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