fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/slab.h>
  43 #include <linux/sunrpc/clnt.h>
  44 #include <linux/sunrpc/gss_api.h>
  45 #include <linux/nfs.h>
  46 #include <linux/nfs4.h>
  47 #include <linux/nfs_fs.h>
  48 #include <linux/nfs_page.h>
  49 #include <linux/namei.h>
  50 #include <linux/mount.h>
  51 #include <linux/module.h>
  52 #include <linux/sunrpc/bc_xprt.h>
  53 #include <linux/xattr.h>
  54 #include <linux/utsname.h>
  55
  56 #include "nfs4_fs.h"
  57 #include "delegation.h"
  58 #include "internal.h"
  59 #include "iostat.h"
  60 #include "callback.h"
  61 #include "pnfs.h"
  62
  63 #define NFSDBG_FACILITY         NFSDBG_PROC
  64
  65 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  66 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  67
  68 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  69
  70 struct nfs4_opendata;
  71 static int _nfs4_proc_open(struct nfs4_opendata *data);
  72 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  73 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  74 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  75 static int _nfs4_proc_lookup(struct rpc_clnt *client, struct inode *dir,
  76                              const struct qstr *name, struct nfs_fh *fhandle,
  77                              struct nfs_fattr *fattr);
  78 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  79 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  80                             struct nfs_fattr *fattr, struct iattr *sattr,
  81                             struct nfs4_state *state);
  82
  83 /* Prevent leaks of NFSv4 errors into userland */
  84 static int nfs4_map_errors(int err)
  85 {
  86         if (err >= -1000)
  87                 return err;
  88         switch (err) {
  89         case -NFS4ERR_RESOURCE:
  90                 return -EREMOTEIO;
  91         case -NFS4ERR_WRONGSEC:
  92                 return -EPERM;
  93         case -NFS4ERR_BADOWNER:
  94         case -NFS4ERR_BADNAME:
  95                 return -EINVAL;
  96         default:
  97                 dprintk("%s could not handle NFSv4 error %d\n",
  98                                 __func__, -err);
  99                 break;
 100         }
 101         return -EIO;
 102 }
 103
 104 /*
 105  * This is our standard bitmap for GETATTR requests.
 106  */
 107 const u32 nfs4_fattr_bitmap[2] = {
 108         FATTR4_WORD0_TYPE
 109         | FATTR4_WORD0_CHANGE
 110         | FATTR4_WORD0_SIZE
 111         | FATTR4_WORD0_FSID
 112         | FATTR4_WORD0_FILEID,
 113         FATTR4_WORD1_MODE
 114         | FATTR4_WORD1_NUMLINKS
 115         | FATTR4_WORD1_OWNER
 116         | FATTR4_WORD1_OWNER_GROUP
 117         | FATTR4_WORD1_RAWDEV
 118         | FATTR4_WORD1_SPACE_USED
 119         | FATTR4_WORD1_TIME_ACCESS
 120         | FATTR4_WORD1_TIME_METADATA
 121         | FATTR4_WORD1_TIME_MODIFY
 122 };
 123
 124 const u32 nfs4_statfs_bitmap[2] = {
 125         FATTR4_WORD0_FILES_AVAIL
 126         | FATTR4_WORD0_FILES_FREE
 127         | FATTR4_WORD0_FILES_TOTAL,
 128         FATTR4_WORD1_SPACE_AVAIL
 129         | FATTR4_WORD1_SPACE_FREE
 130         | FATTR4_WORD1_SPACE_TOTAL
 131 };
 132
 133 const u32 nfs4_pathconf_bitmap[2] = {
 134         FATTR4_WORD0_MAXLINK
 135         | FATTR4_WORD0_MAXNAME,
 136         0
 137 };
 138
 139 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
 140                         | FATTR4_WORD0_MAXREAD
 141                         | FATTR4_WORD0_MAXWRITE
 142                         | FATTR4_WORD0_LEASE_TIME,
 143                         FATTR4_WORD1_TIME_DELTA
 144                         | FATTR4_WORD1_FS_LAYOUT_TYPES
 145 };
 146
 147 const u32 nfs4_fs_locations_bitmap[2] = {
 148         FATTR4_WORD0_TYPE
 149         | FATTR4_WORD0_CHANGE
 150         | FATTR4_WORD0_SIZE
 151         | FATTR4_WORD0_FSID
 152         | FATTR4_WORD0_FILEID
 153         | FATTR4_WORD0_FS_LOCATIONS,
 154         FATTR4_WORD1_MODE
 155         | FATTR4_WORD1_NUMLINKS
 156         | FATTR4_WORD1_OWNER
 157         | FATTR4_WORD1_OWNER_GROUP
 158         | FATTR4_WORD1_RAWDEV
 159         | FATTR4_WORD1_SPACE_USED
 160         | FATTR4_WORD1_TIME_ACCESS
 161         | FATTR4_WORD1_TIME_METADATA
 162         | FATTR4_WORD1_TIME_MODIFY
 163         | FATTR4_WORD1_MOUNTED_ON_FILEID
 164 };
 165
 166 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 167                 struct nfs4_readdir_arg *readdir)
 168 {
 169         __be32 *start, *p;
 170
 171         BUG_ON(readdir->count < 80);
 172         if (cookie > 2) {
 173                 readdir->cookie = cookie;
 174                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 175                 return;
 176         }
 177
 178         readdir->cookie = 0;
 179         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 180         if (cookie == 2)
 181                 return;
 182
 183         /*
 184          * NFSv4 servers do not return entries for '.' and '..'
 185          * Therefore, we fake these entries here.  We let '.'
 186          * have cookie 0 and '..' have cookie 1.  Note that
 187          * when talking to the server, we always send cookie 0
 188          * instead of 1 or 2.
 189          */
 190         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 191
 192         if (cookie == 0) {
 193                 *p++ = xdr_one;                                  /* next */
 194                 *p++ = xdr_zero;                   /* cookie, first word */
 195                 *p++ = xdr_one;                   /* cookie, second word */
 196                 *p++ = xdr_one;                             /* entry len */
 197                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 198                 p++;
 199                 *p++ = xdr_one;                         /* bitmap length */
 200                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 201                 *p++ = htonl(8);              /* attribute buffer length */
 202                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 203         }
 204
 205         *p++ = xdr_one;                                  /* next */
 206         *p++ = xdr_zero;                   /* cookie, first word */
 207         *p++ = xdr_two;                   /* cookie, second word */
 208         *p++ = xdr_two;                             /* entry len */
 209         memcpy(p, "..\0\0", 4);                         /* entry */
 210         p++;
 211         *p++ = xdr_one;                         /* bitmap length */
 212         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 213         *p++ = htonl(8);              /* attribute buffer length */
 214         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 215
 216         readdir->pgbase = (char *)p - (char *)start;
 217         readdir->count -= readdir->pgbase;
 218         kunmap_atomic(start, KM_USER0);
 219 }
 220
 221 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 222 {
 223         int res;
 224
 225         might_sleep();
 226
 227         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 228                         nfs_wait_bit_killable, TASK_KILLABLE);
 229         return res;
 230 }
 231
 232 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 233 {
 234         int res = 0;
 235
 236         might_sleep();
 237
 238         if (*timeout <= 0)
 239                 *timeout = NFS4_POLL_RETRY_MIN;
 240         if (*timeout > NFS4_POLL_RETRY_MAX)
 241                 *timeout = NFS4_POLL_RETRY_MAX;
 242         schedule_timeout_killable(*timeout);
 243         if (fatal_signal_pending(current))
 244                 res = -ERESTARTSYS;
 245         *timeout <<= 1;
 246         return res;
 247 }
 248
 249 /* This is the error handling routine for processes that are allowed
 250  * to sleep.
 251  */
 252 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 253 {
 254         struct nfs_client *clp = server->nfs_client;
 255         struct nfs4_state *state = exception->state;
 256         int ret = errorcode;
 257
 258         exception->retry = 0;
 259         switch(errorcode) {
 260                 case 0:
 261                         return 0;
 262                 case -NFS4ERR_ADMIN_REVOKED:
 263                 case -NFS4ERR_BAD_STATEID:
 264                 case -NFS4ERR_OPENMODE:
 265                         if (state == NULL)
 266                                 break;
 267                         nfs4_schedule_stateid_recovery(server, state);
 268                         goto wait_on_recovery;
 269                 case -NFS4ERR_STALE_STATEID:
 270                 case -NFS4ERR_STALE_CLIENTID:
 271                 case -NFS4ERR_EXPIRED:
 272                         nfs4_schedule_lease_recovery(clp);
 273                         goto wait_on_recovery;
 274 #if defined(CONFIG_NFS_V4_1)
 275                 case -NFS4ERR_BADSESSION:
 276                 case -NFS4ERR_BADSLOT:
 277                 case -NFS4ERR_BAD_HIGH_SLOT:
 278                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 279                 case -NFS4ERR_DEADSESSION:
 280                 case -NFS4ERR_SEQ_FALSE_RETRY:
 281                 case -NFS4ERR_SEQ_MISORDERED:
 282                         dprintk("%s ERROR: %d Reset session\n", __func__,
 283                                 errorcode);
 284                         nfs4_schedule_session_recovery(clp->cl_session);
 285                         exception->retry = 1;
 286                         break;
 287 #endif /* defined(CONFIG_NFS_V4_1) */
 288                 case -NFS4ERR_FILE_OPEN:
 289                         if (exception->timeout > HZ) {
 290                                 /* We have retried a decent amount, time to
 291                                  * fail
 292                                  */
 293                                 ret = -EBUSY;
 294                                 break;
 295                         }
 296                 case -NFS4ERR_GRACE:
 297                 case -NFS4ERR_DELAY:
 298                 case -EKEYEXPIRED:
 299                         ret = nfs4_delay(server->client, &exception->timeout);
 300                         if (ret != 0)
 301                                 break;
 302                 case -NFS4ERR_OLD_STATEID:
 303                         exception->retry = 1;
 304                         break;
 305                 case -NFS4ERR_BADOWNER:
 306                         /* The following works around a Linux server bug! */
 307                 case -NFS4ERR_BADNAME:
 308                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 309                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 310                                 exception->retry = 1;
 311                                 printk(KERN_WARNING "NFS: v4 server %s "
 312                                                 "does not accept raw "
 313                                                 "uid/gids. "
 314                                                 "Reenabling the idmapper.\n",
 315                                                 server->nfs_client->cl_hostname);
 316                         }
 317         }
 318         /* We failed to handle the error */
 319         return nfs4_map_errors(ret);
 320 wait_on_recovery:
 321         ret = nfs4_wait_clnt_recover(clp);
 322         if (ret == 0)
 323                 exception->retry = 1;
 324         return ret;
 325 }
 326
 327
 328 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 329 {
 330         spin_lock(&clp->cl_lock);
 331         if (time_before(clp->cl_last_renewal,timestamp))
 332                 clp->cl_last_renewal = timestamp;
 333         spin_unlock(&clp->cl_lock);
 334 }
 335
 336 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 337 {
 338         do_renew_lease(server->nfs_client, timestamp);
 339 }
 340
 341 #if defined(CONFIG_NFS_V4_1)
 342
 343 /*
 344  * nfs4_free_slot - free a slot and efficiently update slot table.
 345  *
 346  * freeing a slot is trivially done by clearing its respective bit
 347  * in the bitmap.
 348  * If the freed slotid equals highest_used_slotid we want to update it
 349  * so that the server would be able to size down the slot table if needed,
 350  * otherwise we know that the highest_used_slotid is still in use.
 351  * When updating highest_used_slotid there may be "holes" in the bitmap
 352  * so we need to scan down from highest_used_slotid to 0 looking for the now
 353  * highest slotid in use.
 354  * If none found, highest_used_slotid is set to -1.
 355  *
 356  * Must be called while holding tbl->slot_tbl_lock
 357  */
 358 static void
 359 nfs4_free_slot(struct nfs4_slot_table *tbl, struct nfs4_slot *free_slot)
 360 {
 361         int free_slotid = free_slot - tbl->slots;
 362         int slotid = free_slotid;
 363
 364         BUG_ON(slotid < 0 || slotid >= NFS4_MAX_SLOT_TABLE);
 365         /* clear used bit in bitmap */
 366         __clear_bit(slotid, tbl->used_slots);
 367
 368         /* update highest_used_slotid when it is freed */
 369         if (slotid == tbl->highest_used_slotid) {
 370                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 371                 if (slotid < tbl->max_slots)
 372                         tbl->highest_used_slotid = slotid;
 373                 else
 374                         tbl->highest_used_slotid = -1;
 375         }
 376         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
 377                 free_slotid, tbl->highest_used_slotid);
 378 }
 379
 380 /*
 381  * Signal state manager thread if session fore channel is drained
 382  */
 383 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 384 {
 385         struct rpc_task *task;
 386
 387         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 388                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
 389                 if (task)
 390                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 391                 return;
 392         }
 393
 394         if (ses->fc_slot_table.highest_used_slotid != -1)
 395                 return;
 396
 397         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 398         complete(&ses->fc_slot_table.complete);
 399 }
 400
 401 /*
 402  * Signal state manager thread if session back channel is drained
 403  */
 404 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 405 {
 406         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 407             ses->bc_slot_table.highest_used_slotid != -1)
 408                 return;
 409         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 410         complete(&ses->bc_slot_table.complete);
 411 }
 412
 413 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 414 {
 415         struct nfs4_slot_table *tbl;
 416
 417         tbl = &res->sr_session->fc_slot_table;
 418         if (!res->sr_slot) {
 419                 /* just wake up the next guy waiting since
 420                  * we may have not consumed a slot after all */
 421                 dprintk("%s: No slot\n", __func__);
 422                 return;
 423         }
 424
 425         spin_lock(&tbl->slot_tbl_lock);
 426         nfs4_free_slot(tbl, res->sr_slot);
 427         nfs4_check_drain_fc_complete(res->sr_session);
 428         spin_unlock(&tbl->slot_tbl_lock);
 429         res->sr_slot = NULL;
 430 }
 431
 432 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 433 {
 434         unsigned long timestamp;
 435         struct nfs_client *clp;
 436
 437         /*
 438          * sr_status remains 1 if an RPC level error occurred. The server
 439          * may or may not have processed the sequence operation..
 440          * Proceed as if the server received and processed the sequence
 441          * operation.
 442          */
 443         if (res->sr_status == 1)
 444                 res->sr_status = NFS_OK;
 445
 446         /* -ERESTARTSYS can result in skipping nfs41_sequence_setup */
 447         if (!res->sr_slot)
 448                 goto out;
 449
 450         /* Check the SEQUENCE operation status */
 451         switch (res->sr_status) {
 452         case 0:
 453                 /* Update the slot's sequence and clientid lease timer */
 454                 ++res->sr_slot->seq_nr;
 455                 timestamp = res->sr_renewal_time;
 456                 clp = res->sr_session->clp;
 457                 do_renew_lease(clp, timestamp);
 458                 /* Check sequence flags */
 459                 if (res->sr_status_flags != 0)
 460                         nfs4_schedule_lease_recovery(clp);
 461                 break;
 462         case -NFS4ERR_DELAY:
 463                 /* The server detected a resend of the RPC call and
 464                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 465                  * of RFC5661.
 466                  */
 467                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 468                         __func__,
 469                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 470                         res->sr_slot->seq_nr);
 471                 goto out_retry;
 472         default:
 473                 /* Just update the slot sequence no. */
 474                 ++res->sr_slot->seq_nr;
 475         }
 476 out:
 477         /* The session may be reset by one of the error handlers. */
 478         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 479         nfs41_sequence_free_slot(res);
 480         return 1;
 481 out_retry:
 482         if (!rpc_restart_call(task))
 483                 goto out;
 484         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 485         return 0;
 486 }
 487
 488 static int nfs4_sequence_done(struct rpc_task *task,
 489                                struct nfs4_sequence_res *res)
 490 {
 491         if (res->sr_session == NULL)
 492                 return 1;
 493         return nfs41_sequence_done(task, res);
 494 }
 495
 496 /*
 497  * nfs4_find_slot - efficiently look for a free slot
 498  *
 499  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 500  * If found, we mark the slot as used, update the highest_used_slotid,
 501  * and respectively set up the sequence operation args.
 502  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
 503  *
 504  * Note: must be called with under the slot_tbl_lock.
 505  */
 506 static u8
 507 nfs4_find_slot(struct nfs4_slot_table *tbl)
 508 {
 509         int slotid;
 510         u8 ret_id = NFS4_MAX_SLOT_TABLE;
 511         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
 512
 513         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
 514                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 515                 tbl->max_slots);
 516         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 517         if (slotid >= tbl->max_slots)
 518                 goto out;
 519         __set_bit(slotid, tbl->used_slots);
 520         if (slotid > tbl->highest_used_slotid)
 521                 tbl->highest_used_slotid = slotid;
 522         ret_id = slotid;
 523 out:
 524         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 525                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 526         return ret_id;
 527 }
 528
 529 int nfs41_setup_sequence(struct nfs4_session *session,
 530                                 struct nfs4_sequence_args *args,
 531                                 struct nfs4_sequence_res *res,
 532                                 int cache_reply,
 533                                 struct rpc_task *task)
 534 {
 535         struct nfs4_slot *slot;
 536         struct nfs4_slot_table *tbl;
 537         u8 slotid;
 538
 539         dprintk("--> %s\n", __func__);
 540         /* slot already allocated? */
 541         if (res->sr_slot != NULL)
 542                 return 0;
 543
 544         tbl = &session->fc_slot_table;
 545
 546         spin_lock(&tbl->slot_tbl_lock);
 547         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 548             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 549                 /*
 550                  * The state manager will wait until the slot table is empty.
 551                  * Schedule the reset thread
 552                  */
 553                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 554                 spin_unlock(&tbl->slot_tbl_lock);
 555                 dprintk("%s Schedule Session Reset\n", __func__);
 556                 return -EAGAIN;
 557         }
 558
 559         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 560             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 561                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 562                 spin_unlock(&tbl->slot_tbl_lock);
 563                 dprintk("%s enforce FIFO order\n", __func__);
 564                 return -EAGAIN;
 565         }
 566
 567         slotid = nfs4_find_slot(tbl);
 568         if (slotid == NFS4_MAX_SLOT_TABLE) {
 569                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 570                 spin_unlock(&tbl->slot_tbl_lock);
 571                 dprintk("<-- %s: no free slots\n", __func__);
 572                 return -EAGAIN;
 573         }
 574         spin_unlock(&tbl->slot_tbl_lock);
 575
 576         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 577         slot = tbl->slots + slotid;
 578         args->sa_session = session;
 579         args->sa_slotid = slotid;
 580         args->sa_cache_this = cache_reply;
 581
 582         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 583
 584         res->sr_session = session;
 585         res->sr_slot = slot;
 586         res->sr_renewal_time = jiffies;
 587         res->sr_status_flags = 0;
 588         /*
 589          * sr_status is only set in decode_sequence, and so will remain
 590          * set to 1 if an rpc level failure occurs.
 591          */
 592         res->sr_status = 1;
 593         return 0;
 594 }
 595 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 596
 597 int nfs4_setup_sequence(const struct nfs_server *server,
 598                         struct nfs4_sequence_args *args,
 599                         struct nfs4_sequence_res *res,
 600                         int cache_reply,
 601                         struct rpc_task *task)
 602 {
 603         struct nfs4_session *session = nfs4_get_session(server);
 604         int ret = 0;
 605
 606         if (session == NULL) {
 607                 args->sa_session = NULL;
 608                 res->sr_session = NULL;
 609                 goto out;
 610         }
 611
 612         dprintk("--> %s clp %p session %p sr_slot %td\n",
 613                 __func__, session->clp, session, res->sr_slot ?
 614                         res->sr_slot - session->fc_slot_table.slots : -1);
 615
 616         ret = nfs41_setup_sequence(session, args, res, cache_reply,
 617                                    task);
 618 out:
 619         dprintk("<-- %s status=%d\n", __func__, ret);
 620         return ret;
 621 }
 622
 623 struct nfs41_call_sync_data {
 624         const struct nfs_server *seq_server;
 625         struct nfs4_sequence_args *seq_args;
 626         struct nfs4_sequence_res *seq_res;
 627         int cache_reply;
 628 };
 629
 630 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 631 {
 632         struct nfs41_call_sync_data *data = calldata;
 633
 634         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 635
 636         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 637                                 data->seq_res, data->cache_reply, task))
 638                 return;
 639         rpc_call_start(task);
 640 }
 641
 642 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 643 {
 644         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 645         nfs41_call_sync_prepare(task, calldata);
 646 }
 647
 648 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 649 {
 650         struct nfs41_call_sync_data *data = calldata;
 651
 652         nfs41_sequence_done(task, data->seq_res);
 653 }
 654
 655 struct rpc_call_ops nfs41_call_sync_ops = {
 656         .rpc_call_prepare = nfs41_call_sync_prepare,
 657         .rpc_call_done = nfs41_call_sync_done,
 658 };
 659
 660 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 661         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 662         .rpc_call_done = nfs41_call_sync_done,
 663 };
 664
 665 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 666                                    struct nfs_server *server,
 667                                    struct rpc_message *msg,
 668                                    struct nfs4_sequence_args *args,
 669                                    struct nfs4_sequence_res *res,
 670                                    int cache_reply,
 671                                    int privileged)
 672 {
 673         int ret;
 674         struct rpc_task *task;
 675         struct nfs41_call_sync_data data = {
 676                 .seq_server = server,
 677                 .seq_args = args,
 678                 .seq_res = res,
 679                 .cache_reply = cache_reply,
 680         };
 681         struct rpc_task_setup task_setup = {
 682                 .rpc_client = clnt,
 683                 .rpc_message = msg,
 684                 .callback_ops = &nfs41_call_sync_ops,
 685                 .callback_data = &data
 686         };
 687
 688         res->sr_slot = NULL;
 689         if (privileged)
 690                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 691         task = rpc_run_task(&task_setup);
 692         if (IS_ERR(task))
 693                 ret = PTR_ERR(task);
 694         else {
 695                 ret = task->tk_status;
 696                 rpc_put_task(task);
 697         }
 698         return ret;
 699 }
 700
 701 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 702                             struct nfs_server *server,
 703                             struct rpc_message *msg,
 704                             struct nfs4_sequence_args *args,
 705                             struct nfs4_sequence_res *res,
 706                             int cache_reply)
 707 {
 708         return nfs4_call_sync_sequence(clnt, server, msg, args, res, cache_reply, 0);
 709 }
 710
 711 #else
 712 static int nfs4_sequence_done(struct rpc_task *task,
 713                                struct nfs4_sequence_res *res)
 714 {
 715         return 1;
 716 }
 717 #endif /* CONFIG_NFS_V4_1 */
 718
 719 int _nfs4_call_sync(struct rpc_clnt *clnt,
 720                     struct nfs_server *server,
 721                     struct rpc_message *msg,
 722                     struct nfs4_sequence_args *args,
 723                     struct nfs4_sequence_res *res,
 724                     int cache_reply)
 725 {
 726         args->sa_session = res->sr_session = NULL;
 727         return rpc_call_sync(clnt, msg, 0);
 728 }
 729
 730 static inline
 731 int nfs4_call_sync(struct rpc_clnt *clnt,
 732                    struct nfs_server *server,
 733                    struct rpc_message *msg,
 734                    struct nfs4_sequence_args *args,
 735                    struct nfs4_sequence_res *res,
 736                    int cache_reply)
 737 {
 738         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 739                                                 args, res, cache_reply);
 740 }
 741
 742 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 743 {
 744         struct nfs_inode *nfsi = NFS_I(dir);
 745
 746         spin_lock(&dir->i_lock);
 747         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 748         if (!cinfo->atomic || cinfo->before != nfsi->change_attr)
 749                 nfs_force_lookup_revalidate(dir);
 750         nfsi->change_attr = cinfo->after;
 751         spin_unlock(&dir->i_lock);
 752 }
 753
 754 struct nfs4_opendata {
 755         struct kref kref;
 756         struct nfs_openargs o_arg;
 757         struct nfs_openres o_res;
 758         struct nfs_open_confirmargs c_arg;
 759         struct nfs_open_confirmres c_res;
 760         struct nfs_fattr f_attr;
 761         struct nfs_fattr dir_attr;
 762         struct path path;
 763         struct dentry *dir;
 764         struct nfs4_state_owner *owner;
 765         struct nfs4_state *state;
 766         struct iattr attrs;
 767         unsigned long timestamp;
 768         unsigned int rpc_done : 1;
 769         int rpc_status;
 770         int cancelled;
 771 };
 772
 773
 774 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 775 {
 776         p->o_res.f_attr = &p->f_attr;
 777         p->o_res.dir_attr = &p->dir_attr;
 778         p->o_res.seqid = p->o_arg.seqid;
 779         p->c_res.seqid = p->c_arg.seqid;
 780         p->o_res.server = p->o_arg.server;
 781         nfs_fattr_init(&p->f_attr);
 782         nfs_fattr_init(&p->dir_attr);
 783 }
 784
 785 static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
 786                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 787                 const struct iattr *attrs,
 788                 gfp_t gfp_mask)
 789 {
 790         struct dentry *parent = dget_parent(path->dentry);
 791         struct inode *dir = parent->d_inode;
 792         struct nfs_server *server = NFS_SERVER(dir);
 793         struct nfs4_opendata *p;
 794
 795         p = kzalloc(sizeof(*p), gfp_mask);
 796         if (p == NULL)
 797                 goto err;
 798         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 799         if (p->o_arg.seqid == NULL)
 800                 goto err_free;
 801         path_get(path);
 802         p->path = *path;
 803         p->dir = parent;
 804         p->owner = sp;
 805         atomic_inc(&sp->so_count);
 806         p->o_arg.fh = NFS_FH(dir);
 807         p->o_arg.open_flags = flags;
 808         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 809         p->o_arg.clientid = server->nfs_client->cl_clientid;
 810         p->o_arg.id = sp->so_owner_id.id;
 811         p->o_arg.name = &p->path.dentry->d_name;
 812         p->o_arg.server = server;
 813         p->o_arg.bitmask = server->attr_bitmask;
 814         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 815         if (flags & O_CREAT) {
 816                 u32 *s;
 817
 818                 p->o_arg.u.attrs = &p->attrs;
 819                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 820                 s = (u32 *) p->o_arg.u.verifier.data;
 821                 s[0] = jiffies;
 822                 s[1] = current->pid;
 823         }
 824         p->c_arg.fh = &p->o_res.fh;
 825         p->c_arg.stateid = &p->o_res.stateid;
 826         p->c_arg.seqid = p->o_arg.seqid;
 827         nfs4_init_opendata_res(p);
 828         kref_init(&p->kref);
 829         return p;
 830 err_free:
 831         kfree(p);
 832 err:
 833         dput(parent);
 834         return NULL;
 835 }
 836
 837 static void nfs4_opendata_free(struct kref *kref)
 838 {
 839         struct nfs4_opendata *p = container_of(kref,
 840                         struct nfs4_opendata, kref);
 841
 842         nfs_free_seqid(p->o_arg.seqid);
 843         if (p->state != NULL)
 844                 nfs4_put_open_state(p->state);
 845         nfs4_put_state_owner(p->owner);
 846         dput(p->dir);
 847         path_put(&p->path);
 848         kfree(p);
 849 }
 850
 851 static void nfs4_opendata_put(struct nfs4_opendata *p)
 852 {
 853         if (p != NULL)
 854                 kref_put(&p->kref, nfs4_opendata_free);
 855 }
 856
 857 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 858 {
 859         int ret;
 860
 861         ret = rpc_wait_for_completion_task(task);
 862         return ret;
 863 }
 864
 865 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 866 {
 867         int ret = 0;
 868
 869         if (open_mode & O_EXCL)
 870                 goto out;
 871         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 872                 case FMODE_READ:
 873                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 874                                 && state->n_rdonly != 0;
 875                         break;
 876                 case FMODE_WRITE:
 877                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 878                                 && state->n_wronly != 0;
 879                         break;
 880                 case FMODE_READ|FMODE_WRITE:
 881                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 882                                 && state->n_rdwr != 0;
 883         }
 884 out:
 885         return ret;
 886 }
 887
 888 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 889 {
 890         if ((delegation->type & fmode) != fmode)
 891                 return 0;
 892         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 893                 return 0;
 894         nfs_mark_delegation_referenced(delegation);
 895         return 1;
 896 }
 897
 898 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 899 {
 900         switch (fmode) {
 901                 case FMODE_WRITE:
 902                         state->n_wronly++;
 903                         break;
 904                 case FMODE_READ:
 905                         state->n_rdonly++;
 906                         break;
 907                 case FMODE_READ|FMODE_WRITE:
 908                         state->n_rdwr++;
 909         }
 910         nfs4_state_set_mode_locked(state, state->state | fmode);
 911 }
 912
 913 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 914 {
 915         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 916                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 917         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 918         switch (fmode) {
 919                 case FMODE_READ:
 920                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 921                         break;
 922                 case FMODE_WRITE:
 923                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 924                         break;
 925                 case FMODE_READ|FMODE_WRITE:
 926                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 927         }
 928 }
 929
 930 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 931 {
 932         write_seqlock(&state->seqlock);
 933         nfs_set_open_stateid_locked(state, stateid, fmode);
 934         write_sequnlock(&state->seqlock);
 935 }
 936
 937 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 938 {
 939         /*
 940          * Protect the call to nfs4_state_set_mode_locked and
 941          * serialise the stateid update
 942          */
 943         write_seqlock(&state->seqlock);
 944         if (deleg_stateid != NULL) {
 945                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 946                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 947         }
 948         if (open_stateid != NULL)
 949                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 950         write_sequnlock(&state->seqlock);
 951         spin_lock(&state->owner->so_lock);
 952         update_open_stateflags(state, fmode);
 953         spin_unlock(&state->owner->so_lock);
 954 }
 955
 956 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 957 {
 958         struct nfs_inode *nfsi = NFS_I(state->inode);
 959         struct nfs_delegation *deleg_cur;
 960         int ret = 0;
 961
 962         fmode &= (FMODE_READ|FMODE_WRITE);
 963
 964         rcu_read_lock();
 965         deleg_cur = rcu_dereference(nfsi->delegation);
 966         if (deleg_cur == NULL)
 967                 goto no_delegation;
 968
 969         spin_lock(&deleg_cur->lock);
 970         if (nfsi->delegation != deleg_cur ||
 971             (deleg_cur->type & fmode) != fmode)
 972                 goto no_delegation_unlock;
 973
 974         if (delegation == NULL)
 975                 delegation = &deleg_cur->stateid;
 976         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
 977                 goto no_delegation_unlock;
 978
 979         nfs_mark_delegation_referenced(deleg_cur);
 980         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
 981         ret = 1;
 982 no_delegation_unlock:
 983         spin_unlock(&deleg_cur->lock);
 984 no_delegation:
 985         rcu_read_unlock();
 986
 987         if (!ret && open_stateid != NULL) {
 988                 __update_open_stateid(state, open_stateid, NULL, fmode);
 989                 ret = 1;
 990         }
 991
 992         return ret;
 993 }
 994
 995
 996 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
 997 {
 998         struct nfs_delegation *delegation;
 999
1000         rcu_read_lock();
1001         delegation = rcu_dereference(NFS_I(inode)->delegation);
1002         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1003                 rcu_read_unlock();
1004                 return;
1005         }
1006         rcu_read_unlock();
1007         nfs_inode_return_delegation(inode);
1008 }
1009
1010 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1011 {
1012         struct nfs4_state *state = opendata->state;
1013         struct nfs_inode *nfsi = NFS_I(state->inode);
1014         struct nfs_delegation *delegation;
1015         int open_mode = opendata->o_arg.open_flags & O_EXCL;
1016         fmode_t fmode = opendata->o_arg.fmode;
1017         nfs4_stateid stateid;
1018         int ret = -EAGAIN;
1019
1020         for (;;) {
1021                 if (can_open_cached(state, fmode, open_mode)) {
1022                         spin_lock(&state->owner->so_lock);
1023                         if (can_open_cached(state, fmode, open_mode)) {
1024                                 update_open_stateflags(state, fmode);
1025                                 spin_unlock(&state->owner->so_lock);
1026                                 goto out_return_state;
1027                         }
1028                         spin_unlock(&state->owner->so_lock);
1029                 }
1030                 rcu_read_lock();
1031                 delegation = rcu_dereference(nfsi->delegation);
1032                 if (delegation == NULL ||
1033                     !can_open_delegated(delegation, fmode)) {
1034                         rcu_read_unlock();
1035                         break;
1036                 }
1037                 /* Save the delegation */
1038                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1039                 rcu_read_unlock();
1040                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1041                 if (ret != 0)
1042                         goto out;
1043                 ret = -EAGAIN;
1044
1045                 /* Try to update the stateid using the delegation */
1046                 if (update_open_stateid(state, NULL, &stateid, fmode))
1047                         goto out_return_state;
1048         }
1049 out:
1050         return ERR_PTR(ret);
1051 out_return_state:
1052         atomic_inc(&state->count);
1053         return state;
1054 }
1055
1056 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1057 {
1058         struct inode *inode;
1059         struct nfs4_state *state = NULL;
1060         struct nfs_delegation *delegation;
1061         int ret;
1062
1063         if (!data->rpc_done) {
1064                 state = nfs4_try_open_cached(data);
1065                 goto out;
1066         }
1067
1068         ret = -EAGAIN;
1069         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1070                 goto err;
1071         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1072         ret = PTR_ERR(inode);
1073         if (IS_ERR(inode))
1074                 goto err;
1075         ret = -ENOMEM;
1076         state = nfs4_get_open_state(inode, data->owner);
1077         if (state == NULL)
1078                 goto err_put_inode;
1079         if (data->o_res.delegation_type != 0) {
1080                 int delegation_flags = 0;
1081
1082                 rcu_read_lock();
1083                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1084                 if (delegation)
1085                         delegation_flags = delegation->flags;
1086                 rcu_read_unlock();
1087                 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1088                         nfs_inode_set_delegation(state->inode,
1089                                         data->owner->so_cred,
1090                                         &data->o_res);
1091                 else
1092                         nfs_inode_reclaim_delegation(state->inode,
1093                                         data->owner->so_cred,
1094                                         &data->o_res);
1095         }
1096
1097         update_open_stateid(state, &data->o_res.stateid, NULL,
1098                         data->o_arg.fmode);
1099         iput(inode);
1100 out:
1101         return state;
1102 err_put_inode:
1103         iput(inode);
1104 err:
1105         return ERR_PTR(ret);
1106 }
1107
1108 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1109 {
1110         struct nfs_inode *nfsi = NFS_I(state->inode);
1111         struct nfs_open_context *ctx;
1112
1113         spin_lock(&state->inode->i_lock);
1114         list_for_each_entry(ctx, &nfsi->open_files, list) {
1115                 if (ctx->state != state)
1116                         continue;
1117                 get_nfs_open_context(ctx);
1118                 spin_unlock(&state->inode->i_lock);
1119                 return ctx;
1120         }
1121         spin_unlock(&state->inode->i_lock);
1122         return ERR_PTR(-ENOENT);
1123 }
1124
1125 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1126 {
1127         struct nfs4_opendata *opendata;
1128
1129         opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, 0, NULL, GFP_NOFS);
1130         if (opendata == NULL)
1131                 return ERR_PTR(-ENOMEM);
1132         opendata->state = state;
1133         atomic_inc(&state->count);
1134         return opendata;
1135 }
1136
1137 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1138 {
1139         struct nfs4_state *newstate;
1140         int ret;
1141
1142         opendata->o_arg.open_flags = 0;
1143         opendata->o_arg.fmode = fmode;
1144         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1145         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1146         nfs4_init_opendata_res(opendata);
1147         ret = _nfs4_recover_proc_open(opendata);
1148         if (ret != 0)
1149                 return ret;
1150         newstate = nfs4_opendata_to_nfs4_state(opendata);
1151         if (IS_ERR(newstate))
1152                 return PTR_ERR(newstate);
1153         nfs4_close_state(&opendata->path, newstate, fmode);
1154         *res = newstate;
1155         return 0;
1156 }
1157
1158 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1159 {
1160         struct nfs4_state *newstate;
1161         int ret;
1162
1163         /* memory barrier prior to reading state->n_* */
1164         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1165         smp_rmb();
1166         if (state->n_rdwr != 0) {
1167                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1168                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1169                 if (ret != 0)
1170                         return ret;
1171                 if (newstate != state)
1172                         return -ESTALE;
1173         }
1174         if (state->n_wronly != 0) {
1175                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1176                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1177                 if (ret != 0)
1178                         return ret;
1179                 if (newstate != state)
1180                         return -ESTALE;
1181         }
1182         if (state->n_rdonly != 0) {
1183                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1184                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1185                 if (ret != 0)
1186                         return ret;
1187                 if (newstate != state)
1188                         return -ESTALE;
1189         }
1190         /*
1191          * We may have performed cached opens for all three recoveries.
1192          * Check if we need to update the current stateid.
1193          */
1194         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1195             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1196                 write_seqlock(&state->seqlock);
1197                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1198                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1199                 write_sequnlock(&state->seqlock);
1200         }
1201         return 0;
1202 }
1203
1204 /*
1205  * OPEN_RECLAIM:
1206  *      reclaim state on the server after a reboot.
1207  */
1208 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1209 {
1210         struct nfs_delegation *delegation;
1211         struct nfs4_opendata *opendata;
1212         fmode_t delegation_type = 0;
1213         int status;
1214
1215         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1216         if (IS_ERR(opendata))
1217                 return PTR_ERR(opendata);
1218         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1219         opendata->o_arg.fh = NFS_FH(state->inode);
1220         rcu_read_lock();
1221         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1222         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1223                 delegation_type = delegation->type;
1224         rcu_read_unlock();
1225         opendata->o_arg.u.delegation_type = delegation_type;
1226         status = nfs4_open_recover(opendata, state);
1227         nfs4_opendata_put(opendata);
1228         return status;
1229 }
1230
1231 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1232 {
1233         struct nfs_server *server = NFS_SERVER(state->inode);
1234         struct nfs4_exception exception = { };
1235         int err;
1236         do {
1237                 err = _nfs4_do_open_reclaim(ctx, state);
1238                 if (err != -NFS4ERR_DELAY)
1239                         break;
1240                 nfs4_handle_exception(server, err, &exception);
1241         } while (exception.retry);
1242         return err;
1243 }
1244
1245 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1246 {
1247         struct nfs_open_context *ctx;
1248         int ret;
1249
1250         ctx = nfs4_state_find_open_context(state);
1251         if (IS_ERR(ctx))
1252                 return PTR_ERR(ctx);
1253         ret = nfs4_do_open_reclaim(ctx, state);
1254         put_nfs_open_context(ctx);
1255         return ret;
1256 }
1257
1258 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1259 {
1260         struct nfs4_opendata *opendata;
1261         int ret;
1262
1263         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1264         if (IS_ERR(opendata))
1265                 return PTR_ERR(opendata);
1266         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1267         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1268                         sizeof(opendata->o_arg.u.delegation.data));
1269         ret = nfs4_open_recover(opendata, state);
1270         nfs4_opendata_put(opendata);
1271         return ret;
1272 }
1273
1274 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1275 {
1276         struct nfs4_exception exception = { };
1277         struct nfs_server *server = NFS_SERVER(state->inode);
1278         int err;
1279         do {
1280                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1281                 switch (err) {
1282                         case 0:
1283                         case -ENOENT:
1284                         case -ESTALE:
1285                                 goto out;
1286                         case -NFS4ERR_BADSESSION:
1287                         case -NFS4ERR_BADSLOT:
1288                         case -NFS4ERR_BAD_HIGH_SLOT:
1289                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1290                         case -NFS4ERR_DEADSESSION:
1291                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1292                                 goto out;
1293                         case -NFS4ERR_STALE_CLIENTID:
1294                         case -NFS4ERR_STALE_STATEID:
1295                         case -NFS4ERR_EXPIRED:
1296                                 /* Don't recall a delegation if it was lost */
1297                                 nfs4_schedule_lease_recovery(server->nfs_client);
1298                                 goto out;
1299                         case -ERESTARTSYS:
1300                                 /*
1301                                  * The show must go on: exit, but mark the
1302                                  * stateid as needing recovery.
1303                                  */
1304                         case -NFS4ERR_ADMIN_REVOKED:
1305                         case -NFS4ERR_BAD_STATEID:
1306                                 nfs4_schedule_stateid_recovery(server, state);
1307                         case -EKEYEXPIRED:
1308                                 /*
1309                                  * User RPCSEC_GSS context has expired.
1310                                  * We cannot recover this stateid now, so
1311                                  * skip it and allow recovery thread to
1312                                  * proceed.
1313                                  */
1314                         case -ENOMEM:
1315                                 err = 0;
1316                                 goto out;
1317                 }
1318                 err = nfs4_handle_exception(server, err, &exception);
1319         } while (exception.retry);
1320 out:
1321         return err;
1322 }
1323
1324 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1325 {
1326         struct nfs4_opendata *data = calldata;
1327
1328         data->rpc_status = task->tk_status;
1329         if (data->rpc_status == 0) {
1330                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1331                                 sizeof(data->o_res.stateid.data));
1332                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1333                 renew_lease(data->o_res.server, data->timestamp);
1334                 data->rpc_done = 1;
1335         }
1336 }
1337
1338 static void nfs4_open_confirm_release(void *calldata)
1339 {
1340         struct nfs4_opendata *data = calldata;
1341         struct nfs4_state *state = NULL;
1342
1343         /* If this request hasn't been cancelled, do nothing */
1344         if (data->cancelled == 0)
1345                 goto out_free;
1346         /* In case of error, no cleanup! */
1347         if (!data->rpc_done)
1348                 goto out_free;
1349         state = nfs4_opendata_to_nfs4_state(data);
1350         if (!IS_ERR(state))
1351                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1352 out_free:
1353         nfs4_opendata_put(data);
1354 }
1355
1356 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1357         .rpc_call_done = nfs4_open_confirm_done,
1358         .rpc_release = nfs4_open_confirm_release,
1359 };
1360
1361 /*
1362  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1363  */
1364 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1365 {
1366         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1367         struct rpc_task *task;
1368         struct  rpc_message msg = {
1369                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1370                 .rpc_argp = &data->c_arg,
1371                 .rpc_resp = &data->c_res,
1372                 .rpc_cred = data->owner->so_cred,
1373         };
1374         struct rpc_task_setup task_setup_data = {
1375                 .rpc_client = server->client,
1376                 .rpc_message = &msg,
1377                 .callback_ops = &nfs4_open_confirm_ops,
1378                 .callback_data = data,
1379                 .workqueue = nfsiod_workqueue,
1380                 .flags = RPC_TASK_ASYNC,
1381         };
1382         int status;
1383
1384         kref_get(&data->kref);
1385         data->rpc_done = 0;
1386         data->rpc_status = 0;
1387         data->timestamp = jiffies;
1388         task = rpc_run_task(&task_setup_data);
1389         if (IS_ERR(task))
1390                 return PTR_ERR(task);
1391         status = nfs4_wait_for_completion_rpc_task(task);
1392         if (status != 0) {
1393                 data->cancelled = 1;
1394                 smp_wmb();
1395         } else
1396                 status = data->rpc_status;
1397         rpc_put_task(task);
1398         return status;
1399 }
1400
1401 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1402 {
1403         struct nfs4_opendata *data = calldata;
1404         struct nfs4_state_owner *sp = data->owner;
1405
1406         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1407                 return;
1408         /*
1409          * Check if we still need to send an OPEN call, or if we can use
1410          * a delegation instead.
1411          */
1412         if (data->state != NULL) {
1413                 struct nfs_delegation *delegation;
1414
1415                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1416                         goto out_no_action;
1417                 rcu_read_lock();
1418                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1419                 if (delegation != NULL &&
1420                     test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) == 0) {
1421                         rcu_read_unlock();
1422                         goto out_no_action;
1423                 }
1424                 rcu_read_unlock();
1425         }
1426         /* Update sequence id. */
1427         data->o_arg.id = sp->so_owner_id.id;
1428         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1429         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1430                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1431                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1432         }
1433         data->timestamp = jiffies;
1434         if (nfs4_setup_sequence(data->o_arg.server,
1435                                 &data->o_arg.seq_args,
1436                                 &data->o_res.seq_res, 1, task))
1437                 return;
1438         rpc_call_start(task);
1439         return;
1440 out_no_action:
1441         task->tk_action = NULL;
1442
1443 }
1444
1445 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1446 {
1447         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1448         nfs4_open_prepare(task, calldata);
1449 }
1450
1451 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1452 {
1453         struct nfs4_opendata *data = calldata;
1454
1455         data->rpc_status = task->tk_status;
1456
1457         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1458                 return;
1459
1460         if (task->tk_status == 0) {
1461                 switch (data->o_res.f_attr->mode & S_IFMT) {
1462                         case S_IFREG:
1463                                 break;
1464                         case S_IFLNK:
1465                                 data->rpc_status = -ELOOP;
1466                                 break;
1467                         case S_IFDIR:
1468                                 data->rpc_status = -EISDIR;
1469                                 break;
1470                         default:
1471                                 data->rpc_status = -ENOTDIR;
1472                 }
1473                 renew_lease(data->o_res.server, data->timestamp);
1474                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1475                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1476         }
1477         data->rpc_done = 1;
1478 }
1479
1480 static void nfs4_open_release(void *calldata)
1481 {
1482         struct nfs4_opendata *data = calldata;
1483         struct nfs4_state *state = NULL;
1484
1485         /* If this request hasn't been cancelled, do nothing */
1486         if (data->cancelled == 0)
1487                 goto out_free;
1488         /* In case of error, no cleanup! */
1489         if (data->rpc_status != 0 || !data->rpc_done)
1490                 goto out_free;
1491         /* In case we need an open_confirm, no cleanup! */
1492         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1493                 goto out_free;
1494         state = nfs4_opendata_to_nfs4_state(data);
1495         if (!IS_ERR(state))
1496                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1497 out_free:
1498         nfs4_opendata_put(data);
1499 }
1500
1501 static const struct rpc_call_ops nfs4_open_ops = {
1502         .rpc_call_prepare = nfs4_open_prepare,
1503         .rpc_call_done = nfs4_open_done,
1504         .rpc_release = nfs4_open_release,
1505 };
1506
1507 static const struct rpc_call_ops nfs4_recover_open_ops = {
1508         .rpc_call_prepare = nfs4_recover_open_prepare,
1509         .rpc_call_done = nfs4_open_done,
1510         .rpc_release = nfs4_open_release,
1511 };
1512
1513 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1514 {
1515         struct inode *dir = data->dir->d_inode;
1516         struct nfs_server *server = NFS_SERVER(dir);
1517         struct nfs_openargs *o_arg = &data->o_arg;
1518         struct nfs_openres *o_res = &data->o_res;
1519         struct rpc_task *task;
1520         struct rpc_message msg = {
1521                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1522                 .rpc_argp = o_arg,
1523                 .rpc_resp = o_res,
1524                 .rpc_cred = data->owner->so_cred,
1525         };
1526         struct rpc_task_setup task_setup_data = {
1527                 .rpc_client = server->client,
1528                 .rpc_message = &msg,
1529                 .callback_ops = &nfs4_open_ops,
1530                 .callback_data = data,
1531                 .workqueue = nfsiod_workqueue,
1532                 .flags = RPC_TASK_ASYNC,
1533         };
1534         int status;
1535
1536         kref_get(&data->kref);
1537         data->rpc_done = 0;
1538         data->rpc_status = 0;
1539         data->cancelled = 0;
1540         if (isrecover)
1541                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1542         task = rpc_run_task(&task_setup_data);
1543         if (IS_ERR(task))
1544                 return PTR_ERR(task);
1545         status = nfs4_wait_for_completion_rpc_task(task);
1546         if (status != 0) {
1547                 data->cancelled = 1;
1548                 smp_wmb();
1549         } else
1550                 status = data->rpc_status;
1551         rpc_put_task(task);
1552
1553         return status;
1554 }
1555
1556 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1557 {
1558         struct inode *dir = data->dir->d_inode;
1559         struct nfs_openres *o_res = &data->o_res;
1560         int status;
1561
1562         status = nfs4_run_open_task(data, 1);
1563         if (status != 0 || !data->rpc_done)
1564                 return status;
1565
1566         nfs_refresh_inode(dir, o_res->dir_attr);
1567
1568         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1569                 status = _nfs4_proc_open_confirm(data);
1570                 if (status != 0)
1571                         return status;
1572         }
1573
1574         return status;
1575 }
1576
1577 /*
1578  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1579  */
1580 static int _nfs4_proc_open(struct nfs4_opendata *data)
1581 {
1582         struct inode *dir = data->dir->d_inode;
1583         struct nfs_server *server = NFS_SERVER(dir);
1584         struct nfs_openargs *o_arg = &data->o_arg;
1585         struct nfs_openres *o_res = &data->o_res;
1586         int status;
1587
1588         status = nfs4_run_open_task(data, 0);
1589         if (status != 0 || !data->rpc_done)
1590                 return status;
1591
1592         if (o_arg->open_flags & O_CREAT) {
1593                 update_changeattr(dir, &o_res->cinfo);
1594                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1595         } else
1596                 nfs_refresh_inode(dir, o_res->dir_attr);
1597         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1598                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1599         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1600                 status = _nfs4_proc_open_confirm(data);
1601                 if (status != 0)
1602                         return status;
1603         }
1604         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1605                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1606         return 0;
1607 }
1608
1609 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1610 {
1611         unsigned int loop;
1612         int ret;
1613
1614         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1615                 ret = nfs4_wait_clnt_recover(clp);
1616                 if (ret != 0)
1617                         break;
1618                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1619                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1620                         break;
1621                 nfs4_schedule_state_manager(clp);
1622                 ret = -EIO;
1623         }
1624         return ret;
1625 }
1626
1627 static int nfs4_recover_expired_lease(struct nfs_server *server)
1628 {
1629         return nfs4_client_recover_expired_lease(server->nfs_client);
1630 }
1631
1632 /*
1633  * OPEN_EXPIRED:
1634  *      reclaim state on the server after a network partition.
1635  *      Assumes caller holds the appropriate lock
1636  */
1637 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1638 {
1639         struct nfs4_opendata *opendata;
1640         int ret;
1641
1642         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1643         if (IS_ERR(opendata))
1644                 return PTR_ERR(opendata);
1645         ret = nfs4_open_recover(opendata, state);
1646         if (ret == -ESTALE)
1647                 d_drop(ctx->path.dentry);
1648         nfs4_opendata_put(opendata);
1649         return ret;
1650 }
1651
1652 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1653 {
1654         struct nfs_server *server = NFS_SERVER(state->inode);
1655         struct nfs4_exception exception = { };
1656         int err;
1657
1658         do {
1659                 err = _nfs4_open_expired(ctx, state);
1660                 switch (err) {
1661                 default:
1662                         goto out;
1663                 case -NFS4ERR_GRACE:
1664                 case -NFS4ERR_DELAY:
1665                         nfs4_handle_exception(server, err, &exception);
1666                         err = 0;
1667                 }
1668         } while (exception.retry);
1669 out:
1670         return err;
1671 }
1672
1673 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1674 {
1675         struct nfs_open_context *ctx;
1676         int ret;
1677
1678         ctx = nfs4_state_find_open_context(state);
1679         if (IS_ERR(ctx))
1680                 return PTR_ERR(ctx);
1681         ret = nfs4_do_open_expired(ctx, state);
1682         put_nfs_open_context(ctx);
1683         return ret;
1684 }
1685
1686 /*
1687  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1688  * fields corresponding to attributes that were used to store the verifier.
1689  * Make sure we clobber those fields in the later setattr call
1690  */
1691 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1692 {
1693         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1694             !(sattr->ia_valid & ATTR_ATIME_SET))
1695                 sattr->ia_valid |= ATTR_ATIME;
1696
1697         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1698             !(sattr->ia_valid & ATTR_MTIME_SET))
1699                 sattr->ia_valid |= ATTR_MTIME;
1700 }
1701
1702 /*
1703  * Returns a referenced nfs4_state
1704  */
1705 static int _nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1706 {
1707         struct nfs4_state_owner  *sp;
1708         struct nfs4_state     *state = NULL;
1709         struct nfs_server       *server = NFS_SERVER(dir);
1710         struct nfs4_opendata *opendata;
1711         int status;
1712
1713         /* Protect against reboot recovery conflicts */
1714         status = -ENOMEM;
1715         if (!(sp = nfs4_get_state_owner(server, cred))) {
1716                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1717                 goto out_err;
1718         }
1719         status = nfs4_recover_expired_lease(server);
1720         if (status != 0)
1721                 goto err_put_state_owner;
1722         if (path->dentry->d_inode != NULL)
1723                 nfs4_return_incompatible_delegation(path->dentry->d_inode, fmode);
1724         status = -ENOMEM;
1725         opendata = nfs4_opendata_alloc(path, sp, fmode, flags, sattr, GFP_KERNEL);
1726         if (opendata == NULL)
1727                 goto err_put_state_owner;
1728
1729         if (path->dentry->d_inode != NULL)
1730                 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1731
1732         status = _nfs4_proc_open(opendata);
1733         if (status != 0)
1734                 goto err_opendata_put;
1735
1736         state = nfs4_opendata_to_nfs4_state(opendata);
1737         status = PTR_ERR(state);
1738         if (IS_ERR(state))
1739                 goto err_opendata_put;
1740         if (server->caps & NFS_CAP_POSIX_LOCK)
1741                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1742
1743         if (opendata->o_arg.open_flags & O_EXCL) {
1744                 nfs4_exclusive_attrset(opendata, sattr);
1745
1746                 nfs_fattr_init(opendata->o_res.f_attr);
1747                 status = nfs4_do_setattr(state->inode, cred,
1748                                 opendata->o_res.f_attr, sattr,
1749                                 state);
1750                 if (status == 0)
1751                         nfs_setattr_update_inode(state->inode, sattr);
1752                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1753         }
1754         nfs4_opendata_put(opendata);
1755         nfs4_put_state_owner(sp);
1756         *res = state;
1757         return 0;
1758 err_opendata_put:
1759         nfs4_opendata_put(opendata);
1760 err_put_state_owner:
1761         nfs4_put_state_owner(sp);
1762 out_err:
1763         *res = NULL;
1764         return status;
1765 }
1766
1767
1768 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1769 {
1770         struct nfs4_exception exception = { };
1771         struct nfs4_state *res;
1772         int status;
1773
1774         do {
1775                 status = _nfs4_do_open(dir, path, fmode, flags, sattr, cred, &res);
1776                 if (status == 0)
1777                         break;
1778                 /* NOTE: BAD_SEQID means the server and client disagree about the
1779                  * book-keeping w.r.t. state-changing operations
1780                  * (OPEN/CLOSE/LOCK/LOCKU...)
1781                  * It is actually a sign of a bug on the client or on the server.
1782                  *
1783                  * If we receive a BAD_SEQID error in the particular case of
1784                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1785                  * have unhashed the old state_owner for us, and that we can
1786                  * therefore safely retry using a new one. We should still warn
1787                  * the user though...
1788                  */
1789                 if (status == -NFS4ERR_BAD_SEQID) {
1790                         printk(KERN_WARNING "NFS: v4 server %s "
1791                                         " returned a bad sequence-id error!\n",
1792                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1793                         exception.retry = 1;
1794                         continue;
1795                 }
1796                 /*
1797                  * BAD_STATEID on OPEN means that the server cancelled our
1798                  * state before it received the OPEN_CONFIRM.
1799                  * Recover by retrying the request as per the discussion
1800                  * on Page 181 of RFC3530.
1801                  */
1802                 if (status == -NFS4ERR_BAD_STATEID) {
1803                         exception.retry = 1;
1804                         continue;
1805                 }
1806                 if (status == -EAGAIN) {
1807                         /* We must have found a delegation */
1808                         exception.retry = 1;
1809                         continue;
1810                 }
1811                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1812                                         status, &exception));
1813         } while (exception.retry);
1814         return res;
1815 }
1816
1817 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1818                             struct nfs_fattr *fattr, struct iattr *sattr,
1819                             struct nfs4_state *state)
1820 {
1821         struct nfs_server *server = NFS_SERVER(inode);
1822         struct nfs_setattrargs  arg = {
1823                 .fh             = NFS_FH(inode),
1824                 .iap            = sattr,
1825                 .server         = server,
1826                 .bitmask = server->attr_bitmask,
1827         };
1828         struct nfs_setattrres  res = {
1829                 .fattr          = fattr,
1830                 .server         = server,
1831         };
1832         struct rpc_message msg = {
1833                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1834                 .rpc_argp       = &arg,
1835                 .rpc_resp       = &res,
1836                 .rpc_cred       = cred,
1837         };
1838         unsigned long timestamp = jiffies;
1839         int status;
1840
1841         nfs_fattr_init(fattr);
1842
1843         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1844                 /* Use that stateid */
1845         } else if (state != NULL) {
1846                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1847         } else
1848                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1849
1850         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1851         if (status == 0 && state != NULL)
1852                 renew_lease(server, timestamp);
1853         return status;
1854 }
1855
1856 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1857                            struct nfs_fattr *fattr, struct iattr *sattr,
1858                            struct nfs4_state *state)
1859 {
1860         struct nfs_server *server = NFS_SERVER(inode);
1861         struct nfs4_exception exception = { };
1862         int err;
1863         do {
1864                 err = nfs4_handle_exception(server,
1865                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1866                                 &exception);
1867         } while (exception.retry);
1868         return err;
1869 }
1870
1871 struct nfs4_closedata {
1872         struct path path;
1873         struct inode *inode;
1874         struct nfs4_state *state;
1875         struct nfs_closeargs arg;
1876         struct nfs_closeres res;
1877         struct nfs_fattr fattr;
1878         unsigned long timestamp;
1879         bool roc;
1880         u32 roc_barrier;
1881 };
1882
1883 static void nfs4_free_closedata(void *data)
1884 {
1885         struct nfs4_closedata *calldata = data;
1886         struct nfs4_state_owner *sp = calldata->state->owner;
1887
1888         if (calldata->roc)
1889                 pnfs_roc_release(calldata->state->inode);
1890         nfs4_put_open_state(calldata->state);
1891         nfs_free_seqid(calldata->arg.seqid);
1892         nfs4_put_state_owner(sp);
1893         path_put(&calldata->path);
1894         kfree(calldata);
1895 }
1896
1897 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1898                 fmode_t fmode)
1899 {
1900         spin_lock(&state->owner->so_lock);
1901         if (!(fmode & FMODE_READ))
1902                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1903         if (!(fmode & FMODE_WRITE))
1904                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1905         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1906         spin_unlock(&state->owner->so_lock);
1907 }
1908
1909 static void nfs4_close_done(struct rpc_task *task, void *data)
1910 {
1911         struct nfs4_closedata *calldata = data;
1912         struct nfs4_state *state = calldata->state;
1913         struct nfs_server *server = NFS_SERVER(calldata->inode);
1914
1915         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1916                 return;
1917         /* hmm. we are done with the inode, and in the process of freeing
1918          * the state_owner. we keep this around to process errors
1919          */
1920         switch (task->tk_status) {
1921                 case 0:
1922                         if (calldata->roc)
1923                                 pnfs_roc_set_barrier(state->inode,
1924                                                      calldata->roc_barrier);
1925                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1926                         renew_lease(server, calldata->timestamp);
1927                         nfs4_close_clear_stateid_flags(state,
1928                                         calldata->arg.fmode);
1929                         break;
1930                 case -NFS4ERR_STALE_STATEID:
1931                 case -NFS4ERR_OLD_STATEID:
1932                 case -NFS4ERR_BAD_STATEID:
1933                 case -NFS4ERR_EXPIRED:
1934                         if (calldata->arg.fmode == 0)
1935                                 break;
1936                 default:
1937                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1938                                 rpc_restart_call_prepare(task);
1939         }
1940         nfs_release_seqid(calldata->arg.seqid);
1941         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1942 }
1943
1944 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1945 {
1946         struct nfs4_closedata *calldata = data;
1947         struct nfs4_state *state = calldata->state;
1948         int call_close = 0;
1949
1950         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1951                 return;
1952
1953         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1954         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1955         spin_lock(&state->owner->so_lock);
1956         /* Calculate the change in open mode */
1957         if (state->n_rdwr == 0) {
1958                 if (state->n_rdonly == 0) {
1959                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
1960                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1961                         calldata->arg.fmode &= ~FMODE_READ;
1962                 }
1963                 if (state->n_wronly == 0) {
1964                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
1965                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1966                         calldata->arg.fmode &= ~FMODE_WRITE;
1967                 }
1968         }
1969         spin_unlock(&state->owner->so_lock);
1970
1971         if (!call_close) {
1972                 /* Note: exit _without_ calling nfs4_close_done */
1973                 task->tk_action = NULL;
1974                 return;
1975         }
1976
1977         if (calldata->arg.fmode == 0) {
1978                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
1979                 if (calldata->roc &&
1980                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
1981                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
1982                                      task, NULL);
1983                         return;
1984                 }
1985         }
1986
1987         nfs_fattr_init(calldata->res.fattr);
1988         calldata->timestamp = jiffies;
1989         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
1990                                 &calldata->arg.seq_args, &calldata->res.seq_res,
1991                                 1, task))
1992                 return;
1993         rpc_call_start(task);
1994 }
1995
1996 static const struct rpc_call_ops nfs4_close_ops = {
1997         .rpc_call_prepare = nfs4_close_prepare,
1998         .rpc_call_done = nfs4_close_done,
1999         .rpc_release = nfs4_free_closedata,
2000 };
2001
2002 /*
2003  * It is possible for data to be read/written from a mem-mapped file
2004  * after the sys_close call (which hits the vfs layer as a flush).
2005  * This means that we can't safely call nfsv4 close on a file until
2006  * the inode is cleared. This in turn means that we are not good
2007  * NFSv4 citizens - we do not indicate to the server to update the file's
2008  * share state even when we are done with one of the three share
2009  * stateid's in the inode.
2010  *
2011  * NOTE: Caller must be holding the sp->so_owner semaphore!
2012  */
2013 int nfs4_do_close(struct path *path, struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2014 {
2015         struct nfs_server *server = NFS_SERVER(state->inode);
2016         struct nfs4_closedata *calldata;
2017         struct nfs4_state_owner *sp = state->owner;
2018         struct rpc_task *task;
2019         struct rpc_message msg = {
2020                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2021                 .rpc_cred = state->owner->so_cred,
2022         };
2023         struct rpc_task_setup task_setup_data = {
2024                 .rpc_client = server->client,
2025                 .rpc_message = &msg,
2026                 .callback_ops = &nfs4_close_ops,
2027                 .workqueue = nfsiod_workqueue,
2028                 .flags = RPC_TASK_ASYNC,
2029         };
2030         int status = -ENOMEM;
2031
2032         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2033         if (calldata == NULL)
2034                 goto out;
2035         calldata->inode = state->inode;
2036         calldata->state = state;
2037         calldata->arg.fh = NFS_FH(state->inode);
2038         calldata->arg.stateid = &state->open_stateid;
2039         /* Serialization for the sequence id */
2040         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2041         if (calldata->arg.seqid == NULL)
2042                 goto out_free_calldata;
2043         calldata->arg.fmode = 0;
2044         calldata->arg.bitmask = server->cache_consistency_bitmask;
2045         calldata->res.fattr = &calldata->fattr;
2046         calldata->res.seqid = calldata->arg.seqid;
2047         calldata->res.server = server;
2048         calldata->roc = roc;
2049         path_get(path);
2050         calldata->path = *path;
2051
2052         msg.rpc_argp = &calldata->arg;
2053         msg.rpc_resp = &calldata->res;
2054         task_setup_data.callback_data = calldata;
2055         task = rpc_run_task(&task_setup_data);
2056         if (IS_ERR(task))
2057                 return PTR_ERR(task);
2058         status = 0;
2059         if (wait)
2060                 status = rpc_wait_for_completion_task(task);
2061         rpc_put_task(task);
2062         return status;
2063 out_free_calldata:
2064         kfree(calldata);
2065 out:
2066         if (roc)
2067                 pnfs_roc_release(state->inode);
2068         nfs4_put_open_state(state);
2069         nfs4_put_state_owner(sp);
2070         return status;
2071 }
2072
2073 static struct inode *
2074 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2075 {
2076         struct nfs4_state *state;
2077
2078         /* Protect against concurrent sillydeletes */
2079         state = nfs4_do_open(dir, &ctx->path, ctx->mode, open_flags, attr, ctx->cred);
2080         if (IS_ERR(state))
2081                 return ERR_CAST(state);
2082         ctx->state = state;
2083         return igrab(state->inode);
2084 }
2085
2086 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2087 {
2088         if (ctx->state == NULL)
2089                 return;
2090         if (is_sync)
2091                 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
2092         else
2093                 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
2094 }
2095
2096 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2097 {
2098         struct nfs4_server_caps_arg args = {
2099                 .fhandle = fhandle,
2100         };
2101         struct nfs4_server_caps_res res = {};
2102         struct rpc_message msg = {
2103                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2104                 .rpc_argp = &args,
2105                 .rpc_resp = &res,
2106         };
2107         int status;
2108
2109         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2110         if (status == 0) {
2111                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2112                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2113                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2114                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2115                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2116                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2117                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2118                         server->caps |= NFS_CAP_ACLS;
2119                 if (res.has_links != 0)
2120                         server->caps |= NFS_CAP_HARDLINKS;
2121                 if (res.has_symlinks != 0)
2122                         server->caps |= NFS_CAP_SYMLINKS;
2123                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2124                         server->caps |= NFS_CAP_FILEID;
2125                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2126                         server->caps |= NFS_CAP_MODE;
2127                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2128                         server->caps |= NFS_CAP_NLINK;
2129                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2130                         server->caps |= NFS_CAP_OWNER;
2131                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2132                         server->caps |= NFS_CAP_OWNER_GROUP;
2133                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2134                         server->caps |= NFS_CAP_ATIME;
2135                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2136                         server->caps |= NFS_CAP_CTIME;
2137                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2138                         server->caps |= NFS_CAP_MTIME;
2139
2140                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2141                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2142                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2143                 server->acl_bitmask = res.acl_bitmask;
2144         }
2145
2146         return status;
2147 }
2148
2149 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2150 {
2151         struct nfs4_exception exception = { };
2152         int err;
2153         do {
2154                 err = nfs4_handle_exception(server,
2155                                 _nfs4_server_capabilities(server, fhandle),
2156                                 &exception);
2157         } while (exception.retry);
2158         return err;
2159 }
2160
2161 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2162                 struct nfs_fsinfo *info)
2163 {
2164         struct nfs4_lookup_root_arg args = {
2165                 .bitmask = nfs4_fattr_bitmap,
2166         };
2167         struct nfs4_lookup_res res = {
2168                 .server = server,
2169                 .fattr = info->fattr,
2170                 .fh = fhandle,
2171         };
2172         struct rpc_message msg = {
2173                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2174                 .rpc_argp = &args,
2175                 .rpc_resp = &res,
2176         };
2177
2178         nfs_fattr_init(info->fattr);
2179         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2180 }
2181
2182 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2183                 struct nfs_fsinfo *info)
2184 {
2185         struct nfs4_exception exception = { };
2186         int err;
2187         do {
2188                 err = nfs4_handle_exception(server,
2189                                 _nfs4_lookup_root(server, fhandle, info),
2190                                 &exception);
2191         } while (exception.retry);
2192         return err;
2193 }
2194
2195 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2196                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2197 {
2198         struct rpc_auth *auth;
2199         int ret;
2200
2201         auth = rpcauth_create(flavor, server->client);
2202         if (!auth) {
2203                 ret = -EIO;
2204                 goto out;
2205         }
2206         ret = nfs4_lookup_root(server, fhandle, info);
2207 out:
2208         return ret;
2209 }
2210
2211 /*
2212  * get the file handle for the "/" directory on the server
2213  */
2214 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2215                               struct nfs_fsinfo *info)
2216 {
2217         int i, len, status = 0;
2218         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS + 2];
2219
2220         flav_array[0] = RPC_AUTH_UNIX;
2221         len = gss_mech_list_pseudoflavors(&flav_array[1]);
2222         flav_array[1+len] = RPC_AUTH_NULL;
2223         len += 2;
2224
2225         for (i = 0; i < len; i++) {
2226                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2227                 if (status != -EPERM)
2228                         break;
2229         }
2230         if (status == 0)
2231                 status = nfs4_server_capabilities(server, fhandle);
2232         if (status == 0)
2233                 status = nfs4_do_fsinfo(server, fhandle, info);
2234         return nfs4_map_errors(status);
2235 }
2236
2237 /*
2238  * Get locations and (maybe) other attributes of a referral.
2239  * Note that we'll actually follow the referral later when
2240  * we detect fsid mismatch in inode revalidation
2241  */
2242 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2243 {
2244         int status = -ENOMEM;
2245         struct page *page = NULL;
2246         struct nfs4_fs_locations *locations = NULL;
2247
2248         page = alloc_page(GFP_KERNEL);
2249         if (page == NULL)
2250                 goto out;
2251         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2252         if (locations == NULL)
2253                 goto out;
2254
2255         status = nfs4_proc_fs_locations(dir, name, locations, page);
2256         if (status != 0)
2257                 goto out;
2258         /* Make sure server returned a different fsid for the referral */
2259         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2260                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2261                 status = -EIO;
2262                 goto out;
2263         }
2264
2265         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2266         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2267         if (!fattr->mode)
2268                 fattr->mode = S_IFDIR;
2269         memset(fhandle, 0, sizeof(struct nfs_fh));
2270 out:
2271         if (page)
2272                 __free_page(page);
2273         kfree(locations);
2274         return status;
2275 }
2276
2277 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2278 {
2279         struct nfs4_getattr_arg args = {
2280                 .fh = fhandle,
2281                 .bitmask = server->attr_bitmask,
2282         };
2283         struct nfs4_getattr_res res = {
2284                 .fattr = fattr,
2285                 .server = server,
2286         };
2287         struct rpc_message msg = {
2288                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2289                 .rpc_argp = &args,
2290                 .rpc_resp = &res,
2291         };
2292
2293         nfs_fattr_init(fattr);
2294         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2295 }
2296
2297 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2298 {
2299         struct nfs4_exception exception = { };
2300         int err;
2301         do {
2302                 err = nfs4_handle_exception(server,
2303                                 _nfs4_proc_getattr(server, fhandle, fattr),
2304                                 &exception);
2305         } while (exception.retry);
2306         return err;
2307 }
2308
2309 /*
2310  * The file is not closed if it is opened due to the a request to change
2311  * the size of the file. The open call will not be needed once the
2312  * VFS layer lookup-intents are implemented.
2313  *
2314  * Close is called when the inode is destroyed.
2315  * If we haven't opened the file for O_WRONLY, we
2316  * need to in the size_change case to obtain a stateid.
2317  *
2318  * Got race?
2319  * Because OPEN is always done by name in nfsv4, it is
2320  * possible that we opened a different file by the same
2321  * name.  We can recognize this race condition, but we
2322  * can't do anything about it besides returning an error.
2323  *
2324  * This will be fixed with VFS changes (lookup-intent).
2325  */
2326 static int
2327 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2328                   struct iattr *sattr)
2329 {
2330         struct inode *inode = dentry->d_inode;
2331         struct rpc_cred *cred = NULL;
2332         struct nfs4_state *state = NULL;
2333         int status;
2334
2335         nfs_fattr_init(fattr);
2336
2337         /* Search for an existing open(O_WRITE) file */
2338         if (sattr->ia_valid & ATTR_FILE) {
2339                 struct nfs_open_context *ctx;
2340
2341                 ctx = nfs_file_open_context(sattr->ia_file);
2342                 if (ctx) {
2343                         cred = ctx->cred;
2344                         state = ctx->state;
2345                 }
2346         }
2347
2348         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2349         if (status == 0)
2350                 nfs_setattr_update_inode(inode, sattr);
2351         return status;
2352 }
2353
2354 static int _nfs4_proc_lookupfh(struct rpc_clnt *clnt, struct nfs_server *server,
2355                 const struct nfs_fh *dirfh, const struct qstr *name,
2356                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2357 {
2358         int                    status;
2359         struct nfs4_lookup_arg args = {
2360                 .bitmask = server->attr_bitmask,
2361                 .dir_fh = dirfh,
2362                 .name = name,
2363         };
2364         struct nfs4_lookup_res res = {
2365                 .server = server,
2366                 .fattr = fattr,
2367                 .fh = fhandle,
2368         };
2369         struct rpc_message msg = {
2370                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2371                 .rpc_argp = &args,
2372                 .rpc_resp = &res,
2373         };
2374
2375         nfs_fattr_init(fattr);
2376
2377         dprintk("NFS call  lookupfh %s\n", name->name);
2378         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2379         dprintk("NFS reply lookupfh: %d\n", status);
2380         return status;
2381 }
2382
2383 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2384                               struct qstr *name, struct nfs_fh *fhandle,
2385                               struct nfs_fattr *fattr)
2386 {
2387         struct nfs4_exception exception = { };
2388         int err;
2389         do {
2390                 err = _nfs4_proc_lookupfh(server->client, server, dirfh, name, fhandle, fattr);
2391                 /* FIXME: !!!! */
2392                 if (err == -NFS4ERR_MOVED) {
2393                         err = -EREMOTE;
2394                         break;
2395                 }
2396                 err = nfs4_handle_exception(server, err, &exception);
2397         } while (exception.retry);
2398         return err;
2399 }
2400
2401 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2402                 const struct qstr *name, struct nfs_fh *fhandle,
2403                 struct nfs_fattr *fattr)
2404 {
2405         int status;
2406
2407         dprintk("NFS call  lookup %s\n", name->name);
2408         status = _nfs4_proc_lookupfh(clnt, NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2409         if (status == -NFS4ERR_MOVED)
2410                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2411         dprintk("NFS reply lookup: %d\n", status);
2412         return status;
2413 }
2414
2415 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2416 {
2417         memset(fh, 0, sizeof(struct nfs_fh));
2418         fattr->fsid.major = 1;
2419         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2420                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2421         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2422         fattr->nlink = 2;
2423 }
2424
2425 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2426                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2427 {
2428         struct nfs4_exception exception = { };
2429         int err;
2430         do {
2431                 err = nfs4_handle_exception(NFS_SERVER(dir),
2432                                 _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr),
2433                                 &exception);
2434                 if (err == -EPERM)
2435                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2436         } while (exception.retry);
2437         return err;
2438 }
2439
2440 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2441 {
2442         struct nfs_server *server = NFS_SERVER(inode);
2443         struct nfs4_accessargs args = {
2444                 .fh = NFS_FH(inode),
2445                 .bitmask = server->attr_bitmask,
2446         };
2447         struct nfs4_accessres res = {
2448                 .server = server,
2449         };
2450         struct rpc_message msg = {
2451                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2452                 .rpc_argp = &args,
2453                 .rpc_resp = &res,
2454                 .rpc_cred = entry->cred,
2455         };
2456         int mode = entry->mask;
2457         int status;
2458
2459         /*
2460          * Determine which access bits we want to ask for...
2461          */
2462         if (mode & MAY_READ)
2463                 args.access |= NFS4_ACCESS_READ;
2464         if (S_ISDIR(inode->i_mode)) {
2465                 if (mode & MAY_WRITE)
2466                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2467                 if (mode & MAY_EXEC)
2468                         args.access |= NFS4_ACCESS_LOOKUP;
2469         } else {
2470                 if (mode & MAY_WRITE)
2471                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2472                 if (mode & MAY_EXEC)
2473                         args.access |= NFS4_ACCESS_EXECUTE;
2474         }
2475
2476         res.fattr = nfs_alloc_fattr();
2477         if (res.fattr == NULL)
2478                 return -ENOMEM;
2479
2480         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2481         if (!status) {
2482                 entry->mask = 0;
2483                 if (res.access & NFS4_ACCESS_READ)
2484                         entry->mask |= MAY_READ;
2485                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2486                         entry->mask |= MAY_WRITE;
2487                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2488                         entry->mask |= MAY_EXEC;
2489                 nfs_refresh_inode(inode, res.fattr);
2490         }
2491         nfs_free_fattr(res.fattr);
2492         return status;
2493 }
2494
2495 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2496 {
2497         struct nfs4_exception exception = { };
2498         int err;
2499         do {
2500                 err = nfs4_handle_exception(NFS_SERVER(inode),
2501                                 _nfs4_proc_access(inode, entry),
2502                                 &exception);
2503         } while (exception.retry);
2504         return err;
2505 }
2506
2507 /*
2508  * TODO: For the time being, we don't try to get any attributes
2509  * along with any of the zero-copy operations READ, READDIR,
2510  * READLINK, WRITE.
2511  *
2512  * In the case of the first three, we want to put the GETATTR
2513  * after the read-type operation -- this is because it is hard
2514  * to predict the length of a GETATTR response in v4, and thus
2515  * align the READ data correctly.  This means that the GETATTR
2516  * may end up partially falling into the page cache, and we should
2517  * shift it into the 'tail' of the xdr_buf before processing.
2518  * To do this efficiently, we need to know the total length
2519  * of data received, which doesn't seem to be available outside
2520  * of the RPC layer.
2521  *
2522  * In the case of WRITE, we also want to put the GETATTR after
2523  * the operation -- in this case because we want to make sure
2524  * we get the post-operation mtime and size.  This means that
2525  * we can't use xdr_encode_pages() as written: we need a variant
2526  * of it which would leave room in the 'tail' iovec.
2527  *
2528  * Both of these changes to the XDR layer would in fact be quite
2529  * minor, but I decided to leave them for a subsequent patch.
2530  */
2531 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2532                 unsigned int pgbase, unsigned int pglen)
2533 {
2534         struct nfs4_readlink args = {
2535                 .fh       = NFS_FH(inode),
2536                 .pgbase   = pgbase,
2537                 .pglen    = pglen,
2538                 .pages    = &page,
2539         };
2540         struct nfs4_readlink_res res;
2541         struct rpc_message msg = {
2542                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2543                 .rpc_argp = &args,
2544                 .rpc_resp = &res,
2545         };
2546
2547         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2548 }
2549
2550 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2551                 unsigned int pgbase, unsigned int pglen)
2552 {
2553         struct nfs4_exception exception = { };
2554         int err;
2555         do {
2556                 err = nfs4_handle_exception(NFS_SERVER(inode),
2557                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2558                                 &exception);
2559         } while (exception.retry);
2560         return err;
2561 }
2562
2563 /*
2564  * Got race?
2565  * We will need to arrange for the VFS layer to provide an atomic open.
2566  * Until then, this create/open method is prone to inefficiency and race
2567  * conditions due to the lookup, create, and open VFS calls from sys_open()
2568  * placed on the wire.
2569  *
2570  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2571  * The file will be opened again in the subsequent VFS open call
2572  * (nfs4_proc_file_open).
2573  *
2574  * The open for read will just hang around to be used by any process that
2575  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2576  */
2577
2578 static int
2579 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2580                  int flags, struct nfs_open_context *ctx)
2581 {
2582         struct path my_path = {
2583                 .dentry = dentry,
2584         };
2585         struct path *path = &my_path;
2586         struct nfs4_state *state;
2587         struct rpc_cred *cred = NULL;
2588         fmode_t fmode = 0;
2589         int status = 0;
2590
2591         if (ctx != NULL) {
2592                 cred = ctx->cred;
2593                 path = &ctx->path;
2594                 fmode = ctx->mode;
2595         }
2596         sattr->ia_mode &= ~current_umask();
2597         state = nfs4_do_open(dir, path, fmode, flags, sattr, cred);
2598         d_drop(dentry);
2599         if (IS_ERR(state)) {
2600                 status = PTR_ERR(state);
2601                 goto out;
2602         }
2603         d_add(dentry, igrab(state->inode));
2604         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2605         if (ctx != NULL)
2606                 ctx->state = state;
2607         else
2608                 nfs4_close_sync(path, state, fmode);
2609 out:
2610         return status;
2611 }
2612
2613 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2614 {
2615         struct nfs_server *server = NFS_SERVER(dir);
2616         struct nfs_removeargs args = {
2617                 .fh = NFS_FH(dir),
2618                 .name.len = name->len,
2619                 .name.name = name->name,
2620                 .bitmask = server->attr_bitmask,
2621         };
2622         struct nfs_removeres res = {
2623                 .server = server,
2624         };
2625         struct rpc_message msg = {
2626                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2627                 .rpc_argp = &args,
2628                 .rpc_resp = &res,
2629         };
2630         int status = -ENOMEM;
2631
2632         res.dir_attr = nfs_alloc_fattr();
2633         if (res.dir_attr == NULL)
2634                 goto out;
2635
2636         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2637         if (status == 0) {
2638                 update_changeattr(dir, &res.cinfo);
2639                 nfs_post_op_update_inode(dir, res.dir_attr);
2640         }
2641         nfs_free_fattr(res.dir_attr);
2642 out:
2643         return status;
2644 }
2645
2646 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2647 {
2648         struct nfs4_exception exception = { };
2649         int err;
2650         do {
2651                 err = nfs4_handle_exception(NFS_SERVER(dir),
2652                                 _nfs4_proc_remove(dir, name),
2653                                 &exception);
2654         } while (exception.retry);
2655         return err;
2656 }
2657
2658 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2659 {
2660         struct nfs_server *server = NFS_SERVER(dir);
2661         struct nfs_removeargs *args = msg->rpc_argp;
2662         struct nfs_removeres *res = msg->rpc_resp;
2663
2664         args->bitmask = server->cache_consistency_bitmask;
2665         res->server = server;
2666         res->seq_res.sr_slot = NULL;
2667         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2668 }
2669
2670 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2671 {
2672         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2673
2674         if (!nfs4_sequence_done(task, &res->seq_res))
2675                 return 0;
2676         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2677                 return 0;
2678         update_changeattr(dir, &res->cinfo);
2679         nfs_post_op_update_inode(dir, res->dir_attr);
2680         return 1;
2681 }
2682
2683 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2684 {
2685         struct nfs_server *server = NFS_SERVER(dir);
2686         struct nfs_renameargs *arg = msg->rpc_argp;
2687         struct nfs_renameres *res = msg->rpc_resp;
2688
2689         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2690         arg->bitmask = server->attr_bitmask;
2691         res->server = server;
2692 }
2693
2694 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2695                                  struct inode *new_dir)
2696 {
2697         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2698
2699         if (!nfs4_sequence_done(task, &res->seq_res))
2700                 return 0;
2701         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2702                 return 0;
2703
2704         update_changeattr(old_dir, &res->old_cinfo);
2705         nfs_post_op_update_inode(old_dir, res->old_fattr);
2706         update_changeattr(new_dir, &res->new_cinfo);
2707         nfs_post_op_update_inode(new_dir, res->new_fattr);
2708         return 1;
2709 }
2710
2711 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2712                 struct inode *new_dir, struct qstr *new_name)
2713 {
2714         struct nfs_server *server = NFS_SERVER(old_dir);
2715         struct nfs_renameargs arg = {
2716                 .old_dir = NFS_FH(old_dir),
2717                 .new_dir = NFS_FH(new_dir),
2718                 .old_name = old_name,
2719                 .new_name = new_name,
2720                 .bitmask = server->attr_bitmask,
2721         };
2722         struct nfs_renameres res = {
2723                 .server = server,
2724         };
2725         struct rpc_message msg = {
2726                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2727                 .rpc_argp = &arg,
2728                 .rpc_resp = &res,
2729         };
2730         int status = -ENOMEM;
2731
2732         res.old_fattr = nfs_alloc_fattr();
2733         res.new_fattr = nfs_alloc_fattr();
2734         if (res.old_fattr == NULL || res.new_fattr == NULL)
2735                 goto out;
2736
2737         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2738         if (!status) {
2739                 update_changeattr(old_dir, &res.old_cinfo);
2740                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2741                 update_changeattr(new_dir, &res.new_cinfo);
2742                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2743         }
2744 out:
2745         nfs_free_fattr(res.new_fattr);
2746         nfs_free_fattr(res.old_fattr);
2747         return status;
2748 }
2749
2750 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2751                 struct inode *new_dir, struct qstr *new_name)
2752 {
2753         struct nfs4_exception exception = { };
2754         int err;
2755         do {
2756                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2757                                 _nfs4_proc_rename(old_dir, old_name,
2758                                         new_dir, new_name),
2759                                 &exception);
2760         } while (exception.retry);
2761         return err;
2762 }
2763
2764 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2765 {
2766         struct nfs_server *server = NFS_SERVER(inode);
2767         struct nfs4_link_arg arg = {
2768                 .fh     = NFS_FH(inode),
2769                 .dir_fh = NFS_FH(dir),
2770                 .name   = name,
2771                 .bitmask = server->attr_bitmask,
2772         };
2773         struct nfs4_link_res res = {
2774                 .server = server,
2775         };
2776         struct rpc_message msg = {
2777                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2778                 .rpc_argp = &arg,
2779                 .rpc_resp = &res,
2780         };
2781         int status = -ENOMEM;
2782
2783         res.fattr = nfs_alloc_fattr();
2784         res.dir_attr = nfs_alloc_fattr();
2785         if (res.fattr == NULL || res.dir_attr == NULL)
2786                 goto out;
2787
2788         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2789         if (!status) {
2790                 update_changeattr(dir, &res.cinfo);
2791                 nfs_post_op_update_inode(dir, res.dir_attr);
2792                 nfs_post_op_update_inode(inode, res.fattr);
2793         }
2794 out:
2795         nfs_free_fattr(res.dir_attr);
2796         nfs_free_fattr(res.fattr);
2797         return status;
2798 }
2799
2800 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2801 {
2802         struct nfs4_exception exception = { };
2803         int err;
2804         do {
2805                 err = nfs4_handle_exception(NFS_SERVER(inode),
2806                                 _nfs4_proc_link(inode, dir, name),
2807                                 &exception);
2808         } while (exception.retry);
2809         return err;
2810 }
2811
2812 struct nfs4_createdata {
2813         struct rpc_message msg;
2814         struct nfs4_create_arg arg;
2815         struct nfs4_create_res res;
2816         struct nfs_fh fh;
2817         struct nfs_fattr fattr;
2818         struct nfs_fattr dir_fattr;
2819 };
2820
2821 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2822                 struct qstr *name, struct iattr *sattr, u32 ftype)
2823 {
2824         struct nfs4_createdata *data;
2825
2826         data = kzalloc(sizeof(*data), GFP_KERNEL);
2827         if (data != NULL) {
2828                 struct nfs_server *server = NFS_SERVER(dir);
2829
2830                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2831                 data->msg.rpc_argp = &data->arg;
2832                 data->msg.rpc_resp = &data->res;
2833                 data->arg.dir_fh = NFS_FH(dir);
2834                 data->arg.server = server;
2835                 data->arg.name = name;
2836                 data->arg.attrs = sattr;
2837                 data->arg.ftype = ftype;
2838                 data->arg.bitmask = server->attr_bitmask;
2839                 data->res.server = server;
2840                 data->res.fh = &data->fh;
2841                 data->res.fattr = &data->fattr;
2842                 data->res.dir_fattr = &data->dir_fattr;
2843                 nfs_fattr_init(data->res.fattr);
2844                 nfs_fattr_init(data->res.dir_fattr);
2845         }
2846         return data;
2847 }
2848
2849 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2850 {
2851         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2852                                     &data->arg.seq_args, &data->res.seq_res, 1);
2853         if (status == 0) {
2854                 update_changeattr(dir, &data->res.dir_cinfo);
2855                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2856                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2857         }
2858         return status;
2859 }
2860
2861 static void nfs4_free_createdata(struct nfs4_createdata *data)
2862 {
2863         kfree(data);
2864 }
2865
2866 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2867                 struct page *page, unsigned int len, struct iattr *sattr)
2868 {
2869         struct nfs4_createdata *data;
2870         int status = -ENAMETOOLONG;
2871
2872         if (len > NFS4_MAXPATHLEN)
2873                 goto out;
2874
2875         status = -ENOMEM;
2876         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2877         if (data == NULL)
2878                 goto out;
2879
2880         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2881         data->arg.u.symlink.pages = &page;
2882         data->arg.u.symlink.len = len;
2883
2884         status = nfs4_do_create(dir, dentry, data);
2885
2886         nfs4_free_createdata(data);
2887 out:
2888         return status;
2889 }
2890
2891 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2892                 struct page *page, unsigned int len, struct iattr *sattr)
2893 {
2894         struct nfs4_exception exception = { };
2895         int err;
2896         do {
2897                 err = nfs4_handle_exception(NFS_SERVER(dir),
2898                                 _nfs4_proc_symlink(dir, dentry, page,
2899                                                         len, sattr),
2900                                 &exception);
2901         } while (exception.retry);
2902         return err;
2903 }
2904
2905 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2906                 struct iattr *sattr)
2907 {
2908         struct nfs4_createdata *data;
2909         int status = -ENOMEM;
2910
2911         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2912         if (data == NULL)
2913                 goto out;
2914
2915         status = nfs4_do_create(dir, dentry, data);
2916
2917         nfs4_free_createdata(data);
2918 out:
2919         return status;
2920 }
2921
2922 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2923                 struct iattr *sattr)
2924 {
2925         struct nfs4_exception exception = { };
2926         int err;
2927
2928         sattr->ia_mode &= ~current_umask();
2929         do {
2930                 err = nfs4_handle_exception(NFS_SERVER(dir),
2931                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2932                                 &exception);
2933         } while (exception.retry);
2934         return err;
2935 }
2936
2937 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2938                 u64 cookie, struct page **pages, unsigned int count, int plus)
2939 {
2940         struct inode            *dir = dentry->d_inode;
2941         struct nfs4_readdir_arg args = {
2942                 .fh = NFS_FH(dir),
2943                 .pages = pages,
2944                 .pgbase = 0,
2945                 .count = count,
2946                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2947                 .plus = plus,
2948         };
2949         struct nfs4_readdir_res res;
2950         struct rpc_message msg = {
2951                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2952                 .rpc_argp = &args,
2953                 .rpc_resp = &res,
2954                 .rpc_cred = cred,
2955         };
2956         int                     status;
2957
2958         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2959                         dentry->d_parent->d_name.name,
2960                         dentry->d_name.name,
2961                         (unsigned long long)cookie);
2962         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2963         res.pgbase = args.pgbase;
2964         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
2965         if (status >= 0) {
2966                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2967                 status += args.pgbase;
2968         }
2969
2970         nfs_invalidate_atime(dir);
2971
2972         dprintk("%s: returns %d\n", __func__, status);
2973         return status;
2974 }
2975
2976 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2977                 u64 cookie, struct page **pages, unsigned int count, int plus)
2978 {
2979         struct nfs4_exception exception = { };
2980         int err;
2981         do {
2982                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
2983                                 _nfs4_proc_readdir(dentry, cred, cookie,
2984                                         pages, count, plus),
2985                                 &exception);
2986         } while (exception.retry);
2987         return err;
2988 }
2989
2990 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2991                 struct iattr *sattr, dev_t rdev)
2992 {
2993         struct nfs4_createdata *data;
2994         int mode = sattr->ia_mode;
2995         int status = -ENOMEM;
2996
2997         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
2998         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
2999
3000         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3001         if (data == NULL)
3002                 goto out;
3003
3004         if (S_ISFIFO(mode))
3005                 data->arg.ftype = NF4FIFO;
3006         else if (S_ISBLK(mode)) {
3007                 data->arg.ftype = NF4BLK;
3008                 data->arg.u.device.specdata1 = MAJOR(rdev);
3009                 data->arg.u.device.specdata2 = MINOR(rdev);
3010         }
3011         else if (S_ISCHR(mode)) {
3012                 data->arg.ftype = NF4CHR;
3013                 data->arg.u.device.specdata1 = MAJOR(rdev);
3014                 data->arg.u.device.specdata2 = MINOR(rdev);
3015         }
3016
3017         status = nfs4_do_create(dir, dentry, data);
3018
3019         nfs4_free_createdata(data);
3020 out:
3021         return status;
3022 }
3023
3024 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3025                 struct iattr *sattr, dev_t rdev)
3026 {
3027         struct nfs4_exception exception = { };
3028         int err;
3029
3030         sattr->ia_mode &= ~current_umask();
3031         do {
3032                 err = nfs4_handle_exception(NFS_SERVER(dir),
3033                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3034                                 &exception);
3035         } while (exception.retry);
3036         return err;
3037 }
3038
3039 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3040                  struct nfs_fsstat *fsstat)
3041 {
3042         struct nfs4_statfs_arg args = {
3043                 .fh = fhandle,
3044                 .bitmask = server->attr_bitmask,
3045         };
3046         struct nfs4_statfs_res res = {
3047                 .fsstat = fsstat,
3048         };
3049         struct rpc_message msg = {
3050                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3051                 .rpc_argp = &args,
3052                 .rpc_resp = &res,
3053         };
3054
3055         nfs_fattr_init(fsstat->fattr);
3056         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3057 }
3058
3059 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3060 {
3061         struct nfs4_exception exception = { };
3062         int err;
3063         do {
3064                 err = nfs4_handle_exception(server,
3065                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3066                                 &exception);
3067         } while (exception.retry);
3068         return err;
3069 }
3070
3071 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3072                 struct nfs_fsinfo *fsinfo)
3073 {
3074         struct nfs4_fsinfo_arg args = {
3075                 .fh = fhandle,
3076                 .bitmask = server->attr_bitmask,
3077         };
3078         struct nfs4_fsinfo_res res = {
3079                 .fsinfo = fsinfo,
3080         };
3081         struct rpc_message msg = {
3082                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3083                 .rpc_argp = &args,
3084                 .rpc_resp = &res,
3085         };
3086
3087         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3088 }
3089
3090 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3091 {
3092         struct nfs4_exception exception = { };
3093         int err;
3094
3095         do {
3096                 err = nfs4_handle_exception(server,
3097                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3098                                 &exception);
3099         } while (exception.retry);
3100         return err;
3101 }
3102
3103 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3104 {
3105         nfs_fattr_init(fsinfo->fattr);
3106         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3107 }
3108
3109 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3110                 struct nfs_pathconf *pathconf)
3111 {
3112         struct nfs4_pathconf_arg args = {
3113                 .fh = fhandle,
3114                 .bitmask = server->attr_bitmask,
3115         };
3116         struct nfs4_pathconf_res res = {
3117                 .pathconf = pathconf,
3118         };
3119         struct rpc_message msg = {
3120                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3121                 .rpc_argp = &args,
3122                 .rpc_resp = &res,
3123         };
3124
3125         /* None of the pathconf attributes are mandatory to implement */
3126         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3127                 memset(pathconf, 0, sizeof(*pathconf));
3128                 return 0;
3129         }
3130
3131         nfs_fattr_init(pathconf->fattr);
3132         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3133 }
3134
3135 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3136                 struct nfs_pathconf *pathconf)
3137 {
3138         struct nfs4_exception exception = { };
3139         int err;
3140
3141         do {
3142                 err = nfs4_handle_exception(server,
3143                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3144                                 &exception);
3145         } while (exception.retry);
3146         return err;
3147 }
3148
3149 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3150 {
3151         struct nfs_server *server = NFS_SERVER(data->inode);
3152
3153         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3154                 nfs_restart_rpc(task, server->nfs_client);
3155                 return -EAGAIN;
3156         }
3157
3158         nfs_invalidate_atime(data->inode);
3159         if (task->tk_status > 0)
3160                 renew_lease(server, data->timestamp);
3161         return 0;
3162 }
3163
3164 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3165 {
3166
3167         dprintk("--> %s\n", __func__);
3168
3169         if (!nfs4_sequence_done(task, &data->res.seq_res))
3170                 return -EAGAIN;
3171
3172         return data->read_done_cb(task, data);
3173 }
3174
3175 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3176 {
3177         data->timestamp   = jiffies;
3178         data->read_done_cb = nfs4_read_done_cb;
3179         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3180 }
3181
3182 /* Reset the the nfs_read_data to send the read to the MDS. */
3183 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3184 {
3185         dprintk("%s Reset task for i/o through\n", __func__);
3186         put_lseg(data->lseg);
3187         data->lseg = NULL;
3188         /* offsets will differ in the dense stripe case */
3189         data->args.offset = data->mds_offset;
3190         data->ds_clp = NULL;
3191         data->args.fh     = NFS_FH(data->inode);
3192         data->read_done_cb = nfs4_read_done_cb;
3193         task->tk_ops = data->mds_ops;
3194         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3195 }
3196 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3197
3198 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3199 {
3200         struct inode *inode = data->inode;
3201
3202         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3203                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3204                 return -EAGAIN;
3205         }
3206         if (task->tk_status >= 0) {
3207                 renew_lease(NFS_SERVER(inode), data->timestamp);
3208                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3209         }
3210         return 0;
3211 }
3212
3213 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3214 {
3215         if (!nfs4_sequence_done(task, &data->res.seq_res))
3216                 return -EAGAIN;
3217         return data->write_done_cb(task, data);
3218 }
3219
3220 /* Reset the the nfs_write_data to send the write to the MDS. */
3221 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3222 {
3223         dprintk("%s Reset task for i/o through\n", __func__);
3224         put_lseg(data->lseg);
3225         data->lseg          = NULL;
3226         data->ds_clp        = NULL;
3227         data->write_done_cb = nfs4_write_done_cb;
3228         data->args.fh       = NFS_FH(data->inode);
3229         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3230         data->args.offset   = data->mds_offset;
3231         data->res.fattr     = &data->fattr;
3232         task->tk_ops        = data->mds_ops;
3233         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3234 }
3235 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3236
3237 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3238 {
3239         struct nfs_server *server = NFS_SERVER(data->inode);
3240
3241         if (data->lseg) {
3242                 data->args.bitmask = NULL;
3243                 data->res.fattr = NULL;
3244         } else
3245                 data->args.bitmask = server->cache_consistency_bitmask;
3246         if (!data->write_done_cb)
3247                 data->write_done_cb = nfs4_write_done_cb;
3248         data->res.server = server;
3249         data->timestamp   = jiffies;
3250
3251         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3252 }
3253
3254 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3255 {
3256         struct inode *inode = data->inode;
3257
3258         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3259                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3260                 return -EAGAIN;
3261         }
3262         nfs_refresh_inode(inode, data->res.fattr);
3263         return 0;
3264 }
3265
3266 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3267 {
3268         if (!nfs4_sequence_done(task, &data->res.seq_res))
3269                 return -EAGAIN;
3270         return data->write_done_cb(task, data);
3271 }
3272
3273 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3274 {
3275         struct nfs_server *server = NFS_SERVER(data->inode);
3276
3277         if (data->lseg) {
3278                 data->args.bitmask = NULL;
3279                 data->res.fattr = NULL;
3280         } else
3281                 data->args.bitmask = server->cache_consistency_bitmask;
3282         if (!data->write_done_cb)
3283                 data->write_done_cb = nfs4_commit_done_cb;
3284         data->res.server = server;
3285         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3286 }
3287
3288 struct nfs4_renewdata {
3289         struct nfs_client       *client;
3290         unsigned long           timestamp;
3291 };
3292
3293 /*
3294  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3295  * standalone procedure for queueing an asynchronous RENEW.
3296  */
3297 static void nfs4_renew_release(void *calldata)
3298 {
3299         struct nfs4_renewdata *data = calldata;
3300         struct nfs_client *clp = data->client;
3301
3302         if (atomic_read(&clp->cl_count) > 1)
3303                 nfs4_schedule_state_renewal(clp);
3304         nfs_put_client(clp);
3305         kfree(data);
3306 }
3307
3308 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3309 {
3310         struct nfs4_renewdata *data = calldata;
3311         struct nfs_client *clp = data->client;
3312         unsigned long timestamp = data->timestamp;
3313
3314         if (task->tk_status < 0) {
3315                 /* Unless we're shutting down, schedule state recovery! */
3316                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
3317                         nfs4_schedule_lease_recovery(clp);
3318                 return;
3319         }
3320         do_renew_lease(clp, timestamp);
3321 }
3322
3323 static const struct rpc_call_ops nfs4_renew_ops = {
3324         .rpc_call_done = nfs4_renew_done,
3325         .rpc_release = nfs4_renew_release,
3326 };
3327
3328 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
3329 {
3330         struct rpc_message msg = {
3331                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3332                 .rpc_argp       = clp,
3333                 .rpc_cred       = cred,
3334         };
3335         struct nfs4_renewdata *data;
3336
3337         if (!atomic_inc_not_zero(&clp->cl_count))
3338                 return -EIO;
3339         data = kmalloc(sizeof(*data), GFP_KERNEL);
3340         if (data == NULL)
3341                 return -ENOMEM;
3342         data->client = clp;
3343         data->timestamp = jiffies;
3344         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3345                         &nfs4_renew_ops, data);
3346 }
3347
3348 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3349 {
3350         struct rpc_message msg = {
3351                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3352                 .rpc_argp       = clp,
3353                 .rpc_cred       = cred,
3354         };
3355         unsigned long now = jiffies;
3356         int status;
3357
3358         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3359         if (status < 0)
3360                 return status;
3361         do_renew_lease(clp, now);
3362         return 0;
3363 }
3364
3365 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3366 {
3367         return (server->caps & NFS_CAP_ACLS)
3368                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3369                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3370 }
3371
3372 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3373  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3374  * the stack.
3375  */
3376 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3377
3378 static void buf_to_pages(const void *buf, size_t buflen,
3379                 struct page **pages, unsigned int *pgbase)
3380 {
3381         const void *p = buf;
3382
3383         *pgbase = offset_in_page(buf);
3384         p -= *pgbase;
3385         while (p < buf + buflen) {
3386                 *(pages++) = virt_to_page(p);
3387                 p += PAGE_CACHE_SIZE;
3388         }
3389 }
3390
3391 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3392                 struct page **pages, unsigned int *pgbase)
3393 {
3394         struct page *newpage, **spages;
3395         int rc = 0;
3396         size_t len;
3397         spages = pages;
3398
3399         do {
3400                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3401                 newpage = alloc_page(GFP_KERNEL);
3402
3403                 if (newpage == NULL)
3404                         goto unwind;
3405                 memcpy(page_address(newpage), buf, len);
3406                 buf += len;
3407                 buflen -= len;
3408                 *pages++ = newpage;
3409                 rc++;
3410         } while (buflen != 0);
3411
3412         return rc;
3413
3414 unwind:
3415         for(; rc > 0; rc--)
3416                 __free_page(spages[rc-1]);
3417         return -ENOMEM;
3418 }
3419
3420 struct nfs4_cached_acl {
3421         int cached;
3422         size_t len;
3423         char data[0];
3424 };
3425
3426 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3427 {
3428         struct nfs_inode *nfsi = NFS_I(inode);
3429
3430         spin_lock(&inode->i_lock);
3431         kfree(nfsi->nfs4_acl);
3432         nfsi->nfs4_acl = acl;
3433         spin_unlock(&inode->i_lock);
3434 }
3435
3436 static void nfs4_zap_acl_attr(struct inode *inode)
3437 {
3438         nfs4_set_cached_acl(inode, NULL);
3439 }
3440
3441 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3442 {
3443         struct nfs_inode *nfsi = NFS_I(inode);
3444         struct nfs4_cached_acl *acl;
3445         int ret = -ENOENT;
3446
3447         spin_lock(&inode->i_lock);
3448         acl = nfsi->nfs4_acl;
3449         if (acl == NULL)
3450                 goto out;
3451         if (buf == NULL) /* user is just asking for length */
3452                 goto out_len;
3453         if (acl->cached == 0)
3454                 goto out;
3455         ret = -ERANGE; /* see getxattr(2) man page */
3456         if (acl->len > buflen)
3457                 goto out;
3458         memcpy(buf, acl->data, acl->len);
3459 out_len:
3460         ret = acl->len;
3461 out:
3462         spin_unlock(&inode->i_lock);
3463         return ret;
3464 }
3465
3466 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3467 {
3468         struct nfs4_cached_acl *acl;
3469
3470         if (buf && acl_len <= PAGE_SIZE) {
3471                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3472                 if (acl == NULL)
3473                         goto out;
3474                 acl->cached = 1;
3475                 memcpy(acl->data, buf, acl_len);
3476         } else {
3477                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3478                 if (acl == NULL)
3479                         goto out;
3480                 acl->cached = 0;
3481         }
3482         acl->len = acl_len;
3483 out:
3484         nfs4_set_cached_acl(inode, acl);
3485 }
3486
3487 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3488 {
3489         struct page *pages[NFS4ACL_MAXPAGES];
3490         struct nfs_getaclargs args = {
3491                 .fh = NFS_FH(inode),
3492                 .acl_pages = pages,
3493                 .acl_len = buflen,
3494         };
3495         struct nfs_getaclres res = {
3496                 .acl_len = buflen,
3497         };
3498         void *resp_buf;
3499         struct rpc_message msg = {
3500                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3501                 .rpc_argp = &args,
3502                 .rpc_resp = &res,
3503         };
3504         struct page *localpage = NULL;
3505         int ret;
3506
3507         if (buflen < PAGE_SIZE) {
3508                 /* As long as we're doing a round trip to the server anyway,
3509                  * let's be prepared for a page of acl data. */
3510                 localpage = alloc_page(GFP_KERNEL);
3511                 resp_buf = page_address(localpage);
3512                 if (localpage == NULL)
3513                         return -ENOMEM;
3514                 args.acl_pages[0] = localpage;
3515                 args.acl_pgbase = 0;
3516                 args.acl_len = PAGE_SIZE;
3517         } else {
3518                 resp_buf = buf;
3519                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3520         }
3521         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3522         if (ret)
3523                 goto out_free;
3524         if (res.acl_len > args.acl_len)
3525                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3526         else
3527                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3528         if (buf) {
3529                 ret = -ERANGE;
3530                 if (res.acl_len > buflen)
3531                         goto out_free;
3532                 if (localpage)
3533                         memcpy(buf, resp_buf, res.acl_len);
3534         }
3535         ret = res.acl_len;
3536 out_free:
3537         if (localpage)
3538                 __free_page(localpage);
3539         return ret;
3540 }
3541
3542 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3543 {
3544         struct nfs4_exception exception = { };
3545         ssize_t ret;
3546         do {
3547                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3548                 if (ret >= 0)
3549                         break;
3550                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3551         } while (exception.retry);
3552         return ret;
3553 }
3554
3555 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3556 {
3557         struct nfs_server *server = NFS_SERVER(inode);
3558         int ret;
3559
3560         if (!nfs4_server_supports_acls(server))
3561                 return -EOPNOTSUPP;
3562         ret = nfs_revalidate_inode(server, inode);
3563         if (ret < 0)
3564                 return ret;
3565         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3566                 nfs_zap_acl_cache(inode);
3567         ret = nfs4_read_cached_acl(inode, buf, buflen);
3568         if (ret != -ENOENT)
3569                 return ret;
3570         return nfs4_get_acl_uncached(inode, buf, buflen);
3571 }
3572
3573 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3574 {
3575         struct nfs_server *server = NFS_SERVER(inode);
3576         struct page *pages[NFS4ACL_MAXPAGES];
3577         struct nfs_setaclargs arg = {
3578                 .fh             = NFS_FH(inode),
3579                 .acl_pages      = pages,
3580                 .acl_len        = buflen,
3581         };
3582         struct nfs_setaclres res;
3583         struct rpc_message msg = {
3584                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3585                 .rpc_argp       = &arg,
3586                 .rpc_resp       = &res,
3587         };
3588         int ret, i;
3589
3590         if (!nfs4_server_supports_acls(server))
3591                 return -EOPNOTSUPP;
3592         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3593         if (i < 0)
3594                 return i;
3595         nfs_inode_return_delegation(inode);
3596         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3597
3598         /*
3599          * Free each page after tx, so the only ref left is
3600          * held by the network stack
3601          */
3602         for (; i > 0; i--)
3603                 put_page(pages[i-1]);
3604
3605         /*
3606          * Acl update can result in inode attribute update.
3607          * so mark the attribute cache invalid.
3608          */
3609         spin_lock(&inode->i_lock);
3610         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3611         spin_unlock(&inode->i_lock);
3612         nfs_access_zap_cache(inode);
3613         nfs_zap_acl_cache(inode);
3614         return ret;
3615 }
3616
3617 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3618 {
3619         struct nfs4_exception exception = { };
3620         int err;
3621         do {
3622                 err = nfs4_handle_exception(NFS_SERVER(inode),
3623                                 __nfs4_proc_set_acl(inode, buf, buflen),
3624                                 &exception);
3625         } while (exception.retry);
3626         return err;
3627 }
3628
3629 static int
3630 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3631 {
3632         struct nfs_client *clp = server->nfs_client;
3633
3634         if (task->tk_status >= 0)
3635                 return 0;
3636         switch(task->tk_status) {
3637                 case -NFS4ERR_ADMIN_REVOKED:
3638                 case -NFS4ERR_BAD_STATEID:
3639                 case -NFS4ERR_OPENMODE:
3640                         if (state == NULL)
3641                                 break;
3642                         nfs4_schedule_stateid_recovery(server, state);
3643                         goto wait_on_recovery;
3644                 case -NFS4ERR_STALE_STATEID:
3645                 case -NFS4ERR_STALE_CLIENTID:
3646                 case -NFS4ERR_EXPIRED:
3647                         nfs4_schedule_lease_recovery(clp);
3648                         goto wait_on_recovery;
3649 #if defined(CONFIG_NFS_V4_1)
3650                 case -NFS4ERR_BADSESSION:
3651                 case -NFS4ERR_BADSLOT:
3652                 case -NFS4ERR_BAD_HIGH_SLOT:
3653                 case -NFS4ERR_DEADSESSION:
3654                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3655                 case -NFS4ERR_SEQ_FALSE_RETRY:
3656                 case -NFS4ERR_SEQ_MISORDERED:
3657                         dprintk("%s ERROR %d, Reset session\n", __func__,
3658                                 task->tk_status);
3659                         nfs4_schedule_session_recovery(clp->cl_session);
3660                         task->tk_status = 0;
3661                         return -EAGAIN;
3662 #endif /* CONFIG_NFS_V4_1 */
3663                 case -NFS4ERR_DELAY:
3664                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3665                 case -NFS4ERR_GRACE:
3666                 case -EKEYEXPIRED:
3667                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3668                         task->tk_status = 0;
3669                         return -EAGAIN;
3670                 case -NFS4ERR_OLD_STATEID:
3671                         task->tk_status = 0;
3672                         return -EAGAIN;
3673         }
3674         task->tk_status = nfs4_map_errors(task->tk_status);
3675         return 0;
3676 wait_on_recovery:
3677         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3678         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3679                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3680         task->tk_status = 0;
3681         return -EAGAIN;
3682 }
3683
3684 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3685                 unsigned short port, struct rpc_cred *cred,
3686                 struct nfs4_setclientid_res *res)
3687 {
3688         nfs4_verifier sc_verifier;
3689         struct nfs4_setclientid setclientid = {
3690                 .sc_verifier = &sc_verifier,
3691                 .sc_prog = program,
3692                 .sc_cb_ident = clp->cl_cb_ident,
3693         };
3694         struct rpc_message msg = {
3695                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3696                 .rpc_argp = &setclientid,
3697                 .rpc_resp = res,
3698                 .rpc_cred = cred,
3699         };
3700         __be32 *p;
3701         int loop = 0;
3702         int status;
3703
3704         p = (__be32*)sc_verifier.data;
3705         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3706         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3707
3708         for(;;) {
3709                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3710                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3711                                 clp->cl_ipaddr,
3712                                 rpc_peeraddr2str(clp->cl_rpcclient,
3713                                                         RPC_DISPLAY_ADDR),
3714                                 rpc_peeraddr2str(clp->cl_rpcclient,
3715                                                         RPC_DISPLAY_PROTO),
3716                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3717                                 clp->cl_id_uniquifier);
3718                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3719                                 sizeof(setclientid.sc_netid),
3720                                 rpc_peeraddr2str(clp->cl_rpcclient,
3721                                                         RPC_DISPLAY_NETID));
3722                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3723                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3724                                 clp->cl_ipaddr, port >> 8, port & 255);
3725
3726                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3727                 if (status != -NFS4ERR_CLID_INUSE)
3728                         break;
3729                 if (signalled())
3730                         break;
3731                 if (loop++ & 1)
3732                         ssleep(clp->cl_lease_time / HZ + 1);
3733                 else
3734                         if (++clp->cl_id_uniquifier == 0)
3735                                 break;
3736         }
3737         return status;
3738 }
3739
3740 static int _nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3741                 struct nfs4_setclientid_res *arg,
3742                 struct rpc_cred *cred)
3743 {
3744         struct nfs_fsinfo fsinfo;
3745         struct rpc_message msg = {
3746                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3747                 .rpc_argp = arg,
3748                 .rpc_resp = &fsinfo,
3749                 .rpc_cred = cred,
3750         };
3751         unsigned long now;
3752         int status;
3753
3754         now = jiffies;
3755         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3756         if (status == 0) {
3757                 spin_lock(&clp->cl_lock);
3758                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3759                 clp->cl_last_renewal = now;
3760                 spin_unlock(&clp->cl_lock);
3761         }
3762         return status;
3763 }
3764
3765 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3766                 struct nfs4_setclientid_res *arg,
3767                 struct rpc_cred *cred)
3768 {
3769         long timeout = 0;
3770         int err;
3771         do {
3772                 err = _nfs4_proc_setclientid_confirm(clp, arg, cred);
3773                 switch (err) {
3774                         case 0:
3775                                 return err;
3776                         case -NFS4ERR_RESOURCE:
3777                                 /* The IBM lawyers misread another document! */
3778                         case -NFS4ERR_DELAY:
3779                                 err = nfs4_delay(clp->cl_rpcclient, &timeout);
3780                 }
3781         } while (err == 0);
3782         return err;
3783 }
3784
3785 struct nfs4_delegreturndata {
3786         struct nfs4_delegreturnargs args;
3787         struct nfs4_delegreturnres res;
3788         struct nfs_fh fh;
3789         nfs4_stateid stateid;
3790         unsigned long timestamp;
3791         struct nfs_fattr fattr;
3792         int rpc_status;
3793 };
3794
3795 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3796 {
3797         struct nfs4_delegreturndata *data = calldata;
3798
3799         if (!nfs4_sequence_done(task, &data->res.seq_res))
3800                 return;
3801
3802         switch (task->tk_status) {
3803         case -NFS4ERR_STALE_STATEID:
3804         case -NFS4ERR_EXPIRED:
3805         case 0:
3806                 renew_lease(data->res.server, data->timestamp);
3807                 break;
3808         default:
3809                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3810                                 -EAGAIN) {
3811                         nfs_restart_rpc(task, data->res.server->nfs_client);
3812                         return;
3813                 }
3814         }
3815         data->rpc_status = task->tk_status;
3816 }
3817
3818 static void nfs4_delegreturn_release(void *calldata)
3819 {
3820         kfree(calldata);
3821 }
3822
3823 #if defined(CONFIG_NFS_V4_1)
3824 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3825 {
3826         struct nfs4_delegreturndata *d_data;
3827
3828         d_data = (struct nfs4_delegreturndata *)data;
3829
3830         if (nfs4_setup_sequence(d_data->res.server,
3831                                 &d_data->args.seq_args,
3832                                 &d_data->res.seq_res, 1, task))
3833                 return;
3834         rpc_call_start(task);
3835 }
3836 #endif /* CONFIG_NFS_V4_1 */
3837
3838 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3839 #if defined(CONFIG_NFS_V4_1)
3840         .rpc_call_prepare = nfs4_delegreturn_prepare,
3841 #endif /* CONFIG_NFS_V4_1 */
3842         .rpc_call_done = nfs4_delegreturn_done,
3843         .rpc_release = nfs4_delegreturn_release,
3844 };
3845
3846 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3847 {
3848         struct nfs4_delegreturndata *data;
3849         struct nfs_server *server = NFS_SERVER(inode);
3850         struct rpc_task *task;
3851         struct rpc_message msg = {
3852                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3853                 .rpc_cred = cred,
3854         };
3855         struct rpc_task_setup task_setup_data = {
3856                 .rpc_client = server->client,
3857                 .rpc_message = &msg,
3858                 .callback_ops = &nfs4_delegreturn_ops,
3859                 .flags = RPC_TASK_ASYNC,
3860         };
3861         int status = 0;
3862
3863         data = kzalloc(sizeof(*data), GFP_NOFS);
3864         if (data == NULL)
3865                 return -ENOMEM;
3866         data->args.fhandle = &data->fh;
3867         data->args.stateid = &data->stateid;
3868         data->args.bitmask = server->attr_bitmask;
3869         nfs_copy_fh(&data->fh, NFS_FH(inode));
3870         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3871         data->res.fattr = &data->fattr;
3872         data->res.server = server;
3873         nfs_fattr_init(data->res.fattr);
3874         data->timestamp = jiffies;
3875         data->rpc_status = 0;
3876
3877         task_setup_data.callback_data = data;
3878         msg.rpc_argp = &data->args;
3879         msg.rpc_resp = &data->res;
3880         task = rpc_run_task(&task_setup_data);
3881         if (IS_ERR(task))
3882                 return PTR_ERR(task);
3883         if (!issync)
3884                 goto out;
3885         status = nfs4_wait_for_completion_rpc_task(task);
3886         if (status != 0)
3887                 goto out;
3888         status = data->rpc_status;
3889         if (status != 0)
3890                 goto out;
3891         nfs_refresh_inode(inode, &data->fattr);
3892 out:
3893         rpc_put_task(task);
3894         return status;
3895 }
3896
3897 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3898 {
3899         struct nfs_server *server = NFS_SERVER(inode);
3900         struct nfs4_exception exception = { };
3901         int err;
3902         do {
3903                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3904                 switch (err) {
3905                         case -NFS4ERR_STALE_STATEID:
3906                         case -NFS4ERR_EXPIRED:
3907                         case 0:
3908                                 return 0;
3909                 }
3910                 err = nfs4_handle_exception(server, err, &exception);
3911         } while (exception.retry);
3912         return err;
3913 }
3914
3915 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3916 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3917
3918 /*
3919  * sleep, with exponential backoff, and retry the LOCK operation.
3920  */
3921 static unsigned long
3922 nfs4_set_lock_task_retry(unsigned long timeout)
3923 {
3924         schedule_timeout_killable(timeout);
3925         timeout <<= 1;
3926         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3927                 return NFS4_LOCK_MAXTIMEOUT;
3928         return timeout;
3929 }
3930
3931 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3932 {
3933         struct inode *inode = state->inode;
3934         struct nfs_server *server = NFS_SERVER(inode);
3935         struct nfs_client *clp = server->nfs_client;
3936         struct nfs_lockt_args arg = {
3937                 .fh = NFS_FH(inode),
3938                 .fl = request,
3939         };
3940         struct nfs_lockt_res res = {
3941                 .denied = request,
3942         };
3943         struct rpc_message msg = {
3944                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3945                 .rpc_argp       = &arg,
3946                 .rpc_resp       = &res,
3947                 .rpc_cred       = state->owner->so_cred,
3948         };
3949         struct nfs4_lock_state *lsp;
3950         int status;
3951
3952         arg.lock_owner.clientid = clp->cl_clientid;
3953         status = nfs4_set_lock_state(state, request);
3954         if (status != 0)
3955                 goto out;
3956         lsp = request->fl_u.nfs4_fl.owner;
3957         arg.lock_owner.id = lsp->ls_id.id;
3958         arg.lock_owner.s_dev = server->s_dev;
3959         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3960         switch (status) {
3961                 case 0:
3962                         request->fl_type = F_UNLCK;
3963                         break;
3964                 case -NFS4ERR_DENIED:
3965                         status = 0;
3966         }
3967         request->fl_ops->fl_release_private(request);
3968 out:
3969         return status;
3970 }
3971
3972 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3973 {
3974         struct nfs4_exception exception = { };
3975         int err;
3976
3977         do {
3978                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3979                                 _nfs4_proc_getlk(state, cmd, request),
3980                                 &exception);
3981         } while (exception.retry);
3982         return err;
3983 }
3984
3985 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3986 {
3987         int res = 0;
3988         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3989                 case FL_POSIX:
3990                         res = posix_lock_file_wait(file, fl);
3991                         break;
3992                 case FL_FLOCK:
3993                         res = flock_lock_file_wait(file, fl);
3994                         break;
3995                 default:
3996                         BUG();
3997         }
3998         return res;
3999 }
4000
4001 struct nfs4_unlockdata {
4002         struct nfs_locku_args arg;
4003         struct nfs_locku_res res;
4004         struct nfs4_lock_state *lsp;
4005         struct nfs_open_context *ctx;
4006         struct file_lock fl;
4007         const struct nfs_server *server;
4008         unsigned long timestamp;
4009 };
4010
4011 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4012                 struct nfs_open_context *ctx,
4013                 struct nfs4_lock_state *lsp,
4014                 struct nfs_seqid *seqid)
4015 {
4016         struct nfs4_unlockdata *p;
4017         struct inode *inode = lsp->ls_state->inode;
4018
4019         p = kzalloc(sizeof(*p), GFP_NOFS);
4020         if (p == NULL)
4021                 return NULL;
4022         p->arg.fh = NFS_FH(inode);
4023         p->arg.fl = &p->fl;
4024         p->arg.seqid = seqid;
4025         p->res.seqid = seqid;
4026         p->arg.stateid = &lsp->ls_stateid;
4027         p->lsp = lsp;
4028         atomic_inc(&lsp->ls_count);
4029         /* Ensure we don't close file until we're done freeing locks! */
4030         p->ctx = get_nfs_open_context(ctx);
4031         memcpy(&p->fl, fl, sizeof(p->fl));
4032         p->server = NFS_SERVER(inode);
4033         return p;
4034 }
4035
4036 static void nfs4_locku_release_calldata(void *data)
4037 {
4038         struct nfs4_unlockdata *calldata = data;
4039         nfs_free_seqid(calldata->arg.seqid);
4040         nfs4_put_lock_state(calldata->lsp);
4041         put_nfs_open_context(calldata->ctx);
4042         kfree(calldata);
4043 }
4044
4045 static void nfs4_locku_done(struct rpc_task *task, void *data)
4046 {
4047         struct nfs4_unlockdata *calldata = data;
4048
4049         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4050                 return;
4051         switch (task->tk_status) {
4052                 case 0:
4053                         memcpy(calldata->lsp->ls_stateid.data,
4054                                         calldata->res.stateid.data,
4055                                         sizeof(calldata->lsp->ls_stateid.data));
4056                         renew_lease(calldata->server, calldata->timestamp);
4057                         break;
4058                 case -NFS4ERR_BAD_STATEID:
4059                 case -NFS4ERR_OLD_STATEID:
4060                 case -NFS4ERR_STALE_STATEID:
4061                 case -NFS4ERR_EXPIRED:
4062                         break;
4063                 default:
4064                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4065                                 nfs_restart_rpc(task,
4066                                                  calldata->server->nfs_client);
4067         }
4068 }
4069
4070 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4071 {
4072         struct nfs4_unlockdata *calldata = data;
4073
4074         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4075                 return;
4076         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4077                 /* Note: exit _without_ running nfs4_locku_done */
4078                 task->tk_action = NULL;
4079                 return;
4080         }
4081         calldata->timestamp = jiffies;
4082         if (nfs4_setup_sequence(calldata->server,
4083                                 &calldata->arg.seq_args,
4084                                 &calldata->res.seq_res, 1, task))
4085                 return;
4086         rpc_call_start(task);
4087 }
4088
4089 static const struct rpc_call_ops nfs4_locku_ops = {
4090         .rpc_call_prepare = nfs4_locku_prepare,
4091         .rpc_call_done = nfs4_locku_done,
4092         .rpc_release = nfs4_locku_release_calldata,
4093 };
4094
4095 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4096                 struct nfs_open_context *ctx,
4097                 struct nfs4_lock_state *lsp,
4098                 struct nfs_seqid *seqid)
4099 {
4100         struct nfs4_unlockdata *data;
4101         struct rpc_message msg = {
4102                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4103                 .rpc_cred = ctx->cred,
4104         };
4105         struct rpc_task_setup task_setup_data = {
4106                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4107                 .rpc_message = &msg,
4108                 .callback_ops = &nfs4_locku_ops,
4109                 .workqueue = nfsiod_workqueue,
4110                 .flags = RPC_TASK_ASYNC,
4111         };
4112
4113         /* Ensure this is an unlock - when canceling a lock, the
4114          * canceled lock is passed in, and it won't be an unlock.
4115          */
4116         fl->fl_type = F_UNLCK;
4117
4118         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4119         if (data == NULL) {
4120                 nfs_free_seqid(seqid);
4121                 return ERR_PTR(-ENOMEM);
4122         }
4123
4124         msg.rpc_argp = &data->arg;
4125         msg.rpc_resp = &data->res;
4126         task_setup_data.callback_data = data;
4127         return rpc_run_task(&task_setup_data);
4128 }
4129
4130 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4131 {
4132         struct nfs_inode *nfsi = NFS_I(state->inode);
4133         struct nfs_seqid *seqid;
4134         struct nfs4_lock_state *lsp;
4135         struct rpc_task *task;
4136         int status = 0;
4137         unsigned char fl_flags = request->fl_flags;
4138
4139         status = nfs4_set_lock_state(state, request);
4140         /* Unlock _before_ we do the RPC call */
4141         request->fl_flags |= FL_EXISTS;
4142         down_read(&nfsi->rwsem);
4143         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4144                 up_read(&nfsi->rwsem);
4145                 goto out;
4146         }
4147         up_read(&nfsi->rwsem);
4148         if (status != 0)
4149                 goto out;
4150         /* Is this a delegated lock? */
4151         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4152                 goto out;
4153         lsp = request->fl_u.nfs4_fl.owner;
4154         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4155         status = -ENOMEM;
4156         if (seqid == NULL)
4157                 goto out;
4158         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4159         status = PTR_ERR(task);
4160         if (IS_ERR(task))
4161                 goto out;
4162         status = nfs4_wait_for_completion_rpc_task(task);
4163         rpc_put_task(task);
4164 out:
4165         request->fl_flags = fl_flags;
4166         return status;
4167 }
4168
4169 struct nfs4_lockdata {
4170         struct nfs_lock_args arg;
4171         struct nfs_lock_res res;
4172         struct nfs4_lock_state *lsp;
4173         struct nfs_open_context *ctx;
4174         struct file_lock fl;
4175         unsigned long timestamp;
4176         int rpc_status;
4177         int cancelled;
4178         struct nfs_server *server;
4179 };
4180
4181 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4182                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4183                 gfp_t gfp_mask)
4184 {
4185         struct nfs4_lockdata *p;
4186         struct inode *inode = lsp->ls_state->inode;
4187         struct nfs_server *server = NFS_SERVER(inode);
4188
4189         p = kzalloc(sizeof(*p), gfp_mask);
4190         if (p == NULL)
4191                 return NULL;
4192
4193         p->arg.fh = NFS_FH(inode);
4194         p->arg.fl = &p->fl;
4195         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4196         if (p->arg.open_seqid == NULL)
4197                 goto out_free;
4198         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4199         if (p->arg.lock_seqid == NULL)
4200                 goto out_free_seqid;
4201         p->arg.lock_stateid = &lsp->ls_stateid;
4202         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4203         p->arg.lock_owner.id = lsp->ls_id.id;
4204         p->arg.lock_owner.s_dev = server->s_dev;
4205         p->res.lock_seqid = p->arg.lock_seqid;
4206         p->lsp = lsp;
4207         p->server = server;
4208         atomic_inc(&lsp->ls_count);
4209         p->ctx = get_nfs_open_context(ctx);
4210         memcpy(&p->fl, fl, sizeof(p->fl));
4211         return p;
4212 out_free_seqid:
4213         nfs_free_seqid(p->arg.open_seqid);
4214 out_free:
4215         kfree(p);
4216         return NULL;
4217 }
4218
4219 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4220 {
4221         struct nfs4_lockdata *data = calldata;
4222         struct nfs4_state *state = data->lsp->ls_state;
4223
4224         dprintk("%s: begin!\n", __func__);
4225         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4226                 return;
4227         /* Do we need to do an open_to_lock_owner? */
4228         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4229                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4230                         return;
4231                 data->arg.open_stateid = &state->stateid;
4232                 data->arg.new_lock_owner = 1;
4233                 data->res.open_seqid = data->arg.open_seqid;
4234         } else
4235                 data->arg.new_lock_owner = 0;
4236         data->timestamp = jiffies;
4237         if (nfs4_setup_sequence(data->server,
4238                                 &data->arg.seq_args,
4239                                 &data->res.seq_res, 1, task))
4240                 return;
4241         rpc_call_start(task);
4242         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4243 }
4244
4245 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4246 {
4247         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4248         nfs4_lock_prepare(task, calldata);
4249 }
4250
4251 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4252 {
4253         struct nfs4_lockdata *data = calldata;
4254
4255         dprintk("%s: begin!\n", __func__);
4256
4257         if (!nfs4_sequence_done(task, &data->res.seq_res))
4258                 return;
4259
4260         data->rpc_status = task->tk_status;
4261         if (data->arg.new_lock_owner != 0) {
4262                 if (data->rpc_status == 0)
4263                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4264                 else
4265                         goto out;
4266         }
4267         if (data->rpc_status == 0) {
4268                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4269                                         sizeof(data->lsp->ls_stateid.data));
4270                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4271                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
4272         }
4273 out:
4274         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4275 }
4276
4277 static void nfs4_lock_release(void *calldata)
4278 {
4279         struct nfs4_lockdata *data = calldata;
4280
4281         dprintk("%s: begin!\n", __func__);
4282         nfs_free_seqid(data->arg.open_seqid);
4283         if (data->cancelled != 0) {
4284                 struct rpc_task *task;
4285                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4286                                 data->arg.lock_seqid);
4287                 if (!IS_ERR(task))
4288                         rpc_put_task_async(task);
4289                 dprintk("%s: cancelling lock!\n", __func__);
4290         } else
4291                 nfs_free_seqid(data->arg.lock_seqid);
4292         nfs4_put_lock_state(data->lsp);
4293         put_nfs_open_context(data->ctx);
4294         kfree(data);
4295         dprintk("%s: done!\n", __func__);
4296 }
4297
4298 static const struct rpc_call_ops nfs4_lock_ops = {
4299         .rpc_call_prepare = nfs4_lock_prepare,
4300         .rpc_call_done = nfs4_lock_done,
4301         .rpc_release = nfs4_lock_release,
4302 };
4303
4304 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4305         .rpc_call_prepare = nfs4_recover_lock_prepare,
4306         .rpc_call_done = nfs4_lock_done,
4307         .rpc_release = nfs4_lock_release,
4308 };
4309
4310 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4311 {
4312         switch (error) {
4313         case -NFS4ERR_ADMIN_REVOKED:
4314         case -NFS4ERR_BAD_STATEID:
4315                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4316                 if (new_lock_owner != 0 ||
4317                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4318                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4319                 break;
4320         case -NFS4ERR_STALE_STATEID:
4321                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4322         case -NFS4ERR_EXPIRED:
4323                 nfs4_schedule_lease_recovery(server->nfs_client);
4324         };
4325 }
4326
4327 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4328 {
4329         struct nfs4_lockdata *data;
4330         struct rpc_task *task;
4331         struct rpc_message msg = {
4332                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4333                 .rpc_cred = state->owner->so_cred,
4334         };
4335         struct rpc_task_setup task_setup_data = {
4336                 .rpc_client = NFS_CLIENT(state->inode),
4337                 .rpc_message = &msg,
4338                 .callback_ops = &nfs4_lock_ops,
4339                 .workqueue = nfsiod_workqueue,
4340                 .flags = RPC_TASK_ASYNC,
4341         };
4342         int ret;
4343
4344         dprintk("%s: begin!\n", __func__);
4345         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4346                         fl->fl_u.nfs4_fl.owner,
4347                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4348         if (data == NULL)
4349                 return -ENOMEM;
4350         if (IS_SETLKW(cmd))
4351                 data->arg.block = 1;
4352         if (recovery_type > NFS_LOCK_NEW) {
4353                 if (recovery_type == NFS_LOCK_RECLAIM)
4354                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4355                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4356         }
4357         msg.rpc_argp = &data->arg;
4358         msg.rpc_resp = &data->res;
4359         task_setup_data.callback_data = data;
4360         task = rpc_run_task(&task_setup_data);
4361         if (IS_ERR(task))
4362                 return PTR_ERR(task);
4363         ret = nfs4_wait_for_completion_rpc_task(task);
4364         if (ret == 0) {
4365                 ret = data->rpc_status;
4366                 if (ret)
4367                         nfs4_handle_setlk_error(data->server, data->lsp,
4368                                         data->arg.new_lock_owner, ret);
4369         } else
4370                 data->cancelled = 1;
4371         rpc_put_task(task);
4372         dprintk("%s: done, ret = %d!\n", __func__, ret);
4373         return ret;
4374 }
4375
4376 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4377 {
4378         struct nfs_server *server = NFS_SERVER(state->inode);
4379         struct nfs4_exception exception = { };
4380         int err;
4381
4382         do {
4383                 /* Cache the lock if possible... */
4384                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4385                         return 0;
4386                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4387                 if (err != -NFS4ERR_DELAY)
4388                         break;
4389                 nfs4_handle_exception(server, err, &exception);
4390         } while (exception.retry);
4391         return err;
4392 }
4393
4394 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4395 {
4396         struct nfs_server *server = NFS_SERVER(state->inode);
4397         struct nfs4_exception exception = { };
4398         int err;
4399
4400         err = nfs4_set_lock_state(state, request);
4401         if (err != 0)
4402                 return err;
4403         do {
4404                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4405                         return 0;
4406                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4407                 switch (err) {
4408                 default:
4409                         goto out;
4410                 case -NFS4ERR_GRACE:
4411                 case -NFS4ERR_DELAY:
4412                         nfs4_handle_exception(server, err, &exception);
4413                         err = 0;
4414                 }
4415         } while (exception.retry);
4416 out:
4417         return err;
4418 }
4419
4420 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4421 {
4422         struct nfs_inode *nfsi = NFS_I(state->inode);
4423         unsigned char fl_flags = request->fl_flags;
4424         int status = -ENOLCK;
4425
4426         if ((fl_flags & FL_POSIX) &&
4427                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4428                 goto out;
4429         /* Is this a delegated open? */
4430         status = nfs4_set_lock_state(state, request);
4431         if (status != 0)
4432                 goto out;
4433         request->fl_flags |= FL_ACCESS;
4434         status = do_vfs_lock(request->fl_file, request);
4435         if (status < 0)
4436                 goto out;
4437         down_read(&nfsi->rwsem);
4438         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4439                 /* Yes: cache locks! */
4440                 /* ...but avoid races with delegation recall... */
4441                 request->fl_flags = fl_flags & ~FL_SLEEP;
4442                 status = do_vfs_lock(request->fl_file, request);
4443                 goto out_unlock;
4444         }
4445         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4446         if (status != 0)
4447                 goto out_unlock;
4448         /* Note: we always want to sleep here! */
4449         request->fl_flags = fl_flags | FL_SLEEP;
4450         if (do_vfs_lock(request->fl_file, request) < 0)
4451                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4452 out_unlock:
4453         up_read(&nfsi->rwsem);
4454 out:
4455         request->fl_flags = fl_flags;
4456         return status;
4457 }
4458
4459 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4460 {
4461         struct nfs4_exception exception = { };
4462         int err;
4463
4464         do {
4465                 err = _nfs4_proc_setlk(state, cmd, request);
4466                 if (err == -NFS4ERR_DENIED)
4467                         err = -EAGAIN;
4468                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4469                                 err, &exception);
4470         } while (exception.retry);
4471         return err;
4472 }
4473
4474 static int
4475 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4476 {
4477         struct nfs_open_context *ctx;
4478         struct nfs4_state *state;
4479         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4480         int status;
4481
4482         /* verify open state */
4483         ctx = nfs_file_open_context(filp);
4484         state = ctx->state;
4485
4486         if (request->fl_start < 0 || request->fl_end < 0)
4487                 return -EINVAL;
4488
4489         if (IS_GETLK(cmd)) {
4490                 if (state != NULL)
4491                         return nfs4_proc_getlk(state, F_GETLK, request);
4492                 return 0;
4493         }
4494
4495         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4496                 return -EINVAL;
4497
4498         if (request->fl_type == F_UNLCK) {
4499                 if (state != NULL)
4500                         return nfs4_proc_unlck(state, cmd, request);
4501                 return 0;
4502         }
4503
4504         if (state == NULL)
4505                 return -ENOLCK;
4506         do {
4507                 status = nfs4_proc_setlk(state, cmd, request);
4508                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4509                         break;
4510                 timeout = nfs4_set_lock_task_retry(timeout);
4511                 status = -ERESTARTSYS;
4512                 if (signalled())
4513                         break;
4514         } while(status < 0);
4515         return status;
4516 }
4517
4518 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4519 {
4520         struct nfs_server *server = NFS_SERVER(state->inode);
4521         struct nfs4_exception exception = { };
4522         int err;
4523
4524         err = nfs4_set_lock_state(state, fl);
4525         if (err != 0)
4526                 goto out;
4527         do {
4528                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4529                 switch (err) {
4530                         default:
4531                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4532                                                 __func__, err);
4533                         case 0:
4534                         case -ESTALE:
4535                                 goto out;
4536                         case -NFS4ERR_EXPIRED:
4537                         case -NFS4ERR_STALE_CLIENTID:
4538                         case -NFS4ERR_STALE_STATEID:
4539                                 nfs4_schedule_lease_recovery(server->nfs_client);
4540                                 goto out;
4541                         case -NFS4ERR_BADSESSION:
4542                         case -NFS4ERR_BADSLOT:
4543                         case -NFS4ERR_BAD_HIGH_SLOT:
4544                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4545                         case -NFS4ERR_DEADSESSION:
4546                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4547                                 goto out;
4548                         case -ERESTARTSYS:
4549                                 /*
4550                                  * The show must go on: exit, but mark the
4551                                  * stateid as needing recovery.
4552                                  */
4553                         case -NFS4ERR_ADMIN_REVOKED:
4554                         case -NFS4ERR_BAD_STATEID:
4555                         case -NFS4ERR_OPENMODE:
4556                                 nfs4_schedule_stateid_recovery(server, state);
4557                                 err = 0;
4558                                 goto out;
4559                         case -EKEYEXPIRED:
4560                                 /*
4561                                  * User RPCSEC_GSS context has expired.
4562                                  * We cannot recover this stateid now, so
4563                                  * skip it and allow recovery thread to
4564                                  * proceed.
4565                                  */
4566                                 err = 0;
4567                                 goto out;
4568                         case -ENOMEM:
4569                         case -NFS4ERR_DENIED:
4570                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4571                                 err = 0;
4572                                 goto out;
4573                         case -NFS4ERR_DELAY:
4574                                 break;
4575                 }
4576                 err = nfs4_handle_exception(server, err, &exception);
4577         } while (exception.retry);
4578 out:
4579         return err;
4580 }
4581
4582 static void nfs4_release_lockowner_release(void *calldata)
4583 {
4584         kfree(calldata);
4585 }
4586
4587 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4588         .rpc_release = nfs4_release_lockowner_release,
4589 };
4590
4591 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4592 {
4593         struct nfs_server *server = lsp->ls_state->owner->so_server;
4594         struct nfs_release_lockowner_args *args;
4595         struct rpc_message msg = {
4596                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4597         };
4598
4599         if (server->nfs_client->cl_mvops->minor_version != 0)
4600                 return;
4601         args = kmalloc(sizeof(*args), GFP_NOFS);
4602         if (!args)
4603                 return;
4604         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4605         args->lock_owner.id = lsp->ls_id.id;
4606         args->lock_owner.s_dev = server->s_dev;
4607         msg.rpc_argp = args;
4608         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4609 }
4610
4611 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4612
4613 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4614                                    const void *buf, size_t buflen,
4615                                    int flags, int type)
4616 {
4617         if (strcmp(key, "") != 0)
4618                 return -EINVAL;
4619
4620         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4621 }
4622
4623 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4624                                    void *buf, size_t buflen, int type)
4625 {
4626         if (strcmp(key, "") != 0)
4627                 return -EINVAL;
4628
4629         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4630 }
4631
4632 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4633                                        size_t list_len, const char *name,
4634                                        size_t name_len, int type)
4635 {
4636         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4637
4638         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4639                 return 0;
4640
4641         if (list && len <= list_len)
4642                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4643         return len;
4644 }
4645
4646 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4647 {
4648         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4649                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4650                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4651                 return;
4652
4653         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4654                 NFS_ATTR_FATTR_NLINK;
4655         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4656         fattr->nlink = 2;
4657 }
4658
4659 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4660                 struct nfs4_fs_locations *fs_locations, struct page *page)
4661 {
4662         struct nfs_server *server = NFS_SERVER(dir);
4663         u32 bitmask[2] = {
4664                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4665                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4666         };
4667         struct nfs4_fs_locations_arg args = {
4668                 .dir_fh = NFS_FH(dir),
4669                 .name = name,
4670                 .page = page,
4671                 .bitmask = bitmask,
4672         };
4673         struct nfs4_fs_locations_res res = {
4674                 .fs_locations = fs_locations,
4675         };
4676         struct rpc_message msg = {
4677                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4678                 .rpc_argp = &args,
4679                 .rpc_resp = &res,
4680         };
4681         int status;
4682
4683         dprintk("%s: start\n", __func__);
4684         nfs_fattr_init(&fs_locations->fattr);
4685         fs_locations->server = server;
4686         fs_locations->nlocations = 0;
4687         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4688         nfs_fixup_referral_attributes(&fs_locations->fattr);
4689         dprintk("%s: returned status = %d\n", __func__, status);
4690         return status;
4691 }
4692
4693 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4694 {
4695         int status;
4696         struct nfs4_secinfo_arg args = {
4697                 .dir_fh = NFS_FH(dir),
4698                 .name   = name,
4699         };
4700         struct nfs4_secinfo_res res = {
4701                 .flavors     = flavors,
4702         };
4703         struct rpc_message msg = {
4704                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4705                 .rpc_argp = &args,
4706                 .rpc_resp = &res,
4707         };
4708
4709         dprintk("NFS call  secinfo %s\n", name->name);
4710         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4711         dprintk("NFS reply  secinfo: %d\n", status);
4712         return status;
4713 }
4714
4715 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4716 {
4717         struct nfs4_exception exception = { };
4718         int err;
4719         do {
4720                 err = nfs4_handle_exception(NFS_SERVER(dir),
4721                                 _nfs4_proc_secinfo(dir, name, flavors),
4722                                 &exception);
4723         } while (exception.retry);
4724         return err;
4725 }
4726
4727 #ifdef CONFIG_NFS_V4_1
4728 /*
4729  * Check the exchange flags returned by the server for invalid flags, having
4730  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4731  * DS flags set.
4732  */
4733 static int nfs4_check_cl_exchange_flags(u32 flags)
4734 {
4735         if (flags & ~EXCHGID4_FLAG_MASK_R)
4736                 goto out_inval;
4737         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4738             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4739                 goto out_inval;
4740         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4741                 goto out_inval;
4742         return NFS_OK;
4743 out_inval:
4744         return -NFS4ERR_INVAL;
4745 }
4746
4747 /*
4748  * nfs4_proc_exchange_id()
4749  *
4750  * Since the clientid has expired, all compounds using sessions
4751  * associated with the stale clientid will be returning
4752  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4753  * be in some phase of session reset.
4754  */
4755 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4756 {
4757         nfs4_verifier verifier;
4758         struct nfs41_exchange_id_args args = {
4759                 .client = clp,
4760                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4761         };
4762         struct nfs41_exchange_id_res res = {
4763                 .client = clp,
4764         };
4765         int status;
4766         struct rpc_message msg = {
4767                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4768                 .rpc_argp = &args,
4769                 .rpc_resp = &res,
4770                 .rpc_cred = cred,
4771         };
4772         __be32 *p;
4773
4774         dprintk("--> %s\n", __func__);
4775         BUG_ON(clp == NULL);
4776
4777         p = (u32 *)verifier.data;
4778         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4779         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4780         args.verifier = &verifier;
4781
4782         args.id_len = scnprintf(args.id, sizeof(args.id),
4783                                 "%s/%s.%s/%u",
4784                                 clp->cl_ipaddr,
4785                                 init_utsname()->nodename,
4786                                 init_utsname()->domainname,
4787                                 clp->cl_rpcclient->cl_auth->au_flavor);
4788
4789         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
4790         if (!status)
4791                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4792         dprintk("<-- %s status= %d\n", __func__, status);
4793         return status;
4794 }
4795
4796 struct nfs4_get_lease_time_data {
4797         struct nfs4_get_lease_time_args *args;
4798         struct nfs4_get_lease_time_res *res;
4799         struct nfs_client *clp;
4800 };
4801
4802 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4803                                         void *calldata)
4804 {
4805         int ret;
4806         struct nfs4_get_lease_time_data *data =
4807                         (struct nfs4_get_lease_time_data *)calldata;
4808
4809         dprintk("--> %s\n", __func__);
4810         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4811         /* just setup sequence, do not trigger session recovery
4812            since we're invoked within one */
4813         ret = nfs41_setup_sequence(data->clp->cl_session,
4814                                    &data->args->la_seq_args,
4815                                    &data->res->lr_seq_res, 0, task);
4816
4817         BUG_ON(ret == -EAGAIN);
4818         rpc_call_start(task);
4819         dprintk("<-- %s\n", __func__);
4820 }
4821
4822 /*
4823  * Called from nfs4_state_manager thread for session setup, so don't recover
4824  * from sequence operation or clientid errors.
4825  */
4826 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4827 {
4828         struct nfs4_get_lease_time_data *data =
4829                         (struct nfs4_get_lease_time_data *)calldata;
4830
4831         dprintk("--> %s\n", __func__);
4832         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
4833                 return;
4834         switch (task->tk_status) {
4835         case -NFS4ERR_DELAY:
4836         case -NFS4ERR_GRACE:
4837                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4838                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4839                 task->tk_status = 0;
4840                 nfs_restart_rpc(task, data->clp);
4841                 return;
4842         }
4843         dprintk("<-- %s\n", __func__);
4844 }
4845
4846 struct rpc_call_ops nfs4_get_lease_time_ops = {
4847         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4848         .rpc_call_done = nfs4_get_lease_time_done,
4849 };
4850
4851 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4852 {
4853         struct rpc_task *task;
4854         struct nfs4_get_lease_time_args args;
4855         struct nfs4_get_lease_time_res res = {
4856                 .lr_fsinfo = fsinfo,
4857         };
4858         struct nfs4_get_lease_time_data data = {
4859                 .args = &args,
4860                 .res = &res,
4861                 .clp = clp,
4862         };
4863         struct rpc_message msg = {
4864                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4865                 .rpc_argp = &args,
4866                 .rpc_resp = &res,
4867         };
4868         struct rpc_task_setup task_setup = {
4869                 .rpc_client = clp->cl_rpcclient,
4870                 .rpc_message = &msg,
4871                 .callback_ops = &nfs4_get_lease_time_ops,
4872                 .callback_data = &data
4873         };
4874         int status;
4875
4876         dprintk("--> %s\n", __func__);
4877         task = rpc_run_task(&task_setup);
4878
4879         if (IS_ERR(task))
4880                 status = PTR_ERR(task);
4881         else {
4882                 status = task->tk_status;
4883                 rpc_put_task(task);
4884         }
4885         dprintk("<-- %s return %d\n", __func__, status);
4886
4887         return status;
4888 }
4889
4890 /*
4891  * Reset a slot table
4892  */
4893 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
4894                                  int ivalue)
4895 {
4896         struct nfs4_slot *new = NULL;
4897         int i;
4898         int ret = 0;
4899
4900         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
4901                 max_reqs, tbl->max_slots);
4902
4903         /* Does the newly negotiated max_reqs match the existing slot table? */
4904         if (max_reqs != tbl->max_slots) {
4905                 ret = -ENOMEM;
4906                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
4907                               GFP_NOFS);
4908                 if (!new)
4909                         goto out;
4910                 ret = 0;
4911                 kfree(tbl->slots);
4912         }
4913         spin_lock(&tbl->slot_tbl_lock);
4914         if (new) {
4915                 tbl->slots = new;
4916                 tbl->max_slots = max_reqs;
4917         }
4918         for (i = 0; i < tbl->max_slots; ++i)
4919                 tbl->slots[i].seq_nr = ivalue;
4920         spin_unlock(&tbl->slot_tbl_lock);
4921         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4922                 tbl, tbl->slots, tbl->max_slots);
4923 out:
4924         dprintk("<-- %s: return %d\n", __func__, ret);
4925         return ret;
4926 }
4927
4928 /*
4929  * Reset the forechannel and backchannel slot tables
4930  */
4931 static int nfs4_reset_slot_tables(struct nfs4_session *session)
4932 {
4933         int status;
4934
4935         status = nfs4_reset_slot_table(&session->fc_slot_table,
4936                         session->fc_attrs.max_reqs, 1);
4937         if (status)
4938                 return status;
4939
4940         status = nfs4_reset_slot_table(&session->bc_slot_table,
4941                         session->bc_attrs.max_reqs, 0);
4942         return status;
4943 }
4944
4945 /* Destroy the slot table */
4946 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
4947 {
4948         if (session->fc_slot_table.slots != NULL) {
4949                 kfree(session->fc_slot_table.slots);
4950                 session->fc_slot_table.slots = NULL;
4951         }
4952         if (session->bc_slot_table.slots != NULL) {
4953                 kfree(session->bc_slot_table.slots);
4954                 session->bc_slot_table.slots = NULL;
4955         }
4956         return;
4957 }
4958
4959 /*
4960  * Initialize slot table
4961  */
4962 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
4963                 int max_slots, int ivalue)
4964 {
4965         struct nfs4_slot *slot;
4966         int ret = -ENOMEM;
4967
4968         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
4969
4970         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
4971
4972         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
4973         if (!slot)
4974                 goto out;
4975         ret = 0;
4976
4977         spin_lock(&tbl->slot_tbl_lock);
4978         tbl->max_slots = max_slots;
4979         tbl->slots = slot;
4980         tbl->highest_used_slotid = -1;  /* no slot is currently used */
4981         spin_unlock(&tbl->slot_tbl_lock);
4982         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4983                 tbl, tbl->slots, tbl->max_slots);
4984 out:
4985         dprintk("<-- %s: return %d\n", __func__, ret);
4986         return ret;
4987 }
4988
4989 /*
4990  * Initialize the forechannel and backchannel tables
4991  */
4992 static int nfs4_init_slot_tables(struct nfs4_session *session)
4993 {
4994         struct nfs4_slot_table *tbl;
4995         int status = 0;
4996
4997         tbl = &session->fc_slot_table;
4998         if (tbl->slots == NULL) {
4999                 status = nfs4_init_slot_table(tbl,
5000                                 session->fc_attrs.max_reqs, 1);
5001                 if (status)
5002                         return status;
5003         }
5004
5005         tbl = &session->bc_slot_table;
5006         if (tbl->slots == NULL) {
5007                 status = nfs4_init_slot_table(tbl,
5008                                 session->bc_attrs.max_reqs, 0);
5009                 if (status)
5010                         nfs4_destroy_slot_tables(session);
5011         }
5012
5013         return status;
5014 }
5015
5016 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5017 {
5018         struct nfs4_session *session;
5019         struct nfs4_slot_table *tbl;
5020
5021         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5022         if (!session)
5023                 return NULL;
5024
5025         tbl = &session->fc_slot_table;
5026         tbl->highest_used_slotid = -1;
5027         spin_lock_init(&tbl->slot_tbl_lock);
5028         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5029         init_completion(&tbl->complete);
5030
5031         tbl = &session->bc_slot_table;
5032         tbl->highest_used_slotid = -1;
5033         spin_lock_init(&tbl->slot_tbl_lock);
5034         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5035         init_completion(&tbl->complete);
5036
5037         session->session_state = 1<<NFS4_SESSION_INITING;
5038
5039         session->clp = clp;
5040         return session;
5041 }
5042
5043 void nfs4_destroy_session(struct nfs4_session *session)
5044 {
5045         nfs4_proc_destroy_session(session);
5046         dprintk("%s Destroy backchannel for xprt %p\n",
5047                 __func__, session->clp->cl_rpcclient->cl_xprt);
5048         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
5049                                 NFS41_BC_MIN_CALLBACKS);
5050         nfs4_destroy_slot_tables(session);
5051         kfree(session);
5052 }
5053
5054 /*
5055  * Initialize the values to be used by the client in CREATE_SESSION
5056  * If nfs4_init_session set the fore channel request and response sizes,
5057  * use them.
5058  *
5059  * Set the back channel max_resp_sz_cached to zero to force the client to
5060  * always set csa_cachethis to FALSE because the current implementation
5061  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5062  */
5063 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5064 {
5065         struct nfs4_session *session = args->client->cl_session;
5066         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5067                      mxresp_sz = session->fc_attrs.max_resp_sz;
5068
5069         if (mxrqst_sz == 0)
5070                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5071         if (mxresp_sz == 0)
5072                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5073         /* Fore channel attributes */
5074         args->fc_attrs.headerpadsz = 0;
5075         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5076         args->fc_attrs.max_resp_sz = mxresp_sz;
5077         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5078         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
5079
5080         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5081                 "max_ops=%u max_reqs=%u\n",
5082                 __func__,
5083                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5084                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5085
5086         /* Back channel attributes */
5087         args->bc_attrs.headerpadsz = 0;
5088         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5089         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5090         args->bc_attrs.max_resp_sz_cached = 0;
5091         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5092         args->bc_attrs.max_reqs = 1;
5093
5094         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5095                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5096                 __func__,
5097                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5098                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5099                 args->bc_attrs.max_reqs);
5100 }
5101
5102 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5103 {
5104         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5105         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5106
5107         if (rcvd->headerpadsz > sent->headerpadsz)
5108                 return -EINVAL;
5109         if (rcvd->max_resp_sz > sent->max_resp_sz)
5110                 return -EINVAL;
5111         /*
5112          * Our requested max_ops is the minimum we need; we're not
5113          * prepared to break up compounds into smaller pieces than that.
5114          * So, no point even trying to continue if the server won't
5115          * cooperate:
5116          */
5117         if (rcvd->max_ops < sent->max_ops)
5118                 return -EINVAL;
5119         if (rcvd->max_reqs == 0)
5120                 return -EINVAL;
5121         return 0;
5122 }
5123
5124 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5125 {
5126         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5127         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5128
5129         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5130                 return -EINVAL;
5131         if (rcvd->max_resp_sz < sent->max_resp_sz)
5132                 return -EINVAL;
5133         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5134                 return -EINVAL;
5135         /* These would render the backchannel useless: */
5136         if (rcvd->max_ops  == 0)
5137                 return -EINVAL;
5138         if (rcvd->max_reqs == 0)
5139                 return -EINVAL;
5140         return 0;
5141 }
5142
5143 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5144                                      struct nfs4_session *session)
5145 {
5146         int ret;
5147
5148         ret = nfs4_verify_fore_channel_attrs(args, session);
5149         if (ret)
5150                 return ret;
5151         return nfs4_verify_back_channel_attrs(args, session);
5152 }
5153
5154 static int _nfs4_proc_create_session(struct nfs_client *clp)
5155 {
5156         struct nfs4_session *session = clp->cl_session;
5157         struct nfs41_create_session_args args = {
5158                 .client = clp,
5159                 .cb_program = NFS4_CALLBACK,
5160         };
5161         struct nfs41_create_session_res res = {
5162                 .client = clp,
5163         };
5164         struct rpc_message msg = {
5165                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5166                 .rpc_argp = &args,
5167                 .rpc_resp = &res,
5168         };
5169         int status;
5170
5171         nfs4_init_channel_attrs(&args);
5172         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5173
5174         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5175
5176         if (!status)
5177                 /* Verify the session's negotiated channel_attrs values */
5178                 status = nfs4_verify_channel_attrs(&args, session);
5179         if (!status) {
5180                 /* Increment the clientid slot sequence id */
5181                 clp->cl_seqid++;
5182         }
5183
5184         return status;
5185 }
5186
5187 /*
5188  * Issues a CREATE_SESSION operation to the server.
5189  * It is the responsibility of the caller to verify the session is
5190  * expired before calling this routine.
5191  */
5192 int nfs4_proc_create_session(struct nfs_client *clp)
5193 {
5194         int status;
5195         unsigned *ptr;
5196         struct nfs4_session *session = clp->cl_session;
5197         long timeout = 0;
5198         int err;
5199
5200         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5201
5202         do {
5203                 status = _nfs4_proc_create_session(clp);
5204                 if (status == -NFS4ERR_DELAY) {
5205                         err = nfs4_delay(clp->cl_rpcclient, &timeout);
5206                         if (err)
5207                                 status = err;
5208                 }
5209         } while (status == -NFS4ERR_DELAY);
5210
5211         if (status)
5212                 goto out;
5213
5214         /* Init and reset the fore channel */
5215         status = nfs4_init_slot_tables(session);
5216         dprintk("slot table initialization returned %d\n", status);
5217         if (status)
5218                 goto out;
5219         status = nfs4_reset_slot_tables(session);
5220         dprintk("slot table reset returned %d\n", status);
5221         if (status)
5222                 goto out;
5223
5224         ptr = (unsigned *)&session->sess_id.data[0];
5225         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5226                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5227 out:
5228         dprintk("<-- %s\n", __func__);
5229         return status;
5230 }
5231
5232 /*
5233  * Issue the over-the-wire RPC DESTROY_SESSION.
5234  * The caller must serialize access to this routine.
5235  */
5236 int nfs4_proc_destroy_session(struct nfs4_session *session)
5237 {
5238         int status = 0;
5239         struct rpc_message msg;
5240
5241         dprintk("--> nfs4_proc_destroy_session\n");
5242
5243         /* session is still being setup */
5244         if (session->clp->cl_cons_state != NFS_CS_READY)
5245                 return status;
5246
5247         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5248         msg.rpc_argp = session;
5249         msg.rpc_resp = NULL;
5250         msg.rpc_cred = NULL;
5251         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5252
5253         if (status)
5254                 printk(KERN_WARNING
5255                         "Got error %d from the server on DESTROY_SESSION. "
5256                         "Session has been destroyed regardless...\n", status);
5257
5258         dprintk("<-- nfs4_proc_destroy_session\n");
5259         return status;
5260 }
5261
5262 int nfs4_init_session(struct nfs_server *server)
5263 {
5264         struct nfs_client *clp = server->nfs_client;
5265         struct nfs4_session *session;
5266         unsigned int rsize, wsize;
5267         int ret;
5268
5269         if (!nfs4_has_session(clp))
5270                 return 0;
5271
5272         session = clp->cl_session;
5273         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5274                 return 0;
5275
5276         rsize = server->rsize;
5277         if (rsize == 0)
5278                 rsize = NFS_MAX_FILE_IO_SIZE;
5279         wsize = server->wsize;
5280         if (wsize == 0)
5281                 wsize = NFS_MAX_FILE_IO_SIZE;
5282
5283         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5284         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5285
5286         ret = nfs4_recover_expired_lease(server);
5287         if (!ret)
5288                 ret = nfs4_check_client_ready(clp);
5289         return ret;
5290 }
5291
5292 int nfs4_init_ds_session(struct nfs_client *clp)
5293 {
5294         struct nfs4_session *session = clp->cl_session;
5295         int ret;
5296
5297         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5298                 return 0;
5299
5300         ret = nfs4_client_recover_expired_lease(clp);
5301         if (!ret)
5302                 /* Test for the DS role */
5303                 if (!is_ds_client(clp))
5304                         ret = -ENODEV;
5305         if (!ret)
5306                 ret = nfs4_check_client_ready(clp);
5307         return ret;
5308
5309 }
5310 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5311
5312
5313 /*
5314  * Renew the cl_session lease.
5315  */
5316 struct nfs4_sequence_data {
5317         struct nfs_client *clp;
5318         struct nfs4_sequence_args args;
5319         struct nfs4_sequence_res res;
5320 };
5321
5322 static void nfs41_sequence_release(void *data)
5323 {
5324         struct nfs4_sequence_data *calldata = data;
5325         struct nfs_client *clp = calldata->clp;
5326
5327         if (atomic_read(&clp->cl_count) > 1)
5328                 nfs4_schedule_state_renewal(clp);
5329         nfs_put_client(clp);
5330         kfree(calldata);
5331 }
5332
5333 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5334 {
5335         switch(task->tk_status) {
5336         case -NFS4ERR_DELAY:
5337                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5338                 return -EAGAIN;
5339         default:
5340                 nfs4_schedule_lease_recovery(clp);
5341         }
5342         return 0;
5343 }
5344
5345 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5346 {
5347         struct nfs4_sequence_data *calldata = data;
5348         struct nfs_client *clp = calldata->clp;
5349
5350         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5351                 return;
5352
5353         if (task->tk_status < 0) {
5354                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5355                 if (atomic_read(&clp->cl_count) == 1)
5356                         goto out;
5357
5358                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5359                         rpc_restart_call_prepare(task);
5360                         return;
5361                 }
5362         }
5363         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5364 out:
5365         dprintk("<-- %s\n", __func__);
5366 }
5367
5368 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5369 {
5370         struct nfs4_sequence_data *calldata = data;
5371         struct nfs_client *clp = calldata->clp;
5372         struct nfs4_sequence_args *args;
5373         struct nfs4_sequence_res *res;
5374
5375         args = task->tk_msg.rpc_argp;
5376         res = task->tk_msg.rpc_resp;
5377
5378         if (nfs41_setup_sequence(clp->cl_session, args, res, 0, task))
5379                 return;
5380         rpc_call_start(task);
5381 }
5382
5383 static const struct rpc_call_ops nfs41_sequence_ops = {
5384         .rpc_call_done = nfs41_sequence_call_done,
5385         .rpc_call_prepare = nfs41_sequence_prepare,
5386         .rpc_release = nfs41_sequence_release,
5387 };
5388
5389 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5390 {
5391         struct nfs4_sequence_data *calldata;
5392         struct rpc_message msg = {
5393                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5394                 .rpc_cred = cred,
5395         };
5396         struct rpc_task_setup task_setup_data = {
5397                 .rpc_client = clp->cl_rpcclient,
5398                 .rpc_message = &msg,
5399                 .callback_ops = &nfs41_sequence_ops,
5400                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5401         };
5402
5403         if (!atomic_inc_not_zero(&clp->cl_count))
5404                 return ERR_PTR(-EIO);
5405         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5406         if (calldata == NULL) {
5407                 nfs_put_client(clp);
5408                 return ERR_PTR(-ENOMEM);
5409         }
5410         msg.rpc_argp = &calldata->args;
5411         msg.rpc_resp = &calldata->res;
5412         calldata->clp = clp;
5413         task_setup_data.callback_data = calldata;
5414
5415         return rpc_run_task(&task_setup_data);
5416 }
5417
5418 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5419 {
5420         struct rpc_task *task;
5421         int ret = 0;
5422
5423         task = _nfs41_proc_sequence(clp, cred);
5424         if (IS_ERR(task))
5425                 ret = PTR_ERR(task);
5426         else
5427                 rpc_put_task_async(task);
5428         dprintk("<-- %s status=%d\n", __func__, ret);
5429         return ret;
5430 }
5431
5432 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5433 {
5434         struct rpc_task *task;
5435         int ret;
5436
5437         task = _nfs41_proc_sequence(clp, cred);
5438         if (IS_ERR(task)) {
5439                 ret = PTR_ERR(task);
5440                 goto out;
5441         }
5442         ret = rpc_wait_for_completion_task(task);
5443         if (!ret) {
5444                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5445
5446                 if (task->tk_status == 0)
5447                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5448                 ret = task->tk_status;
5449         }
5450         rpc_put_task(task);
5451 out:
5452         dprintk("<-- %s status=%d\n", __func__, ret);
5453         return ret;
5454 }
5455
5456 struct nfs4_reclaim_complete_data {
5457         struct nfs_client *clp;
5458         struct nfs41_reclaim_complete_args arg;
5459         struct nfs41_reclaim_complete_res res;
5460 };
5461
5462 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5463 {
5464         struct nfs4_reclaim_complete_data *calldata = data;
5465
5466         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5467         if (nfs41_setup_sequence(calldata->clp->cl_session,
5468                                 &calldata->arg.seq_args,
5469                                 &calldata->res.seq_res, 0, task))
5470                 return;
5471
5472         rpc_call_start(task);
5473 }
5474
5475 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5476 {
5477         switch(task->tk_status) {
5478         case 0:
5479         case -NFS4ERR_COMPLETE_ALREADY:
5480         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5481                 break;
5482         case -NFS4ERR_DELAY:
5483                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5484                 return -EAGAIN;
5485         default:
5486                 nfs4_schedule_lease_recovery(clp);
5487         }
5488         return 0;
5489 }
5490
5491 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5492 {
5493         struct nfs4_reclaim_complete_data *calldata = data;
5494         struct nfs_client *clp = calldata->clp;
5495         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5496
5497         dprintk("--> %s\n", __func__);
5498         if (!nfs41_sequence_done(task, res))
5499                 return;
5500
5501         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5502                 rpc_restart_call_prepare(task);
5503                 return;
5504         }
5505         dprintk("<-- %s\n", __func__);
5506 }
5507
5508 static void nfs4_free_reclaim_complete_data(void *data)
5509 {
5510         struct nfs4_reclaim_complete_data *calldata = data;
5511
5512         kfree(calldata);
5513 }
5514
5515 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5516         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5517         .rpc_call_done = nfs4_reclaim_complete_done,
5518         .rpc_release = nfs4_free_reclaim_complete_data,
5519 };
5520
5521 /*
5522  * Issue a global reclaim complete.
5523  */
5524 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5525 {
5526         struct nfs4_reclaim_complete_data *calldata;
5527         struct rpc_task *task;
5528         struct rpc_message msg = {
5529                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5530         };
5531         struct rpc_task_setup task_setup_data = {
5532                 .rpc_client = clp->cl_rpcclient,
5533                 .rpc_message = &msg,
5534                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5535                 .flags = RPC_TASK_ASYNC,
5536         };
5537         int status = -ENOMEM;
5538
5539         dprintk("--> %s\n", __func__);
5540         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5541         if (calldata == NULL)
5542                 goto out;
5543         calldata->clp = clp;
5544         calldata->arg.one_fs = 0;
5545
5546         msg.rpc_argp = &calldata->arg;
5547         msg.rpc_resp = &calldata->res;
5548         task_setup_data.callback_data = calldata;
5549         task = rpc_run_task(&task_setup_data);
5550         if (IS_ERR(task)) {
5551                 status = PTR_ERR(task);
5552                 goto out;
5553         }
5554         status = nfs4_wait_for_completion_rpc_task(task);
5555         if (status == 0)
5556                 status = task->tk_status;
5557         rpc_put_task(task);
5558         return 0;
5559 out:
5560         dprintk("<-- %s status=%d\n", __func__, status);
5561         return status;
5562 }
5563
5564 static void
5565 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5566 {
5567         struct nfs4_layoutget *lgp = calldata;
5568         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5569
5570         dprintk("--> %s\n", __func__);
5571         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5572          * right now covering the LAYOUTGET we are about to send.
5573          * However, that is not so catastrophic, and there seems
5574          * to be no way to prevent it completely.
5575          */
5576         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5577                                 &lgp->res.seq_res, 0, task))
5578                 return;
5579         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5580                                           NFS_I(lgp->args.inode)->layout,
5581                                           lgp->args.ctx->state)) {
5582                 rpc_exit(task, NFS4_OK);
5583                 return;
5584         }
5585         rpc_call_start(task);
5586 }
5587
5588 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5589 {
5590         struct nfs4_layoutget *lgp = calldata;
5591         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5592
5593         dprintk("--> %s\n", __func__);
5594
5595         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5596                 return;
5597
5598         switch (task->tk_status) {
5599         case 0:
5600                 break;
5601         case -NFS4ERR_LAYOUTTRYLATER:
5602         case -NFS4ERR_RECALLCONFLICT:
5603                 task->tk_status = -NFS4ERR_DELAY;
5604                 /* Fall through */
5605         default:
5606                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5607                         rpc_restart_call_prepare(task);
5608                         return;
5609                 }
5610         }
5611         dprintk("<-- %s\n", __func__);
5612 }
5613
5614 static void nfs4_layoutget_release(void *calldata)
5615 {
5616         struct nfs4_layoutget *lgp = calldata;
5617
5618         dprintk("--> %s\n", __func__);
5619         put_nfs_open_context(lgp->args.ctx);
5620         kfree(calldata);
5621         dprintk("<-- %s\n", __func__);
5622 }
5623
5624 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5625         .rpc_call_prepare = nfs4_layoutget_prepare,
5626         .rpc_call_done = nfs4_layoutget_done,
5627         .rpc_release = nfs4_layoutget_release,
5628 };
5629
5630 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5631 {
5632         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5633         struct rpc_task *task;
5634         struct rpc_message msg = {
5635                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5636                 .rpc_argp = &lgp->args,
5637                 .rpc_resp = &lgp->res,
5638         };
5639         struct rpc_task_setup task_setup_data = {
5640                 .rpc_client = server->client,
5641                 .rpc_message = &msg,
5642                 .callback_ops = &nfs4_layoutget_call_ops,
5643                 .callback_data = lgp,
5644                 .flags = RPC_TASK_ASYNC,
5645         };
5646         int status = 0;
5647
5648         dprintk("--> %s\n", __func__);
5649
5650         lgp->res.layoutp = &lgp->args.layout;
5651         lgp->res.seq_res.sr_slot = NULL;
5652         task = rpc_run_task(&task_setup_data);
5653         if (IS_ERR(task))
5654                 return PTR_ERR(task);
5655         status = nfs4_wait_for_completion_rpc_task(task);
5656         if (status == 0)
5657                 status = task->tk_status;
5658         if (status == 0)
5659                 status = pnfs_layout_process(lgp);
5660         rpc_put_task(task);
5661         dprintk("<-- %s status=%d\n", __func__, status);
5662         return status;
5663 }
5664
5665 static int
5666 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5667 {
5668         struct nfs4_getdeviceinfo_args args = {
5669                 .pdev = pdev,
5670         };
5671         struct nfs4_getdeviceinfo_res res = {
5672                 .pdev = pdev,
5673         };
5674         struct rpc_message msg = {
5675                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5676                 .rpc_argp = &args,
5677                 .rpc_resp = &res,
5678         };
5679         int status;
5680
5681         dprintk("--> %s\n", __func__);
5682         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5683         dprintk("<-- %s status=%d\n", __func__, status);
5684
5685         return status;
5686 }
5687
5688 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5689 {
5690         struct nfs4_exception exception = { };
5691         int err;
5692
5693         do {
5694                 err = nfs4_handle_exception(server,
5695                                         _nfs4_proc_getdeviceinfo(server, pdev),
5696                                         &exception);
5697         } while (exception.retry);
5698         return err;
5699 }
5700 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5701
5702 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
5703 {
5704         struct nfs4_layoutcommit_data *data = calldata;
5705         struct nfs_server *server = NFS_SERVER(data->args.inode);
5706
5707         if (nfs4_setup_sequence(server, &data->args.seq_args,
5708                                 &data->res.seq_res, 1, task))
5709                 return;
5710         rpc_call_start(task);
5711 }
5712
5713 static void
5714 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
5715 {
5716         struct nfs4_layoutcommit_data *data = calldata;
5717         struct nfs_server *server = NFS_SERVER(data->args.inode);
5718
5719         if (!nfs4_sequence_done(task, &data->res.seq_res))
5720                 return;
5721
5722         switch (task->tk_status) { /* Just ignore these failures */
5723         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
5724         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
5725         case NFS4ERR_BADLAYOUT:     /* no layout */
5726         case NFS4ERR_GRACE:         /* loca_recalim always false */
5727                 task->tk_status = 0;
5728         }
5729
5730         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5731                 nfs_restart_rpc(task, server->nfs_client);
5732                 return;
5733         }
5734
5735         if (task->tk_status == 0)
5736                 nfs_post_op_update_inode_force_wcc(data->args.inode,
5737                                                    data->res.fattr);
5738 }
5739
5740 static void nfs4_layoutcommit_release(void *calldata)
5741 {
5742         struct nfs4_layoutcommit_data *data = calldata;
5743
5744         /* Matched by references in pnfs_set_layoutcommit */
5745         put_lseg(data->lseg);
5746         put_rpccred(data->cred);
5747         kfree(data);
5748 }
5749
5750 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
5751         .rpc_call_prepare = nfs4_layoutcommit_prepare,
5752         .rpc_call_done = nfs4_layoutcommit_done,
5753         .rpc_release = nfs4_layoutcommit_release,
5754 };
5755
5756 int
5757 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
5758 {
5759         struct rpc_message msg = {
5760                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
5761                 .rpc_argp = &data->args,
5762                 .rpc_resp = &data->res,
5763                 .rpc_cred = data->cred,
5764         };
5765         struct rpc_task_setup task_setup_data = {
5766                 .task = &data->task,
5767                 .rpc_client = NFS_CLIENT(data->args.inode),
5768                 .rpc_message = &msg,
5769                 .callback_ops = &nfs4_layoutcommit_ops,
5770                 .callback_data = data,
5771                 .flags = RPC_TASK_ASYNC,
5772         };
5773         struct rpc_task *task;
5774         int status = 0;
5775
5776         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
5777                 "lbw: %llu inode %lu\n",
5778                 data->task.tk_pid, sync,
5779                 data->args.lastbytewritten,
5780                 data->args.inode->i_ino);
5781
5782         task = rpc_run_task(&task_setup_data);
5783         if (IS_ERR(task))
5784                 return PTR_ERR(task);
5785         if (sync == false)
5786                 goto out;
5787         status = nfs4_wait_for_completion_rpc_task(task);
5788         if (status != 0)
5789                 goto out;
5790         status = task->tk_status;
5791 out:
5792         dprintk("%s: status %d\n", __func__, status);
5793         rpc_put_task(task);
5794         return status;
5795 }
5796 #endif /* CONFIG_NFS_V4_1 */
5797
5798 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
5799         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5800         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5801         .recover_open   = nfs4_open_reclaim,
5802         .recover_lock   = nfs4_lock_reclaim,
5803         .establish_clid = nfs4_init_clientid,
5804         .get_clid_cred  = nfs4_get_setclientid_cred,
5805 };
5806
5807 #if defined(CONFIG_NFS_V4_1)
5808 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
5809         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5810         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5811         .recover_open   = nfs4_open_reclaim,
5812         .recover_lock   = nfs4_lock_reclaim,
5813         .establish_clid = nfs41_init_clientid,
5814         .get_clid_cred  = nfs4_get_exchange_id_cred,
5815         .reclaim_complete = nfs41_proc_reclaim_complete,
5816 };
5817 #endif /* CONFIG_NFS_V4_1 */
5818
5819 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
5820         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5821         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5822         .recover_open   = nfs4_open_expired,
5823         .recover_lock   = nfs4_lock_expired,
5824         .establish_clid = nfs4_init_clientid,
5825         .get_clid_cred  = nfs4_get_setclientid_cred,
5826 };
5827
5828 #if defined(CONFIG_NFS_V4_1)
5829 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
5830         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5831         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5832         .recover_open   = nfs4_open_expired,
5833         .recover_lock   = nfs4_lock_expired,
5834         .establish_clid = nfs41_init_clientid,
5835         .get_clid_cred  = nfs4_get_exchange_id_cred,
5836 };
5837 #endif /* CONFIG_NFS_V4_1 */
5838
5839 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
5840         .sched_state_renewal = nfs4_proc_async_renew,
5841         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
5842         .renew_lease = nfs4_proc_renew,
5843 };
5844
5845 #if defined(CONFIG_NFS_V4_1)
5846 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
5847         .sched_state_renewal = nfs41_proc_async_sequence,
5848         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
5849         .renew_lease = nfs4_proc_sequence,
5850 };
5851 #endif
5852
5853 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
5854         .minor_version = 0,
5855         .call_sync = _nfs4_call_sync,
5856         .validate_stateid = nfs4_validate_delegation_stateid,
5857         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
5858         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
5859         .state_renewal_ops = &nfs40_state_renewal_ops,
5860 };
5861
5862 #if defined(CONFIG_NFS_V4_1)
5863 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
5864         .minor_version = 1,
5865         .call_sync = _nfs4_call_sync_session,
5866         .validate_stateid = nfs41_validate_delegation_stateid,
5867         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
5868         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
5869         .state_renewal_ops = &nfs41_state_renewal_ops,
5870 };
5871 #endif
5872
5873 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
5874         [0] = &nfs_v4_0_minor_ops,
5875 #if defined(CONFIG_NFS_V4_1)
5876         [1] = &nfs_v4_1_minor_ops,
5877 #endif
5878 };
5879
5880 static const struct inode_operations nfs4_file_inode_operations = {
5881         .permission     = nfs_permission,
5882         .getattr        = nfs_getattr,
5883         .setattr        = nfs_setattr,
5884         .getxattr       = generic_getxattr,
5885         .setxattr       = generic_setxattr,
5886         .listxattr      = generic_listxattr,
5887         .removexattr    = generic_removexattr,
5888 };
5889
5890 const struct nfs_rpc_ops nfs_v4_clientops = {
5891         .version        = 4,                    /* protocol version */
5892         .dentry_ops     = &nfs4_dentry_operations,
5893         .dir_inode_ops  = &nfs4_dir_inode_operations,
5894         .file_inode_ops = &nfs4_file_inode_operations,
5895         .getroot        = nfs4_proc_get_root,
5896         .getattr        = nfs4_proc_getattr,
5897         .setattr        = nfs4_proc_setattr,
5898         .lookupfh       = nfs4_proc_lookupfh,
5899         .lookup         = nfs4_proc_lookup,
5900         .access         = nfs4_proc_access,
5901         .readlink       = nfs4_proc_readlink,
5902         .create         = nfs4_proc_create,
5903         .remove         = nfs4_proc_remove,
5904         .unlink_setup   = nfs4_proc_unlink_setup,
5905         .unlink_done    = nfs4_proc_unlink_done,
5906         .rename         = nfs4_proc_rename,
5907         .rename_setup   = nfs4_proc_rename_setup,
5908         .rename_done    = nfs4_proc_rename_done,
5909         .link           = nfs4_proc_link,
5910         .symlink        = nfs4_proc_symlink,
5911         .mkdir          = nfs4_proc_mkdir,
5912         .rmdir          = nfs4_proc_remove,
5913         .readdir        = nfs4_proc_readdir,
5914         .mknod          = nfs4_proc_mknod,
5915         .statfs         = nfs4_proc_statfs,
5916         .fsinfo         = nfs4_proc_fsinfo,
5917         .pathconf       = nfs4_proc_pathconf,
5918         .set_capabilities = nfs4_server_capabilities,
5919         .decode_dirent  = nfs4_decode_dirent,
5920         .read_setup     = nfs4_proc_read_setup,
5921         .read_done      = nfs4_read_done,
5922         .write_setup    = nfs4_proc_write_setup,
5923         .write_done     = nfs4_write_done,
5924         .commit_setup   = nfs4_proc_commit_setup,
5925         .commit_done    = nfs4_commit_done,
5926         .lock           = nfs4_proc_lock,
5927         .clear_acl_cache = nfs4_zap_acl_attr,
5928         .close_context  = nfs4_close_context,
5929         .open_context   = nfs4_atomic_open,
5930         .init_client    = nfs4_init_client,
5931         .secinfo        = nfs4_proc_secinfo,
5932 };
5933
5934 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
5935         .prefix = XATTR_NAME_NFSV4_ACL,
5936         .list   = nfs4_xattr_list_nfs4_acl,
5937         .get    = nfs4_xattr_get_nfs4_acl,
5938         .set    = nfs4_xattr_set_nfs4_acl,
5939 };
5940
5941 const struct xattr_handler *nfs4_xattr_handlers[] = {
5942         &nfs4_xattr_nfs4_acl_handler,
5943         NULL
5944 };
5945
5946 /*
5947  * Local variables:
5948  *  c-basic-offset: 8
5949  * End:
5950  */