fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/sunrpc/gss_api.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/sunrpc/bc_xprt.h>
  56 #include <linux/xattr.h>
  57 #include <linux/utsname.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65
  66 #define NFSDBG_FACILITY         NFSDBG_PROC
  67
  68 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  69 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  70
  71 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  72
  73 struct nfs4_opendata;
  74 static int _nfs4_proc_open(struct nfs4_opendata *data);
  75 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  76 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  77 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  78 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  79 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  80                             struct nfs_fattr *fattr, struct iattr *sattr,
  81                             struct nfs4_state *state);
  82 #ifdef CONFIG_NFS_V4_1
  83 static int nfs41_test_stateid(struct nfs_server *, struct nfs4_state *);
  84 static int nfs41_free_stateid(struct nfs_server *, struct nfs4_state *);
  85 #endif
  86 /* Prevent leaks of NFSv4 errors into userland */
  87 static int nfs4_map_errors(int err)
  88 {
  89         if (err >= -1000)
  90                 return err;
  91         switch (err) {
  92         case -NFS4ERR_RESOURCE:
  93                 return -EREMOTEIO;
  94         case -NFS4ERR_WRONGSEC:
  95                 return -EPERM;
  96         case -NFS4ERR_BADOWNER:
  97         case -NFS4ERR_BADNAME:
  98                 return -EINVAL;
  99         default:
 100                 dprintk("%s could not handle NFSv4 error %d\n",
 101                                 __func__, -err);
 102                 break;
 103         }
 104         return -EIO;
 105 }
 106
 107 /*
 108  * This is our standard bitmap for GETATTR requests.
 109  */
 110 const u32 nfs4_fattr_bitmap[2] = {
 111         FATTR4_WORD0_TYPE
 112         | FATTR4_WORD0_CHANGE
 113         | FATTR4_WORD0_SIZE
 114         | FATTR4_WORD0_FSID
 115         | FATTR4_WORD0_FILEID,
 116         FATTR4_WORD1_MODE
 117         | FATTR4_WORD1_NUMLINKS
 118         | FATTR4_WORD1_OWNER
 119         | FATTR4_WORD1_OWNER_GROUP
 120         | FATTR4_WORD1_RAWDEV
 121         | FATTR4_WORD1_SPACE_USED
 122         | FATTR4_WORD1_TIME_ACCESS
 123         | FATTR4_WORD1_TIME_METADATA
 124         | FATTR4_WORD1_TIME_MODIFY
 125 };
 126
 127 const u32 nfs4_statfs_bitmap[2] = {
 128         FATTR4_WORD0_FILES_AVAIL
 129         | FATTR4_WORD0_FILES_FREE
 130         | FATTR4_WORD0_FILES_TOTAL,
 131         FATTR4_WORD1_SPACE_AVAIL
 132         | FATTR4_WORD1_SPACE_FREE
 133         | FATTR4_WORD1_SPACE_TOTAL
 134 };
 135
 136 const u32 nfs4_pathconf_bitmap[2] = {
 137         FATTR4_WORD0_MAXLINK
 138         | FATTR4_WORD0_MAXNAME,
 139         0
 140 };
 141
 142 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 143                         | FATTR4_WORD0_MAXREAD
 144                         | FATTR4_WORD0_MAXWRITE
 145                         | FATTR4_WORD0_LEASE_TIME,
 146                         FATTR4_WORD1_TIME_DELTA
 147                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 148                         FATTR4_WORD2_LAYOUT_BLKSIZE
 149 };
 150
 151 const u32 nfs4_fs_locations_bitmap[2] = {
 152         FATTR4_WORD0_TYPE
 153         | FATTR4_WORD0_CHANGE
 154         | FATTR4_WORD0_SIZE
 155         | FATTR4_WORD0_FSID
 156         | FATTR4_WORD0_FILEID
 157         | FATTR4_WORD0_FS_LOCATIONS,
 158         FATTR4_WORD1_MODE
 159         | FATTR4_WORD1_NUMLINKS
 160         | FATTR4_WORD1_OWNER
 161         | FATTR4_WORD1_OWNER_GROUP
 162         | FATTR4_WORD1_RAWDEV
 163         | FATTR4_WORD1_SPACE_USED
 164         | FATTR4_WORD1_TIME_ACCESS
 165         | FATTR4_WORD1_TIME_METADATA
 166         | FATTR4_WORD1_TIME_MODIFY
 167         | FATTR4_WORD1_MOUNTED_ON_FILEID
 168 };
 169
 170 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 171                 struct nfs4_readdir_arg *readdir)
 172 {
 173         __be32 *start, *p;
 174
 175         BUG_ON(readdir->count < 80);
 176         if (cookie > 2) {
 177                 readdir->cookie = cookie;
 178                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 179                 return;
 180         }
 181
 182         readdir->cookie = 0;
 183         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 184         if (cookie == 2)
 185                 return;
 186
 187         /*
 188          * NFSv4 servers do not return entries for '.' and '..'
 189          * Therefore, we fake these entries here.  We let '.'
 190          * have cookie 0 and '..' have cookie 1.  Note that
 191          * when talking to the server, we always send cookie 0
 192          * instead of 1 or 2.
 193          */
 194         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 195
 196         if (cookie == 0) {
 197                 *p++ = xdr_one;                                  /* next */
 198                 *p++ = xdr_zero;                   /* cookie, first word */
 199                 *p++ = xdr_one;                   /* cookie, second word */
 200                 *p++ = xdr_one;                             /* entry len */
 201                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 202                 p++;
 203                 *p++ = xdr_one;                         /* bitmap length */
 204                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 205                 *p++ = htonl(8);              /* attribute buffer length */
 206                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 207         }
 208
 209         *p++ = xdr_one;                                  /* next */
 210         *p++ = xdr_zero;                   /* cookie, first word */
 211         *p++ = xdr_two;                   /* cookie, second word */
 212         *p++ = xdr_two;                             /* entry len */
 213         memcpy(p, "..\0\0", 4);                         /* entry */
 214         p++;
 215         *p++ = xdr_one;                         /* bitmap length */
 216         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 217         *p++ = htonl(8);              /* attribute buffer length */
 218         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 219
 220         readdir->pgbase = (char *)p - (char *)start;
 221         readdir->count -= readdir->pgbase;
 222         kunmap_atomic(start, KM_USER0);
 223 }
 224
 225 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 226 {
 227         int res;
 228
 229         might_sleep();
 230
 231         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 232                         nfs_wait_bit_killable, TASK_KILLABLE);
 233         return res;
 234 }
 235
 236 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 237 {
 238         int res = 0;
 239
 240         might_sleep();
 241
 242         if (*timeout <= 0)
 243                 *timeout = NFS4_POLL_RETRY_MIN;
 244         if (*timeout > NFS4_POLL_RETRY_MAX)
 245                 *timeout = NFS4_POLL_RETRY_MAX;
 246         schedule_timeout_killable(*timeout);
 247         if (fatal_signal_pending(current))
 248                 res = -ERESTARTSYS;
 249         *timeout <<= 1;
 250         return res;
 251 }
 252
 253 /* This is the error handling routine for processes that are allowed
 254  * to sleep.
 255  */
 256 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 257 {
 258         struct nfs_client *clp = server->nfs_client;
 259         struct nfs4_state *state = exception->state;
 260         struct inode *inode = exception->inode;
 261         int ret = errorcode;
 262
 263         exception->retry = 0;
 264         switch(errorcode) {
 265                 case 0:
 266                         return 0;
 267                 case -NFS4ERR_OPENMODE:
 268                         if (nfs_have_delegation(inode, FMODE_READ)) {
 269                                 nfs_inode_return_delegation(inode);
 270                                 exception->retry = 1;
 271                                 return 0;
 272                         }
 273                         if (state == NULL)
 274                                 break;
 275                         nfs4_schedule_stateid_recovery(server, state);
 276                         goto wait_on_recovery;
 277                 case -NFS4ERR_DELEG_REVOKED:
 278                 case -NFS4ERR_ADMIN_REVOKED:
 279                 case -NFS4ERR_BAD_STATEID:
 280                         if (state != NULL)
 281                                 nfs_remove_bad_delegation(state->inode);
 282                         if (state == NULL)
 283                                 break;
 284                         nfs4_schedule_stateid_recovery(server, state);
 285                         goto wait_on_recovery;
 286                 case -NFS4ERR_EXPIRED:
 287                         if (state != NULL)
 288                                 nfs4_schedule_stateid_recovery(server, state);
 289                 case -NFS4ERR_STALE_STATEID:
 290                 case -NFS4ERR_STALE_CLIENTID:
 291                         nfs4_schedule_lease_recovery(clp);
 292                         goto wait_on_recovery;
 293 #if defined(CONFIG_NFS_V4_1)
 294                 case -NFS4ERR_BADSESSION:
 295                 case -NFS4ERR_BADSLOT:
 296                 case -NFS4ERR_BAD_HIGH_SLOT:
 297                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 298                 case -NFS4ERR_DEADSESSION:
 299                 case -NFS4ERR_SEQ_FALSE_RETRY:
 300                 case -NFS4ERR_SEQ_MISORDERED:
 301                         dprintk("%s ERROR: %d Reset session\n", __func__,
 302                                 errorcode);
 303                         nfs4_schedule_session_recovery(clp->cl_session);
 304                         exception->retry = 1;
 305                         break;
 306 #endif /* defined(CONFIG_NFS_V4_1) */
 307                 case -NFS4ERR_FILE_OPEN:
 308                         if (exception->timeout > HZ) {
 309                                 /* We have retried a decent amount, time to
 310                                  * fail
 311                                  */
 312                                 ret = -EBUSY;
 313                                 break;
 314                         }
 315                 case -NFS4ERR_GRACE:
 316                 case -NFS4ERR_DELAY:
 317                 case -EKEYEXPIRED:
 318                         ret = nfs4_delay(server->client, &exception->timeout);
 319                         if (ret != 0)
 320                                 break;
 321                 case -NFS4ERR_RETRY_UNCACHED_REP:
 322                 case -NFS4ERR_OLD_STATEID:
 323                         exception->retry = 1;
 324                         break;
 325                 case -NFS4ERR_BADOWNER:
 326                         /* The following works around a Linux server bug! */
 327                 case -NFS4ERR_BADNAME:
 328                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 329                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 330                                 exception->retry = 1;
 331                                 printk(KERN_WARNING "NFS: v4 server %s "
 332                                                 "does not accept raw "
 333                                                 "uid/gids. "
 334                                                 "Reenabling the idmapper.\n",
 335                                                 server->nfs_client->cl_hostname);
 336                         }
 337         }
 338         /* We failed to handle the error */
 339         return nfs4_map_errors(ret);
 340 wait_on_recovery:
 341         ret = nfs4_wait_clnt_recover(clp);
 342         if (ret == 0)
 343                 exception->retry = 1;
 344         return ret;
 345 }
 346
 347
 348 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 349 {
 350         spin_lock(&clp->cl_lock);
 351         if (time_before(clp->cl_last_renewal,timestamp))
 352                 clp->cl_last_renewal = timestamp;
 353         spin_unlock(&clp->cl_lock);
 354 }
 355
 356 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 357 {
 358         do_renew_lease(server->nfs_client, timestamp);
 359 }
 360
 361 #if defined(CONFIG_NFS_V4_1)
 362
 363 /*
 364  * nfs4_free_slot - free a slot and efficiently update slot table.
 365  *
 366  * freeing a slot is trivially done by clearing its respective bit
 367  * in the bitmap.
 368  * If the freed slotid equals highest_used_slotid we want to update it
 369  * so that the server would be able to size down the slot table if needed,
 370  * otherwise we know that the highest_used_slotid is still in use.
 371  * When updating highest_used_slotid there may be "holes" in the bitmap
 372  * so we need to scan down from highest_used_slotid to 0 looking for the now
 373  * highest slotid in use.
 374  * If none found, highest_used_slotid is set to -1.
 375  *
 376  * Must be called while holding tbl->slot_tbl_lock
 377  */
 378 static void
 379 nfs4_free_slot(struct nfs4_slot_table *tbl, struct nfs4_slot *free_slot)
 380 {
 381         int free_slotid = free_slot - tbl->slots;
 382         int slotid = free_slotid;
 383
 384         BUG_ON(slotid < 0 || slotid >= NFS4_MAX_SLOT_TABLE);
 385         /* clear used bit in bitmap */
 386         __clear_bit(slotid, tbl->used_slots);
 387
 388         /* update highest_used_slotid when it is freed */
 389         if (slotid == tbl->highest_used_slotid) {
 390                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 391                 if (slotid < tbl->max_slots)
 392                         tbl->highest_used_slotid = slotid;
 393                 else
 394                         tbl->highest_used_slotid = -1;
 395         }
 396         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
 397                 free_slotid, tbl->highest_used_slotid);
 398 }
 399
 400 /*
 401  * Signal state manager thread if session fore channel is drained
 402  */
 403 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 404 {
 405         struct rpc_task *task;
 406
 407         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 408                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
 409                 if (task)
 410                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 411                 return;
 412         }
 413
 414         if (ses->fc_slot_table.highest_used_slotid != -1)
 415                 return;
 416
 417         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 418         complete(&ses->fc_slot_table.complete);
 419 }
 420
 421 /*
 422  * Signal state manager thread if session back channel is drained
 423  */
 424 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 425 {
 426         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 427             ses->bc_slot_table.highest_used_slotid != -1)
 428                 return;
 429         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 430         complete(&ses->bc_slot_table.complete);
 431 }
 432
 433 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 434 {
 435         struct nfs4_slot_table *tbl;
 436
 437         tbl = &res->sr_session->fc_slot_table;
 438         if (!res->sr_slot) {
 439                 /* just wake up the next guy waiting since
 440                  * we may have not consumed a slot after all */
 441                 dprintk("%s: No slot\n", __func__);
 442                 return;
 443         }
 444
 445         spin_lock(&tbl->slot_tbl_lock);
 446         nfs4_free_slot(tbl, res->sr_slot);
 447         nfs4_check_drain_fc_complete(res->sr_session);
 448         spin_unlock(&tbl->slot_tbl_lock);
 449         res->sr_slot = NULL;
 450 }
 451
 452 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 453 {
 454         unsigned long timestamp;
 455         struct nfs_client *clp;
 456
 457         /*
 458          * sr_status remains 1 if an RPC level error occurred. The server
 459          * may or may not have processed the sequence operation..
 460          * Proceed as if the server received and processed the sequence
 461          * operation.
 462          */
 463         if (res->sr_status == 1)
 464                 res->sr_status = NFS_OK;
 465
 466         /* don't increment the sequence number if the task wasn't sent */
 467         if (!RPC_WAS_SENT(task))
 468                 goto out;
 469
 470         /* Check the SEQUENCE operation status */
 471         switch (res->sr_status) {
 472         case 0:
 473                 /* Update the slot's sequence and clientid lease timer */
 474                 ++res->sr_slot->seq_nr;
 475                 timestamp = res->sr_renewal_time;
 476                 clp = res->sr_session->clp;
 477                 do_renew_lease(clp, timestamp);
 478                 /* Check sequence flags */
 479                 if (res->sr_status_flags != 0)
 480                         nfs4_schedule_lease_recovery(clp);
 481                 break;
 482         case -NFS4ERR_DELAY:
 483                 /* The server detected a resend of the RPC call and
 484                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 485                  * of RFC5661.
 486                  */
 487                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 488                         __func__,
 489                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 490                         res->sr_slot->seq_nr);
 491                 goto out_retry;
 492         default:
 493                 /* Just update the slot sequence no. */
 494                 ++res->sr_slot->seq_nr;
 495         }
 496 out:
 497         /* The session may be reset by one of the error handlers. */
 498         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 499         nfs41_sequence_free_slot(res);
 500         return 1;
 501 out_retry:
 502         if (!rpc_restart_call(task))
 503                 goto out;
 504         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 505         return 0;
 506 }
 507
 508 static int nfs4_sequence_done(struct rpc_task *task,
 509                                struct nfs4_sequence_res *res)
 510 {
 511         if (res->sr_session == NULL)
 512                 return 1;
 513         return nfs41_sequence_done(task, res);
 514 }
 515
 516 /*
 517  * nfs4_find_slot - efficiently look for a free slot
 518  *
 519  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 520  * If found, we mark the slot as used, update the highest_used_slotid,
 521  * and respectively set up the sequence operation args.
 522  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
 523  *
 524  * Note: must be called with under the slot_tbl_lock.
 525  */
 526 static u8
 527 nfs4_find_slot(struct nfs4_slot_table *tbl)
 528 {
 529         int slotid;
 530         u8 ret_id = NFS4_MAX_SLOT_TABLE;
 531         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
 532
 533         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
 534                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 535                 tbl->max_slots);
 536         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 537         if (slotid >= tbl->max_slots)
 538                 goto out;
 539         __set_bit(slotid, tbl->used_slots);
 540         if (slotid > tbl->highest_used_slotid)
 541                 tbl->highest_used_slotid = slotid;
 542         ret_id = slotid;
 543 out:
 544         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 545                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 546         return ret_id;
 547 }
 548
 549 int nfs41_setup_sequence(struct nfs4_session *session,
 550                                 struct nfs4_sequence_args *args,
 551                                 struct nfs4_sequence_res *res,
 552                                 int cache_reply,
 553                                 struct rpc_task *task)
 554 {
 555         struct nfs4_slot *slot;
 556         struct nfs4_slot_table *tbl;
 557         u8 slotid;
 558
 559         dprintk("--> %s\n", __func__);
 560         /* slot already allocated? */
 561         if (res->sr_slot != NULL)
 562                 return 0;
 563
 564         tbl = &session->fc_slot_table;
 565
 566         spin_lock(&tbl->slot_tbl_lock);
 567         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 568             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 569                 /*
 570                  * The state manager will wait until the slot table is empty.
 571                  * Schedule the reset thread
 572                  */
 573                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 574                 spin_unlock(&tbl->slot_tbl_lock);
 575                 dprintk("%s Schedule Session Reset\n", __func__);
 576                 return -EAGAIN;
 577         }
 578
 579         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 580             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 581                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 582                 spin_unlock(&tbl->slot_tbl_lock);
 583                 dprintk("%s enforce FIFO order\n", __func__);
 584                 return -EAGAIN;
 585         }
 586
 587         slotid = nfs4_find_slot(tbl);
 588         if (slotid == NFS4_MAX_SLOT_TABLE) {
 589                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 590                 spin_unlock(&tbl->slot_tbl_lock);
 591                 dprintk("<-- %s: no free slots\n", __func__);
 592                 return -EAGAIN;
 593         }
 594         spin_unlock(&tbl->slot_tbl_lock);
 595
 596         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 597         slot = tbl->slots + slotid;
 598         args->sa_session = session;
 599         args->sa_slotid = slotid;
 600         args->sa_cache_this = cache_reply;
 601
 602         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 603
 604         res->sr_session = session;
 605         res->sr_slot = slot;
 606         res->sr_renewal_time = jiffies;
 607         res->sr_status_flags = 0;
 608         /*
 609          * sr_status is only set in decode_sequence, and so will remain
 610          * set to 1 if an rpc level failure occurs.
 611          */
 612         res->sr_status = 1;
 613         return 0;
 614 }
 615 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 616
 617 int nfs4_setup_sequence(const struct nfs_server *server,
 618                         struct nfs4_sequence_args *args,
 619                         struct nfs4_sequence_res *res,
 620                         int cache_reply,
 621                         struct rpc_task *task)
 622 {
 623         struct nfs4_session *session = nfs4_get_session(server);
 624         int ret = 0;
 625
 626         if (session == NULL) {
 627                 args->sa_session = NULL;
 628                 res->sr_session = NULL;
 629                 goto out;
 630         }
 631
 632         dprintk("--> %s clp %p session %p sr_slot %td\n",
 633                 __func__, session->clp, session, res->sr_slot ?
 634                         res->sr_slot - session->fc_slot_table.slots : -1);
 635
 636         ret = nfs41_setup_sequence(session, args, res, cache_reply,
 637                                    task);
 638 out:
 639         dprintk("<-- %s status=%d\n", __func__, ret);
 640         return ret;
 641 }
 642
 643 struct nfs41_call_sync_data {
 644         const struct nfs_server *seq_server;
 645         struct nfs4_sequence_args *seq_args;
 646         struct nfs4_sequence_res *seq_res;
 647         int cache_reply;
 648 };
 649
 650 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 651 {
 652         struct nfs41_call_sync_data *data = calldata;
 653
 654         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 655
 656         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 657                                 data->seq_res, data->cache_reply, task))
 658                 return;
 659         rpc_call_start(task);
 660 }
 661
 662 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 663 {
 664         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 665         nfs41_call_sync_prepare(task, calldata);
 666 }
 667
 668 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 669 {
 670         struct nfs41_call_sync_data *data = calldata;
 671
 672         nfs41_sequence_done(task, data->seq_res);
 673 }
 674
 675 struct rpc_call_ops nfs41_call_sync_ops = {
 676         .rpc_call_prepare = nfs41_call_sync_prepare,
 677         .rpc_call_done = nfs41_call_sync_done,
 678 };
 679
 680 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 681         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 682         .rpc_call_done = nfs41_call_sync_done,
 683 };
 684
 685 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 686                                    struct nfs_server *server,
 687                                    struct rpc_message *msg,
 688                                    struct nfs4_sequence_args *args,
 689                                    struct nfs4_sequence_res *res,
 690                                    int cache_reply,
 691                                    int privileged)
 692 {
 693         int ret;
 694         struct rpc_task *task;
 695         struct nfs41_call_sync_data data = {
 696                 .seq_server = server,
 697                 .seq_args = args,
 698                 .seq_res = res,
 699                 .cache_reply = cache_reply,
 700         };
 701         struct rpc_task_setup task_setup = {
 702                 .rpc_client = clnt,
 703                 .rpc_message = msg,
 704                 .callback_ops = &nfs41_call_sync_ops,
 705                 .callback_data = &data
 706         };
 707
 708         res->sr_slot = NULL;
 709         if (privileged)
 710                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 711         task = rpc_run_task(&task_setup);
 712         if (IS_ERR(task))
 713                 ret = PTR_ERR(task);
 714         else {
 715                 ret = task->tk_status;
 716                 rpc_put_task(task);
 717         }
 718         return ret;
 719 }
 720
 721 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 722                             struct nfs_server *server,
 723                             struct rpc_message *msg,
 724                             struct nfs4_sequence_args *args,
 725                             struct nfs4_sequence_res *res,
 726                             int cache_reply)
 727 {
 728         return nfs4_call_sync_sequence(clnt, server, msg, args, res, cache_reply, 0);
 729 }
 730
 731 #else
 732 static int nfs4_sequence_done(struct rpc_task *task,
 733                                struct nfs4_sequence_res *res)
 734 {
 735         return 1;
 736 }
 737 #endif /* CONFIG_NFS_V4_1 */
 738
 739 int _nfs4_call_sync(struct rpc_clnt *clnt,
 740                     struct nfs_server *server,
 741                     struct rpc_message *msg,
 742                     struct nfs4_sequence_args *args,
 743                     struct nfs4_sequence_res *res,
 744                     int cache_reply)
 745 {
 746         args->sa_session = res->sr_session = NULL;
 747         return rpc_call_sync(clnt, msg, 0);
 748 }
 749
 750 static inline
 751 int nfs4_call_sync(struct rpc_clnt *clnt,
 752                    struct nfs_server *server,
 753                    struct rpc_message *msg,
 754                    struct nfs4_sequence_args *args,
 755                    struct nfs4_sequence_res *res,
 756                    int cache_reply)
 757 {
 758         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 759                                                 args, res, cache_reply);
 760 }
 761
 762 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 763 {
 764         struct nfs_inode *nfsi = NFS_I(dir);
 765
 766         spin_lock(&dir->i_lock);
 767         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 768         if (!cinfo->atomic || cinfo->before != dir->i_version)
 769                 nfs_force_lookup_revalidate(dir);
 770         dir->i_version = cinfo->after;
 771         spin_unlock(&dir->i_lock);
 772 }
 773
 774 struct nfs4_opendata {
 775         struct kref kref;
 776         struct nfs_openargs o_arg;
 777         struct nfs_openres o_res;
 778         struct nfs_open_confirmargs c_arg;
 779         struct nfs_open_confirmres c_res;
 780         struct nfs_fattr f_attr;
 781         struct nfs_fattr dir_attr;
 782         struct dentry *dir;
 783         struct dentry *dentry;
 784         struct nfs4_state_owner *owner;
 785         struct nfs4_state *state;
 786         struct iattr attrs;
 787         unsigned long timestamp;
 788         unsigned int rpc_done : 1;
 789         int rpc_status;
 790         int cancelled;
 791 };
 792
 793
 794 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 795 {
 796         p->o_res.f_attr = &p->f_attr;
 797         p->o_res.dir_attr = &p->dir_attr;
 798         p->o_res.seqid = p->o_arg.seqid;
 799         p->c_res.seqid = p->c_arg.seqid;
 800         p->o_res.server = p->o_arg.server;
 801         nfs_fattr_init(&p->f_attr);
 802         nfs_fattr_init(&p->dir_attr);
 803 }
 804
 805 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 806                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 807                 const struct iattr *attrs,
 808                 gfp_t gfp_mask)
 809 {
 810         struct dentry *parent = dget_parent(dentry);
 811         struct inode *dir = parent->d_inode;
 812         struct nfs_server *server = NFS_SERVER(dir);
 813         struct nfs4_opendata *p;
 814
 815         p = kzalloc(sizeof(*p), gfp_mask);
 816         if (p == NULL)
 817                 goto err;
 818         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 819         if (p->o_arg.seqid == NULL)
 820                 goto err_free;
 821         nfs_sb_active(dentry->d_sb);
 822         p->dentry = dget(dentry);
 823         p->dir = parent;
 824         p->owner = sp;
 825         atomic_inc(&sp->so_count);
 826         p->o_arg.fh = NFS_FH(dir);
 827         p->o_arg.open_flags = flags;
 828         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 829         p->o_arg.clientid = server->nfs_client->cl_clientid;
 830         p->o_arg.id = sp->so_owner_id.id;
 831         p->o_arg.name = &dentry->d_name;
 832         p->o_arg.server = server;
 833         p->o_arg.bitmask = server->attr_bitmask;
 834         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 835         if (flags & O_CREAT) {
 836                 u32 *s;
 837
 838                 p->o_arg.u.attrs = &p->attrs;
 839                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 840                 s = (u32 *) p->o_arg.u.verifier.data;
 841                 s[0] = jiffies;
 842                 s[1] = current->pid;
 843         }
 844         p->c_arg.fh = &p->o_res.fh;
 845         p->c_arg.stateid = &p->o_res.stateid;
 846         p->c_arg.seqid = p->o_arg.seqid;
 847         nfs4_init_opendata_res(p);
 848         kref_init(&p->kref);
 849         return p;
 850 err_free:
 851         kfree(p);
 852 err:
 853         dput(parent);
 854         return NULL;
 855 }
 856
 857 static void nfs4_opendata_free(struct kref *kref)
 858 {
 859         struct nfs4_opendata *p = container_of(kref,
 860                         struct nfs4_opendata, kref);
 861         struct super_block *sb = p->dentry->d_sb;
 862
 863         nfs_free_seqid(p->o_arg.seqid);
 864         if (p->state != NULL)
 865                 nfs4_put_open_state(p->state);
 866         nfs4_put_state_owner(p->owner);
 867         dput(p->dir);
 868         dput(p->dentry);
 869         nfs_sb_deactive(sb);
 870         kfree(p);
 871 }
 872
 873 static void nfs4_opendata_put(struct nfs4_opendata *p)
 874 {
 875         if (p != NULL)
 876                 kref_put(&p->kref, nfs4_opendata_free);
 877 }
 878
 879 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 880 {
 881         int ret;
 882
 883         ret = rpc_wait_for_completion_task(task);
 884         return ret;
 885 }
 886
 887 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 888 {
 889         int ret = 0;
 890
 891         if (open_mode & O_EXCL)
 892                 goto out;
 893         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 894                 case FMODE_READ:
 895                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 896                                 && state->n_rdonly != 0;
 897                         break;
 898                 case FMODE_WRITE:
 899                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 900                                 && state->n_wronly != 0;
 901                         break;
 902                 case FMODE_READ|FMODE_WRITE:
 903                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 904                                 && state->n_rdwr != 0;
 905         }
 906 out:
 907         return ret;
 908 }
 909
 910 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 911 {
 912         if (delegation == NULL)
 913                 return 0;
 914         if ((delegation->type & fmode) != fmode)
 915                 return 0;
 916         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 917                 return 0;
 918         nfs_mark_delegation_referenced(delegation);
 919         return 1;
 920 }
 921
 922 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 923 {
 924         switch (fmode) {
 925                 case FMODE_WRITE:
 926                         state->n_wronly++;
 927                         break;
 928                 case FMODE_READ:
 929                         state->n_rdonly++;
 930                         break;
 931                 case FMODE_READ|FMODE_WRITE:
 932                         state->n_rdwr++;
 933         }
 934         nfs4_state_set_mode_locked(state, state->state | fmode);
 935 }
 936
 937 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 938 {
 939         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 940                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 941         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 942         switch (fmode) {
 943                 case FMODE_READ:
 944                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 945                         break;
 946                 case FMODE_WRITE:
 947                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 948                         break;
 949                 case FMODE_READ|FMODE_WRITE:
 950                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 951         }
 952 }
 953
 954 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 955 {
 956         write_seqlock(&state->seqlock);
 957         nfs_set_open_stateid_locked(state, stateid, fmode);
 958         write_sequnlock(&state->seqlock);
 959 }
 960
 961 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 962 {
 963         /*
 964          * Protect the call to nfs4_state_set_mode_locked and
 965          * serialise the stateid update
 966          */
 967         write_seqlock(&state->seqlock);
 968         if (deleg_stateid != NULL) {
 969                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 970                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 971         }
 972         if (open_stateid != NULL)
 973                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 974         write_sequnlock(&state->seqlock);
 975         spin_lock(&state->owner->so_lock);
 976         update_open_stateflags(state, fmode);
 977         spin_unlock(&state->owner->so_lock);
 978 }
 979
 980 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 981 {
 982         struct nfs_inode *nfsi = NFS_I(state->inode);
 983         struct nfs_delegation *deleg_cur;
 984         int ret = 0;
 985
 986         fmode &= (FMODE_READ|FMODE_WRITE);
 987
 988         rcu_read_lock();
 989         deleg_cur = rcu_dereference(nfsi->delegation);
 990         if (deleg_cur == NULL)
 991                 goto no_delegation;
 992
 993         spin_lock(&deleg_cur->lock);
 994         if (nfsi->delegation != deleg_cur ||
 995             (deleg_cur->type & fmode) != fmode)
 996                 goto no_delegation_unlock;
 997
 998         if (delegation == NULL)
 999                 delegation = &deleg_cur->stateid;
1000         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
1001                 goto no_delegation_unlock;
1002
1003         nfs_mark_delegation_referenced(deleg_cur);
1004         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1005         ret = 1;
1006 no_delegation_unlock:
1007         spin_unlock(&deleg_cur->lock);
1008 no_delegation:
1009         rcu_read_unlock();
1010
1011         if (!ret && open_stateid != NULL) {
1012                 __update_open_stateid(state, open_stateid, NULL, fmode);
1013                 ret = 1;
1014         }
1015
1016         return ret;
1017 }
1018
1019
1020 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1021 {
1022         struct nfs_delegation *delegation;
1023
1024         rcu_read_lock();
1025         delegation = rcu_dereference(NFS_I(inode)->delegation);
1026         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1027                 rcu_read_unlock();
1028                 return;
1029         }
1030         rcu_read_unlock();
1031         nfs_inode_return_delegation(inode);
1032 }
1033
1034 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1035 {
1036         struct nfs4_state *state = opendata->state;
1037         struct nfs_inode *nfsi = NFS_I(state->inode);
1038         struct nfs_delegation *delegation;
1039         int open_mode = opendata->o_arg.open_flags & O_EXCL;
1040         fmode_t fmode = opendata->o_arg.fmode;
1041         nfs4_stateid stateid;
1042         int ret = -EAGAIN;
1043
1044         for (;;) {
1045                 if (can_open_cached(state, fmode, open_mode)) {
1046                         spin_lock(&state->owner->so_lock);
1047                         if (can_open_cached(state, fmode, open_mode)) {
1048                                 update_open_stateflags(state, fmode);
1049                                 spin_unlock(&state->owner->so_lock);
1050                                 goto out_return_state;
1051                         }
1052                         spin_unlock(&state->owner->so_lock);
1053                 }
1054                 rcu_read_lock();
1055                 delegation = rcu_dereference(nfsi->delegation);
1056                 if (!can_open_delegated(delegation, fmode)) {
1057                         rcu_read_unlock();
1058                         break;
1059                 }
1060                 /* Save the delegation */
1061                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1062                 rcu_read_unlock();
1063                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1064                 if (ret != 0)
1065                         goto out;
1066                 ret = -EAGAIN;
1067
1068                 /* Try to update the stateid using the delegation */
1069                 if (update_open_stateid(state, NULL, &stateid, fmode))
1070                         goto out_return_state;
1071         }
1072 out:
1073         return ERR_PTR(ret);
1074 out_return_state:
1075         atomic_inc(&state->count);
1076         return state;
1077 }
1078
1079 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1080 {
1081         struct inode *inode;
1082         struct nfs4_state *state = NULL;
1083         struct nfs_delegation *delegation;
1084         int ret;
1085
1086         if (!data->rpc_done) {
1087                 state = nfs4_try_open_cached(data);
1088                 goto out;
1089         }
1090
1091         ret = -EAGAIN;
1092         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1093                 goto err;
1094         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1095         ret = PTR_ERR(inode);
1096         if (IS_ERR(inode))
1097                 goto err;
1098         ret = -ENOMEM;
1099         state = nfs4_get_open_state(inode, data->owner);
1100         if (state == NULL)
1101                 goto err_put_inode;
1102         if (data->o_res.delegation_type != 0) {
1103                 int delegation_flags = 0;
1104
1105                 rcu_read_lock();
1106                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1107                 if (delegation)
1108                         delegation_flags = delegation->flags;
1109                 rcu_read_unlock();
1110                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1111                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1112                                         "returning a delegation for "
1113                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1114                                         NFS_CLIENT(inode)->cl_server);
1115                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1116                         nfs_inode_set_delegation(state->inode,
1117                                         data->owner->so_cred,
1118                                         &data->o_res);
1119                 else
1120                         nfs_inode_reclaim_delegation(state->inode,
1121                                         data->owner->so_cred,
1122                                         &data->o_res);
1123         }
1124
1125         update_open_stateid(state, &data->o_res.stateid, NULL,
1126                         data->o_arg.fmode);
1127         iput(inode);
1128 out:
1129         return state;
1130 err_put_inode:
1131         iput(inode);
1132 err:
1133         return ERR_PTR(ret);
1134 }
1135
1136 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1137 {
1138         struct nfs_inode *nfsi = NFS_I(state->inode);
1139         struct nfs_open_context *ctx;
1140
1141         spin_lock(&state->inode->i_lock);
1142         list_for_each_entry(ctx, &nfsi->open_files, list) {
1143                 if (ctx->state != state)
1144                         continue;
1145                 get_nfs_open_context(ctx);
1146                 spin_unlock(&state->inode->i_lock);
1147                 return ctx;
1148         }
1149         spin_unlock(&state->inode->i_lock);
1150         return ERR_PTR(-ENOENT);
1151 }
1152
1153 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1154 {
1155         struct nfs4_opendata *opendata;
1156
1157         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1158         if (opendata == NULL)
1159                 return ERR_PTR(-ENOMEM);
1160         opendata->state = state;
1161         atomic_inc(&state->count);
1162         return opendata;
1163 }
1164
1165 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1166 {
1167         struct nfs4_state *newstate;
1168         int ret;
1169
1170         opendata->o_arg.open_flags = 0;
1171         opendata->o_arg.fmode = fmode;
1172         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1173         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1174         nfs4_init_opendata_res(opendata);
1175         ret = _nfs4_recover_proc_open(opendata);
1176         if (ret != 0)
1177                 return ret;
1178         newstate = nfs4_opendata_to_nfs4_state(opendata);
1179         if (IS_ERR(newstate))
1180                 return PTR_ERR(newstate);
1181         nfs4_close_state(newstate, fmode);
1182         *res = newstate;
1183         return 0;
1184 }
1185
1186 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1187 {
1188         struct nfs4_state *newstate;
1189         int ret;
1190
1191         /* memory barrier prior to reading state->n_* */
1192         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1193         smp_rmb();
1194         if (state->n_rdwr != 0) {
1195                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1196                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1197                 if (ret != 0)
1198                         return ret;
1199                 if (newstate != state)
1200                         return -ESTALE;
1201         }
1202         if (state->n_wronly != 0) {
1203                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1204                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1205                 if (ret != 0)
1206                         return ret;
1207                 if (newstate != state)
1208                         return -ESTALE;
1209         }
1210         if (state->n_rdonly != 0) {
1211                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1212                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1213                 if (ret != 0)
1214                         return ret;
1215                 if (newstate != state)
1216                         return -ESTALE;
1217         }
1218         /*
1219          * We may have performed cached opens for all three recoveries.
1220          * Check if we need to update the current stateid.
1221          */
1222         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1223             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1224                 write_seqlock(&state->seqlock);
1225                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1226                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1227                 write_sequnlock(&state->seqlock);
1228         }
1229         return 0;
1230 }
1231
1232 /*
1233  * OPEN_RECLAIM:
1234  *      reclaim state on the server after a reboot.
1235  */
1236 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1237 {
1238         struct nfs_delegation *delegation;
1239         struct nfs4_opendata *opendata;
1240         fmode_t delegation_type = 0;
1241         int status;
1242
1243         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1244         if (IS_ERR(opendata))
1245                 return PTR_ERR(opendata);
1246         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1247         opendata->o_arg.fh = NFS_FH(state->inode);
1248         rcu_read_lock();
1249         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1250         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1251                 delegation_type = delegation->type;
1252         rcu_read_unlock();
1253         opendata->o_arg.u.delegation_type = delegation_type;
1254         status = nfs4_open_recover(opendata, state);
1255         nfs4_opendata_put(opendata);
1256         return status;
1257 }
1258
1259 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1260 {
1261         struct nfs_server *server = NFS_SERVER(state->inode);
1262         struct nfs4_exception exception = { };
1263         int err;
1264         do {
1265                 err = _nfs4_do_open_reclaim(ctx, state);
1266                 if (err != -NFS4ERR_DELAY)
1267                         break;
1268                 nfs4_handle_exception(server, err, &exception);
1269         } while (exception.retry);
1270         return err;
1271 }
1272
1273 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1274 {
1275         struct nfs_open_context *ctx;
1276         int ret;
1277
1278         ctx = nfs4_state_find_open_context(state);
1279         if (IS_ERR(ctx))
1280                 return PTR_ERR(ctx);
1281         ret = nfs4_do_open_reclaim(ctx, state);
1282         put_nfs_open_context(ctx);
1283         return ret;
1284 }
1285
1286 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1287 {
1288         struct nfs4_opendata *opendata;
1289         int ret;
1290
1291         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1292         if (IS_ERR(opendata))
1293                 return PTR_ERR(opendata);
1294         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1295         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1296                         sizeof(opendata->o_arg.u.delegation.data));
1297         ret = nfs4_open_recover(opendata, state);
1298         nfs4_opendata_put(opendata);
1299         return ret;
1300 }
1301
1302 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1303 {
1304         struct nfs4_exception exception = { };
1305         struct nfs_server *server = NFS_SERVER(state->inode);
1306         int err;
1307         do {
1308                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1309                 switch (err) {
1310                         case 0:
1311                         case -ENOENT:
1312                         case -ESTALE:
1313                                 goto out;
1314                         case -NFS4ERR_BADSESSION:
1315                         case -NFS4ERR_BADSLOT:
1316                         case -NFS4ERR_BAD_HIGH_SLOT:
1317                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1318                         case -NFS4ERR_DEADSESSION:
1319                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1320                                 goto out;
1321                         case -NFS4ERR_STALE_CLIENTID:
1322                         case -NFS4ERR_STALE_STATEID:
1323                         case -NFS4ERR_EXPIRED:
1324                                 /* Don't recall a delegation if it was lost */
1325                                 nfs4_schedule_lease_recovery(server->nfs_client);
1326                                 goto out;
1327                         case -ERESTARTSYS:
1328                                 /*
1329                                  * The show must go on: exit, but mark the
1330                                  * stateid as needing recovery.
1331                                  */
1332                         case -NFS4ERR_DELEG_REVOKED:
1333                         case -NFS4ERR_ADMIN_REVOKED:
1334                         case -NFS4ERR_BAD_STATEID:
1335                                 nfs_inode_find_state_and_recover(state->inode,
1336                                                 stateid);
1337                                 nfs4_schedule_stateid_recovery(server, state);
1338                         case -EKEYEXPIRED:
1339                                 /*
1340                                  * User RPCSEC_GSS context has expired.
1341                                  * We cannot recover this stateid now, so
1342                                  * skip it and allow recovery thread to
1343                                  * proceed.
1344                                  */
1345                         case -ENOMEM:
1346                                 err = 0;
1347                                 goto out;
1348                 }
1349                 err = nfs4_handle_exception(server, err, &exception);
1350         } while (exception.retry);
1351 out:
1352         return err;
1353 }
1354
1355 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1356 {
1357         struct nfs4_opendata *data = calldata;
1358
1359         data->rpc_status = task->tk_status;
1360         if (data->rpc_status == 0) {
1361                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1362                                 sizeof(data->o_res.stateid.data));
1363                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1364                 renew_lease(data->o_res.server, data->timestamp);
1365                 data->rpc_done = 1;
1366         }
1367 }
1368
1369 static void nfs4_open_confirm_release(void *calldata)
1370 {
1371         struct nfs4_opendata *data = calldata;
1372         struct nfs4_state *state = NULL;
1373
1374         /* If this request hasn't been cancelled, do nothing */
1375         if (data->cancelled == 0)
1376                 goto out_free;
1377         /* In case of error, no cleanup! */
1378         if (!data->rpc_done)
1379                 goto out_free;
1380         state = nfs4_opendata_to_nfs4_state(data);
1381         if (!IS_ERR(state))
1382                 nfs4_close_state(state, data->o_arg.fmode);
1383 out_free:
1384         nfs4_opendata_put(data);
1385 }
1386
1387 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1388         .rpc_call_done = nfs4_open_confirm_done,
1389         .rpc_release = nfs4_open_confirm_release,
1390 };
1391
1392 /*
1393  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1394  */
1395 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1396 {
1397         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1398         struct rpc_task *task;
1399         struct  rpc_message msg = {
1400                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1401                 .rpc_argp = &data->c_arg,
1402                 .rpc_resp = &data->c_res,
1403                 .rpc_cred = data->owner->so_cred,
1404         };
1405         struct rpc_task_setup task_setup_data = {
1406                 .rpc_client = server->client,
1407                 .rpc_message = &msg,
1408                 .callback_ops = &nfs4_open_confirm_ops,
1409                 .callback_data = data,
1410                 .workqueue = nfsiod_workqueue,
1411                 .flags = RPC_TASK_ASYNC,
1412         };
1413         int status;
1414
1415         kref_get(&data->kref);
1416         data->rpc_done = 0;
1417         data->rpc_status = 0;
1418         data->timestamp = jiffies;
1419         task = rpc_run_task(&task_setup_data);
1420         if (IS_ERR(task))
1421                 return PTR_ERR(task);
1422         status = nfs4_wait_for_completion_rpc_task(task);
1423         if (status != 0) {
1424                 data->cancelled = 1;
1425                 smp_wmb();
1426         } else
1427                 status = data->rpc_status;
1428         rpc_put_task(task);
1429         return status;
1430 }
1431
1432 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1433 {
1434         struct nfs4_opendata *data = calldata;
1435         struct nfs4_state_owner *sp = data->owner;
1436
1437         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1438                 return;
1439         /*
1440          * Check if we still need to send an OPEN call, or if we can use
1441          * a delegation instead.
1442          */
1443         if (data->state != NULL) {
1444                 struct nfs_delegation *delegation;
1445
1446                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1447                         goto out_no_action;
1448                 rcu_read_lock();
1449                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1450                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1451                     can_open_delegated(delegation, data->o_arg.fmode))
1452                         goto unlock_no_action;
1453                 rcu_read_unlock();
1454         }
1455         /* Update sequence id. */
1456         data->o_arg.id = sp->so_owner_id.id;
1457         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1458         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1459                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1460                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1461         }
1462         data->timestamp = jiffies;
1463         if (nfs4_setup_sequence(data->o_arg.server,
1464                                 &data->o_arg.seq_args,
1465                                 &data->o_res.seq_res, 1, task))
1466                 return;
1467         rpc_call_start(task);
1468         return;
1469 unlock_no_action:
1470         rcu_read_unlock();
1471 out_no_action:
1472         task->tk_action = NULL;
1473
1474 }
1475
1476 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1477 {
1478         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1479         nfs4_open_prepare(task, calldata);
1480 }
1481
1482 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1483 {
1484         struct nfs4_opendata *data = calldata;
1485
1486         data->rpc_status = task->tk_status;
1487
1488         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1489                 return;
1490
1491         if (task->tk_status == 0) {
1492                 switch (data->o_res.f_attr->mode & S_IFMT) {
1493                         case S_IFREG:
1494                                 break;
1495                         case S_IFLNK:
1496                                 data->rpc_status = -ELOOP;
1497                                 break;
1498                         case S_IFDIR:
1499                                 data->rpc_status = -EISDIR;
1500                                 break;
1501                         default:
1502                                 data->rpc_status = -ENOTDIR;
1503                 }
1504                 renew_lease(data->o_res.server, data->timestamp);
1505                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1506                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1507         }
1508         data->rpc_done = 1;
1509 }
1510
1511 static void nfs4_open_release(void *calldata)
1512 {
1513         struct nfs4_opendata *data = calldata;
1514         struct nfs4_state *state = NULL;
1515
1516         /* If this request hasn't been cancelled, do nothing */
1517         if (data->cancelled == 0)
1518                 goto out_free;
1519         /* In case of error, no cleanup! */
1520         if (data->rpc_status != 0 || !data->rpc_done)
1521                 goto out_free;
1522         /* In case we need an open_confirm, no cleanup! */
1523         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1524                 goto out_free;
1525         state = nfs4_opendata_to_nfs4_state(data);
1526         if (!IS_ERR(state))
1527                 nfs4_close_state(state, data->o_arg.fmode);
1528 out_free:
1529         nfs4_opendata_put(data);
1530 }
1531
1532 static const struct rpc_call_ops nfs4_open_ops = {
1533         .rpc_call_prepare = nfs4_open_prepare,
1534         .rpc_call_done = nfs4_open_done,
1535         .rpc_release = nfs4_open_release,
1536 };
1537
1538 static const struct rpc_call_ops nfs4_recover_open_ops = {
1539         .rpc_call_prepare = nfs4_recover_open_prepare,
1540         .rpc_call_done = nfs4_open_done,
1541         .rpc_release = nfs4_open_release,
1542 };
1543
1544 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1545 {
1546         struct inode *dir = data->dir->d_inode;
1547         struct nfs_server *server = NFS_SERVER(dir);
1548         struct nfs_openargs *o_arg = &data->o_arg;
1549         struct nfs_openres *o_res = &data->o_res;
1550         struct rpc_task *task;
1551         struct rpc_message msg = {
1552                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1553                 .rpc_argp = o_arg,
1554                 .rpc_resp = o_res,
1555                 .rpc_cred = data->owner->so_cred,
1556         };
1557         struct rpc_task_setup task_setup_data = {
1558                 .rpc_client = server->client,
1559                 .rpc_message = &msg,
1560                 .callback_ops = &nfs4_open_ops,
1561                 .callback_data = data,
1562                 .workqueue = nfsiod_workqueue,
1563                 .flags = RPC_TASK_ASYNC,
1564         };
1565         int status;
1566
1567         kref_get(&data->kref);
1568         data->rpc_done = 0;
1569         data->rpc_status = 0;
1570         data->cancelled = 0;
1571         if (isrecover)
1572                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1573         task = rpc_run_task(&task_setup_data);
1574         if (IS_ERR(task))
1575                 return PTR_ERR(task);
1576         status = nfs4_wait_for_completion_rpc_task(task);
1577         if (status != 0) {
1578                 data->cancelled = 1;
1579                 smp_wmb();
1580         } else
1581                 status = data->rpc_status;
1582         rpc_put_task(task);
1583
1584         return status;
1585 }
1586
1587 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1588 {
1589         struct inode *dir = data->dir->d_inode;
1590         struct nfs_openres *o_res = &data->o_res;
1591         int status;
1592
1593         status = nfs4_run_open_task(data, 1);
1594         if (status != 0 || !data->rpc_done)
1595                 return status;
1596
1597         nfs_refresh_inode(dir, o_res->dir_attr);
1598
1599         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1600                 status = _nfs4_proc_open_confirm(data);
1601                 if (status != 0)
1602                         return status;
1603         }
1604
1605         return status;
1606 }
1607
1608 /*
1609  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1610  */
1611 static int _nfs4_proc_open(struct nfs4_opendata *data)
1612 {
1613         struct inode *dir = data->dir->d_inode;
1614         struct nfs_server *server = NFS_SERVER(dir);
1615         struct nfs_openargs *o_arg = &data->o_arg;
1616         struct nfs_openres *o_res = &data->o_res;
1617         int status;
1618
1619         status = nfs4_run_open_task(data, 0);
1620         if (!data->rpc_done)
1621                 return status;
1622         if (status != 0) {
1623                 if (status == -NFS4ERR_BADNAME &&
1624                                 !(o_arg->open_flags & O_CREAT))
1625                         return -ENOENT;
1626                 return status;
1627         }
1628
1629         if (o_arg->open_flags & O_CREAT) {
1630                 update_changeattr(dir, &o_res->cinfo);
1631                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1632         } else
1633                 nfs_refresh_inode(dir, o_res->dir_attr);
1634         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1635                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1636         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1637                 status = _nfs4_proc_open_confirm(data);
1638                 if (status != 0)
1639                         return status;
1640         }
1641         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1642                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1643         return 0;
1644 }
1645
1646 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1647 {
1648         unsigned int loop;
1649         int ret;
1650
1651         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1652                 ret = nfs4_wait_clnt_recover(clp);
1653                 if (ret != 0)
1654                         break;
1655                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1656                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1657                         break;
1658                 nfs4_schedule_state_manager(clp);
1659                 ret = -EIO;
1660         }
1661         return ret;
1662 }
1663
1664 static int nfs4_recover_expired_lease(struct nfs_server *server)
1665 {
1666         return nfs4_client_recover_expired_lease(server->nfs_client);
1667 }
1668
1669 /*
1670  * OPEN_EXPIRED:
1671  *      reclaim state on the server after a network partition.
1672  *      Assumes caller holds the appropriate lock
1673  */
1674 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1675 {
1676         struct nfs4_opendata *opendata;
1677         int ret;
1678
1679         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1680         if (IS_ERR(opendata))
1681                 return PTR_ERR(opendata);
1682         ret = nfs4_open_recover(opendata, state);
1683         if (ret == -ESTALE)
1684                 d_drop(ctx->dentry);
1685         nfs4_opendata_put(opendata);
1686         return ret;
1687 }
1688
1689 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1690 {
1691         struct nfs_server *server = NFS_SERVER(state->inode);
1692         struct nfs4_exception exception = { };
1693         int err;
1694
1695         do {
1696                 err = _nfs4_open_expired(ctx, state);
1697                 switch (err) {
1698                 default:
1699                         goto out;
1700                 case -NFS4ERR_GRACE:
1701                 case -NFS4ERR_DELAY:
1702                         nfs4_handle_exception(server, err, &exception);
1703                         err = 0;
1704                 }
1705         } while (exception.retry);
1706 out:
1707         return err;
1708 }
1709
1710 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1711 {
1712         struct nfs_open_context *ctx;
1713         int ret;
1714
1715         ctx = nfs4_state_find_open_context(state);
1716         if (IS_ERR(ctx))
1717                 return PTR_ERR(ctx);
1718         ret = nfs4_do_open_expired(ctx, state);
1719         put_nfs_open_context(ctx);
1720         return ret;
1721 }
1722
1723 #if defined(CONFIG_NFS_V4_1)
1724 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1725 {
1726         int status;
1727         struct nfs_server *server = NFS_SERVER(state->inode);
1728
1729         status = nfs41_test_stateid(server, state);
1730         if (status == NFS_OK)
1731                 return 0;
1732         nfs41_free_stateid(server, state);
1733         return nfs4_open_expired(sp, state);
1734 }
1735 #endif
1736
1737 /*
1738  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1739  * fields corresponding to attributes that were used to store the verifier.
1740  * Make sure we clobber those fields in the later setattr call
1741  */
1742 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1743 {
1744         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1745             !(sattr->ia_valid & ATTR_ATIME_SET))
1746                 sattr->ia_valid |= ATTR_ATIME;
1747
1748         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1749             !(sattr->ia_valid & ATTR_MTIME_SET))
1750                 sattr->ia_valid |= ATTR_MTIME;
1751 }
1752
1753 /*
1754  * Returns a referenced nfs4_state
1755  */
1756 static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1757 {
1758         struct nfs4_state_owner  *sp;
1759         struct nfs4_state     *state = NULL;
1760         struct nfs_server       *server = NFS_SERVER(dir);
1761         struct nfs4_opendata *opendata;
1762         int status;
1763
1764         /* Protect against reboot recovery conflicts */
1765         status = -ENOMEM;
1766         if (!(sp = nfs4_get_state_owner(server, cred))) {
1767                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1768                 goto out_err;
1769         }
1770         status = nfs4_recover_expired_lease(server);
1771         if (status != 0)
1772                 goto err_put_state_owner;
1773         if (dentry->d_inode != NULL)
1774                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1775         status = -ENOMEM;
1776         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1777         if (opendata == NULL)
1778                 goto err_put_state_owner;
1779
1780         if (dentry->d_inode != NULL)
1781                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1782
1783         status = _nfs4_proc_open(opendata);
1784         if (status != 0)
1785                 goto err_opendata_put;
1786
1787         state = nfs4_opendata_to_nfs4_state(opendata);
1788         status = PTR_ERR(state);
1789         if (IS_ERR(state))
1790                 goto err_opendata_put;
1791         if (server->caps & NFS_CAP_POSIX_LOCK)
1792                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1793
1794         if (opendata->o_arg.open_flags & O_EXCL) {
1795                 nfs4_exclusive_attrset(opendata, sattr);
1796
1797                 nfs_fattr_init(opendata->o_res.f_attr);
1798                 status = nfs4_do_setattr(state->inode, cred,
1799                                 opendata->o_res.f_attr, sattr,
1800                                 state);
1801                 if (status == 0)
1802                         nfs_setattr_update_inode(state->inode, sattr);
1803                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1804         }
1805         nfs_revalidate_inode(server, state->inode);
1806         nfs4_opendata_put(opendata);
1807         nfs4_put_state_owner(sp);
1808         *res = state;
1809         return 0;
1810 err_opendata_put:
1811         nfs4_opendata_put(opendata);
1812 err_put_state_owner:
1813         nfs4_put_state_owner(sp);
1814 out_err:
1815         *res = NULL;
1816         return status;
1817 }
1818
1819
1820 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1821 {
1822         struct nfs4_exception exception = { };
1823         struct nfs4_state *res;
1824         int status;
1825
1826         do {
1827                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred, &res);
1828                 if (status == 0)
1829                         break;
1830                 /* NOTE: BAD_SEQID means the server and client disagree about the
1831                  * book-keeping w.r.t. state-changing operations
1832                  * (OPEN/CLOSE/LOCK/LOCKU...)
1833                  * It is actually a sign of a bug on the client or on the server.
1834                  *
1835                  * If we receive a BAD_SEQID error in the particular case of
1836                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1837                  * have unhashed the old state_owner for us, and that we can
1838                  * therefore safely retry using a new one. We should still warn
1839                  * the user though...
1840                  */
1841                 if (status == -NFS4ERR_BAD_SEQID) {
1842                         pr_warn_ratelimited("NFS: v4 server %s "
1843                                         " returned a bad sequence-id error!\n",
1844                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1845                         exception.retry = 1;
1846                         continue;
1847                 }
1848                 /*
1849                  * BAD_STATEID on OPEN means that the server cancelled our
1850                  * state before it received the OPEN_CONFIRM.
1851                  * Recover by retrying the request as per the discussion
1852                  * on Page 181 of RFC3530.
1853                  */
1854                 if (status == -NFS4ERR_BAD_STATEID) {
1855                         exception.retry = 1;
1856                         continue;
1857                 }
1858                 if (status == -EAGAIN) {
1859                         /* We must have found a delegation */
1860                         exception.retry = 1;
1861                         continue;
1862                 }
1863                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1864                                         status, &exception));
1865         } while (exception.retry);
1866         return res;
1867 }
1868
1869 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1870                             struct nfs_fattr *fattr, struct iattr *sattr,
1871                             struct nfs4_state *state)
1872 {
1873         struct nfs_server *server = NFS_SERVER(inode);
1874         struct nfs_setattrargs  arg = {
1875                 .fh             = NFS_FH(inode),
1876                 .iap            = sattr,
1877                 .server         = server,
1878                 .bitmask = server->attr_bitmask,
1879         };
1880         struct nfs_setattrres  res = {
1881                 .fattr          = fattr,
1882                 .server         = server,
1883         };
1884         struct rpc_message msg = {
1885                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1886                 .rpc_argp       = &arg,
1887                 .rpc_resp       = &res,
1888                 .rpc_cred       = cred,
1889         };
1890         unsigned long timestamp = jiffies;
1891         int status;
1892
1893         nfs_fattr_init(fattr);
1894
1895         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1896                 /* Use that stateid */
1897         } else if (state != NULL) {
1898                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1899         } else
1900                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1901
1902         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1903         if (status == 0 && state != NULL)
1904                 renew_lease(server, timestamp);
1905         return status;
1906 }
1907
1908 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1909                            struct nfs_fattr *fattr, struct iattr *sattr,
1910                            struct nfs4_state *state)
1911 {
1912         struct nfs_server *server = NFS_SERVER(inode);
1913         struct nfs4_exception exception = {
1914                 .state = state,
1915                 .inode = inode,
1916         };
1917         int err;
1918         do {
1919                 err = nfs4_handle_exception(server,
1920                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1921                                 &exception);
1922         } while (exception.retry);
1923         return err;
1924 }
1925
1926 struct nfs4_closedata {
1927         struct inode *inode;
1928         struct nfs4_state *state;
1929         struct nfs_closeargs arg;
1930         struct nfs_closeres res;
1931         struct nfs_fattr fattr;
1932         unsigned long timestamp;
1933         bool roc;
1934         u32 roc_barrier;
1935 };
1936
1937 static void nfs4_free_closedata(void *data)
1938 {
1939         struct nfs4_closedata *calldata = data;
1940         struct nfs4_state_owner *sp = calldata->state->owner;
1941         struct super_block *sb = calldata->state->inode->i_sb;
1942
1943         if (calldata->roc)
1944                 pnfs_roc_release(calldata->state->inode);
1945         nfs4_put_open_state(calldata->state);
1946         nfs_free_seqid(calldata->arg.seqid);
1947         nfs4_put_state_owner(sp);
1948         nfs_sb_deactive(sb);
1949         kfree(calldata);
1950 }
1951
1952 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1953                 fmode_t fmode)
1954 {
1955         spin_lock(&state->owner->so_lock);
1956         if (!(fmode & FMODE_READ))
1957                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1958         if (!(fmode & FMODE_WRITE))
1959                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1960         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1961         spin_unlock(&state->owner->so_lock);
1962 }
1963
1964 static void nfs4_close_done(struct rpc_task *task, void *data)
1965 {
1966         struct nfs4_closedata *calldata = data;
1967         struct nfs4_state *state = calldata->state;
1968         struct nfs_server *server = NFS_SERVER(calldata->inode);
1969
1970         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1971                 return;
1972         /* hmm. we are done with the inode, and in the process of freeing
1973          * the state_owner. we keep this around to process errors
1974          */
1975         switch (task->tk_status) {
1976                 case 0:
1977                         if (calldata->roc)
1978                                 pnfs_roc_set_barrier(state->inode,
1979                                                      calldata->roc_barrier);
1980                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1981                         renew_lease(server, calldata->timestamp);
1982                         nfs4_close_clear_stateid_flags(state,
1983                                         calldata->arg.fmode);
1984                         break;
1985                 case -NFS4ERR_STALE_STATEID:
1986                 case -NFS4ERR_OLD_STATEID:
1987                 case -NFS4ERR_BAD_STATEID:
1988                 case -NFS4ERR_EXPIRED:
1989                         if (calldata->arg.fmode == 0)
1990                                 break;
1991                 default:
1992                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1993                                 rpc_restart_call_prepare(task);
1994         }
1995         nfs_release_seqid(calldata->arg.seqid);
1996         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1997 }
1998
1999 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2000 {
2001         struct nfs4_closedata *calldata = data;
2002         struct nfs4_state *state = calldata->state;
2003         int call_close = 0;
2004
2005         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2006                 return;
2007
2008         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2009         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2010         spin_lock(&state->owner->so_lock);
2011         /* Calculate the change in open mode */
2012         if (state->n_rdwr == 0) {
2013                 if (state->n_rdonly == 0) {
2014                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2015                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2016                         calldata->arg.fmode &= ~FMODE_READ;
2017                 }
2018                 if (state->n_wronly == 0) {
2019                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2020                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2021                         calldata->arg.fmode &= ~FMODE_WRITE;
2022                 }
2023         }
2024         spin_unlock(&state->owner->so_lock);
2025
2026         if (!call_close) {
2027                 /* Note: exit _without_ calling nfs4_close_done */
2028                 task->tk_action = NULL;
2029                 return;
2030         }
2031
2032         if (calldata->arg.fmode == 0) {
2033                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2034                 if (calldata->roc &&
2035                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2036                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2037                                      task, NULL);
2038                         return;
2039                 }
2040         }
2041
2042         nfs_fattr_init(calldata->res.fattr);
2043         calldata->timestamp = jiffies;
2044         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2045                                 &calldata->arg.seq_args, &calldata->res.seq_res,
2046                                 1, task))
2047                 return;
2048         rpc_call_start(task);
2049 }
2050
2051 static const struct rpc_call_ops nfs4_close_ops = {
2052         .rpc_call_prepare = nfs4_close_prepare,
2053         .rpc_call_done = nfs4_close_done,
2054         .rpc_release = nfs4_free_closedata,
2055 };
2056
2057 /*
2058  * It is possible for data to be read/written from a mem-mapped file
2059  * after the sys_close call (which hits the vfs layer as a flush).
2060  * This means that we can't safely call nfsv4 close on a file until
2061  * the inode is cleared. This in turn means that we are not good
2062  * NFSv4 citizens - we do not indicate to the server to update the file's
2063  * share state even when we are done with one of the three share
2064  * stateid's in the inode.
2065  *
2066  * NOTE: Caller must be holding the sp->so_owner semaphore!
2067  */
2068 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2069 {
2070         struct nfs_server *server = NFS_SERVER(state->inode);
2071         struct nfs4_closedata *calldata;
2072         struct nfs4_state_owner *sp = state->owner;
2073         struct rpc_task *task;
2074         struct rpc_message msg = {
2075                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2076                 .rpc_cred = state->owner->so_cred,
2077         };
2078         struct rpc_task_setup task_setup_data = {
2079                 .rpc_client = server->client,
2080                 .rpc_message = &msg,
2081                 .callback_ops = &nfs4_close_ops,
2082                 .workqueue = nfsiod_workqueue,
2083                 .flags = RPC_TASK_ASYNC,
2084         };
2085         int status = -ENOMEM;
2086
2087         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2088         if (calldata == NULL)
2089                 goto out;
2090         calldata->inode = state->inode;
2091         calldata->state = state;
2092         calldata->arg.fh = NFS_FH(state->inode);
2093         calldata->arg.stateid = &state->open_stateid;
2094         /* Serialization for the sequence id */
2095         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2096         if (calldata->arg.seqid == NULL)
2097                 goto out_free_calldata;
2098         calldata->arg.fmode = 0;
2099         calldata->arg.bitmask = server->cache_consistency_bitmask;
2100         calldata->res.fattr = &calldata->fattr;
2101         calldata->res.seqid = calldata->arg.seqid;
2102         calldata->res.server = server;
2103         calldata->roc = roc;
2104         nfs_sb_active(calldata->inode->i_sb);
2105
2106         msg.rpc_argp = &calldata->arg;
2107         msg.rpc_resp = &calldata->res;
2108         task_setup_data.callback_data = calldata;
2109         task = rpc_run_task(&task_setup_data);
2110         if (IS_ERR(task))
2111                 return PTR_ERR(task);
2112         status = 0;
2113         if (wait)
2114                 status = rpc_wait_for_completion_task(task);
2115         rpc_put_task(task);
2116         return status;
2117 out_free_calldata:
2118         kfree(calldata);
2119 out:
2120         if (roc)
2121                 pnfs_roc_release(state->inode);
2122         nfs4_put_open_state(state);
2123         nfs4_put_state_owner(sp);
2124         return status;
2125 }
2126
2127 static struct inode *
2128 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2129 {
2130         struct nfs4_state *state;
2131
2132         /* Protect against concurrent sillydeletes */
2133         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2134         if (IS_ERR(state))
2135                 return ERR_CAST(state);
2136         ctx->state = state;
2137         return igrab(state->inode);
2138 }
2139
2140 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2141 {
2142         if (ctx->state == NULL)
2143                 return;
2144         if (is_sync)
2145                 nfs4_close_sync(ctx->state, ctx->mode);
2146         else
2147                 nfs4_close_state(ctx->state, ctx->mode);
2148 }
2149
2150 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2151 {
2152         struct nfs4_server_caps_arg args = {
2153                 .fhandle = fhandle,
2154         };
2155         struct nfs4_server_caps_res res = {};
2156         struct rpc_message msg = {
2157                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2158                 .rpc_argp = &args,
2159                 .rpc_resp = &res,
2160         };
2161         int status;
2162
2163         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2164         if (status == 0) {
2165                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2166                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2167                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2168                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2169                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2170                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2171                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2172                         server->caps |= NFS_CAP_ACLS;
2173                 if (res.has_links != 0)
2174                         server->caps |= NFS_CAP_HARDLINKS;
2175                 if (res.has_symlinks != 0)
2176                         server->caps |= NFS_CAP_SYMLINKS;
2177                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2178                         server->caps |= NFS_CAP_FILEID;
2179                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2180                         server->caps |= NFS_CAP_MODE;
2181                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2182                         server->caps |= NFS_CAP_NLINK;
2183                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2184                         server->caps |= NFS_CAP_OWNER;
2185                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2186                         server->caps |= NFS_CAP_OWNER_GROUP;
2187                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2188                         server->caps |= NFS_CAP_ATIME;
2189                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2190                         server->caps |= NFS_CAP_CTIME;
2191                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2192                         server->caps |= NFS_CAP_MTIME;
2193
2194                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2195                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2196                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2197                 server->acl_bitmask = res.acl_bitmask;
2198         }
2199
2200         return status;
2201 }
2202
2203 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2204 {
2205         struct nfs4_exception exception = { };
2206         int err;
2207         do {
2208                 err = nfs4_handle_exception(server,
2209                                 _nfs4_server_capabilities(server, fhandle),
2210                                 &exception);
2211         } while (exception.retry);
2212         return err;
2213 }
2214
2215 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2216                 struct nfs_fsinfo *info)
2217 {
2218         struct nfs4_lookup_root_arg args = {
2219                 .bitmask = nfs4_fattr_bitmap,
2220         };
2221         struct nfs4_lookup_res res = {
2222                 .server = server,
2223                 .fattr = info->fattr,
2224                 .fh = fhandle,
2225         };
2226         struct rpc_message msg = {
2227                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2228                 .rpc_argp = &args,
2229                 .rpc_resp = &res,
2230         };
2231
2232         nfs_fattr_init(info->fattr);
2233         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2234 }
2235
2236 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2237                 struct nfs_fsinfo *info)
2238 {
2239         struct nfs4_exception exception = { };
2240         int err;
2241         do {
2242                 err = _nfs4_lookup_root(server, fhandle, info);
2243                 switch (err) {
2244                 case 0:
2245                 case -NFS4ERR_WRONGSEC:
2246                         goto out;
2247                 default:
2248                         err = nfs4_handle_exception(server, err, &exception);
2249                 }
2250         } while (exception.retry);
2251 out:
2252         return err;
2253 }
2254
2255 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2256                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2257 {
2258         struct rpc_auth *auth;
2259         int ret;
2260
2261         auth = rpcauth_create(flavor, server->client);
2262         if (!auth) {
2263                 ret = -EIO;
2264                 goto out;
2265         }
2266         ret = nfs4_lookup_root(server, fhandle, info);
2267 out:
2268         return ret;
2269 }
2270
2271 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2272                               struct nfs_fsinfo *info)
2273 {
2274         int i, len, status = 0;
2275         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2276
2277         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2278         flav_array[len] = RPC_AUTH_NULL;
2279         len += 1;
2280
2281         for (i = 0; i < len; i++) {
2282                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2283                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2284                         continue;
2285                 break;
2286         }
2287         /*
2288          * -EACCESS could mean that the user doesn't have correct permissions
2289          * to access the mount.  It could also mean that we tried to mount
2290          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2291          * existing mount programs don't handle -EACCES very well so it should
2292          * be mapped to -EPERM instead.
2293          */
2294         if (status == -EACCES)
2295                 status = -EPERM;
2296         return status;
2297 }
2298
2299 /*
2300  * get the file handle for the "/" directory on the server
2301  */
2302 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2303                               struct nfs_fsinfo *info)
2304 {
2305         int minor_version = server->nfs_client->cl_minorversion;
2306         int status = nfs4_lookup_root(server, fhandle, info);
2307         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2308                 /*
2309                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2310                  * by nfs4_map_errors() as this function exits.
2311                  */
2312                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2313         if (status == 0)
2314                 status = nfs4_server_capabilities(server, fhandle);
2315         if (status == 0)
2316                 status = nfs4_do_fsinfo(server, fhandle, info);
2317         return nfs4_map_errors(status);
2318 }
2319
2320 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
2321 /*
2322  * Get locations and (maybe) other attributes of a referral.
2323  * Note that we'll actually follow the referral later when
2324  * we detect fsid mismatch in inode revalidation
2325  */
2326 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2327                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2328 {
2329         int status = -ENOMEM;
2330         struct page *page = NULL;
2331         struct nfs4_fs_locations *locations = NULL;
2332
2333         page = alloc_page(GFP_KERNEL);
2334         if (page == NULL)
2335                 goto out;
2336         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2337         if (locations == NULL)
2338                 goto out;
2339
2340         status = nfs4_proc_fs_locations(dir, name, locations, page);
2341         if (status != 0)
2342                 goto out;
2343         /* Make sure server returned a different fsid for the referral */
2344         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2345                 dprintk("%s: server did not return a different fsid for"
2346                         " a referral at %s\n", __func__, name->name);
2347                 status = -EIO;
2348                 goto out;
2349         }
2350         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2351         nfs_fixup_referral_attributes(&locations->fattr);
2352
2353         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2354         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2355         memset(fhandle, 0, sizeof(struct nfs_fh));
2356 out:
2357         if (page)
2358                 __free_page(page);
2359         kfree(locations);
2360         return status;
2361 }
2362
2363 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2364 {
2365         struct nfs4_getattr_arg args = {
2366                 .fh = fhandle,
2367                 .bitmask = server->attr_bitmask,
2368         };
2369         struct nfs4_getattr_res res = {
2370                 .fattr = fattr,
2371                 .server = server,
2372         };
2373         struct rpc_message msg = {
2374                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2375                 .rpc_argp = &args,
2376                 .rpc_resp = &res,
2377         };
2378
2379         nfs_fattr_init(fattr);
2380         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2381 }
2382
2383 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2384 {
2385         struct nfs4_exception exception = { };
2386         int err;
2387         do {
2388                 err = nfs4_handle_exception(server,
2389                                 _nfs4_proc_getattr(server, fhandle, fattr),
2390                                 &exception);
2391         } while (exception.retry);
2392         return err;
2393 }
2394
2395 /*
2396  * The file is not closed if it is opened due to the a request to change
2397  * the size of the file. The open call will not be needed once the
2398  * VFS layer lookup-intents are implemented.
2399  *
2400  * Close is called when the inode is destroyed.
2401  * If we haven't opened the file for O_WRONLY, we
2402  * need to in the size_change case to obtain a stateid.
2403  *
2404  * Got race?
2405  * Because OPEN is always done by name in nfsv4, it is
2406  * possible that we opened a different file by the same
2407  * name.  We can recognize this race condition, but we
2408  * can't do anything about it besides returning an error.
2409  *
2410  * This will be fixed with VFS changes (lookup-intent).
2411  */
2412 static int
2413 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2414                   struct iattr *sattr)
2415 {
2416         struct inode *inode = dentry->d_inode;
2417         struct rpc_cred *cred = NULL;
2418         struct nfs4_state *state = NULL;
2419         int status;
2420
2421         if (pnfs_ld_layoutret_on_setattr(inode))
2422                 pnfs_return_layout(inode);
2423
2424         nfs_fattr_init(fattr);
2425
2426         /* Search for an existing open(O_WRITE) file */
2427         if (sattr->ia_valid & ATTR_FILE) {
2428                 struct nfs_open_context *ctx;
2429
2430                 ctx = nfs_file_open_context(sattr->ia_file);
2431                 if (ctx) {
2432                         cred = ctx->cred;
2433                         state = ctx->state;
2434                 }
2435         }
2436
2437         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2438         if (status == 0)
2439                 nfs_setattr_update_inode(inode, sattr);
2440         return status;
2441 }
2442
2443 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2444                 const struct qstr *name, struct nfs_fh *fhandle,
2445                 struct nfs_fattr *fattr)
2446 {
2447         struct nfs_server *server = NFS_SERVER(dir);
2448         int                    status;
2449         struct nfs4_lookup_arg args = {
2450                 .bitmask = server->attr_bitmask,
2451                 .dir_fh = NFS_FH(dir),
2452                 .name = name,
2453         };
2454         struct nfs4_lookup_res res = {
2455                 .server = server,
2456                 .fattr = fattr,
2457                 .fh = fhandle,
2458         };
2459         struct rpc_message msg = {
2460                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2461                 .rpc_argp = &args,
2462                 .rpc_resp = &res,
2463         };
2464
2465         nfs_fattr_init(fattr);
2466
2467         dprintk("NFS call  lookup %s\n", name->name);
2468         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2469         dprintk("NFS reply lookup: %d\n", status);
2470         return status;
2471 }
2472
2473 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2474 {
2475         memset(fh, 0, sizeof(struct nfs_fh));
2476         fattr->fsid.major = 1;
2477         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2478                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2479         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2480         fattr->nlink = 2;
2481 }
2482
2483 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2484                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2485 {
2486         struct nfs4_exception exception = { };
2487         int err;
2488         do {
2489                 int status;
2490
2491                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2492                 switch (status) {
2493                 case -NFS4ERR_BADNAME:
2494                         return -ENOENT;
2495                 case -NFS4ERR_MOVED:
2496                         return nfs4_get_referral(dir, name, fattr, fhandle);
2497                 case -NFS4ERR_WRONGSEC:
2498                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2499                 }
2500                 err = nfs4_handle_exception(NFS_SERVER(dir),
2501                                 status, &exception);
2502         } while (exception.retry);
2503         return err;
2504 }
2505
2506 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2507 {
2508         struct nfs_server *server = NFS_SERVER(inode);
2509         struct nfs4_accessargs args = {
2510                 .fh = NFS_FH(inode),
2511                 .bitmask = server->attr_bitmask,
2512         };
2513         struct nfs4_accessres res = {
2514                 .server = server,
2515         };
2516         struct rpc_message msg = {
2517                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2518                 .rpc_argp = &args,
2519                 .rpc_resp = &res,
2520                 .rpc_cred = entry->cred,
2521         };
2522         int mode = entry->mask;
2523         int status;
2524
2525         /*
2526          * Determine which access bits we want to ask for...
2527          */
2528         if (mode & MAY_READ)
2529                 args.access |= NFS4_ACCESS_READ;
2530         if (S_ISDIR(inode->i_mode)) {
2531                 if (mode & MAY_WRITE)
2532                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2533                 if (mode & MAY_EXEC)
2534                         args.access |= NFS4_ACCESS_LOOKUP;
2535         } else {
2536                 if (mode & MAY_WRITE)
2537                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2538                 if (mode & MAY_EXEC)
2539                         args.access |= NFS4_ACCESS_EXECUTE;
2540         }
2541
2542         res.fattr = nfs_alloc_fattr();
2543         if (res.fattr == NULL)
2544                 return -ENOMEM;
2545
2546         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2547         if (!status) {
2548                 entry->mask = 0;
2549                 if (res.access & NFS4_ACCESS_READ)
2550                         entry->mask |= MAY_READ;
2551                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2552                         entry->mask |= MAY_WRITE;
2553                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2554                         entry->mask |= MAY_EXEC;
2555                 nfs_refresh_inode(inode, res.fattr);
2556         }
2557         nfs_free_fattr(res.fattr);
2558         return status;
2559 }
2560
2561 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2562 {
2563         struct nfs4_exception exception = { };
2564         int err;
2565         do {
2566                 err = nfs4_handle_exception(NFS_SERVER(inode),
2567                                 _nfs4_proc_access(inode, entry),
2568                                 &exception);
2569         } while (exception.retry);
2570         return err;
2571 }
2572
2573 /*
2574  * TODO: For the time being, we don't try to get any attributes
2575  * along with any of the zero-copy operations READ, READDIR,
2576  * READLINK, WRITE.
2577  *
2578  * In the case of the first three, we want to put the GETATTR
2579  * after the read-type operation -- this is because it is hard
2580  * to predict the length of a GETATTR response in v4, and thus
2581  * align the READ data correctly.  This means that the GETATTR
2582  * may end up partially falling into the page cache, and we should
2583  * shift it into the 'tail' of the xdr_buf before processing.
2584  * To do this efficiently, we need to know the total length
2585  * of data received, which doesn't seem to be available outside
2586  * of the RPC layer.
2587  *
2588  * In the case of WRITE, we also want to put the GETATTR after
2589  * the operation -- in this case because we want to make sure
2590  * we get the post-operation mtime and size.  This means that
2591  * we can't use xdr_encode_pages() as written: we need a variant
2592  * of it which would leave room in the 'tail' iovec.
2593  *
2594  * Both of these changes to the XDR layer would in fact be quite
2595  * minor, but I decided to leave them for a subsequent patch.
2596  */
2597 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2598                 unsigned int pgbase, unsigned int pglen)
2599 {
2600         struct nfs4_readlink args = {
2601                 .fh       = NFS_FH(inode),
2602                 .pgbase   = pgbase,
2603                 .pglen    = pglen,
2604                 .pages    = &page,
2605         };
2606         struct nfs4_readlink_res res;
2607         struct rpc_message msg = {
2608                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2609                 .rpc_argp = &args,
2610                 .rpc_resp = &res,
2611         };
2612
2613         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2614 }
2615
2616 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2617                 unsigned int pgbase, unsigned int pglen)
2618 {
2619         struct nfs4_exception exception = { };
2620         int err;
2621         do {
2622                 err = nfs4_handle_exception(NFS_SERVER(inode),
2623                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2624                                 &exception);
2625         } while (exception.retry);
2626         return err;
2627 }
2628
2629 /*
2630  * Got race?
2631  * We will need to arrange for the VFS layer to provide an atomic open.
2632  * Until then, this create/open method is prone to inefficiency and race
2633  * conditions due to the lookup, create, and open VFS calls from sys_open()
2634  * placed on the wire.
2635  *
2636  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2637  * The file will be opened again in the subsequent VFS open call
2638  * (nfs4_proc_file_open).
2639  *
2640  * The open for read will just hang around to be used by any process that
2641  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2642  */
2643
2644 static int
2645 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2646                  int flags, struct nfs_open_context *ctx)
2647 {
2648         struct dentry *de = dentry;
2649         struct nfs4_state *state;
2650         struct rpc_cred *cred = NULL;
2651         fmode_t fmode = 0;
2652         int status = 0;
2653
2654         if (ctx != NULL) {
2655                 cred = ctx->cred;
2656                 de = ctx->dentry;
2657                 fmode = ctx->mode;
2658         }
2659         sattr->ia_mode &= ~current_umask();
2660         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2661         d_drop(dentry);
2662         if (IS_ERR(state)) {
2663                 status = PTR_ERR(state);
2664                 goto out;
2665         }
2666         d_add(dentry, igrab(state->inode));
2667         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2668         if (ctx != NULL)
2669                 ctx->state = state;
2670         else
2671                 nfs4_close_sync(state, fmode);
2672 out:
2673         return status;
2674 }
2675
2676 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2677 {
2678         struct nfs_server *server = NFS_SERVER(dir);
2679         struct nfs_removeargs args = {
2680                 .fh = NFS_FH(dir),
2681                 .name.len = name->len,
2682                 .name.name = name->name,
2683                 .bitmask = server->attr_bitmask,
2684         };
2685         struct nfs_removeres res = {
2686                 .server = server,
2687         };
2688         struct rpc_message msg = {
2689                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2690                 .rpc_argp = &args,
2691                 .rpc_resp = &res,
2692         };
2693         int status = -ENOMEM;
2694
2695         res.dir_attr = nfs_alloc_fattr();
2696         if (res.dir_attr == NULL)
2697                 goto out;
2698
2699         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2700         if (status == 0) {
2701                 update_changeattr(dir, &res.cinfo);
2702                 nfs_post_op_update_inode(dir, res.dir_attr);
2703         }
2704         nfs_free_fattr(res.dir_attr);
2705 out:
2706         return status;
2707 }
2708
2709 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2710 {
2711         struct nfs4_exception exception = { };
2712         int err;
2713         do {
2714                 err = nfs4_handle_exception(NFS_SERVER(dir),
2715                                 _nfs4_proc_remove(dir, name),
2716                                 &exception);
2717         } while (exception.retry);
2718         return err;
2719 }
2720
2721 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2722 {
2723         struct nfs_server *server = NFS_SERVER(dir);
2724         struct nfs_removeargs *args = msg->rpc_argp;
2725         struct nfs_removeres *res = msg->rpc_resp;
2726
2727         args->bitmask = server->cache_consistency_bitmask;
2728         res->server = server;
2729         res->seq_res.sr_slot = NULL;
2730         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2731 }
2732
2733 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2734 {
2735         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2736
2737         if (!nfs4_sequence_done(task, &res->seq_res))
2738                 return 0;
2739         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2740                 return 0;
2741         update_changeattr(dir, &res->cinfo);
2742         nfs_post_op_update_inode(dir, res->dir_attr);
2743         return 1;
2744 }
2745
2746 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2747 {
2748         struct nfs_server *server = NFS_SERVER(dir);
2749         struct nfs_renameargs *arg = msg->rpc_argp;
2750         struct nfs_renameres *res = msg->rpc_resp;
2751
2752         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2753         arg->bitmask = server->attr_bitmask;
2754         res->server = server;
2755 }
2756
2757 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2758                                  struct inode *new_dir)
2759 {
2760         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2761
2762         if (!nfs4_sequence_done(task, &res->seq_res))
2763                 return 0;
2764         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2765                 return 0;
2766
2767         update_changeattr(old_dir, &res->old_cinfo);
2768         nfs_post_op_update_inode(old_dir, res->old_fattr);
2769         update_changeattr(new_dir, &res->new_cinfo);
2770         nfs_post_op_update_inode(new_dir, res->new_fattr);
2771         return 1;
2772 }
2773
2774 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2775                 struct inode *new_dir, struct qstr *new_name)
2776 {
2777         struct nfs_server *server = NFS_SERVER(old_dir);
2778         struct nfs_renameargs arg = {
2779                 .old_dir = NFS_FH(old_dir),
2780                 .new_dir = NFS_FH(new_dir),
2781                 .old_name = old_name,
2782                 .new_name = new_name,
2783                 .bitmask = server->attr_bitmask,
2784         };
2785         struct nfs_renameres res = {
2786                 .server = server,
2787         };
2788         struct rpc_message msg = {
2789                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2790                 .rpc_argp = &arg,
2791                 .rpc_resp = &res,
2792         };
2793         int status = -ENOMEM;
2794
2795         res.old_fattr = nfs_alloc_fattr();
2796         res.new_fattr = nfs_alloc_fattr();
2797         if (res.old_fattr == NULL || res.new_fattr == NULL)
2798                 goto out;
2799
2800         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2801         if (!status) {
2802                 update_changeattr(old_dir, &res.old_cinfo);
2803                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2804                 update_changeattr(new_dir, &res.new_cinfo);
2805                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2806         }
2807 out:
2808         nfs_free_fattr(res.new_fattr);
2809         nfs_free_fattr(res.old_fattr);
2810         return status;
2811 }
2812
2813 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2814                 struct inode *new_dir, struct qstr *new_name)
2815 {
2816         struct nfs4_exception exception = { };
2817         int err;
2818         do {
2819                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2820                                 _nfs4_proc_rename(old_dir, old_name,
2821                                         new_dir, new_name),
2822                                 &exception);
2823         } while (exception.retry);
2824         return err;
2825 }
2826
2827 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2828 {
2829         struct nfs_server *server = NFS_SERVER(inode);
2830         struct nfs4_link_arg arg = {
2831                 .fh     = NFS_FH(inode),
2832                 .dir_fh = NFS_FH(dir),
2833                 .name   = name,
2834                 .bitmask = server->attr_bitmask,
2835         };
2836         struct nfs4_link_res res = {
2837                 .server = server,
2838         };
2839         struct rpc_message msg = {
2840                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2841                 .rpc_argp = &arg,
2842                 .rpc_resp = &res,
2843         };
2844         int status = -ENOMEM;
2845
2846         res.fattr = nfs_alloc_fattr();
2847         res.dir_attr = nfs_alloc_fattr();
2848         if (res.fattr == NULL || res.dir_attr == NULL)
2849                 goto out;
2850
2851         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2852         if (!status) {
2853                 update_changeattr(dir, &res.cinfo);
2854                 nfs_post_op_update_inode(dir, res.dir_attr);
2855                 nfs_post_op_update_inode(inode, res.fattr);
2856         }
2857 out:
2858         nfs_free_fattr(res.dir_attr);
2859         nfs_free_fattr(res.fattr);
2860         return status;
2861 }
2862
2863 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2864 {
2865         struct nfs4_exception exception = { };
2866         int err;
2867         do {
2868                 err = nfs4_handle_exception(NFS_SERVER(inode),
2869                                 _nfs4_proc_link(inode, dir, name),
2870                                 &exception);
2871         } while (exception.retry);
2872         return err;
2873 }
2874
2875 struct nfs4_createdata {
2876         struct rpc_message msg;
2877         struct nfs4_create_arg arg;
2878         struct nfs4_create_res res;
2879         struct nfs_fh fh;
2880         struct nfs_fattr fattr;
2881         struct nfs_fattr dir_fattr;
2882 };
2883
2884 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2885                 struct qstr *name, struct iattr *sattr, u32 ftype)
2886 {
2887         struct nfs4_createdata *data;
2888
2889         data = kzalloc(sizeof(*data), GFP_KERNEL);
2890         if (data != NULL) {
2891                 struct nfs_server *server = NFS_SERVER(dir);
2892
2893                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2894                 data->msg.rpc_argp = &data->arg;
2895                 data->msg.rpc_resp = &data->res;
2896                 data->arg.dir_fh = NFS_FH(dir);
2897                 data->arg.server = server;
2898                 data->arg.name = name;
2899                 data->arg.attrs = sattr;
2900                 data->arg.ftype = ftype;
2901                 data->arg.bitmask = server->attr_bitmask;
2902                 data->res.server = server;
2903                 data->res.fh = &data->fh;
2904                 data->res.fattr = &data->fattr;
2905                 data->res.dir_fattr = &data->dir_fattr;
2906                 nfs_fattr_init(data->res.fattr);
2907                 nfs_fattr_init(data->res.dir_fattr);
2908         }
2909         return data;
2910 }
2911
2912 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2913 {
2914         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2915                                     &data->arg.seq_args, &data->res.seq_res, 1);
2916         if (status == 0) {
2917                 update_changeattr(dir, &data->res.dir_cinfo);
2918                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2919                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2920         }
2921         return status;
2922 }
2923
2924 static void nfs4_free_createdata(struct nfs4_createdata *data)
2925 {
2926         kfree(data);
2927 }
2928
2929 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2930                 struct page *page, unsigned int len, struct iattr *sattr)
2931 {
2932         struct nfs4_createdata *data;
2933         int status = -ENAMETOOLONG;
2934
2935         if (len > NFS4_MAXPATHLEN)
2936                 goto out;
2937
2938         status = -ENOMEM;
2939         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2940         if (data == NULL)
2941                 goto out;
2942
2943         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2944         data->arg.u.symlink.pages = &page;
2945         data->arg.u.symlink.len = len;
2946
2947         status = nfs4_do_create(dir, dentry, data);
2948
2949         nfs4_free_createdata(data);
2950 out:
2951         return status;
2952 }
2953
2954 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2955                 struct page *page, unsigned int len, struct iattr *sattr)
2956 {
2957         struct nfs4_exception exception = { };
2958         int err;
2959         do {
2960                 err = nfs4_handle_exception(NFS_SERVER(dir),
2961                                 _nfs4_proc_symlink(dir, dentry, page,
2962                                                         len, sattr),
2963                                 &exception);
2964         } while (exception.retry);
2965         return err;
2966 }
2967
2968 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2969                 struct iattr *sattr)
2970 {
2971         struct nfs4_createdata *data;
2972         int status = -ENOMEM;
2973
2974         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2975         if (data == NULL)
2976                 goto out;
2977
2978         status = nfs4_do_create(dir, dentry, data);
2979
2980         nfs4_free_createdata(data);
2981 out:
2982         return status;
2983 }
2984
2985 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2986                 struct iattr *sattr)
2987 {
2988         struct nfs4_exception exception = { };
2989         int err;
2990
2991         sattr->ia_mode &= ~current_umask();
2992         do {
2993                 err = nfs4_handle_exception(NFS_SERVER(dir),
2994                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2995                                 &exception);
2996         } while (exception.retry);
2997         return err;
2998 }
2999
3000 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3001                 u64 cookie, struct page **pages, unsigned int count, int plus)
3002 {
3003         struct inode            *dir = dentry->d_inode;
3004         struct nfs4_readdir_arg args = {
3005                 .fh = NFS_FH(dir),
3006                 .pages = pages,
3007                 .pgbase = 0,
3008                 .count = count,
3009                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3010                 .plus = plus,
3011         };
3012         struct nfs4_readdir_res res;
3013         struct rpc_message msg = {
3014                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3015                 .rpc_argp = &args,
3016                 .rpc_resp = &res,
3017                 .rpc_cred = cred,
3018         };
3019         int                     status;
3020
3021         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3022                         dentry->d_parent->d_name.name,
3023                         dentry->d_name.name,
3024                         (unsigned long long)cookie);
3025         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3026         res.pgbase = args.pgbase;
3027         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3028         if (status >= 0) {
3029                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3030                 status += args.pgbase;
3031         }
3032
3033         nfs_invalidate_atime(dir);
3034
3035         dprintk("%s: returns %d\n", __func__, status);
3036         return status;
3037 }
3038
3039 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3040                 u64 cookie, struct page **pages, unsigned int count, int plus)
3041 {
3042         struct nfs4_exception exception = { };
3043         int err;
3044         do {
3045                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3046                                 _nfs4_proc_readdir(dentry, cred, cookie,
3047                                         pages, count, plus),
3048                                 &exception);
3049         } while (exception.retry);
3050         return err;
3051 }
3052
3053 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3054                 struct iattr *sattr, dev_t rdev)
3055 {
3056         struct nfs4_createdata *data;
3057         int mode = sattr->ia_mode;
3058         int status = -ENOMEM;
3059
3060         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3061         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3062
3063         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3064         if (data == NULL)
3065                 goto out;
3066
3067         if (S_ISFIFO(mode))
3068                 data->arg.ftype = NF4FIFO;
3069         else if (S_ISBLK(mode)) {
3070                 data->arg.ftype = NF4BLK;
3071                 data->arg.u.device.specdata1 = MAJOR(rdev);
3072                 data->arg.u.device.specdata2 = MINOR(rdev);
3073         }
3074         else if (S_ISCHR(mode)) {
3075                 data->arg.ftype = NF4CHR;
3076                 data->arg.u.device.specdata1 = MAJOR(rdev);
3077                 data->arg.u.device.specdata2 = MINOR(rdev);
3078         }
3079
3080         status = nfs4_do_create(dir, dentry, data);
3081
3082         nfs4_free_createdata(data);
3083 out:
3084         return status;
3085 }
3086
3087 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3088                 struct iattr *sattr, dev_t rdev)
3089 {
3090         struct nfs4_exception exception = { };
3091         int err;
3092
3093         sattr->ia_mode &= ~current_umask();
3094         do {
3095                 err = nfs4_handle_exception(NFS_SERVER(dir),
3096                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3097                                 &exception);
3098         } while (exception.retry);
3099         return err;
3100 }
3101
3102 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3103                  struct nfs_fsstat *fsstat)
3104 {
3105         struct nfs4_statfs_arg args = {
3106                 .fh = fhandle,
3107                 .bitmask = server->attr_bitmask,
3108         };
3109         struct nfs4_statfs_res res = {
3110                 .fsstat = fsstat,
3111         };
3112         struct rpc_message msg = {
3113                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3114                 .rpc_argp = &args,
3115                 .rpc_resp = &res,
3116         };
3117
3118         nfs_fattr_init(fsstat->fattr);
3119         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3120 }
3121
3122 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3123 {
3124         struct nfs4_exception exception = { };
3125         int err;
3126         do {
3127                 err = nfs4_handle_exception(server,
3128                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3129                                 &exception);
3130         } while (exception.retry);
3131         return err;
3132 }
3133
3134 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3135                 struct nfs_fsinfo *fsinfo)
3136 {
3137         struct nfs4_fsinfo_arg args = {
3138                 .fh = fhandle,
3139                 .bitmask = server->attr_bitmask,
3140         };
3141         struct nfs4_fsinfo_res res = {
3142                 .fsinfo = fsinfo,
3143         };
3144         struct rpc_message msg = {
3145                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3146                 .rpc_argp = &args,
3147                 .rpc_resp = &res,
3148         };
3149
3150         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3151 }
3152
3153 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3154 {
3155         struct nfs4_exception exception = { };
3156         int err;
3157
3158         do {
3159                 err = nfs4_handle_exception(server,
3160                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3161                                 &exception);
3162         } while (exception.retry);
3163         return err;
3164 }
3165
3166 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3167 {
3168         nfs_fattr_init(fsinfo->fattr);
3169         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3170 }
3171
3172 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3173                 struct nfs_pathconf *pathconf)
3174 {
3175         struct nfs4_pathconf_arg args = {
3176                 .fh = fhandle,
3177                 .bitmask = server->attr_bitmask,
3178         };
3179         struct nfs4_pathconf_res res = {
3180                 .pathconf = pathconf,
3181         };
3182         struct rpc_message msg = {
3183                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3184                 .rpc_argp = &args,
3185                 .rpc_resp = &res,
3186         };
3187
3188         /* None of the pathconf attributes are mandatory to implement */
3189         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3190                 memset(pathconf, 0, sizeof(*pathconf));
3191                 return 0;
3192         }
3193
3194         nfs_fattr_init(pathconf->fattr);
3195         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3196 }
3197
3198 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3199                 struct nfs_pathconf *pathconf)
3200 {
3201         struct nfs4_exception exception = { };
3202         int err;
3203
3204         do {
3205                 err = nfs4_handle_exception(server,
3206                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3207                                 &exception);
3208         } while (exception.retry);
3209         return err;
3210 }
3211
3212 void __nfs4_read_done_cb(struct nfs_read_data *data)
3213 {
3214         nfs_invalidate_atime(data->inode);
3215 }
3216
3217 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3218 {
3219         struct nfs_server *server = NFS_SERVER(data->inode);
3220
3221         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3222                 rpc_restart_call_prepare(task);
3223                 return -EAGAIN;
3224         }
3225
3226         __nfs4_read_done_cb(data);
3227         if (task->tk_status > 0)
3228                 renew_lease(server, data->timestamp);
3229         return 0;
3230 }
3231
3232 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3233 {
3234
3235         dprintk("--> %s\n", __func__);
3236
3237         if (!nfs4_sequence_done(task, &data->res.seq_res))
3238                 return -EAGAIN;
3239
3240         return data->read_done_cb ? data->read_done_cb(task, data) :
3241                                     nfs4_read_done_cb(task, data);
3242 }
3243
3244 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3245 {
3246         data->timestamp   = jiffies;
3247         data->read_done_cb = nfs4_read_done_cb;
3248         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3249 }
3250
3251 /* Reset the the nfs_read_data to send the read to the MDS. */
3252 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3253 {
3254         dprintk("%s Reset task for i/o through\n", __func__);
3255         put_lseg(data->lseg);
3256         data->lseg = NULL;
3257         /* offsets will differ in the dense stripe case */
3258         data->args.offset = data->mds_offset;
3259         data->ds_clp = NULL;
3260         data->args.fh     = NFS_FH(data->inode);
3261         data->read_done_cb = nfs4_read_done_cb;
3262         task->tk_ops = data->mds_ops;
3263         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3264 }
3265 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3266
3267 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3268 {
3269         struct inode *inode = data->inode;
3270
3271         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3272                 rpc_restart_call_prepare(task);
3273                 return -EAGAIN;
3274         }
3275         if (task->tk_status >= 0) {
3276                 renew_lease(NFS_SERVER(inode), data->timestamp);
3277                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3278         }
3279         return 0;
3280 }
3281
3282 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3283 {
3284         if (!nfs4_sequence_done(task, &data->res.seq_res))
3285                 return -EAGAIN;
3286         return data->write_done_cb ? data->write_done_cb(task, data) :
3287                 nfs4_write_done_cb(task, data);
3288 }
3289
3290 /* Reset the the nfs_write_data to send the write to the MDS. */
3291 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3292 {
3293         dprintk("%s Reset task for i/o through\n", __func__);
3294         put_lseg(data->lseg);
3295         data->lseg          = NULL;
3296         data->ds_clp        = NULL;
3297         data->write_done_cb = nfs4_write_done_cb;
3298         data->args.fh       = NFS_FH(data->inode);
3299         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3300         data->args.offset   = data->mds_offset;
3301         data->res.fattr     = &data->fattr;
3302         task->tk_ops        = data->mds_ops;
3303         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3304 }
3305 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3306
3307 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3308 {
3309         struct nfs_server *server = NFS_SERVER(data->inode);
3310
3311         if (data->lseg) {
3312                 data->args.bitmask = NULL;
3313                 data->res.fattr = NULL;
3314         } else
3315                 data->args.bitmask = server->cache_consistency_bitmask;
3316         if (!data->write_done_cb)
3317                 data->write_done_cb = nfs4_write_done_cb;
3318         data->res.server = server;
3319         data->timestamp   = jiffies;
3320
3321         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3322 }
3323
3324 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3325 {
3326         struct inode *inode = data->inode;
3327
3328         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3329                 rpc_restart_call_prepare(task);
3330                 return -EAGAIN;
3331         }
3332         nfs_refresh_inode(inode, data->res.fattr);
3333         return 0;
3334 }
3335
3336 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3337 {
3338         if (!nfs4_sequence_done(task, &data->res.seq_res))
3339                 return -EAGAIN;
3340         return data->write_done_cb(task, data);
3341 }
3342
3343 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3344 {
3345         struct nfs_server *server = NFS_SERVER(data->inode);
3346
3347         if (data->lseg) {
3348                 data->args.bitmask = NULL;
3349                 data->res.fattr = NULL;
3350         } else
3351                 data->args.bitmask = server->cache_consistency_bitmask;
3352         if (!data->write_done_cb)
3353                 data->write_done_cb = nfs4_commit_done_cb;
3354         data->res.server = server;
3355         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3356 }
3357
3358 struct nfs4_renewdata {
3359         struct nfs_client       *client;
3360         unsigned long           timestamp;
3361 };
3362
3363 /*
3364  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3365  * standalone procedure for queueing an asynchronous RENEW.
3366  */
3367 static void nfs4_renew_release(void *calldata)
3368 {
3369         struct nfs4_renewdata *data = calldata;
3370         struct nfs_client *clp = data->client;
3371
3372         if (atomic_read(&clp->cl_count) > 1)
3373                 nfs4_schedule_state_renewal(clp);
3374         nfs_put_client(clp);
3375         kfree(data);
3376 }
3377
3378 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3379 {
3380         struct nfs4_renewdata *data = calldata;
3381         struct nfs_client *clp = data->client;
3382         unsigned long timestamp = data->timestamp;
3383
3384         if (task->tk_status < 0) {
3385                 /* Unless we're shutting down, schedule state recovery! */
3386                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3387                         return;
3388                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3389                         nfs4_schedule_lease_recovery(clp);
3390                         return;
3391                 }
3392                 nfs4_schedule_path_down_recovery(clp);
3393         }
3394         do_renew_lease(clp, timestamp);
3395 }
3396
3397 static const struct rpc_call_ops nfs4_renew_ops = {
3398         .rpc_call_done = nfs4_renew_done,
3399         .rpc_release = nfs4_renew_release,
3400 };
3401
3402 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3403 {
3404         struct rpc_message msg = {
3405                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3406                 .rpc_argp       = clp,
3407                 .rpc_cred       = cred,
3408         };
3409         struct nfs4_renewdata *data;
3410
3411         if (renew_flags == 0)
3412                 return 0;
3413         if (!atomic_inc_not_zero(&clp->cl_count))
3414                 return -EIO;
3415         data = kmalloc(sizeof(*data), GFP_NOFS);
3416         if (data == NULL)
3417                 return -ENOMEM;
3418         data->client = clp;
3419         data->timestamp = jiffies;
3420         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3421                         &nfs4_renew_ops, data);
3422 }
3423
3424 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3425 {
3426         struct rpc_message msg = {
3427                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3428                 .rpc_argp       = clp,
3429                 .rpc_cred       = cred,
3430         };
3431         unsigned long now = jiffies;
3432         int status;
3433
3434         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3435         if (status < 0)
3436                 return status;
3437         do_renew_lease(clp, now);
3438         return 0;
3439 }
3440
3441 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3442 {
3443         return (server->caps & NFS_CAP_ACLS)
3444                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3445                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3446 }
3447
3448 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3449  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3450  * the stack.
3451  */
3452 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3453
3454 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3455                 struct page **pages, unsigned int *pgbase)
3456 {
3457         struct page *newpage, **spages;
3458         int rc = 0;
3459         size_t len;
3460         spages = pages;
3461
3462         do {
3463                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3464                 newpage = alloc_page(GFP_KERNEL);
3465
3466                 if (newpage == NULL)
3467                         goto unwind;
3468                 memcpy(page_address(newpage), buf, len);
3469                 buf += len;
3470                 buflen -= len;
3471                 *pages++ = newpage;
3472                 rc++;
3473         } while (buflen != 0);
3474
3475         return rc;
3476
3477 unwind:
3478         for(; rc > 0; rc--)
3479                 __free_page(spages[rc-1]);
3480         return -ENOMEM;
3481 }
3482
3483 struct nfs4_cached_acl {
3484         int cached;
3485         size_t len;
3486         char data[0];
3487 };
3488
3489 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3490 {
3491         struct nfs_inode *nfsi = NFS_I(inode);
3492
3493         spin_lock(&inode->i_lock);
3494         kfree(nfsi->nfs4_acl);
3495         nfsi->nfs4_acl = acl;
3496         spin_unlock(&inode->i_lock);
3497 }
3498
3499 static void nfs4_zap_acl_attr(struct inode *inode)
3500 {
3501         nfs4_set_cached_acl(inode, NULL);
3502 }
3503
3504 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3505 {
3506         struct nfs_inode *nfsi = NFS_I(inode);
3507         struct nfs4_cached_acl *acl;
3508         int ret = -ENOENT;
3509
3510         spin_lock(&inode->i_lock);
3511         acl = nfsi->nfs4_acl;
3512         if (acl == NULL)
3513                 goto out;
3514         if (buf == NULL) /* user is just asking for length */
3515                 goto out_len;
3516         if (acl->cached == 0)
3517                 goto out;
3518         ret = -ERANGE; /* see getxattr(2) man page */
3519         if (acl->len > buflen)
3520                 goto out;
3521         memcpy(buf, acl->data, acl->len);
3522 out_len:
3523         ret = acl->len;
3524 out:
3525         spin_unlock(&inode->i_lock);
3526         return ret;
3527 }
3528
3529 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3530 {
3531         struct nfs4_cached_acl *acl;
3532
3533         if (pages && acl_len <= PAGE_SIZE) {
3534                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3535                 if (acl == NULL)
3536                         goto out;
3537                 acl->cached = 1;
3538                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3539         } else {
3540                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3541                 if (acl == NULL)
3542                         goto out;
3543                 acl->cached = 0;
3544         }
3545         acl->len = acl_len;
3546 out:
3547         nfs4_set_cached_acl(inode, acl);
3548 }
3549
3550 /*
3551  * The getxattr API returns the required buffer length when called with a
3552  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3553  * the required buf.  On a NULL buf, we send a page of data to the server
3554  * guessing that the ACL request can be serviced by a page. If so, we cache
3555  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3556  * the cache. If not so, we throw away the page, and cache the required
3557  * length. The next getxattr call will then produce another round trip to
3558  * the server, this time with the input buf of the required size.
3559  */
3560 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3561 {
3562         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3563         struct nfs_getaclargs args = {
3564                 .fh = NFS_FH(inode),
3565                 .acl_pages = pages,
3566                 .acl_len = buflen,
3567         };
3568         struct nfs_getaclres res = {
3569                 .acl_len = buflen,
3570         };
3571         struct rpc_message msg = {
3572                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3573                 .rpc_argp = &args,
3574                 .rpc_resp = &res,
3575         };
3576         int ret = -ENOMEM, npages, i, acl_len = 0;
3577
3578         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3579         /* As long as we're doing a round trip to the server anyway,
3580          * let's be prepared for a page of acl data. */
3581         if (npages == 0)
3582                 npages = 1;
3583
3584         /* Add an extra page to handle the bitmap returned */
3585         npages++;
3586
3587         for (i = 0; i < npages; i++) {
3588                 pages[i] = alloc_page(GFP_KERNEL);
3589                 if (!pages[i])
3590                         goto out_free;
3591         }
3592
3593         /* for decoding across pages */
3594         res.acl_scratch = alloc_page(GFP_KERNEL);
3595         if (!res.acl_scratch)
3596                 goto out_free;
3597
3598         args.acl_len = npages * PAGE_SIZE;
3599         args.acl_pgbase = 0;
3600
3601         /* Let decode_getfacl know not to fail if the ACL data is larger than
3602          * the page we send as a guess */
3603         if (buf == NULL)
3604                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3605
3606         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3607                 __func__, buf, buflen, npages, args.acl_len);
3608         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3609                              &msg, &args.seq_args, &res.seq_res, 0);
3610         if (ret)
3611                 goto out_free;
3612
3613         acl_len = res.acl_len - res.acl_data_offset;
3614         if (acl_len > args.acl_len)
3615                 nfs4_write_cached_acl(inode, NULL, 0, acl_len);
3616         else
3617                 nfs4_write_cached_acl(inode, pages, res.acl_data_offset,
3618                                       acl_len);
3619         if (buf) {
3620                 ret = -ERANGE;
3621                 if (acl_len > buflen)
3622                         goto out_free;
3623                 _copy_from_pages(buf, pages, res.acl_data_offset,
3624                                 acl_len);
3625         }
3626         ret = acl_len;
3627 out_free:
3628         for (i = 0; i < npages; i++)
3629                 if (pages[i])
3630                         __free_page(pages[i]);
3631         if (res.acl_scratch)
3632                 __free_page(res.acl_scratch);
3633         return ret;
3634 }
3635
3636 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3637 {
3638         struct nfs4_exception exception = { };
3639         ssize_t ret;
3640         do {
3641                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3642                 if (ret >= 0)
3643                         break;
3644                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3645         } while (exception.retry);
3646         return ret;
3647 }
3648
3649 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3650 {
3651         struct nfs_server *server = NFS_SERVER(inode);
3652         int ret;
3653
3654         if (!nfs4_server_supports_acls(server))
3655                 return -EOPNOTSUPP;
3656         ret = nfs_revalidate_inode(server, inode);
3657         if (ret < 0)
3658                 return ret;
3659         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3660                 nfs_zap_acl_cache(inode);
3661         ret = nfs4_read_cached_acl(inode, buf, buflen);
3662         if (ret != -ENOENT)
3663                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3664                  * but no cached acl data, just the acl length */
3665                 return ret;
3666         return nfs4_get_acl_uncached(inode, buf, buflen);
3667 }
3668
3669 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3670 {
3671         struct nfs_server *server = NFS_SERVER(inode);
3672         struct page *pages[NFS4ACL_MAXPAGES];
3673         struct nfs_setaclargs arg = {
3674                 .fh             = NFS_FH(inode),
3675                 .acl_pages      = pages,
3676                 .acl_len        = buflen,
3677         };
3678         struct nfs_setaclres res;
3679         struct rpc_message msg = {
3680                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3681                 .rpc_argp       = &arg,
3682                 .rpc_resp       = &res,
3683         };
3684         int ret, i;
3685
3686         if (!nfs4_server_supports_acls(server))
3687                 return -EOPNOTSUPP;
3688         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3689         if (i < 0)
3690                 return i;
3691         nfs_inode_return_delegation(inode);
3692         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3693
3694         /*
3695          * Free each page after tx, so the only ref left is
3696          * held by the network stack
3697          */
3698         for (; i > 0; i--)
3699                 put_page(pages[i-1]);
3700
3701         /*
3702          * Acl update can result in inode attribute update.
3703          * so mark the attribute cache invalid.
3704          */
3705         spin_lock(&inode->i_lock);
3706         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3707         spin_unlock(&inode->i_lock);
3708         nfs_access_zap_cache(inode);
3709         nfs_zap_acl_cache(inode);
3710         return ret;
3711 }
3712
3713 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3714 {
3715         struct nfs4_exception exception = { };
3716         int err;
3717         do {
3718                 err = nfs4_handle_exception(NFS_SERVER(inode),
3719                                 __nfs4_proc_set_acl(inode, buf, buflen),
3720                                 &exception);
3721         } while (exception.retry);
3722         return err;
3723 }
3724
3725 static int
3726 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3727 {
3728         struct nfs_client *clp = server->nfs_client;
3729
3730         if (task->tk_status >= 0)
3731                 return 0;
3732         switch(task->tk_status) {
3733                 case -NFS4ERR_DELEG_REVOKED:
3734                 case -NFS4ERR_ADMIN_REVOKED:
3735                 case -NFS4ERR_BAD_STATEID:
3736                         if (state != NULL)
3737                                 nfs_remove_bad_delegation(state->inode);
3738                 case -NFS4ERR_OPENMODE:
3739                         if (state == NULL)
3740                                 break;
3741                         nfs4_schedule_stateid_recovery(server, state);
3742                         goto wait_on_recovery;
3743                 case -NFS4ERR_EXPIRED:
3744                         if (state != NULL)
3745                                 nfs4_schedule_stateid_recovery(server, state);
3746                 case -NFS4ERR_STALE_STATEID:
3747                 case -NFS4ERR_STALE_CLIENTID:
3748                         nfs4_schedule_lease_recovery(clp);
3749                         goto wait_on_recovery;
3750 #if defined(CONFIG_NFS_V4_1)
3751                 case -NFS4ERR_BADSESSION:
3752                 case -NFS4ERR_BADSLOT:
3753                 case -NFS4ERR_BAD_HIGH_SLOT:
3754                 case -NFS4ERR_DEADSESSION:
3755                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3756                 case -NFS4ERR_SEQ_FALSE_RETRY:
3757                 case -NFS4ERR_SEQ_MISORDERED:
3758                         dprintk("%s ERROR %d, Reset session\n", __func__,
3759                                 task->tk_status);
3760                         nfs4_schedule_session_recovery(clp->cl_session);
3761                         task->tk_status = 0;
3762                         return -EAGAIN;
3763 #endif /* CONFIG_NFS_V4_1 */
3764                 case -NFS4ERR_DELAY:
3765                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3766                 case -NFS4ERR_GRACE:
3767                 case -EKEYEXPIRED:
3768                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3769                         task->tk_status = 0;
3770                         return -EAGAIN;
3771                 case -NFS4ERR_RETRY_UNCACHED_REP:
3772                 case -NFS4ERR_OLD_STATEID:
3773                         task->tk_status = 0;
3774                         return -EAGAIN;
3775         }
3776         task->tk_status = nfs4_map_errors(task->tk_status);
3777         return 0;
3778 wait_on_recovery:
3779         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3780         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3781                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3782         task->tk_status = 0;
3783         return -EAGAIN;
3784 }
3785
3786 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3787                 unsigned short port, struct rpc_cred *cred,
3788                 struct nfs4_setclientid_res *res)
3789 {
3790         nfs4_verifier sc_verifier;
3791         struct nfs4_setclientid setclientid = {
3792                 .sc_verifier = &sc_verifier,
3793                 .sc_prog = program,
3794                 .sc_cb_ident = clp->cl_cb_ident,
3795         };
3796         struct rpc_message msg = {
3797                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3798                 .rpc_argp = &setclientid,
3799                 .rpc_resp = res,
3800                 .rpc_cred = cred,
3801         };
3802         __be32 *p;
3803         int loop = 0;
3804         int status;
3805
3806         p = (__be32*)sc_verifier.data;
3807         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3808         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3809
3810         for(;;) {
3811                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3812                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3813                                 clp->cl_ipaddr,
3814                                 rpc_peeraddr2str(clp->cl_rpcclient,
3815                                                         RPC_DISPLAY_ADDR),
3816                                 rpc_peeraddr2str(clp->cl_rpcclient,
3817                                                         RPC_DISPLAY_PROTO),
3818                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3819                                 clp->cl_id_uniquifier);
3820                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3821                                 sizeof(setclientid.sc_netid),
3822                                 rpc_peeraddr2str(clp->cl_rpcclient,
3823                                                         RPC_DISPLAY_NETID));
3824                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3825                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3826                                 clp->cl_ipaddr, port >> 8, port & 255);
3827
3828                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3829                 if (status != -NFS4ERR_CLID_INUSE)
3830                         break;
3831                 if (loop != 0) {
3832                         ++clp->cl_id_uniquifier;
3833                         break;
3834                 }
3835                 ++loop;
3836                 ssleep(clp->cl_lease_time / HZ + 1);
3837         }
3838         return status;
3839 }
3840
3841 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3842                 struct nfs4_setclientid_res *arg,
3843                 struct rpc_cred *cred)
3844 {
3845         struct nfs_fsinfo fsinfo;
3846         struct rpc_message msg = {
3847                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3848                 .rpc_argp = arg,
3849                 .rpc_resp = &fsinfo,
3850                 .rpc_cred = cred,
3851         };
3852         unsigned long now;
3853         int status;
3854
3855         now = jiffies;
3856         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3857         if (status == 0) {
3858                 spin_lock(&clp->cl_lock);
3859                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3860                 clp->cl_last_renewal = now;
3861                 spin_unlock(&clp->cl_lock);
3862         }
3863         return status;
3864 }
3865
3866 struct nfs4_delegreturndata {
3867         struct nfs4_delegreturnargs args;
3868         struct nfs4_delegreturnres res;
3869         struct nfs_fh fh;
3870         nfs4_stateid stateid;
3871         unsigned long timestamp;
3872         struct nfs_fattr fattr;
3873         int rpc_status;
3874 };
3875
3876 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3877 {
3878         struct nfs4_delegreturndata *data = calldata;
3879
3880         if (!nfs4_sequence_done(task, &data->res.seq_res))
3881                 return;
3882
3883         switch (task->tk_status) {
3884         case -NFS4ERR_STALE_STATEID:
3885         case -NFS4ERR_EXPIRED:
3886         case 0:
3887                 renew_lease(data->res.server, data->timestamp);
3888                 break;
3889         default:
3890                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3891                                 -EAGAIN) {
3892                         rpc_restart_call_prepare(task);
3893                         return;
3894                 }
3895         }
3896         data->rpc_status = task->tk_status;
3897 }
3898
3899 static void nfs4_delegreturn_release(void *calldata)
3900 {
3901         kfree(calldata);
3902 }
3903
3904 #if defined(CONFIG_NFS_V4_1)
3905 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3906 {
3907         struct nfs4_delegreturndata *d_data;
3908
3909         d_data = (struct nfs4_delegreturndata *)data;
3910
3911         if (nfs4_setup_sequence(d_data->res.server,
3912                                 &d_data->args.seq_args,
3913                                 &d_data->res.seq_res, 1, task))
3914                 return;
3915         rpc_call_start(task);
3916 }
3917 #endif /* CONFIG_NFS_V4_1 */
3918
3919 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3920 #if defined(CONFIG_NFS_V4_1)
3921         .rpc_call_prepare = nfs4_delegreturn_prepare,
3922 #endif /* CONFIG_NFS_V4_1 */
3923         .rpc_call_done = nfs4_delegreturn_done,
3924         .rpc_release = nfs4_delegreturn_release,
3925 };
3926
3927 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3928 {
3929         struct nfs4_delegreturndata *data;
3930         struct nfs_server *server = NFS_SERVER(inode);
3931         struct rpc_task *task;
3932         struct rpc_message msg = {
3933                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3934                 .rpc_cred = cred,
3935         };
3936         struct rpc_task_setup task_setup_data = {
3937                 .rpc_client = server->client,
3938                 .rpc_message = &msg,
3939                 .callback_ops = &nfs4_delegreturn_ops,
3940                 .flags = RPC_TASK_ASYNC,
3941         };
3942         int status = 0;
3943
3944         data = kzalloc(sizeof(*data), GFP_NOFS);
3945         if (data == NULL)
3946                 return -ENOMEM;
3947         data->args.fhandle = &data->fh;
3948         data->args.stateid = &data->stateid;
3949         data->args.bitmask = server->attr_bitmask;
3950         nfs_copy_fh(&data->fh, NFS_FH(inode));
3951         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3952         data->res.fattr = &data->fattr;
3953         data->res.server = server;
3954         nfs_fattr_init(data->res.fattr);
3955         data->timestamp = jiffies;
3956         data->rpc_status = 0;
3957
3958         task_setup_data.callback_data = data;
3959         msg.rpc_argp = &data->args;
3960         msg.rpc_resp = &data->res;
3961         task = rpc_run_task(&task_setup_data);
3962         if (IS_ERR(task))
3963                 return PTR_ERR(task);
3964         if (!issync)
3965                 goto out;
3966         status = nfs4_wait_for_completion_rpc_task(task);
3967         if (status != 0)
3968                 goto out;
3969         status = data->rpc_status;
3970         if (status != 0)
3971                 goto out;
3972         nfs_refresh_inode(inode, &data->fattr);
3973 out:
3974         rpc_put_task(task);
3975         return status;
3976 }
3977
3978 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3979 {
3980         struct nfs_server *server = NFS_SERVER(inode);
3981         struct nfs4_exception exception = { };
3982         int err;
3983         do {
3984                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3985                 switch (err) {
3986                         case -NFS4ERR_STALE_STATEID:
3987                         case -NFS4ERR_EXPIRED:
3988                         case 0:
3989                                 return 0;
3990                 }
3991                 err = nfs4_handle_exception(server, err, &exception);
3992         } while (exception.retry);
3993         return err;
3994 }
3995
3996 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3997 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3998
3999 /*
4000  * sleep, with exponential backoff, and retry the LOCK operation.
4001  */
4002 static unsigned long
4003 nfs4_set_lock_task_retry(unsigned long timeout)
4004 {
4005         schedule_timeout_killable(timeout);
4006         timeout <<= 1;
4007         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4008                 return NFS4_LOCK_MAXTIMEOUT;
4009         return timeout;
4010 }
4011
4012 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4013 {
4014         struct inode *inode = state->inode;
4015         struct nfs_server *server = NFS_SERVER(inode);
4016         struct nfs_client *clp = server->nfs_client;
4017         struct nfs_lockt_args arg = {
4018                 .fh = NFS_FH(inode),
4019                 .fl = request,
4020         };
4021         struct nfs_lockt_res res = {
4022                 .denied = request,
4023         };
4024         struct rpc_message msg = {
4025                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4026                 .rpc_argp       = &arg,
4027                 .rpc_resp       = &res,
4028                 .rpc_cred       = state->owner->so_cred,
4029         };
4030         struct nfs4_lock_state *lsp;
4031         int status;
4032
4033         arg.lock_owner.clientid = clp->cl_clientid;
4034         status = nfs4_set_lock_state(state, request);
4035         if (status != 0)
4036                 goto out;
4037         lsp = request->fl_u.nfs4_fl.owner;
4038         arg.lock_owner.id = lsp->ls_id.id;
4039         arg.lock_owner.s_dev = server->s_dev;
4040         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4041         switch (status) {
4042                 case 0:
4043                         request->fl_type = F_UNLCK;
4044                         break;
4045                 case -NFS4ERR_DENIED:
4046                         status = 0;
4047         }
4048         request->fl_ops->fl_release_private(request);
4049 out:
4050         return status;
4051 }
4052
4053 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4054 {
4055         struct nfs4_exception exception = { };
4056         int err;
4057
4058         do {
4059                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4060                                 _nfs4_proc_getlk(state, cmd, request),
4061                                 &exception);
4062         } while (exception.retry);
4063         return err;
4064 }
4065
4066 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4067 {
4068         int res = 0;
4069         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4070                 case FL_POSIX:
4071                         res = posix_lock_file_wait(file, fl);
4072                         break;
4073                 case FL_FLOCK:
4074                         res = flock_lock_file_wait(file, fl);
4075                         break;
4076                 default:
4077                         BUG();
4078         }
4079         return res;
4080 }
4081
4082 struct nfs4_unlockdata {
4083         struct nfs_locku_args arg;
4084         struct nfs_locku_res res;
4085         struct nfs4_lock_state *lsp;
4086         struct nfs_open_context *ctx;
4087         struct file_lock fl;
4088         const struct nfs_server *server;
4089         unsigned long timestamp;
4090 };
4091
4092 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4093                 struct nfs_open_context *ctx,
4094                 struct nfs4_lock_state *lsp,
4095                 struct nfs_seqid *seqid)
4096 {
4097         struct nfs4_unlockdata *p;
4098         struct inode *inode = lsp->ls_state->inode;
4099
4100         p = kzalloc(sizeof(*p), GFP_NOFS);
4101         if (p == NULL)
4102                 return NULL;
4103         p->arg.fh = NFS_FH(inode);
4104         p->arg.fl = &p->fl;
4105         p->arg.seqid = seqid;
4106         p->res.seqid = seqid;
4107         p->arg.stateid = &lsp->ls_stateid;
4108         p->lsp = lsp;
4109         atomic_inc(&lsp->ls_count);
4110         /* Ensure we don't close file until we're done freeing locks! */
4111         p->ctx = get_nfs_open_context(ctx);
4112         memcpy(&p->fl, fl, sizeof(p->fl));
4113         p->server = NFS_SERVER(inode);
4114         return p;
4115 }
4116
4117 static void nfs4_locku_release_calldata(void *data)
4118 {
4119         struct nfs4_unlockdata *calldata = data;
4120         nfs_free_seqid(calldata->arg.seqid);
4121         nfs4_put_lock_state(calldata->lsp);
4122         put_nfs_open_context(calldata->ctx);
4123         kfree(calldata);
4124 }
4125
4126 static void nfs4_locku_done(struct rpc_task *task, void *data)
4127 {
4128         struct nfs4_unlockdata *calldata = data;
4129
4130         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4131                 return;
4132         switch (task->tk_status) {
4133                 case 0:
4134                         memcpy(calldata->lsp->ls_stateid.data,
4135                                         calldata->res.stateid.data,
4136                                         sizeof(calldata->lsp->ls_stateid.data));
4137                         renew_lease(calldata->server, calldata->timestamp);
4138                         break;
4139                 case -NFS4ERR_BAD_STATEID:
4140                 case -NFS4ERR_OLD_STATEID:
4141                 case -NFS4ERR_STALE_STATEID:
4142                 case -NFS4ERR_EXPIRED:
4143                         break;
4144                 default:
4145                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4146                                 rpc_restart_call_prepare(task);
4147         }
4148 }
4149
4150 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4151 {
4152         struct nfs4_unlockdata *calldata = data;
4153
4154         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4155                 return;
4156         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4157                 /* Note: exit _without_ running nfs4_locku_done */
4158                 task->tk_action = NULL;
4159                 return;
4160         }
4161         calldata->timestamp = jiffies;
4162         if (nfs4_setup_sequence(calldata->server,
4163                                 &calldata->arg.seq_args,
4164                                 &calldata->res.seq_res, 1, task))
4165                 return;
4166         rpc_call_start(task);
4167 }
4168
4169 static const struct rpc_call_ops nfs4_locku_ops = {
4170         .rpc_call_prepare = nfs4_locku_prepare,
4171         .rpc_call_done = nfs4_locku_done,
4172         .rpc_release = nfs4_locku_release_calldata,
4173 };
4174
4175 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4176                 struct nfs_open_context *ctx,
4177                 struct nfs4_lock_state *lsp,
4178                 struct nfs_seqid *seqid)
4179 {
4180         struct nfs4_unlockdata *data;
4181         struct rpc_message msg = {
4182                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4183                 .rpc_cred = ctx->cred,
4184         };
4185         struct rpc_task_setup task_setup_data = {
4186                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4187                 .rpc_message = &msg,
4188                 .callback_ops = &nfs4_locku_ops,
4189                 .workqueue = nfsiod_workqueue,
4190                 .flags = RPC_TASK_ASYNC,
4191         };
4192
4193         /* Ensure this is an unlock - when canceling a lock, the
4194          * canceled lock is passed in, and it won't be an unlock.
4195          */
4196         fl->fl_type = F_UNLCK;
4197
4198         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4199         if (data == NULL) {
4200                 nfs_free_seqid(seqid);
4201                 return ERR_PTR(-ENOMEM);
4202         }
4203
4204         msg.rpc_argp = &data->arg;
4205         msg.rpc_resp = &data->res;
4206         task_setup_data.callback_data = data;
4207         return rpc_run_task(&task_setup_data);
4208 }
4209
4210 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4211 {
4212         struct nfs_inode *nfsi = NFS_I(state->inode);
4213         struct nfs_seqid *seqid;
4214         struct nfs4_lock_state *lsp;
4215         struct rpc_task *task;
4216         int status = 0;
4217         unsigned char fl_flags = request->fl_flags;
4218
4219         status = nfs4_set_lock_state(state, request);
4220         /* Unlock _before_ we do the RPC call */
4221         request->fl_flags |= FL_EXISTS;
4222         down_read(&nfsi->rwsem);
4223         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4224                 up_read(&nfsi->rwsem);
4225                 goto out;
4226         }
4227         up_read(&nfsi->rwsem);
4228         if (status != 0)
4229                 goto out;
4230         /* Is this a delegated lock? */
4231         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4232                 goto out;
4233         lsp = request->fl_u.nfs4_fl.owner;
4234         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4235         status = -ENOMEM;
4236         if (seqid == NULL)
4237                 goto out;
4238         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4239         status = PTR_ERR(task);
4240         if (IS_ERR(task))
4241                 goto out;
4242         status = nfs4_wait_for_completion_rpc_task(task);
4243         rpc_put_task(task);
4244 out:
4245         request->fl_flags = fl_flags;
4246         return status;
4247 }
4248
4249 struct nfs4_lockdata {
4250         struct nfs_lock_args arg;
4251         struct nfs_lock_res res;
4252         struct nfs4_lock_state *lsp;
4253         struct nfs_open_context *ctx;
4254         struct file_lock fl;
4255         unsigned long timestamp;
4256         int rpc_status;
4257         int cancelled;
4258         struct nfs_server *server;
4259 };
4260
4261 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4262                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4263                 gfp_t gfp_mask)
4264 {
4265         struct nfs4_lockdata *p;
4266         struct inode *inode = lsp->ls_state->inode;
4267         struct nfs_server *server = NFS_SERVER(inode);
4268
4269         p = kzalloc(sizeof(*p), gfp_mask);
4270         if (p == NULL)
4271                 return NULL;
4272
4273         p->arg.fh = NFS_FH(inode);
4274         p->arg.fl = &p->fl;
4275         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4276         if (p->arg.open_seqid == NULL)
4277                 goto out_free;
4278         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4279         if (p->arg.lock_seqid == NULL)
4280                 goto out_free_seqid;
4281         p->arg.lock_stateid = &lsp->ls_stateid;
4282         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4283         p->arg.lock_owner.id = lsp->ls_id.id;
4284         p->arg.lock_owner.s_dev = server->s_dev;
4285         p->res.lock_seqid = p->arg.lock_seqid;
4286         p->lsp = lsp;
4287         p->server = server;
4288         atomic_inc(&lsp->ls_count);
4289         p->ctx = get_nfs_open_context(ctx);
4290         memcpy(&p->fl, fl, sizeof(p->fl));
4291         return p;
4292 out_free_seqid:
4293         nfs_free_seqid(p->arg.open_seqid);
4294 out_free:
4295         kfree(p);
4296         return NULL;
4297 }
4298
4299 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4300 {
4301         struct nfs4_lockdata *data = calldata;
4302         struct nfs4_state *state = data->lsp->ls_state;
4303
4304         dprintk("%s: begin!\n", __func__);
4305         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4306                 return;
4307         /* Do we need to do an open_to_lock_owner? */
4308         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4309                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4310                         return;
4311                 data->arg.open_stateid = &state->stateid;
4312                 data->arg.new_lock_owner = 1;
4313                 data->res.open_seqid = data->arg.open_seqid;
4314         } else
4315                 data->arg.new_lock_owner = 0;
4316         data->timestamp = jiffies;
4317         if (nfs4_setup_sequence(data->server,
4318                                 &data->arg.seq_args,
4319                                 &data->res.seq_res, 1, task))
4320                 return;
4321         rpc_call_start(task);
4322         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4323 }
4324
4325 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4326 {
4327         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4328         nfs4_lock_prepare(task, calldata);
4329 }
4330
4331 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4332 {
4333         struct nfs4_lockdata *data = calldata;
4334
4335         dprintk("%s: begin!\n", __func__);
4336
4337         if (!nfs4_sequence_done(task, &data->res.seq_res))
4338                 return;
4339
4340         data->rpc_status = task->tk_status;
4341         if (data->arg.new_lock_owner != 0) {
4342                 if (data->rpc_status == 0)
4343                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4344                 else
4345                         goto out;
4346         }
4347         if (data->rpc_status == 0) {
4348                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4349                                         sizeof(data->lsp->ls_stateid.data));
4350                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4351                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4352         }
4353 out:
4354         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4355 }
4356
4357 static void nfs4_lock_release(void *calldata)
4358 {
4359         struct nfs4_lockdata *data = calldata;
4360
4361         dprintk("%s: begin!\n", __func__);
4362         nfs_free_seqid(data->arg.open_seqid);
4363         if (data->cancelled != 0) {
4364                 struct rpc_task *task;
4365                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4366                                 data->arg.lock_seqid);
4367                 if (!IS_ERR(task))
4368                         rpc_put_task_async(task);
4369                 dprintk("%s: cancelling lock!\n", __func__);
4370         } else
4371                 nfs_free_seqid(data->arg.lock_seqid);
4372         nfs4_put_lock_state(data->lsp);
4373         put_nfs_open_context(data->ctx);
4374         kfree(data);
4375         dprintk("%s: done!\n", __func__);
4376 }
4377
4378 static const struct rpc_call_ops nfs4_lock_ops = {
4379         .rpc_call_prepare = nfs4_lock_prepare,
4380         .rpc_call_done = nfs4_lock_done,
4381         .rpc_release = nfs4_lock_release,
4382 };
4383
4384 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4385         .rpc_call_prepare = nfs4_recover_lock_prepare,
4386         .rpc_call_done = nfs4_lock_done,
4387         .rpc_release = nfs4_lock_release,
4388 };
4389
4390 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4391 {
4392         switch (error) {
4393         case -NFS4ERR_ADMIN_REVOKED:
4394         case -NFS4ERR_BAD_STATEID:
4395                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4396                 if (new_lock_owner != 0 ||
4397                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4398                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4399                 break;
4400         case -NFS4ERR_STALE_STATEID:
4401                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4402         case -NFS4ERR_EXPIRED:
4403                 nfs4_schedule_lease_recovery(server->nfs_client);
4404         };
4405 }
4406
4407 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4408 {
4409         struct nfs4_lockdata *data;
4410         struct rpc_task *task;
4411         struct rpc_message msg = {
4412                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4413                 .rpc_cred = state->owner->so_cred,
4414         };
4415         struct rpc_task_setup task_setup_data = {
4416                 .rpc_client = NFS_CLIENT(state->inode),
4417                 .rpc_message = &msg,
4418                 .callback_ops = &nfs4_lock_ops,
4419                 .workqueue = nfsiod_workqueue,
4420                 .flags = RPC_TASK_ASYNC,
4421         };
4422         int ret;
4423
4424         dprintk("%s: begin!\n", __func__);
4425         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4426                         fl->fl_u.nfs4_fl.owner,
4427                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4428         if (data == NULL)
4429                 return -ENOMEM;
4430         if (IS_SETLKW(cmd))
4431                 data->arg.block = 1;
4432         if (recovery_type > NFS_LOCK_NEW) {
4433                 if (recovery_type == NFS_LOCK_RECLAIM)
4434                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4435                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4436         }
4437         msg.rpc_argp = &data->arg;
4438         msg.rpc_resp = &data->res;
4439         task_setup_data.callback_data = data;
4440         task = rpc_run_task(&task_setup_data);
4441         if (IS_ERR(task))
4442                 return PTR_ERR(task);
4443         ret = nfs4_wait_for_completion_rpc_task(task);
4444         if (ret == 0) {
4445                 ret = data->rpc_status;
4446                 if (ret)
4447                         nfs4_handle_setlk_error(data->server, data->lsp,
4448                                         data->arg.new_lock_owner, ret);
4449         } else
4450                 data->cancelled = 1;
4451         rpc_put_task(task);
4452         dprintk("%s: done, ret = %d!\n", __func__, ret);
4453         return ret;
4454 }
4455
4456 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4457 {
4458         struct nfs_server *server = NFS_SERVER(state->inode);
4459         struct nfs4_exception exception = {
4460                 .inode = state->inode,
4461         };
4462         int err;
4463
4464         do {
4465                 /* Cache the lock if possible... */
4466                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4467                         return 0;
4468                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4469                 if (err != -NFS4ERR_DELAY)
4470                         break;
4471                 nfs4_handle_exception(server, err, &exception);
4472         } while (exception.retry);
4473         return err;
4474 }
4475
4476 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4477 {
4478         struct nfs_server *server = NFS_SERVER(state->inode);
4479         struct nfs4_exception exception = {
4480                 .inode = state->inode,
4481         };
4482         int err;
4483
4484         err = nfs4_set_lock_state(state, request);
4485         if (err != 0)
4486                 return err;
4487         do {
4488                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4489                         return 0;
4490                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4491                 switch (err) {
4492                 default:
4493                         goto out;
4494                 case -NFS4ERR_GRACE:
4495                 case -NFS4ERR_DELAY:
4496                         nfs4_handle_exception(server, err, &exception);
4497                         err = 0;
4498                 }
4499         } while (exception.retry);
4500 out:
4501         return err;
4502 }
4503
4504 #if defined(CONFIG_NFS_V4_1)
4505 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4506 {
4507         int status;
4508         struct nfs_server *server = NFS_SERVER(state->inode);
4509
4510         status = nfs41_test_stateid(server, state);
4511         if (status == NFS_OK)
4512                 return 0;
4513         nfs41_free_stateid(server, state);
4514         return nfs4_lock_expired(state, request);
4515 }
4516 #endif
4517
4518 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4519 {
4520         struct nfs_inode *nfsi = NFS_I(state->inode);
4521         unsigned char fl_flags = request->fl_flags;
4522         int status = -ENOLCK;
4523
4524         if ((fl_flags & FL_POSIX) &&
4525                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4526                 goto out;
4527         /* Is this a delegated open? */
4528         status = nfs4_set_lock_state(state, request);
4529         if (status != 0)
4530                 goto out;
4531         request->fl_flags |= FL_ACCESS;
4532         status = do_vfs_lock(request->fl_file, request);
4533         if (status < 0)
4534                 goto out;
4535         down_read(&nfsi->rwsem);
4536         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4537                 /* Yes: cache locks! */
4538                 /* ...but avoid races with delegation recall... */
4539                 request->fl_flags = fl_flags & ~FL_SLEEP;
4540                 status = do_vfs_lock(request->fl_file, request);
4541                 goto out_unlock;
4542         }
4543         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4544         if (status != 0)
4545                 goto out_unlock;
4546         /* Note: we always want to sleep here! */
4547         request->fl_flags = fl_flags | FL_SLEEP;
4548         if (do_vfs_lock(request->fl_file, request) < 0)
4549                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4550 out_unlock:
4551         up_read(&nfsi->rwsem);
4552 out:
4553         request->fl_flags = fl_flags;
4554         return status;
4555 }
4556
4557 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4558 {
4559         struct nfs4_exception exception = {
4560                 .state = state,
4561                 .inode = state->inode,
4562         };
4563         int err;
4564
4565         do {
4566                 err = _nfs4_proc_setlk(state, cmd, request);
4567                 if (err == -NFS4ERR_DENIED)
4568                         err = -EAGAIN;
4569                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4570                                 err, &exception);
4571         } while (exception.retry);
4572         return err;
4573 }
4574
4575 static int
4576 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4577 {
4578         struct nfs_open_context *ctx;
4579         struct nfs4_state *state;
4580         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4581         int status;
4582
4583         /* verify open state */
4584         ctx = nfs_file_open_context(filp);
4585         state = ctx->state;
4586
4587         if (request->fl_start < 0 || request->fl_end < 0)
4588                 return -EINVAL;
4589
4590         if (IS_GETLK(cmd)) {
4591                 if (state != NULL)
4592                         return nfs4_proc_getlk(state, F_GETLK, request);
4593                 return 0;
4594         }
4595
4596         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4597                 return -EINVAL;
4598
4599         if (request->fl_type == F_UNLCK) {
4600                 if (state != NULL)
4601                         return nfs4_proc_unlck(state, cmd, request);
4602                 return 0;
4603         }
4604
4605         if (state == NULL)
4606                 return -ENOLCK;
4607         /*
4608          * Don't rely on the VFS having checked the file open mode,
4609          * since it won't do this for flock() locks.
4610          */
4611         switch (request->fl_type & (F_RDLCK|F_WRLCK|F_UNLCK)) {
4612         case F_RDLCK:
4613                 if (!(filp->f_mode & FMODE_READ))
4614                         return -EBADF;
4615                 break;
4616         case F_WRLCK:
4617                 if (!(filp->f_mode & FMODE_WRITE))
4618                         return -EBADF;
4619         }
4620
4621         do {
4622                 status = nfs4_proc_setlk(state, cmd, request);
4623                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4624                         break;
4625                 timeout = nfs4_set_lock_task_retry(timeout);
4626                 status = -ERESTARTSYS;
4627                 if (signalled())
4628                         break;
4629         } while(status < 0);
4630         return status;
4631 }
4632
4633 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4634 {
4635         struct nfs_server *server = NFS_SERVER(state->inode);
4636         struct nfs4_exception exception = { };
4637         int err;
4638
4639         err = nfs4_set_lock_state(state, fl);
4640         if (err != 0)
4641                 goto out;
4642         do {
4643                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4644                 switch (err) {
4645                         default:
4646                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4647                                                 __func__, err);
4648                         case 0:
4649                         case -ESTALE:
4650                                 goto out;
4651                         case -NFS4ERR_EXPIRED:
4652                                 nfs4_schedule_stateid_recovery(server, state);
4653                         case -NFS4ERR_STALE_CLIENTID:
4654                         case -NFS4ERR_STALE_STATEID:
4655                                 nfs4_schedule_lease_recovery(server->nfs_client);
4656                                 goto out;
4657                         case -NFS4ERR_BADSESSION:
4658                         case -NFS4ERR_BADSLOT:
4659                         case -NFS4ERR_BAD_HIGH_SLOT:
4660                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4661                         case -NFS4ERR_DEADSESSION:
4662                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4663                                 goto out;
4664                         case -ERESTARTSYS:
4665                                 /*
4666                                  * The show must go on: exit, but mark the
4667                                  * stateid as needing recovery.
4668                                  */
4669                         case -NFS4ERR_DELEG_REVOKED:
4670                         case -NFS4ERR_ADMIN_REVOKED:
4671                         case -NFS4ERR_BAD_STATEID:
4672                         case -NFS4ERR_OPENMODE:
4673                                 nfs4_schedule_stateid_recovery(server, state);
4674                                 err = 0;
4675                                 goto out;
4676                         case -EKEYEXPIRED:
4677                                 /*
4678                                  * User RPCSEC_GSS context has expired.
4679                                  * We cannot recover this stateid now, so
4680                                  * skip it and allow recovery thread to
4681                                  * proceed.
4682                                  */
4683                                 err = 0;
4684                                 goto out;
4685                         case -ENOMEM:
4686                         case -NFS4ERR_DENIED:
4687                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4688                                 err = 0;
4689                                 goto out;
4690                         case -NFS4ERR_DELAY:
4691                                 break;
4692                 }
4693                 err = nfs4_handle_exception(server, err, &exception);
4694         } while (exception.retry);
4695 out:
4696         return err;
4697 }
4698
4699 static void nfs4_release_lockowner_release(void *calldata)
4700 {
4701         kfree(calldata);
4702 }
4703
4704 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4705         .rpc_release = nfs4_release_lockowner_release,
4706 };
4707
4708 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4709 {
4710         struct nfs_server *server = lsp->ls_state->owner->so_server;
4711         struct nfs_release_lockowner_args *args;
4712         struct rpc_message msg = {
4713                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4714         };
4715
4716         if (server->nfs_client->cl_mvops->minor_version != 0)
4717                 return;
4718         args = kmalloc(sizeof(*args), GFP_NOFS);
4719         if (!args)
4720                 return;
4721         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4722         args->lock_owner.id = lsp->ls_id.id;
4723         args->lock_owner.s_dev = server->s_dev;
4724         msg.rpc_argp = args;
4725         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4726 }
4727
4728 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4729
4730 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4731                                    const void *buf, size_t buflen,
4732                                    int flags, int type)
4733 {
4734         if (strcmp(key, "") != 0)
4735                 return -EINVAL;
4736
4737         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4738 }
4739
4740 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4741                                    void *buf, size_t buflen, int type)
4742 {
4743         if (strcmp(key, "") != 0)
4744                 return -EINVAL;
4745
4746         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4747 }
4748
4749 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4750                                        size_t list_len, const char *name,
4751                                        size_t name_len, int type)
4752 {
4753         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4754
4755         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4756                 return 0;
4757
4758         if (list && len <= list_len)
4759                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4760         return len;
4761 }
4762
4763 /*
4764  * nfs_fhget will use either the mounted_on_fileid or the fileid
4765  */
4766 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4767 {
4768         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4769                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4770               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4771               (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4772                 return;
4773
4774         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4775                 NFS_ATTR_FATTR_NLINK;
4776         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4777         fattr->nlink = 2;
4778 }
4779
4780 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4781                 struct nfs4_fs_locations *fs_locations, struct page *page)
4782 {
4783         struct nfs_server *server = NFS_SERVER(dir);
4784         u32 bitmask[2] = {
4785                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4786         };
4787         struct nfs4_fs_locations_arg args = {
4788                 .dir_fh = NFS_FH(dir),
4789                 .name = name,
4790                 .page = page,
4791                 .bitmask = bitmask,
4792         };
4793         struct nfs4_fs_locations_res res = {
4794                 .fs_locations = fs_locations,
4795         };
4796         struct rpc_message msg = {
4797                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4798                 .rpc_argp = &args,
4799                 .rpc_resp = &res,
4800         };
4801         int status;
4802
4803         dprintk("%s: start\n", __func__);
4804
4805         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4806          * is not supported */
4807         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4808                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4809         else
4810                 bitmask[0] |= FATTR4_WORD0_FILEID;
4811
4812         nfs_fattr_init(&fs_locations->fattr);
4813         fs_locations->server = server;
4814         fs_locations->nlocations = 0;
4815         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4816         dprintk("%s: returned status = %d\n", __func__, status);
4817         return status;
4818 }
4819
4820 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4821 {
4822         int status;
4823         struct nfs4_secinfo_arg args = {
4824                 .dir_fh = NFS_FH(dir),
4825                 .name   = name,
4826         };
4827         struct nfs4_secinfo_res res = {
4828                 .flavors     = flavors,
4829         };
4830         struct rpc_message msg = {
4831                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4832                 .rpc_argp = &args,
4833                 .rpc_resp = &res,
4834         };
4835
4836         dprintk("NFS call  secinfo %s\n", name->name);
4837         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4838         dprintk("NFS reply  secinfo: %d\n", status);
4839         return status;
4840 }
4841
4842 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4843 {
4844         struct nfs4_exception exception = { };
4845         int err;
4846         do {
4847                 err = nfs4_handle_exception(NFS_SERVER(dir),
4848                                 _nfs4_proc_secinfo(dir, name, flavors),
4849                                 &exception);
4850         } while (exception.retry);
4851         return err;
4852 }
4853
4854 #ifdef CONFIG_NFS_V4_1
4855 /*
4856  * Check the exchange flags returned by the server for invalid flags, having
4857  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4858  * DS flags set.
4859  */
4860 static int nfs4_check_cl_exchange_flags(u32 flags)
4861 {
4862         if (flags & ~EXCHGID4_FLAG_MASK_R)
4863                 goto out_inval;
4864         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4865             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4866                 goto out_inval;
4867         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4868                 goto out_inval;
4869         return NFS_OK;
4870 out_inval:
4871         return -NFS4ERR_INVAL;
4872 }
4873
4874 static bool
4875 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
4876 {
4877         if (a->server_scope_sz == b->server_scope_sz &&
4878             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
4879                 return true;
4880
4881         return false;
4882 }
4883
4884 /*
4885  * nfs4_proc_exchange_id()
4886  *
4887  * Since the clientid has expired, all compounds using sessions
4888  * associated with the stale clientid will be returning
4889  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4890  * be in some phase of session reset.
4891  */
4892 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4893 {
4894         nfs4_verifier verifier;
4895         struct nfs41_exchange_id_args args = {
4896                 .client = clp,
4897                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4898         };
4899         struct nfs41_exchange_id_res res = {
4900                 .client = clp,
4901         };
4902         int status;
4903         struct rpc_message msg = {
4904                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4905                 .rpc_argp = &args,
4906                 .rpc_resp = &res,
4907                 .rpc_cred = cred,
4908         };
4909         __be32 *p;
4910
4911         dprintk("--> %s\n", __func__);
4912         BUG_ON(clp == NULL);
4913
4914         p = (u32 *)verifier.data;
4915         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4916         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4917         args.verifier = &verifier;
4918
4919         args.id_len = scnprintf(args.id, sizeof(args.id),
4920                                 "%s/%s.%s/%u",
4921                                 clp->cl_ipaddr,
4922                                 init_utsname()->nodename,
4923                                 init_utsname()->domainname,
4924                                 clp->cl_rpcclient->cl_auth->au_flavor);
4925
4926         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
4927         if (unlikely(!res.server_scope)) {
4928                 status = -ENOMEM;
4929                 goto out;
4930         }
4931
4932         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4933         if (!status)
4934                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4935
4936         if (!status) {
4937                 if (clp->server_scope &&
4938                     !nfs41_same_server_scope(clp->server_scope,
4939                                              res.server_scope)) {
4940                         dprintk("%s: server_scope mismatch detected\n",
4941                                 __func__);
4942                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
4943                         kfree(clp->server_scope);
4944                         clp->server_scope = NULL;
4945                 }
4946
4947                 if (!clp->server_scope) {
4948                         clp->server_scope = res.server_scope;
4949                         goto out;
4950                 }
4951         }
4952         kfree(res.server_scope);
4953 out:
4954         dprintk("<-- %s status= %d\n", __func__, status);
4955         return status;
4956 }
4957
4958 struct nfs4_get_lease_time_data {
4959         struct nfs4_get_lease_time_args *args;
4960         struct nfs4_get_lease_time_res *res;
4961         struct nfs_client *clp;
4962 };
4963
4964 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4965                                         void *calldata)
4966 {
4967         int ret;
4968         struct nfs4_get_lease_time_data *data =
4969                         (struct nfs4_get_lease_time_data *)calldata;
4970
4971         dprintk("--> %s\n", __func__);
4972         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4973         /* just setup sequence, do not trigger session recovery
4974            since we're invoked within one */
4975         ret = nfs41_setup_sequence(data->clp->cl_session,
4976                                    &data->args->la_seq_args,
4977                                    &data->res->lr_seq_res, 0, task);
4978
4979         BUG_ON(ret == -EAGAIN);
4980         rpc_call_start(task);
4981         dprintk("<-- %s\n", __func__);
4982 }
4983
4984 /*
4985  * Called from nfs4_state_manager thread for session setup, so don't recover
4986  * from sequence operation or clientid errors.
4987  */
4988 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4989 {
4990         struct nfs4_get_lease_time_data *data =
4991                         (struct nfs4_get_lease_time_data *)calldata;
4992
4993         dprintk("--> %s\n", __func__);
4994         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
4995                 return;
4996         switch (task->tk_status) {
4997         case -NFS4ERR_DELAY:
4998         case -NFS4ERR_GRACE:
4999                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5000                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5001                 task->tk_status = 0;
5002                 /* fall through */
5003         case -NFS4ERR_RETRY_UNCACHED_REP:
5004                 rpc_restart_call_prepare(task);
5005                 return;
5006         }
5007         dprintk("<-- %s\n", __func__);
5008 }
5009
5010 struct rpc_call_ops nfs4_get_lease_time_ops = {
5011         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5012         .rpc_call_done = nfs4_get_lease_time_done,
5013 };
5014
5015 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5016 {
5017         struct rpc_task *task;
5018         struct nfs4_get_lease_time_args args;
5019         struct nfs4_get_lease_time_res res = {
5020                 .lr_fsinfo = fsinfo,
5021         };
5022         struct nfs4_get_lease_time_data data = {
5023                 .args = &args,
5024                 .res = &res,
5025                 .clp = clp,
5026         };
5027         struct rpc_message msg = {
5028                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5029                 .rpc_argp = &args,
5030                 .rpc_resp = &res,
5031         };
5032         struct rpc_task_setup task_setup = {
5033                 .rpc_client = clp->cl_rpcclient,
5034                 .rpc_message = &msg,
5035                 .callback_ops = &nfs4_get_lease_time_ops,
5036                 .callback_data = &data,
5037                 .flags = RPC_TASK_TIMEOUT,
5038         };
5039         int status;
5040
5041         dprintk("--> %s\n", __func__);
5042         task = rpc_run_task(&task_setup);
5043
5044         if (IS_ERR(task))
5045                 status = PTR_ERR(task);
5046         else {
5047                 status = task->tk_status;
5048                 rpc_put_task(task);
5049         }
5050         dprintk("<-- %s return %d\n", __func__, status);
5051
5052         return status;
5053 }
5054
5055 /*
5056  * Reset a slot table
5057  */
5058 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5059                                  int ivalue)
5060 {
5061         struct nfs4_slot *new = NULL;
5062         int i;
5063         int ret = 0;
5064
5065         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5066                 max_reqs, tbl->max_slots);
5067
5068         /* Does the newly negotiated max_reqs match the existing slot table? */
5069         if (max_reqs != tbl->max_slots) {
5070                 ret = -ENOMEM;
5071                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
5072                               GFP_NOFS);
5073                 if (!new)
5074                         goto out;
5075                 ret = 0;
5076                 kfree(tbl->slots);
5077         }
5078         spin_lock(&tbl->slot_tbl_lock);
5079         if (new) {
5080                 tbl->slots = new;
5081                 tbl->max_slots = max_reqs;
5082         }
5083         for (i = 0; i < tbl->max_slots; ++i)
5084                 tbl->slots[i].seq_nr = ivalue;
5085         spin_unlock(&tbl->slot_tbl_lock);
5086         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5087                 tbl, tbl->slots, tbl->max_slots);
5088 out:
5089         dprintk("<-- %s: return %d\n", __func__, ret);
5090         return ret;
5091 }
5092
5093 /*
5094  * Reset the forechannel and backchannel slot tables
5095  */
5096 static int nfs4_reset_slot_tables(struct nfs4_session *session)
5097 {
5098         int status;
5099
5100         status = nfs4_reset_slot_table(&session->fc_slot_table,
5101                         session->fc_attrs.max_reqs, 1);
5102         if (status)
5103                 return status;
5104
5105         status = nfs4_reset_slot_table(&session->bc_slot_table,
5106                         session->bc_attrs.max_reqs, 0);
5107         return status;
5108 }
5109
5110 /* Destroy the slot table */
5111 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5112 {
5113         if (session->fc_slot_table.slots != NULL) {
5114                 kfree(session->fc_slot_table.slots);
5115                 session->fc_slot_table.slots = NULL;
5116         }
5117         if (session->bc_slot_table.slots != NULL) {
5118                 kfree(session->bc_slot_table.slots);
5119                 session->bc_slot_table.slots = NULL;
5120         }
5121         return;
5122 }
5123
5124 /*
5125  * Initialize slot table
5126  */
5127 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
5128                 int max_slots, int ivalue)
5129 {
5130         struct nfs4_slot *slot;
5131         int ret = -ENOMEM;
5132
5133         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
5134
5135         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
5136
5137         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
5138         if (!slot)
5139                 goto out;
5140         ret = 0;
5141
5142         spin_lock(&tbl->slot_tbl_lock);
5143         tbl->max_slots = max_slots;
5144         tbl->slots = slot;
5145         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5146         spin_unlock(&tbl->slot_tbl_lock);
5147         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5148                 tbl, tbl->slots, tbl->max_slots);
5149 out:
5150         dprintk("<-- %s: return %d\n", __func__, ret);
5151         return ret;
5152 }
5153
5154 /*
5155  * Initialize the forechannel and backchannel tables
5156  */
5157 static int nfs4_init_slot_tables(struct nfs4_session *session)
5158 {
5159         struct nfs4_slot_table *tbl;
5160         int status = 0;
5161
5162         tbl = &session->fc_slot_table;
5163         if (tbl->slots == NULL) {
5164                 status = nfs4_init_slot_table(tbl,
5165                                 session->fc_attrs.max_reqs, 1);
5166                 if (status)
5167                         return status;
5168         }
5169
5170         tbl = &session->bc_slot_table;
5171         if (tbl->slots == NULL) {
5172                 status = nfs4_init_slot_table(tbl,
5173                                 session->bc_attrs.max_reqs, 0);
5174                 if (status)
5175                         nfs4_destroy_slot_tables(session);
5176         }
5177
5178         return status;
5179 }
5180
5181 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5182 {
5183         struct nfs4_session *session;
5184         struct nfs4_slot_table *tbl;
5185
5186         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5187         if (!session)
5188                 return NULL;
5189
5190         tbl = &session->fc_slot_table;
5191         tbl->highest_used_slotid = -1;
5192         spin_lock_init(&tbl->slot_tbl_lock);
5193         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5194         init_completion(&tbl->complete);
5195
5196         tbl = &session->bc_slot_table;
5197         tbl->highest_used_slotid = -1;
5198         spin_lock_init(&tbl->slot_tbl_lock);
5199         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5200         init_completion(&tbl->complete);
5201
5202         session->session_state = 1<<NFS4_SESSION_INITING;
5203
5204         session->clp = clp;
5205         return session;
5206 }
5207
5208 void nfs4_destroy_session(struct nfs4_session *session)
5209 {
5210         nfs4_proc_destroy_session(session);
5211         dprintk("%s Destroy backchannel for xprt %p\n",
5212                 __func__, session->clp->cl_rpcclient->cl_xprt);
5213         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
5214                                 NFS41_BC_MIN_CALLBACKS);
5215         nfs4_destroy_slot_tables(session);
5216         kfree(session);
5217 }
5218
5219 /*
5220  * Initialize the values to be used by the client in CREATE_SESSION
5221  * If nfs4_init_session set the fore channel request and response sizes,
5222  * use them.
5223  *
5224  * Set the back channel max_resp_sz_cached to zero to force the client to
5225  * always set csa_cachethis to FALSE because the current implementation
5226  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5227  */
5228 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5229 {
5230         struct nfs4_session *session = args->client->cl_session;
5231         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5232                      mxresp_sz = session->fc_attrs.max_resp_sz;
5233
5234         if (mxrqst_sz == 0)
5235                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5236         if (mxresp_sz == 0)
5237                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5238         /* Fore channel attributes */
5239         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5240         args->fc_attrs.max_resp_sz = mxresp_sz;
5241         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5242         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
5243
5244         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5245                 "max_ops=%u max_reqs=%u\n",
5246                 __func__,
5247                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5248                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5249
5250         /* Back channel attributes */
5251         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5252         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5253         args->bc_attrs.max_resp_sz_cached = 0;
5254         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5255         args->bc_attrs.max_reqs = 1;
5256
5257         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5258                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5259                 __func__,
5260                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5261                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5262                 args->bc_attrs.max_reqs);
5263 }
5264
5265 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5266 {
5267         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5268         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5269
5270         if (rcvd->max_resp_sz > sent->max_resp_sz)
5271                 return -EINVAL;
5272         /*
5273          * Our requested max_ops is the minimum we need; we're not
5274          * prepared to break up compounds into smaller pieces than that.
5275          * So, no point even trying to continue if the server won't
5276          * cooperate:
5277          */
5278         if (rcvd->max_ops < sent->max_ops)
5279                 return -EINVAL;
5280         if (rcvd->max_reqs == 0)
5281                 return -EINVAL;
5282         return 0;
5283 }
5284
5285 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5286 {
5287         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5288         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5289
5290         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5291                 return -EINVAL;
5292         if (rcvd->max_resp_sz < sent->max_resp_sz)
5293                 return -EINVAL;
5294         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5295                 return -EINVAL;
5296         /* These would render the backchannel useless: */
5297         if (rcvd->max_ops  == 0)
5298                 return -EINVAL;
5299         if (rcvd->max_reqs == 0)
5300                 return -EINVAL;
5301         return 0;
5302 }
5303
5304 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5305                                      struct nfs4_session *session)
5306 {
5307         int ret;
5308
5309         ret = nfs4_verify_fore_channel_attrs(args, session);
5310         if (ret)
5311                 return ret;
5312         return nfs4_verify_back_channel_attrs(args, session);
5313 }
5314
5315 static int _nfs4_proc_create_session(struct nfs_client *clp)
5316 {
5317         struct nfs4_session *session = clp->cl_session;
5318         struct nfs41_create_session_args args = {
5319                 .client = clp,
5320                 .cb_program = NFS4_CALLBACK,
5321         };
5322         struct nfs41_create_session_res res = {
5323                 .client = clp,
5324         };
5325         struct rpc_message msg = {
5326                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5327                 .rpc_argp = &args,
5328                 .rpc_resp = &res,
5329         };
5330         int status;
5331
5332         nfs4_init_channel_attrs(&args);
5333         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5334
5335         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5336
5337         if (!status)
5338                 /* Verify the session's negotiated channel_attrs values */
5339                 status = nfs4_verify_channel_attrs(&args, session);
5340         if (!status) {
5341                 /* Increment the clientid slot sequence id */
5342                 clp->cl_seqid++;
5343         }
5344
5345         return status;
5346 }
5347
5348 /*
5349  * Issues a CREATE_SESSION operation to the server.
5350  * It is the responsibility of the caller to verify the session is
5351  * expired before calling this routine.
5352  */
5353 int nfs4_proc_create_session(struct nfs_client *clp)
5354 {
5355         int status;
5356         unsigned *ptr;
5357         struct nfs4_session *session = clp->cl_session;
5358
5359         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5360
5361         status = _nfs4_proc_create_session(clp);
5362         if (status)
5363                 goto out;
5364
5365         /* Init and reset the fore channel */
5366         status = nfs4_init_slot_tables(session);
5367         dprintk("slot table initialization returned %d\n", status);
5368         if (status)
5369                 goto out;
5370         status = nfs4_reset_slot_tables(session);
5371         dprintk("slot table reset returned %d\n", status);
5372         if (status)
5373                 goto out;
5374
5375         ptr = (unsigned *)&session->sess_id.data[0];
5376         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5377                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5378 out:
5379         dprintk("<-- %s\n", __func__);
5380         return status;
5381 }
5382
5383 /*
5384  * Issue the over-the-wire RPC DESTROY_SESSION.
5385  * The caller must serialize access to this routine.
5386  */
5387 int nfs4_proc_destroy_session(struct nfs4_session *session)
5388 {
5389         int status = 0;
5390         struct rpc_message msg;
5391
5392         dprintk("--> nfs4_proc_destroy_session\n");
5393
5394         /* session is still being setup */
5395         if (session->clp->cl_cons_state != NFS_CS_READY)
5396                 return status;
5397
5398         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5399         msg.rpc_argp = session;
5400         msg.rpc_resp = NULL;
5401         msg.rpc_cred = NULL;
5402         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5403
5404         if (status)
5405                 printk(KERN_WARNING
5406                         "Got error %d from the server on DESTROY_SESSION. "
5407                         "Session has been destroyed regardless...\n", status);
5408
5409         dprintk("<-- nfs4_proc_destroy_session\n");
5410         return status;
5411 }
5412
5413 int nfs4_init_session(struct nfs_server *server)
5414 {
5415         struct nfs_client *clp = server->nfs_client;
5416         struct nfs4_session *session;
5417         unsigned int rsize, wsize;
5418         int ret;
5419
5420         if (!nfs4_has_session(clp))
5421                 return 0;
5422
5423         session = clp->cl_session;
5424         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5425                 return 0;
5426
5427         rsize = server->rsize;
5428         if (rsize == 0)
5429                 rsize = NFS_MAX_FILE_IO_SIZE;
5430         wsize = server->wsize;
5431         if (wsize == 0)
5432                 wsize = NFS_MAX_FILE_IO_SIZE;
5433
5434         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5435         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5436
5437         ret = nfs4_recover_expired_lease(server);
5438         if (!ret)
5439                 ret = nfs4_check_client_ready(clp);
5440         return ret;
5441 }
5442
5443 int nfs4_init_ds_session(struct nfs_client *clp)
5444 {
5445         struct nfs4_session *session = clp->cl_session;
5446         int ret;
5447
5448         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5449                 return 0;
5450
5451         ret = nfs4_client_recover_expired_lease(clp);
5452         if (!ret)
5453                 /* Test for the DS role */
5454                 if (!is_ds_client(clp))
5455                         ret = -ENODEV;
5456         if (!ret)
5457                 ret = nfs4_check_client_ready(clp);
5458         return ret;
5459
5460 }
5461 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5462
5463
5464 /*
5465  * Renew the cl_session lease.
5466  */
5467 struct nfs4_sequence_data {
5468         struct nfs_client *clp;
5469         struct nfs4_sequence_args args;
5470         struct nfs4_sequence_res res;
5471 };
5472
5473 static void nfs41_sequence_release(void *data)
5474 {
5475         struct nfs4_sequence_data *calldata = data;
5476         struct nfs_client *clp = calldata->clp;
5477
5478         if (atomic_read(&clp->cl_count) > 1)
5479                 nfs4_schedule_state_renewal(clp);
5480         nfs_put_client(clp);
5481         kfree(calldata);
5482 }
5483
5484 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5485 {
5486         switch(task->tk_status) {
5487         case -NFS4ERR_DELAY:
5488                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5489                 return -EAGAIN;
5490         default:
5491                 nfs4_schedule_lease_recovery(clp);
5492         }
5493         return 0;
5494 }
5495
5496 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5497 {
5498         struct nfs4_sequence_data *calldata = data;
5499         struct nfs_client *clp = calldata->clp;
5500
5501         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5502                 return;
5503
5504         if (task->tk_status < 0) {
5505                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5506                 if (atomic_read(&clp->cl_count) == 1)
5507                         goto out;
5508
5509                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5510                         rpc_restart_call_prepare(task);
5511                         return;
5512                 }
5513         }
5514         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5515 out:
5516         dprintk("<-- %s\n", __func__);
5517 }
5518
5519 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5520 {
5521         struct nfs4_sequence_data *calldata = data;
5522         struct nfs_client *clp = calldata->clp;
5523         struct nfs4_sequence_args *args;
5524         struct nfs4_sequence_res *res;
5525
5526         args = task->tk_msg.rpc_argp;
5527         res = task->tk_msg.rpc_resp;
5528
5529         if (nfs41_setup_sequence(clp->cl_session, args, res, 0, task))
5530                 return;
5531         rpc_call_start(task);
5532 }
5533
5534 static const struct rpc_call_ops nfs41_sequence_ops = {
5535         .rpc_call_done = nfs41_sequence_call_done,
5536         .rpc_call_prepare = nfs41_sequence_prepare,
5537         .rpc_release = nfs41_sequence_release,
5538 };
5539
5540 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5541 {
5542         struct nfs4_sequence_data *calldata;
5543         struct rpc_message msg = {
5544                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5545                 .rpc_cred = cred,
5546         };
5547         struct rpc_task_setup task_setup_data = {
5548                 .rpc_client = clp->cl_rpcclient,
5549                 .rpc_message = &msg,
5550                 .callback_ops = &nfs41_sequence_ops,
5551                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5552         };
5553
5554         if (!atomic_inc_not_zero(&clp->cl_count))
5555                 return ERR_PTR(-EIO);
5556         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5557         if (calldata == NULL) {
5558                 nfs_put_client(clp);
5559                 return ERR_PTR(-ENOMEM);
5560         }
5561         msg.rpc_argp = &calldata->args;
5562         msg.rpc_resp = &calldata->res;
5563         calldata->clp = clp;
5564         task_setup_data.callback_data = calldata;
5565
5566         return rpc_run_task(&task_setup_data);
5567 }
5568
5569 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5570 {
5571         struct rpc_task *task;
5572         int ret = 0;
5573
5574         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5575                 return 0;
5576         task = _nfs41_proc_sequence(clp, cred);
5577         if (IS_ERR(task))
5578                 ret = PTR_ERR(task);
5579         else
5580                 rpc_put_task_async(task);
5581         dprintk("<-- %s status=%d\n", __func__, ret);
5582         return ret;
5583 }
5584
5585 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5586 {
5587         struct rpc_task *task;
5588         int ret;
5589
5590         task = _nfs41_proc_sequence(clp, cred);
5591         if (IS_ERR(task)) {
5592                 ret = PTR_ERR(task);
5593                 goto out;
5594         }
5595         ret = rpc_wait_for_completion_task(task);
5596         if (!ret) {
5597                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5598
5599                 if (task->tk_status == 0)
5600                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5601                 ret = task->tk_status;
5602         }
5603         rpc_put_task(task);
5604 out:
5605         dprintk("<-- %s status=%d\n", __func__, ret);
5606         return ret;
5607 }
5608
5609 struct nfs4_reclaim_complete_data {
5610         struct nfs_client *clp;
5611         struct nfs41_reclaim_complete_args arg;
5612         struct nfs41_reclaim_complete_res res;
5613 };
5614
5615 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5616 {
5617         struct nfs4_reclaim_complete_data *calldata = data;
5618
5619         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5620         if (nfs41_setup_sequence(calldata->clp->cl_session,
5621                                 &calldata->arg.seq_args,
5622                                 &calldata->res.seq_res, 0, task))
5623                 return;
5624
5625         rpc_call_start(task);
5626 }
5627
5628 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5629 {
5630         switch(task->tk_status) {
5631         case 0:
5632         case -NFS4ERR_COMPLETE_ALREADY:
5633         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5634                 break;
5635         case -NFS4ERR_DELAY:
5636                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5637                 /* fall through */
5638         case -NFS4ERR_RETRY_UNCACHED_REP:
5639                 return -EAGAIN;
5640         default:
5641                 nfs4_schedule_lease_recovery(clp);
5642         }
5643         return 0;
5644 }
5645
5646 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5647 {
5648         struct nfs4_reclaim_complete_data *calldata = data;
5649         struct nfs_client *clp = calldata->clp;
5650         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5651
5652         dprintk("--> %s\n", __func__);
5653         if (!nfs41_sequence_done(task, res))
5654                 return;
5655
5656         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5657                 rpc_restart_call_prepare(task);
5658                 return;
5659         }
5660         dprintk("<-- %s\n", __func__);
5661 }
5662
5663 static void nfs4_free_reclaim_complete_data(void *data)
5664 {
5665         struct nfs4_reclaim_complete_data *calldata = data;
5666
5667         kfree(calldata);
5668 }
5669
5670 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5671         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5672         .rpc_call_done = nfs4_reclaim_complete_done,
5673         .rpc_release = nfs4_free_reclaim_complete_data,
5674 };
5675
5676 /*
5677  * Issue a global reclaim complete.
5678  */
5679 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5680 {
5681         struct nfs4_reclaim_complete_data *calldata;
5682         struct rpc_task *task;
5683         struct rpc_message msg = {
5684                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5685         };
5686         struct rpc_task_setup task_setup_data = {
5687                 .rpc_client = clp->cl_rpcclient,
5688                 .rpc_message = &msg,
5689                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5690                 .flags = RPC_TASK_ASYNC,
5691         };
5692         int status = -ENOMEM;
5693
5694         dprintk("--> %s\n", __func__);
5695         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5696         if (calldata == NULL)
5697                 goto out;
5698         calldata->clp = clp;
5699         calldata->arg.one_fs = 0;
5700
5701         msg.rpc_argp = &calldata->arg;
5702         msg.rpc_resp = &calldata->res;
5703         task_setup_data.callback_data = calldata;
5704         task = rpc_run_task(&task_setup_data);
5705         if (IS_ERR(task)) {
5706                 status = PTR_ERR(task);
5707                 goto out;
5708         }
5709         status = nfs4_wait_for_completion_rpc_task(task);
5710         if (status == 0)
5711                 status = task->tk_status;
5712         rpc_put_task(task);
5713         return 0;
5714 out:
5715         dprintk("<-- %s status=%d\n", __func__, status);
5716         return status;
5717 }
5718
5719 static void
5720 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5721 {
5722         struct nfs4_layoutget *lgp = calldata;
5723         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5724
5725         dprintk("--> %s\n", __func__);
5726         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5727          * right now covering the LAYOUTGET we are about to send.
5728          * However, that is not so catastrophic, and there seems
5729          * to be no way to prevent it completely.
5730          */
5731         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5732                                 &lgp->res.seq_res, 0, task))
5733                 return;
5734         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5735                                           NFS_I(lgp->args.inode)->layout,
5736                                           lgp->args.ctx->state)) {
5737                 rpc_exit(task, NFS4_OK);
5738                 return;
5739         }
5740         rpc_call_start(task);
5741 }
5742
5743 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5744 {
5745         struct nfs4_layoutget *lgp = calldata;
5746         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5747
5748         dprintk("--> %s\n", __func__);
5749
5750         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5751                 return;
5752
5753         switch (task->tk_status) {
5754         case 0:
5755                 break;
5756         case -NFS4ERR_LAYOUTTRYLATER:
5757         case -NFS4ERR_RECALLCONFLICT:
5758                 task->tk_status = -NFS4ERR_DELAY;
5759                 /* Fall through */
5760         default:
5761                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5762                         rpc_restart_call_prepare(task);
5763                         return;
5764                 }
5765         }
5766         dprintk("<-- %s\n", __func__);
5767 }
5768
5769 static void nfs4_layoutget_release(void *calldata)
5770 {
5771         struct nfs4_layoutget *lgp = calldata;
5772
5773         dprintk("--> %s\n", __func__);
5774         put_nfs_open_context(lgp->args.ctx);
5775         kfree(calldata);
5776         dprintk("<-- %s\n", __func__);
5777 }
5778
5779 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5780         .rpc_call_prepare = nfs4_layoutget_prepare,
5781         .rpc_call_done = nfs4_layoutget_done,
5782         .rpc_release = nfs4_layoutget_release,
5783 };
5784
5785 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5786 {
5787         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5788         struct rpc_task *task;
5789         struct rpc_message msg = {
5790                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5791                 .rpc_argp = &lgp->args,
5792                 .rpc_resp = &lgp->res,
5793         };
5794         struct rpc_task_setup task_setup_data = {
5795                 .rpc_client = server->client,
5796                 .rpc_message = &msg,
5797                 .callback_ops = &nfs4_layoutget_call_ops,
5798                 .callback_data = lgp,
5799                 .flags = RPC_TASK_ASYNC,
5800         };
5801         int status = 0;
5802
5803         dprintk("--> %s\n", __func__);
5804
5805         lgp->res.layoutp = &lgp->args.layout;
5806         lgp->res.seq_res.sr_slot = NULL;
5807         task = rpc_run_task(&task_setup_data);
5808         if (IS_ERR(task))
5809                 return PTR_ERR(task);
5810         status = nfs4_wait_for_completion_rpc_task(task);
5811         if (status == 0)
5812                 status = task->tk_status;
5813         if (status == 0)
5814                 status = pnfs_layout_process(lgp);
5815         rpc_put_task(task);
5816         dprintk("<-- %s status=%d\n", __func__, status);
5817         return status;
5818 }
5819
5820 static void
5821 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5822 {
5823         struct nfs4_layoutreturn *lrp = calldata;
5824
5825         dprintk("--> %s\n", __func__);
5826         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5827                                 &lrp->res.seq_res, 0, task))
5828                 return;
5829         rpc_call_start(task);
5830 }
5831
5832 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5833 {
5834         struct nfs4_layoutreturn *lrp = calldata;
5835         struct nfs_server *server;
5836         struct pnfs_layout_hdr *lo = lrp->args.layout;
5837
5838         dprintk("--> %s\n", __func__);
5839
5840         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5841                 return;
5842
5843         server = NFS_SERVER(lrp->args.inode);
5844         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5845                 rpc_restart_call_prepare(task);
5846                 return;
5847         }
5848         spin_lock(&lo->plh_inode->i_lock);
5849         if (task->tk_status == 0) {
5850                 if (lrp->res.lrs_present) {
5851                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5852                 } else
5853                         BUG_ON(!list_empty(&lo->plh_segs));
5854         }
5855         lo->plh_block_lgets--;
5856         spin_unlock(&lo->plh_inode->i_lock);
5857         dprintk("<-- %s\n", __func__);
5858 }
5859
5860 static void nfs4_layoutreturn_release(void *calldata)
5861 {
5862         struct nfs4_layoutreturn *lrp = calldata;
5863
5864         dprintk("--> %s\n", __func__);
5865         put_layout_hdr(lrp->args.layout);
5866         kfree(calldata);
5867         dprintk("<-- %s\n", __func__);
5868 }
5869
5870 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
5871         .rpc_call_prepare = nfs4_layoutreturn_prepare,
5872         .rpc_call_done = nfs4_layoutreturn_done,
5873         .rpc_release = nfs4_layoutreturn_release,
5874 };
5875
5876 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
5877 {
5878         struct rpc_task *task;
5879         struct rpc_message msg = {
5880                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
5881                 .rpc_argp = &lrp->args,
5882                 .rpc_resp = &lrp->res,
5883         };
5884         struct rpc_task_setup task_setup_data = {
5885                 .rpc_client = lrp->clp->cl_rpcclient,
5886                 .rpc_message = &msg,
5887                 .callback_ops = &nfs4_layoutreturn_call_ops,
5888                 .callback_data = lrp,
5889         };
5890         int status;
5891
5892         dprintk("--> %s\n", __func__);
5893         task = rpc_run_task(&task_setup_data);
5894         if (IS_ERR(task))
5895                 return PTR_ERR(task);
5896         status = task->tk_status;
5897         dprintk("<-- %s status=%d\n", __func__, status);
5898         rpc_put_task(task);
5899         return status;
5900 }
5901
5902 /*
5903  * Retrieve the list of Data Server devices from the MDS.
5904  */
5905 static int _nfs4_getdevicelist(struct nfs_server *server,
5906                                     const struct nfs_fh *fh,
5907                                     struct pnfs_devicelist *devlist)
5908 {
5909         struct nfs4_getdevicelist_args args = {
5910                 .fh = fh,
5911                 .layoutclass = server->pnfs_curr_ld->id,
5912         };
5913         struct nfs4_getdevicelist_res res = {
5914                 .devlist = devlist,
5915         };
5916         struct rpc_message msg = {
5917                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
5918                 .rpc_argp = &args,
5919                 .rpc_resp = &res,
5920         };
5921         int status;
5922
5923         dprintk("--> %s\n", __func__);
5924         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5925                                 &res.seq_res, 0);
5926         dprintk("<-- %s status=%d\n", __func__, status);
5927         return status;
5928 }
5929
5930 int nfs4_proc_getdevicelist(struct nfs_server *server,
5931                             const struct nfs_fh *fh,
5932                             struct pnfs_devicelist *devlist)
5933 {
5934         struct nfs4_exception exception = { };
5935         int err;
5936
5937         do {
5938                 err = nfs4_handle_exception(server,
5939                                 _nfs4_getdevicelist(server, fh, devlist),
5940                                 &exception);
5941         } while (exception.retry);
5942
5943         dprintk("%s: err=%d, num_devs=%u\n", __func__,
5944                 err, devlist->num_devs);
5945
5946         return err;
5947 }
5948 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
5949
5950 static int
5951 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5952 {
5953         struct nfs4_getdeviceinfo_args args = {
5954                 .pdev = pdev,
5955         };
5956         struct nfs4_getdeviceinfo_res res = {
5957                 .pdev = pdev,
5958         };
5959         struct rpc_message msg = {
5960                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5961                 .rpc_argp = &args,
5962                 .rpc_resp = &res,
5963         };
5964         int status;
5965
5966         dprintk("--> %s\n", __func__);
5967         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5968         dprintk("<-- %s status=%d\n", __func__, status);
5969
5970         return status;
5971 }
5972
5973 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5974 {
5975         struct nfs4_exception exception = { };
5976         int err;
5977
5978         do {
5979                 err = nfs4_handle_exception(server,
5980                                         _nfs4_proc_getdeviceinfo(server, pdev),
5981                                         &exception);
5982         } while (exception.retry);
5983         return err;
5984 }
5985 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5986
5987 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
5988 {
5989         struct nfs4_layoutcommit_data *data = calldata;
5990         struct nfs_server *server = NFS_SERVER(data->args.inode);
5991
5992         if (nfs4_setup_sequence(server, &data->args.seq_args,
5993                                 &data->res.seq_res, 1, task))
5994                 return;
5995         rpc_call_start(task);
5996 }
5997
5998 static void
5999 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6000 {
6001         struct nfs4_layoutcommit_data *data = calldata;
6002         struct nfs_server *server = NFS_SERVER(data->args.inode);
6003
6004         if (!nfs4_sequence_done(task, &data->res.seq_res))
6005                 return;
6006
6007         switch (task->tk_status) { /* Just ignore these failures */
6008         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6009         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6010         case -NFS4ERR_BADLAYOUT:     /* no layout */
6011         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6012                 task->tk_status = 0;
6013                 break;
6014         case 0:
6015                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6016                                                    data->res.fattr);
6017                 break;
6018         default:
6019                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6020                         rpc_restart_call_prepare(task);
6021                         return;
6022                 }
6023         }
6024 }
6025
6026 static void nfs4_layoutcommit_release(void *calldata)
6027 {
6028         struct nfs4_layoutcommit_data *data = calldata;
6029         struct pnfs_layout_segment *lseg, *tmp;
6030         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6031
6032         pnfs_cleanup_layoutcommit(data);
6033         /* Matched by references in pnfs_set_layoutcommit */
6034         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6035                 list_del_init(&lseg->pls_lc_list);
6036                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6037                                        &lseg->pls_flags))
6038                         put_lseg(lseg);
6039         }
6040
6041         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6042         smp_mb__after_clear_bit();
6043         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6044
6045         put_rpccred(data->cred);
6046         kfree(data);
6047 }
6048
6049 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6050         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6051         .rpc_call_done = nfs4_layoutcommit_done,
6052         .rpc_release = nfs4_layoutcommit_release,
6053 };
6054
6055 int
6056 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6057 {
6058         struct rpc_message msg = {
6059                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6060                 .rpc_argp = &data->args,
6061                 .rpc_resp = &data->res,
6062                 .rpc_cred = data->cred,
6063         };
6064         struct rpc_task_setup task_setup_data = {
6065                 .task = &data->task,
6066                 .rpc_client = NFS_CLIENT(data->args.inode),
6067                 .rpc_message = &msg,
6068                 .callback_ops = &nfs4_layoutcommit_ops,
6069                 .callback_data = data,
6070                 .flags = RPC_TASK_ASYNC,
6071         };
6072         struct rpc_task *task;
6073         int status = 0;
6074
6075         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6076                 "lbw: %llu inode %lu\n",
6077                 data->task.tk_pid, sync,
6078                 data->args.lastbytewritten,
6079                 data->args.inode->i_ino);
6080
6081         task = rpc_run_task(&task_setup_data);
6082         if (IS_ERR(task))
6083                 return PTR_ERR(task);
6084         if (sync == false)
6085                 goto out;
6086         status = nfs4_wait_for_completion_rpc_task(task);
6087         if (status != 0)
6088                 goto out;
6089         status = task->tk_status;
6090 out:
6091         dprintk("%s: status %d\n", __func__, status);
6092         rpc_put_task(task);
6093         return status;
6094 }
6095
6096 static int
6097 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6098                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6099 {
6100         struct nfs41_secinfo_no_name_args args = {
6101                 .style = SECINFO_STYLE_CURRENT_FH,
6102         };
6103         struct nfs4_secinfo_res res = {
6104                 .flavors = flavors,
6105         };
6106         struct rpc_message msg = {
6107                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6108                 .rpc_argp = &args,
6109                 .rpc_resp = &res,
6110         };
6111         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6112 }
6113
6114 static int
6115 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6116                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6117 {
6118         struct nfs4_exception exception = { };
6119         int err;
6120         do {
6121                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6122                 switch (err) {
6123                 case 0:
6124                 case -NFS4ERR_WRONGSEC:
6125                 case -NFS4ERR_NOTSUPP:
6126                         goto out;
6127                 default:
6128                         err = nfs4_handle_exception(server, err, &exception);
6129                 }
6130         } while (exception.retry);
6131 out:
6132         return err;
6133 }
6134
6135 static int
6136 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6137                     struct nfs_fsinfo *info)
6138 {
6139         int err;
6140         struct page *page;
6141         rpc_authflavor_t flavor;
6142         struct nfs4_secinfo_flavors *flavors;
6143
6144         page = alloc_page(GFP_KERNEL);
6145         if (!page) {
6146                 err = -ENOMEM;
6147                 goto out;
6148         }
6149
6150         flavors = page_address(page);
6151         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6152
6153         /*
6154          * Fall back on "guess and check" method if
6155          * the server doesn't support SECINFO_NO_NAME
6156          */
6157         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6158                 err = nfs4_find_root_sec(server, fhandle, info);
6159                 goto out_freepage;
6160         }
6161         if (err)
6162                 goto out_freepage;
6163
6164         flavor = nfs_find_best_sec(flavors);
6165         if (err == 0)
6166                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6167
6168 out_freepage:
6169         put_page(page);
6170         if (err == -EACCES)
6171                 return -EPERM;
6172 out:
6173         return err;
6174 }
6175 static int _nfs41_test_stateid(struct nfs_server *server, struct nfs4_state *state)
6176 {
6177         int status;
6178         struct nfs41_test_stateid_args args = {
6179                 .stateid = &state->stateid,
6180         };
6181         struct nfs41_test_stateid_res res;
6182         struct rpc_message msg = {
6183                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6184                 .rpc_argp = &args,
6185                 .rpc_resp = &res,
6186         };
6187         args.seq_args.sa_session = res.seq_res.sr_session = NULL;
6188         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 0, 1);
6189         return status;
6190 }
6191
6192 static int nfs41_test_stateid(struct nfs_server *server, struct nfs4_state *state)
6193 {
6194         struct nfs4_exception exception = { };
6195         int err;
6196         do {
6197                 err = nfs4_handle_exception(server,
6198                                 _nfs41_test_stateid(server, state),
6199                                 &exception);
6200         } while (exception.retry);
6201         return err;
6202 }
6203
6204 static int _nfs4_free_stateid(struct nfs_server *server, struct nfs4_state *state)
6205 {
6206         int status;
6207         struct nfs41_free_stateid_args args = {
6208                 .stateid = &state->stateid,
6209         };
6210         struct nfs41_free_stateid_res res;
6211         struct rpc_message msg = {
6212                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6213                 .rpc_argp = &args,
6214                 .rpc_resp = &res,
6215         };
6216
6217         args.seq_args.sa_session = res.seq_res.sr_session = NULL;
6218         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 0, 1);
6219         return status;
6220 }
6221
6222 static int nfs41_free_stateid(struct nfs_server *server, struct nfs4_state *state)
6223 {
6224         struct nfs4_exception exception = { };
6225         int err;
6226         do {
6227                 err = nfs4_handle_exception(server,
6228                                 _nfs4_free_stateid(server, state),
6229                                 &exception);
6230         } while (exception.retry);
6231         return err;
6232 }
6233 #endif /* CONFIG_NFS_V4_1 */
6234
6235 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6236         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6237         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6238         .recover_open   = nfs4_open_reclaim,
6239         .recover_lock   = nfs4_lock_reclaim,
6240         .establish_clid = nfs4_init_clientid,
6241         .get_clid_cred  = nfs4_get_setclientid_cred,
6242 };
6243
6244 #if defined(CONFIG_NFS_V4_1)
6245 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6246         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6247         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6248         .recover_open   = nfs4_open_reclaim,
6249         .recover_lock   = nfs4_lock_reclaim,
6250         .establish_clid = nfs41_init_clientid,
6251         .get_clid_cred  = nfs4_get_exchange_id_cred,
6252         .reclaim_complete = nfs41_proc_reclaim_complete,
6253 };
6254 #endif /* CONFIG_NFS_V4_1 */
6255
6256 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6257         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6258         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6259         .recover_open   = nfs4_open_expired,
6260         .recover_lock   = nfs4_lock_expired,
6261         .establish_clid = nfs4_init_clientid,
6262         .get_clid_cred  = nfs4_get_setclientid_cred,
6263 };
6264
6265 #if defined(CONFIG_NFS_V4_1)
6266 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6267         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6268         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6269         .recover_open   = nfs41_open_expired,
6270         .recover_lock   = nfs41_lock_expired,
6271         .establish_clid = nfs41_init_clientid,
6272         .get_clid_cred  = nfs4_get_exchange_id_cred,
6273 };
6274 #endif /* CONFIG_NFS_V4_1 */
6275
6276 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6277         .sched_state_renewal = nfs4_proc_async_renew,
6278         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6279         .renew_lease = nfs4_proc_renew,
6280 };
6281
6282 #if defined(CONFIG_NFS_V4_1)
6283 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6284         .sched_state_renewal = nfs41_proc_async_sequence,
6285         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6286         .renew_lease = nfs4_proc_sequence,
6287 };
6288 #endif
6289
6290 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6291         .minor_version = 0,
6292         .call_sync = _nfs4_call_sync,
6293         .validate_stateid = nfs4_validate_delegation_stateid,
6294         .find_root_sec = nfs4_find_root_sec,
6295         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6296         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6297         .state_renewal_ops = &nfs40_state_renewal_ops,
6298 };
6299
6300 #if defined(CONFIG_NFS_V4_1)
6301 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6302         .minor_version = 1,
6303         .call_sync = _nfs4_call_sync_session,
6304         .validate_stateid = nfs41_validate_delegation_stateid,
6305         .find_root_sec = nfs41_find_root_sec,
6306         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6307         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6308         .state_renewal_ops = &nfs41_state_renewal_ops,
6309 };
6310 #endif
6311
6312 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6313         [0] = &nfs_v4_0_minor_ops,
6314 #if defined(CONFIG_NFS_V4_1)
6315         [1] = &nfs_v4_1_minor_ops,
6316 #endif
6317 };
6318
6319 static const struct inode_operations nfs4_file_inode_operations = {
6320         .permission     = nfs_permission,
6321         .getattr        = nfs_getattr,
6322         .setattr        = nfs_setattr,
6323         .getxattr       = generic_getxattr,
6324         .setxattr       = generic_setxattr,
6325         .listxattr      = generic_listxattr,
6326         .removexattr    = generic_removexattr,
6327 };
6328
6329 const struct nfs_rpc_ops nfs_v4_clientops = {
6330         .version        = 4,                    /* protocol version */
6331         .dentry_ops     = &nfs4_dentry_operations,
6332         .dir_inode_ops  = &nfs4_dir_inode_operations,
6333         .file_inode_ops = &nfs4_file_inode_operations,
6334         .file_ops       = &nfs4_file_operations,
6335         .getroot        = nfs4_proc_get_root,
6336         .getattr        = nfs4_proc_getattr,
6337         .setattr        = nfs4_proc_setattr,
6338         .lookup         = nfs4_proc_lookup,
6339         .access         = nfs4_proc_access,
6340         .readlink       = nfs4_proc_readlink,
6341         .create         = nfs4_proc_create,
6342         .remove         = nfs4_proc_remove,
6343         .unlink_setup   = nfs4_proc_unlink_setup,
6344         .unlink_done    = nfs4_proc_unlink_done,
6345         .rename         = nfs4_proc_rename,
6346         .rename_setup   = nfs4_proc_rename_setup,
6347         .rename_done    = nfs4_proc_rename_done,
6348         .link           = nfs4_proc_link,
6349         .symlink        = nfs4_proc_symlink,
6350         .mkdir          = nfs4_proc_mkdir,
6351         .rmdir          = nfs4_proc_remove,
6352         .readdir        = nfs4_proc_readdir,
6353         .mknod          = nfs4_proc_mknod,
6354         .statfs         = nfs4_proc_statfs,
6355         .fsinfo         = nfs4_proc_fsinfo,
6356         .pathconf       = nfs4_proc_pathconf,
6357         .set_capabilities = nfs4_server_capabilities,
6358         .decode_dirent  = nfs4_decode_dirent,
6359         .read_setup     = nfs4_proc_read_setup,
6360         .read_done      = nfs4_read_done,
6361         .write_setup    = nfs4_proc_write_setup,
6362         .write_done     = nfs4_write_done,
6363         .commit_setup   = nfs4_proc_commit_setup,
6364         .commit_done    = nfs4_commit_done,
6365         .lock           = nfs4_proc_lock,
6366         .clear_acl_cache = nfs4_zap_acl_attr,
6367         .close_context  = nfs4_close_context,
6368         .open_context   = nfs4_atomic_open,
6369         .init_client    = nfs4_init_client,
6370         .secinfo        = nfs4_proc_secinfo,
6371 };
6372
6373 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6374         .prefix = XATTR_NAME_NFSV4_ACL,
6375         .list   = nfs4_xattr_list_nfs4_acl,
6376         .get    = nfs4_xattr_get_nfs4_acl,
6377         .set    = nfs4_xattr_set_nfs4_acl,
6378 };
6379
6380 const struct xattr_handler *nfs4_xattr_handlers[] = {
6381         &nfs4_xattr_nfs4_acl_handler,
6382         NULL
6383 };
6384
6385 /*
6386  * Local variables:
6387  *  c-basic-offset: 8
6388  * End:
6389  */