
5145
5146 static int _nfs4_proc_create_session(struct nfs_client *clp)
5147 {
5148         struct nfs4_session *session = clp->cl_session;
5149         struct nfs41_create_session_args args = {
5150                 .client = clp,
5151                 .cb_program = NFS4_CALLBACK,
5152         };
5153         struct nfs41_create_session_res res = {
5154                 .client = clp,
5155         };
5156         struct rpc_message msg = {
5157                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5158                 .rpc_argp = &args,
5159                 .rpc_resp = &res,
5160         };
5161         int status;
5162
5163         nfs4_init_channel_attrs(&args);
5164         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5165
5166         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5167
5168         if (!status)
5169                 /* Verify the session's negotiated channel_attrs values */
5170                 status = nfs4_verify_channel_attrs(&args, session);
5171         if (!status) {
5172                 /* Increment the clientid slot sequence id */
5173                 clp->cl_seqid++;
5174         }
5175
5176         return status;
5177 }
5178
5179 /*
5180  * Issues a CREATE_SESSION operation to the server.
5181  * It is the responsibility of the caller to verify the session is
5182  * expired before calling this routine.
5183  */
5184 int nfs4_proc_create_session(struct nfs_client *clp)
5185 {
5186         int status;
5187         unsigned *ptr;
5188         struct nfs4_session *session = clp->cl_session;
5189         long timeout = 0;
5190         int err;
5191
5192         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5193
5194         do {
5195                 status = _nfs4_proc_create_session(clp);
5196                 if (status == -NFS4ERR_DELAY) {
5197                         err = nfs4_delay(clp->cl_rpcclient, &timeout);
5198                         if (err)
5199                                 status = err;
5200                 }
5201         } while (status == -NFS4ERR_DELAY);
5202
5203         if (status)
5204                 goto out;
5205
5206         /* Init and reset the fore channel */
5207         status = nfs4_init_slot_tables(session);
5208         dprintk("slot table initialization returned %d\n", status);
5209         if (status)
5210                 goto out;
5211         status = nfs4_reset_slot_tables(session);
5212         dprintk("slot table reset returned %d\n", status);
5213         if (status)
5214                 goto out;
5215
5216         ptr = (unsigned *)&session->sess_id.data[0];
5217         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5218                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5219 out:
5220         dprintk("<-- %s\n", __func__);
5221         return status;
5222 }
5223
5224 /*
5225  * Issue the over-the-wire RPC DESTROY_SESSION.
5226  * The caller must serialize access to this routine.
5227  */
5228 int nfs4_proc_destroy_session(struct nfs4_session *session)
5229 {
5230         int status = 0;
5231         struct rpc_message msg;
5232
5233         dprintk("--> nfs4_proc_destroy_session\n");
5234
5235         /* session is still being setup */
5236         if (session->clp->cl_cons_state != NFS_CS_READY)
5237                 return status;
5238
5239         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5240         msg.rpc_argp = session;
5241         msg.rpc_resp = NULL;
5242         msg.rpc_cred = NULL;
5243         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5244
5245         if (status)
5246                 printk(KERN_WARNING
5247                         "Got error %d from the server on DESTROY_SESSION. "
5248                         "Session has been destroyed regardless...\n", status);
5249
5250         dprintk("<-- nfs4_proc_destroy_session\n");
5251         return status;
5252 }
5253
5254 int nfs4_init_session(struct nfs_server *server)
5255 {
5256         struct nfs_client *clp = server->nfs_client;
5257         struct nfs4_session *session;
5258         unsigned int rsize, wsize;
5259         int ret;
5260
5261         if (!nfs4_has_session(clp))
5262                 return 0;
5263
5264         session = clp->cl_session;
5265         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5266                 return 0;
5267
5268         rsize = server->rsize;
5269         if (rsize == 0)
5270                 rsize = NFS_MAX_FILE_IO_SIZE;
5271         wsize = server->wsize;
5272         if (wsize == 0)
5273                 wsize = NFS_MAX_FILE_IO_SIZE;
5274
5275         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5276         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5277
5278         ret = nfs4_recover_expired_lease(server);
5279         if (!ret)
5280                 ret = nfs4_check_client_ready(clp);
5281         return ret;
5282 }
5283
5284 int nfs4_init_ds_session(struct nfs_client *clp)
5285 {
5286         struct nfs4_session *session = clp->cl_session;
5287         int ret;
5288
5289         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5290                 return 0;
5291
5292         ret = nfs4_client_recover_expired_lease(clp);
5293         if (!ret)
5294                 /* Test for the DS role */
5295                 if (!is_ds_client(clp))
5296                         ret = -ENODEV;
5297         if (!ret)
5298                 ret = nfs4_check_client_ready(clp);
5299         return ret;
5300
5301 }
5302 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5303
5304
5305 /*
5306  * Renew the cl_session lease.
5307  */
5308 struct nfs4_sequence_data {
5309         struct nfs_client *clp;
5310         struct nfs4_sequence_args args;
5311         struct nfs4_sequence_res res;
5312 };
5313
5314 static void nfs41_sequence_release(void *data)
5315 {
5316         struct nfs4_sequence_data *calldata = data;
5317         struct nfs_client *clp = calldata->clp;
5318
5319         if (atomic_read(&clp->cl_count) > 1)
5320                 nfs4_schedule_state_renewal(clp);
5321         nfs_put_client(clp);
5322         kfree(calldata);
5323 }
5324
5325 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5326 {
5327         switch(task->tk_status) {
5328         case -NFS4ERR_DELAY:
5329                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5330                 return -EAGAIN;
5331         default:
5332                 nfs4_schedule_lease_recovery(clp);
5333         }
5334         return 0;
5335 }
5336
5337 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5338 {
5339         struct nfs4_sequence_data *calldata = data;
5340         struct nfs_client *clp = calldata->clp;
5341
5342         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5343                 return;
5344
5345         if (task->tk_status < 0) {
5346                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5347                 if (atomic_read(&clp->cl_count) == 1)
5348                         goto out;
5349
5350                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5351                         rpc_restart_call_prepare(task);
5352                         return;
5353                 }
5354         }
5355         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5356 out:
5357         dprintk("<-- %s\n", __func__);
5358 }
5359
5360 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5361 {
5362         struct nfs4_sequence_data *calldata = data;
5363         struct nfs_client *clp = calldata->clp;
5364         struct nfs4_sequence_args *args;
5365         struct nfs4_sequence_res *res;
5366
5367         args = task->tk_msg.rpc_argp;
5368         res = task->tk_msg.rpc_resp;
5369
5370         if (nfs41_setup_sequence(clp->cl_session, args, res, 0, task))
5371                 return;
5372         rpc_call_start(task);
5373 }
5374
5375 static const struct rpc_call_ops nfs41_sequence_ops = {
5376         .rpc_call_done = nfs41_sequence_call_done,
5377         .rpc_call_prepare = nfs41_sequence_prepare,
5378         .rpc_release = nfs41_sequence_release,
5379 };
5380
5381 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5382 {
5383         struct nfs4_sequence_data *calldata;
5384         struct rpc_message msg = {
5385                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5386                 .rpc_cred = cred,
5387         };
5388         struct rpc_task_setup task_setup_data = {
5389                 .rpc_client = clp->cl_rpcclient,
5390                 .rpc_message = &msg,
5391                 .callback_ops = &nfs41_sequence_ops,
5392                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5393         };
5394
5395         if (!atomic_inc_not_zero(&clp->cl_count))
5396                 return ERR_PTR(-EIO);
5397         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5398         if (calldata == NULL) {
5399                 nfs_put_client(clp);
5400                 return ERR_PTR(-ENOMEM);
5401         }
5402         msg.rpc_argp = &calldata->args;
5403         msg.rpc_resp = &calldata->res;
5404         calldata->clp = clp;
5405         task_setup_data.callback_data = calldata;
5406
5407         return rpc_run_task(&task_setup_data);
5408 }
5409
5410 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5411 {
5412         struct rpc_task *task;
5413         int ret = 0;
5414
5415         task = _nfs41_proc_sequence(clp, cred);
5416         if (IS_ERR(task))
5417                 ret = PTR_ERR(task);
5418         else
5419                 rpc_put_task_async(task);
5420         dprintk("<-- %s status=%d\n", __func__, ret);
5421         return ret;
5422 }
5423
5424 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5425 {
5426         struct rpc_task *task;
5427         int ret;
5428
5429         task = _nfs41_proc_sequence(clp, cred);
5430         if (IS_ERR(task)) {
5431                 ret = PTR_ERR(task);
5432                 goto out;
5433         }
5434         ret = rpc_wait_for_completion_task(task);
5435         if (!ret) {
5436                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5437
5438                 if (task->tk_status == 0)
5439                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5440                 ret = task->tk_status;
5441         }
5442         rpc_put_task(task);
5443 out:
5444         dprintk("<-- %s status=%d\n", __func__, ret);
5445         return ret;
5446 }
5447
5448 struct nfs4_reclaim_complete_data {
5449         struct nfs_client *clp;
5450         struct nfs41_reclaim_complete_args arg;
5451         struct nfs41_reclaim_complete_res res;
5452 };
5453
5454 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5455 {
5456         struct nfs4_reclaim_complete_data *calldata = data;
5457
5458         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5459         if (nfs41_setup_sequence(calldata->clp->cl_session,
5460                                 &calldata->arg.seq_args,
5461                                 &calldata->res.seq_res, 0, task))
5462                 return;
5463
5464         rpc_call_start(task);
5465 }
5466
5467 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5468 {
5469         switch(task->tk_status) {
5470         case 0:
5471         case -NFS4ERR_COMPLETE_ALREADY:
5472         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5473                 break;
5474         case -NFS4ERR_DELAY:
5475                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5476                 return -EAGAIN;
5477         default:
5478                 nfs4_schedule_lease_recovery(clp);
5479         }
5480         return 0;
5481 }
5482
5483 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5484 {
5485         struct nfs4_reclaim_complete_data *calldata = data;
5486         struct nfs_client *clp = calldata->clp;
5487         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5488
5489         dprintk("--> %s\n", __func__);
5490         if (!nfs41_sequence_done(task, res))
5491                 return;
5492
5493         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5494                 rpc_restart_call_prepare(task);
5495                 return;
5496         }
5497         dprintk("<-- %s\n", __func__);
5498 }
5499
5500 static void nfs4_free_reclaim_complete_data(void *data)
5501 {
5502         struct nfs4_reclaim_complete_data *calldata = data;
5503
5504         kfree(calldata);
5505 }
5506
5507 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5508         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5509         .rpc_call_done = nfs4_reclaim_complete_done,
5510         .rpc_release = nfs4_free_reclaim_complete_data,
5511 };
5512
5513 /*
5514  * Issue a global reclaim complete.
5515  */
5516 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5517 {
5518         struct nfs4_reclaim_complete_data *calldata;
5519         struct rpc_task *task;
5520         struct rpc_message msg = {
5521                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5522         };
5523         struct rpc_task_setup task_setup_data = {
5524                 .rpc_client = clp->cl_rpcclient,
5525                 .rpc_message = &msg,
5526                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5527                 .flags = RPC_TASK_ASYNC,
5528         };
5529         int status = -ENOMEM;
5530
5531         dprintk("--> %s\n", __func__);
5532         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5533         if (calldata == NULL)
5534                 goto out;
5535         calldata->clp = clp;
5536         calldata->arg.one_fs = 0;
5537
5538         msg.rpc_argp = &calldata->arg;
5539         msg.rpc_resp = &calldata->res;
5540         task_setup_data.callback_data = calldata;
5541         task = rpc_run_task(&task_setup_data);
5542         if (IS_ERR(task)) {
5543                 status = PTR_ERR(task);
5544                 goto out;
5545         }
5546         status = nfs4_wait_for_completion_rpc_task(task);
5547         if (status == 0)
5548                 status = task->tk_status;
5549         rpc_put_task(task);
5550         return 0;
5551 out:
5552         dprintk("<-- %s status=%d\n", __func__, status);
5553         return status;
5554 }
5555
5556 static void
5557 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5558 {
5559         struct nfs4_layoutget *lgp = calldata;
5560         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5561
5562         dprintk("--> %s\n", __func__);
5563         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5564          * right now covering the LAYOUTGET we are about to send.
5565          * However, that is not so catastrophic, and there seems
5566          * to be no way to prevent it completely.
5567          */
5568         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5569                                 &lgp->res.seq_res, 0, task))
5570                 return;
5571         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5572                                           NFS_I(lgp->args.inode)->layout,
5573                                           lgp->args.ctx->state)) {
5574                 rpc_exit(task, NFS4_OK);
5575                 return;
5576         }
5577         rpc_call_start(task);
5578 }
5579
5580 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5581 {
5582         struct nfs4_layoutget *lgp = calldata;
5583         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5584
5585         dprintk("--> %s\n", __func__);
5586
5587         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5588                 return;
5589
5590         switch (task->tk_status) {
5591         case 0:
5592                 break;
5593         case -NFS4ERR_LAYOUTTRYLATER:
5594         case -NFS4ERR_RECALLCONFLICT:
5595                 task->tk_status = -NFS4ERR_DELAY;
5596                 /* Fall through */
5597         default:
5598                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5599                         rpc_restart_call_prepare(task);
5600                         return;
5601                 }
5602         }
5603         dprintk("<-- %s\n", __func__);
5604 }
5605
5606 static void nfs4_layoutget_release(void *calldata)
5607 {
5608         struct nfs4_layoutget *lgp = calldata;
5609
5610         dprintk("--> %s\n", __func__);
5611         if (lgp->res.layout.buf != NULL)
5612                 free_page((unsigned long) lgp->res.layout.buf);
5613         put_nfs_open_context(lgp->args.ctx);
5614         kfree(calldata);
5615         dprintk("<-- %s\n", __func__);
5616 }
5617
5618 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5619         .rpc_call_prepare = nfs4_layoutget_prepare,
5620         .rpc_call_done = nfs4_layoutget_done,
5621         .rpc_release = nfs4_layoutget_release,
5622 };
5623
5624 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5625 {
5626         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5627         struct rpc_task *task;
5628         struct rpc_message msg = {
5629                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5630                 .rpc_argp = &lgp->args,
5631                 .rpc_resp = &lgp->res,
5632         };
5633         struct rpc_task_setup task_setup_data = {
5634                 .rpc_client = server->client,
5635                 .rpc_message = &msg,
5636                 .callback_ops = &nfs4_layoutget_call_ops,
5637                 .callback_data = lgp,
5638                 .flags = RPC_TASK_ASYNC,
5639         };
5640         int status = 0;
5641
5642         dprintk("--> %s\n", __func__);
5643
5644         lgp->res.layout.buf = (void *)__get_free_page(GFP_NOFS);
5645         if (lgp->res.layout.buf == NULL) {
5646                 nfs4_layoutget_release(lgp);
5647                 return -ENOMEM;
5648         }
5649
5650         lgp->res.seq_res.sr_slot = NULL;
5651         task = rpc_run_task(&task_setup_data);
5652         if (IS_ERR(task))
5653                 return PTR_ERR(task);
5654         status = nfs4_wait_for_completion_rpc_task(task);
5655         if (status == 0)
5656                 status = task->tk_status;
5657         if (status == 0)
5658                 status = pnfs_layout_process(lgp);
5659         rpc_put_task(task);
5660         dprintk("<-- %s status=%d\n", __func__, status);
5661         return status;
5662 }
5663
5664 static int
5665 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5666 {
5667         struct nfs4_getdeviceinfo_args args = {
5668                 .pdev = pdev,
5669         };
5670         struct nfs4_getdeviceinfo_res res = {
5671                 .pdev = pdev,
5672         };
5673         struct rpc_message msg = {
5674                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5675                 .rpc_argp = &args,
5676                 .rpc_resp = &res,
5677         };
5678         int status;
5679
5680         dprintk("--> %s\n", __func__);
5681         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5682         dprintk("<-- %s status=%d\n", __func__, status);
5683
5684         return status;
5685 }
5686
5687 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5688 {
5689         struct nfs4_exception exception = { };
5690         int err;
5691
5692         do {
5693                 err = nfs4_handle_exception(server,
5694                                         _nfs4_proc_getdeviceinfo(server, pdev),
5695                                         &exception);
5696         } while (exception.retry);
5697         return err;
5698 }
5699 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5700
5701 #endif /* CONFIG_NFS_V4_1 */
5702
5703 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
5704         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5705         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5706         .recover_open   = nfs4_open_reclaim,
5707         .recover_lock   = nfs4_lock_reclaim,
5708         .establish_clid = nfs4_init_clientid,
5709         .get_clid_cred  = nfs4_get_setclientid_cred,
5710 };
5711
5712 #if defined(CONFIG_NFS_V4_1)
5713 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
5714         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5715         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5716         .recover_open   = nfs4_open_reclaim,
5717         .recover_lock   = nfs4_lock_reclaim,
5718         .establish_clid = nfs41_init_clientid,
5719         .get_clid_cred  = nfs4_get_exchange_id_cred,
5720         .reclaim_complete = nfs41_proc_reclaim_complete,
5721 };
5722 #endif /* CONFIG_NFS_V4_1 */
5723
5724 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
5725         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5726         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5727         .recover_open   = nfs4_open_expired,
5728         .recover_lock   = nfs4_lock_expired,
5729         .establish_clid = nfs4_init_clientid,
5730         .get_clid_cred  = nfs4_get_setclientid_cred,
5731 };
5732
5733 #if defined(CONFIG_NFS_V4_1)
5734 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
5735         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5736         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5737         .recover_open   = nfs4_open_expired,
5738         .recover_lock   = nfs4_lock_expired,
5739         .establish_clid = nfs41_init_clientid,
5740         .get_clid_cred  = nfs4_get_exchange_id_cred,
5741 };
5742 #endif /* CONFIG_NFS_V4_1 */
5743
5744 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
5745         .sched_state_renewal = nfs4_proc_async_renew,
5746         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
5747         .renew_lease = nfs4_proc_renew,
5748 };
5749
5750 #if defined(CONFIG_NFS_V4_1)
5751 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
5752         .sched_state_renewal = nfs41_proc_async_sequence,
5753         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
5754         .renew_lease = nfs4_proc_sequence,
5755 };
5756 #endif
5757
5758 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
5759         .minor_version = 0,
5760         .call_sync = _nfs4_call_sync,
5761         .validate_stateid = nfs4_validate_delegation_stateid,
5762         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
5763         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
5764         .state_renewal_ops = &nfs40_state_renewal_ops,
5765 };
5766
5767 #if defined(CONFIG_NFS_V4_1)
5768 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
5769         .minor_version = 1,
5770         .call_sync = _nfs4_call_sync_session,
5771         .validate_stateid = nfs41_validate_delegation_stateid,
5772         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
5773         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
5774         .state_renewal_ops = &nfs41_state_renewal_ops,
5775 };
5776 #endif
5777
5778 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
5779         [0] = &nfs_v4_0_minor_ops,
5780 #if defined(CONFIG_NFS_V4_1)
5781         [1] = &nfs_v4_1_minor_ops,
5782 #endif
5783 };
5784
5785 static const struct inode_operations nfs4_file_inode_operations = {
5786         .permission     = nfs_permission,
5787         .getattr        = nfs_getattr,
5788         .setattr        = nfs_setattr,
5789         .getxattr       = generic_getxattr,
5790         .setxattr       = generic_setxattr,
5791         .listxattr      = generic_listxattr,
5792         .removexattr    = generic_removexattr,
5793 };
5794
5795 const struct nfs_rpc_ops nfs_v4_clientops = {
5796         .version        = 4,                    /* protocol version */
5797         .dentry_ops     = &nfs4_dentry_operations,
5798         .dir_inode_ops  = &nfs4_dir_inode_operations,
5799         .file_inode_ops = &nfs4_file_inode_operations,
5800         .getroot        = nfs4_proc_get_root,
5801         .getattr        = nfs4_proc_getattr,
5802         .setattr        = nfs4_proc_setattr,
5803         .lookupfh       = nfs4_proc_lookupfh,
5804         .lookup         = nfs4_proc_lookup,
5805         .access         = nfs4_proc_access,
5806         .readlink       = nfs4_proc_readlink,
5807         .create         = nfs4_proc_create,
5808         .remove         = nfs4_proc_remove,
5809         .unlink_setup   = nfs4_proc_unlink_setup,
5810         .unlink_done    = nfs4_proc_unlink_done,
5811         .rename         = nfs4_proc_rename,
5812         .rename_setup   = nfs4_proc_rename_setup,
5813         .rename_done    = nfs4_proc_rename_done,
5814         .link           = nfs4_proc_link,
5815         .symlink        = nfs4_proc_symlink,
5816         .mkdir          = nfs4_proc_mkdir,
5817         .rmdir          = nfs4_proc_remove,
5818         .readdir        = nfs4_proc_readdir,
5819         .mknod          = nfs4_proc_mknod,
5820         .statfs         = nfs4_proc_statfs,
5821         .fsinfo         = nfs4_proc_fsinfo,
5822         .pathconf       = nfs4_proc_pathconf,
5823         .set_capabilities = nfs4_server_capabilities,
5824         .decode_dirent  = nfs4_decode_dirent,
5825         .read_setup     = nfs4_proc_read_setup,
5826         .read_done      = nfs4_read_done,
5827         .write_setup    = nfs4_proc_write_setup,
5828         .write_done     = nfs4_write_done,
5829         .commit_setup   = nfs4_proc_commit_setup,
5830         .commit_done    = nfs4_commit_done,
5831         .lock           = nfs4_proc_lock,
5832         .clear_acl_cache = nfs4_zap_acl_attr,
5833         .close_context  = nfs4_close_context,
5834         .open_context   = nfs4_atomic_open,
5835         .init_client    = nfs4_init_client,
5836         .secinfo        = nfs4_proc_secinfo,
5837 };
5838
5839 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
5840         .prefix = XATTR_NAME_NFSV4_ACL,
5841         .list   = nfs4_xattr_list_nfs4_acl,
5842         .get    = nfs4_xattr_get_nfs4_acl,
5843         .set    = nfs4_xattr_set_nfs4_acl,
5844 };
5845
5846 const struct xattr_handler *nfs4_xattr_handlers[] = {
5847         &nfs4_xattr_nfs4_acl_handler,
5848         NULL
5849 };
5850
5851 /*
5852  * Local variables:
5853  *  c-basic-offset: 8
5854  * End:
5855  */