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/nfs_idmap.h>
  56 #include <linux/sunrpc/bc_xprt.h>
  57 #include <linux/xattr.h>
  58 #include <linux/utsname.h>
  59
  60 #include "nfs4_fs.h"
  61 #include "delegation.h"
  62 #include "internal.h"
  63 #include "iostat.h"
  64 #include "callback.h"
  65 #include "pnfs.h"
  66
  67 #define NFSDBG_FACILITY         NFSDBG_PROC
  68
  69 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  70 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  71
  72 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  73
  74 struct nfs4_opendata;
  75 static int _nfs4_proc_open(struct nfs4_opendata *data);
  76 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  77 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  78 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  79 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  80 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  81                             struct nfs_fattr *fattr, struct iattr *sattr,
  82                             struct nfs4_state *state);
  83 #ifdef CONFIG_NFS_V4_1
  84 static int nfs41_test_stateid(struct nfs_server *, struct nfs4_state *);
  85 static int nfs41_free_stateid(struct nfs_server *, struct nfs4_state *);
  86 #endif
  87 /* Prevent leaks of NFSv4 errors into userland */
  88 static int nfs4_map_errors(int err)
  89 {
  90         if (err >= -1000)
  91                 return err;
  92         switch (err) {
  93         case -NFS4ERR_RESOURCE:
  94                 return -EREMOTEIO;
  95         case -NFS4ERR_WRONGSEC:
  96                 return -EPERM;
  97         case -NFS4ERR_BADOWNER:
  98         case -NFS4ERR_BADNAME:
  99                 return -EINVAL;
 100         default:
 101                 dprintk("%s could not handle NFSv4 error %d\n",
 102                                 __func__, -err);
 103                 break;
 104         }
 105         return -EIO;
 106 }
 107
 108 /*
 109  * This is our standard bitmap for GETATTR requests.
 110  */
 111 const u32 nfs4_fattr_bitmap[2] = {
 112         FATTR4_WORD0_TYPE
 113         | FATTR4_WORD0_CHANGE
 114         | FATTR4_WORD0_SIZE
 115         | FATTR4_WORD0_FSID
 116         | FATTR4_WORD0_FILEID,
 117         FATTR4_WORD1_MODE
 118         | FATTR4_WORD1_NUMLINKS
 119         | FATTR4_WORD1_OWNER
 120         | FATTR4_WORD1_OWNER_GROUP
 121         | FATTR4_WORD1_RAWDEV
 122         | FATTR4_WORD1_SPACE_USED
 123         | FATTR4_WORD1_TIME_ACCESS
 124         | FATTR4_WORD1_TIME_METADATA
 125         | FATTR4_WORD1_TIME_MODIFY
 126 };
 127
 128 const u32 nfs4_statfs_bitmap[2] = {
 129         FATTR4_WORD0_FILES_AVAIL
 130         | FATTR4_WORD0_FILES_FREE
 131         | FATTR4_WORD0_FILES_TOTAL,
 132         FATTR4_WORD1_SPACE_AVAIL
 133         | FATTR4_WORD1_SPACE_FREE
 134         | FATTR4_WORD1_SPACE_TOTAL
 135 };
 136
 137 const u32 nfs4_pathconf_bitmap[2] = {
 138         FATTR4_WORD0_MAXLINK
 139         | FATTR4_WORD0_MAXNAME,
 140         0
 141 };
 142
 143 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 144                         | FATTR4_WORD0_MAXREAD
 145                         | FATTR4_WORD0_MAXWRITE
 146                         | FATTR4_WORD0_LEASE_TIME,
 147                         FATTR4_WORD1_TIME_DELTA
 148                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 149                         FATTR4_WORD2_LAYOUT_BLKSIZE
 150 };
 151
 152 const u32 nfs4_fs_locations_bitmap[2] = {
 153         FATTR4_WORD0_TYPE
 154         | FATTR4_WORD0_CHANGE
 155         | FATTR4_WORD0_SIZE
 156         | FATTR4_WORD0_FSID
 157         | FATTR4_WORD0_FILEID
 158         | FATTR4_WORD0_FS_LOCATIONS,
 159         FATTR4_WORD1_MODE
 160         | FATTR4_WORD1_NUMLINKS
 161         | FATTR4_WORD1_OWNER
 162         | FATTR4_WORD1_OWNER_GROUP
 163         | FATTR4_WORD1_RAWDEV
 164         | FATTR4_WORD1_SPACE_USED
 165         | FATTR4_WORD1_TIME_ACCESS
 166         | FATTR4_WORD1_TIME_METADATA
 167         | FATTR4_WORD1_TIME_MODIFY
 168         | FATTR4_WORD1_MOUNTED_ON_FILEID
 169 };
 170
 171 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 172                 struct nfs4_readdir_arg *readdir)
 173 {
 174         __be32 *start, *p;
 175
 176         BUG_ON(readdir->count < 80);
 177         if (cookie > 2) {
 178                 readdir->cookie = cookie;
 179                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 180                 return;
 181         }
 182
 183         readdir->cookie = 0;
 184         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 185         if (cookie == 2)
 186                 return;
 187
 188         /*
 189          * NFSv4 servers do not return entries for '.' and '..'
 190          * Therefore, we fake these entries here.  We let '.'
 191          * have cookie 0 and '..' have cookie 1.  Note that
 192          * when talking to the server, we always send cookie 0
 193          * instead of 1 or 2.
 194          */
 195         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 196
 197         if (cookie == 0) {
 198                 *p++ = xdr_one;                                  /* next */
 199                 *p++ = xdr_zero;                   /* cookie, first word */
 200                 *p++ = xdr_one;                   /* cookie, second word */
 201                 *p++ = xdr_one;                             /* entry len */
 202                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 203                 p++;
 204                 *p++ = xdr_one;                         /* bitmap length */
 205                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 206                 *p++ = htonl(8);              /* attribute buffer length */
 207                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 208         }
 209
 210         *p++ = xdr_one;                                  /* next */
 211         *p++ = xdr_zero;                   /* cookie, first word */
 212         *p++ = xdr_two;                   /* cookie, second word */
 213         *p++ = xdr_two;                             /* entry len */
 214         memcpy(p, "..\0\0", 4);                         /* entry */
 215         p++;
 216         *p++ = xdr_one;                         /* bitmap length */
 217         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 218         *p++ = htonl(8);              /* attribute buffer length */
 219         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 220
 221         readdir->pgbase = (char *)p - (char *)start;
 222         readdir->count -= readdir->pgbase;
 223         kunmap_atomic(start, KM_USER0);
 224 }
 225
 226 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 227 {
 228         int res;
 229
 230         might_sleep();
 231
 232         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 233                         nfs_wait_bit_killable, TASK_KILLABLE);
 234         return res;
 235 }
 236
 237 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 238 {
 239         int res = 0;
 240
 241         might_sleep();
 242
 243         if (*timeout <= 0)
 244                 *timeout = NFS4_POLL_RETRY_MIN;
 245         if (*timeout > NFS4_POLL_RETRY_MAX)
 246                 *timeout = NFS4_POLL_RETRY_MAX;
 247         schedule_timeout_killable(*timeout);
 248         if (fatal_signal_pending(current))
 249                 res = -ERESTARTSYS;
 250         *timeout <<= 1;
 251         return res;
 252 }
 253
 254 /* This is the error handling routine for processes that are allowed
 255  * to sleep.
 256  */
 257 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 258 {
 259         struct nfs_client *clp = server->nfs_client;
 260         struct nfs4_state *state = exception->state;
 261         int ret = errorcode;
 262
 263         exception->retry = 0;
 264         switch(errorcode) {
 265                 case 0:
 266                         return 0;
 267                 case -NFS4ERR_ADMIN_REVOKED:
 268                 case -NFS4ERR_BAD_STATEID:
 269                 case -NFS4ERR_OPENMODE:
 270                         if (state == NULL)
 271                                 break;
 272                         nfs4_schedule_stateid_recovery(server, state);
 273                         goto wait_on_recovery;
 274                 case -NFS4ERR_EXPIRED:
 275                         if (state != NULL)
 276                                 nfs4_schedule_stateid_recovery(server, state);
 277                 case -NFS4ERR_STALE_STATEID:
 278                 case -NFS4ERR_STALE_CLIENTID:
 279                         nfs4_schedule_lease_recovery(clp);
 280                         goto wait_on_recovery;
 281 #if defined(CONFIG_NFS_V4_1)
 282                 case -NFS4ERR_BADSESSION:
 283                 case -NFS4ERR_BADSLOT:
 284                 case -NFS4ERR_BAD_HIGH_SLOT:
 285                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 286                 case -NFS4ERR_DEADSESSION:
 287                 case -NFS4ERR_SEQ_FALSE_RETRY:
 288                 case -NFS4ERR_SEQ_MISORDERED:
 289                         dprintk("%s ERROR: %d Reset session\n", __func__,
 290                                 errorcode);
 291                         nfs4_schedule_session_recovery(clp->cl_session);
 292                         exception->retry = 1;
 293                         break;
 294 #endif /* defined(CONFIG_NFS_V4_1) */
 295                 case -NFS4ERR_FILE_OPEN:
 296                         if (exception->timeout > HZ) {
 297                                 /* We have retried a decent amount, time to
 298                                  * fail
 299                                  */
 300                                 ret = -EBUSY;
 301                                 break;
 302                         }
 303                 case -NFS4ERR_GRACE:
 304                 case -NFS4ERR_DELAY:
 305                 case -EKEYEXPIRED:
 306                         ret = nfs4_delay(server->client, &exception->timeout);
 307                         if (ret != 0)
 308                                 break;
 309                 case -NFS4ERR_RETRY_UNCACHED_REP:
 310                 case -NFS4ERR_OLD_STATEID:
 311                         exception->retry = 1;
 312                         break;
 313                 case -NFS4ERR_BADOWNER:
 314                         /* The following works around a Linux server bug! */
 315                 case -NFS4ERR_BADNAME:
 316                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 317                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 318                                 exception->retry = 1;
 319                                 printk(KERN_WARNING "NFS: v4 server %s "
 320                                                 "does not accept raw "
 321                                                 "uid/gids. "
 322                                                 "Reenabling the idmapper.\n",
 323                                                 server->nfs_client->cl_hostname);
 324                         }
 325         }
 326         /* We failed to handle the error */
 327         return nfs4_map_errors(ret);
 328 wait_on_recovery:
 329         ret = nfs4_wait_clnt_recover(clp);
 330         if (ret == 0)
 331                 exception->retry = 1;
 332         return ret;
 333 }
 334
 335
 336 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 337 {
 338         spin_lock(&clp->cl_lock);
 339         if (time_before(clp->cl_last_renewal,timestamp))
 340                 clp->cl_last_renewal = timestamp;
 341         spin_unlock(&clp->cl_lock);
 342 }
 343
 344 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 345 {
 346         do_renew_lease(server->nfs_client, timestamp);
 347 }
 348
 349 #if defined(CONFIG_NFS_V4_1)
 350
 351 /*
 352  * nfs4_free_slot - free a slot and efficiently update slot table.
 353  *
 354  * freeing a slot is trivially done by clearing its respective bit
 355  * in the bitmap.
 356  * If the freed slotid equals highest_used_slotid we want to update it
 357  * so that the server would be able to size down the slot table if needed,
 358  * otherwise we know that the highest_used_slotid is still in use.
 359  * When updating highest_used_slotid there may be "holes" in the bitmap
 360  * so we need to scan down from highest_used_slotid to 0 looking for the now
 361  * highest slotid in use.
 362  * If none found, highest_used_slotid is set to -1.
 363  *
 364  * Must be called while holding tbl->slot_tbl_lock
 365  */
 366 static void
 367 nfs4_free_slot(struct nfs4_slot_table *tbl, u8 free_slotid)
 368 {
 369         int slotid = free_slotid;
 370
 371         BUG_ON(slotid < 0 || slotid >= NFS4_MAX_SLOT_TABLE);
 372         /* clear used bit in bitmap */
 373         __clear_bit(slotid, tbl->used_slots);
 374
 375         /* update highest_used_slotid when it is freed */
 376         if (slotid == tbl->highest_used_slotid) {
 377                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 378                 if (slotid < tbl->max_slots)
 379                         tbl->highest_used_slotid = slotid;
 380                 else
 381                         tbl->highest_used_slotid = -1;
 382         }
 383         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
 384                 free_slotid, tbl->highest_used_slotid);
 385 }
 386
 387 /*
 388  * Signal state manager thread if session fore channel is drained
 389  */
 390 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 391 {
 392         struct rpc_task *task;
 393
 394         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 395                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
 396                 if (task)
 397                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 398                 return;
 399         }
 400
 401         if (ses->fc_slot_table.highest_used_slotid != -1)
 402                 return;
 403
 404         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 405         complete(&ses->fc_slot_table.complete);
 406 }
 407
 408 /*
 409  * Signal state manager thread if session back channel is drained
 410  */
 411 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 412 {
 413         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 414             ses->bc_slot_table.highest_used_slotid != -1)
 415                 return;
 416         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 417         complete(&ses->bc_slot_table.complete);
 418 }
 419
 420 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 421 {
 422         struct nfs4_slot_table *tbl;
 423
 424         tbl = &res->sr_session->fc_slot_table;
 425         if (!res->sr_slot) {
 426                 /* just wake up the next guy waiting since
 427                  * we may have not consumed a slot after all */
 428                 dprintk("%s: No slot\n", __func__);
 429                 return;
 430         }
 431
 432         spin_lock(&tbl->slot_tbl_lock);
 433         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 434         nfs4_check_drain_fc_complete(res->sr_session);
 435         spin_unlock(&tbl->slot_tbl_lock);
 436         res->sr_slot = NULL;
 437 }
 438
 439 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 440 {
 441         unsigned long timestamp;
 442         struct nfs_client *clp;
 443
 444         /*
 445          * sr_status remains 1 if an RPC level error occurred. The server
 446          * may or may not have processed the sequence operation..
 447          * Proceed as if the server received and processed the sequence
 448          * operation.
 449          */
 450         if (res->sr_status == 1)
 451                 res->sr_status = NFS_OK;
 452
 453         /* don't increment the sequence number if the task wasn't sent */
 454         if (!RPC_WAS_SENT(task))
 455                 goto out;
 456
 457         /* Check the SEQUENCE operation status */
 458         switch (res->sr_status) {
 459         case 0:
 460                 /* Update the slot's sequence and clientid lease timer */
 461                 ++res->sr_slot->seq_nr;
 462                 timestamp = res->sr_renewal_time;
 463                 clp = res->sr_session->clp;
 464                 do_renew_lease(clp, timestamp);
 465                 /* Check sequence flags */
 466                 if (res->sr_status_flags != 0)
 467                         nfs4_schedule_lease_recovery(clp);
 468                 break;
 469         case -NFS4ERR_DELAY:
 470                 /* The server detected a resend of the RPC call and
 471                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 472                  * of RFC5661.
 473                  */
 474                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 475                         __func__,
 476                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 477                         res->sr_slot->seq_nr);
 478                 goto out_retry;
 479         default:
 480                 /* Just update the slot sequence no. */
 481                 ++res->sr_slot->seq_nr;
 482         }
 483 out:
 484         /* The session may be reset by one of the error handlers. */
 485         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 486         nfs41_sequence_free_slot(res);
 487         return 1;
 488 out_retry:
 489         if (!rpc_restart_call(task))
 490                 goto out;
 491         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 492         return 0;
 493 }
 494
 495 static int nfs4_sequence_done(struct rpc_task *task,
 496                                struct nfs4_sequence_res *res)
 497 {
 498         if (res->sr_session == NULL)
 499                 return 1;
 500         return nfs41_sequence_done(task, res);
 501 }
 502
 503 /*
 504  * nfs4_find_slot - efficiently look for a free slot
 505  *
 506  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 507  * If found, we mark the slot as used, update the highest_used_slotid,
 508  * and respectively set up the sequence operation args.
 509  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
 510  *
 511  * Note: must be called with under the slot_tbl_lock.
 512  */
 513 static u8
 514 nfs4_find_slot(struct nfs4_slot_table *tbl)
 515 {
 516         int slotid;
 517         u8 ret_id = NFS4_MAX_SLOT_TABLE;
 518         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
 519
 520         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
 521                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 522                 tbl->max_slots);
 523         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 524         if (slotid >= tbl->max_slots)
 525                 goto out;
 526         __set_bit(slotid, tbl->used_slots);
 527         if (slotid > tbl->highest_used_slotid)
 528                 tbl->highest_used_slotid = slotid;
 529         ret_id = slotid;
 530 out:
 531         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 532                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 533         return ret_id;
 534 }
 535
 536 int nfs41_setup_sequence(struct nfs4_session *session,
 537                                 struct nfs4_sequence_args *args,
 538                                 struct nfs4_sequence_res *res,
 539                                 int cache_reply,
 540                                 struct rpc_task *task)
 541 {
 542         struct nfs4_slot *slot;
 543         struct nfs4_slot_table *tbl;
 544         u8 slotid;
 545
 546         dprintk("--> %s\n", __func__);
 547         /* slot already allocated? */
 548         if (res->sr_slot != NULL)
 549                 return 0;
 550
 551         tbl = &session->fc_slot_table;
 552
 553         spin_lock(&tbl->slot_tbl_lock);
 554         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 555             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 556                 /* The state manager will wait until the slot table is empty */
 557                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 558                 spin_unlock(&tbl->slot_tbl_lock);
 559                 dprintk("%s session is draining\n", __func__);
 560                 return -EAGAIN;
 561         }
 562
 563         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 564             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 565                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 566                 spin_unlock(&tbl->slot_tbl_lock);
 567                 dprintk("%s enforce FIFO order\n", __func__);
 568                 return -EAGAIN;
 569         }
 570
 571         slotid = nfs4_find_slot(tbl);
 572         if (slotid == NFS4_MAX_SLOT_TABLE) {
 573                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 574                 spin_unlock(&tbl->slot_tbl_lock);
 575                 dprintk("<-- %s: no free slots\n", __func__);
 576                 return -EAGAIN;
 577         }
 578         spin_unlock(&tbl->slot_tbl_lock);
 579
 580         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 581         slot = tbl->slots + slotid;
 582         args->sa_session = session;
 583         args->sa_slotid = slotid;
 584         args->sa_cache_this = cache_reply;
 585
 586         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 587
 588         res->sr_session = session;
 589         res->sr_slot = slot;
 590         res->sr_renewal_time = jiffies;
 591         res->sr_status_flags = 0;
 592         /*
 593          * sr_status is only set in decode_sequence, and so will remain
 594          * set to 1 if an rpc level failure occurs.
 595          */
 596         res->sr_status = 1;
 597         return 0;
 598 }
 599 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 600
 601 int nfs4_setup_sequence(const struct nfs_server *server,
 602                         struct nfs4_sequence_args *args,
 603                         struct nfs4_sequence_res *res,
 604                         int cache_reply,
 605                         struct rpc_task *task)
 606 {
 607         struct nfs4_session *session = nfs4_get_session(server);
 608         int ret = 0;
 609
 610         if (session == NULL) {
 611                 args->sa_session = NULL;
 612                 res->sr_session = NULL;
 613                 goto out;
 614         }
 615
 616         dprintk("--> %s clp %p session %p sr_slot %td\n",
 617                 __func__, session->clp, session, res->sr_slot ?
 618                         res->sr_slot - session->fc_slot_table.slots : -1);
 619
 620         ret = nfs41_setup_sequence(session, args, res, cache_reply,
 621                                    task);
 622 out:
 623         dprintk("<-- %s status=%d\n", __func__, ret);
 624         return ret;
 625 }
 626
 627 struct nfs41_call_sync_data {
 628         const struct nfs_server *seq_server;
 629         struct nfs4_sequence_args *seq_args;
 630         struct nfs4_sequence_res *seq_res;
 631         int cache_reply;
 632 };
 633
 634 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 635 {
 636         struct nfs41_call_sync_data *data = calldata;
 637
 638         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 639
 640         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 641                                 data->seq_res, data->cache_reply, task))
 642                 return;
 643         rpc_call_start(task);
 644 }
 645
 646 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 647 {
 648         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 649         nfs41_call_sync_prepare(task, calldata);
 650 }
 651
 652 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 653 {
 654         struct nfs41_call_sync_data *data = calldata;
 655
 656         nfs41_sequence_done(task, data->seq_res);
 657 }
 658
 659 struct rpc_call_ops nfs41_call_sync_ops = {
 660         .rpc_call_prepare = nfs41_call_sync_prepare,
 661         .rpc_call_done = nfs41_call_sync_done,
 662 };
 663
 664 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 665         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 666         .rpc_call_done = nfs41_call_sync_done,
 667 };
 668
 669 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 670                                    struct nfs_server *server,
 671                                    struct rpc_message *msg,
 672                                    struct nfs4_sequence_args *args,
 673                                    struct nfs4_sequence_res *res,
 674                                    int cache_reply,
 675                                    int privileged)
 676 {
 677         int ret;
 678         struct rpc_task *task;
 679         struct nfs41_call_sync_data data = {
 680                 .seq_server = server,
 681                 .seq_args = args,
 682                 .seq_res = res,
 683                 .cache_reply = cache_reply,
 684         };
 685         struct rpc_task_setup task_setup = {
 686                 .rpc_client = clnt,
 687                 .rpc_message = msg,
 688                 .callback_ops = &nfs41_call_sync_ops,
 689                 .callback_data = &data
 690         };
 691
 692         res->sr_slot = NULL;
 693         if (privileged)
 694                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 695         task = rpc_run_task(&task_setup);
 696         if (IS_ERR(task))
 697                 ret = PTR_ERR(task);
 698         else {
 699                 ret = task->tk_status;
 700                 rpc_put_task(task);
 701         }
 702         return ret;
 703 }
 704
 705 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 706                             struct nfs_server *server,
 707                             struct rpc_message *msg,
 708                             struct nfs4_sequence_args *args,
 709                             struct nfs4_sequence_res *res,
 710                             int cache_reply)
 711 {
 712         return nfs4_call_sync_sequence(clnt, server, msg, args, res, cache_reply, 0);
 713 }
 714
 715 #else
 716 static int nfs4_sequence_done(struct rpc_task *task,
 717                                struct nfs4_sequence_res *res)
 718 {
 719         return 1;
 720 }
 721 #endif /* CONFIG_NFS_V4_1 */
 722
 723 int _nfs4_call_sync(struct rpc_clnt *clnt,
 724                     struct nfs_server *server,
 725                     struct rpc_message *msg,
 726                     struct nfs4_sequence_args *args,
 727                     struct nfs4_sequence_res *res,
 728                     int cache_reply)
 729 {
 730         args->sa_session = res->sr_session = NULL;
 731         return rpc_call_sync(clnt, msg, 0);
 732 }
 733
 734 static inline
 735 int nfs4_call_sync(struct rpc_clnt *clnt,
 736                    struct nfs_server *server,
 737                    struct rpc_message *msg,
 738                    struct nfs4_sequence_args *args,
 739                    struct nfs4_sequence_res *res,
 740                    int cache_reply)
 741 {
 742         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 743                                                 args, res, cache_reply);
 744 }
 745
 746 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 747 {
 748         struct nfs_inode *nfsi = NFS_I(dir);
 749
 750         spin_lock(&dir->i_lock);
 751         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 752         if (!cinfo->atomic || cinfo->before != dir->i_version)
 753                 nfs_force_lookup_revalidate(dir);
 754         dir->i_version = cinfo->after;
 755         spin_unlock(&dir->i_lock);
 756 }
 757
 758 struct nfs4_opendata {
 759         struct kref kref;
 760         struct nfs_openargs o_arg;
 761         struct nfs_openres o_res;
 762         struct nfs_open_confirmargs c_arg;
 763         struct nfs_open_confirmres c_res;
 764         struct nfs4_string owner_name;
 765         struct nfs4_string group_name;
 766         struct nfs_fattr f_attr;
 767         struct nfs_fattr dir_attr;
 768         struct dentry *dir;
 769         struct dentry *dentry;
 770         struct nfs4_state_owner *owner;
 771         struct nfs4_state *state;
 772         struct iattr attrs;
 773         unsigned long timestamp;
 774         unsigned int rpc_done : 1;
 775         int rpc_status;
 776         int cancelled;
 777 };
 778
 779
 780 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 781 {
 782         p->o_res.f_attr = &p->f_attr;
 783         p->o_res.dir_attr = &p->dir_attr;
 784         p->o_res.seqid = p->o_arg.seqid;
 785         p->c_res.seqid = p->c_arg.seqid;
 786         p->o_res.server = p->o_arg.server;
 787         nfs_fattr_init(&p->f_attr);
 788         nfs_fattr_init(&p->dir_attr);
 789         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 790 }
 791
 792 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 793                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 794                 const struct iattr *attrs,
 795                 gfp_t gfp_mask)
 796 {
 797         struct dentry *parent = dget_parent(dentry);
 798         struct inode *dir = parent->d_inode;
 799         struct nfs_server *server = NFS_SERVER(dir);
 800         struct nfs4_opendata *p;
 801
 802         p = kzalloc(sizeof(*p), gfp_mask);
 803         if (p == NULL)
 804                 goto err;
 805         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 806         if (p->o_arg.seqid == NULL)
 807                 goto err_free;
 808         nfs_sb_active(dentry->d_sb);
 809         p->dentry = dget(dentry);
 810         p->dir = parent;
 811         p->owner = sp;
 812         atomic_inc(&sp->so_count);
 813         p->o_arg.fh = NFS_FH(dir);
 814         p->o_arg.open_flags = flags;
 815         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 816         p->o_arg.clientid = server->nfs_client->cl_clientid;
 817         p->o_arg.id = sp->so_owner_id.id;
 818         p->o_arg.name = &dentry->d_name;
 819         p->o_arg.server = server;
 820         p->o_arg.bitmask = server->attr_bitmask;
 821         p->o_arg.dir_bitmask = server->cache_consistency_bitmask;
 822         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 823         if (flags & O_CREAT) {
 824                 u32 *s;
 825
 826                 p->o_arg.u.attrs = &p->attrs;
 827                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 828                 s = (u32 *) p->o_arg.u.verifier.data;
 829                 s[0] = jiffies;
 830                 s[1] = current->pid;
 831         }
 832         p->c_arg.fh = &p->o_res.fh;
 833         p->c_arg.stateid = &p->o_res.stateid;
 834         p->c_arg.seqid = p->o_arg.seqid;
 835         nfs4_init_opendata_res(p);
 836         kref_init(&p->kref);
 837         return p;
 838 err_free:
 839         kfree(p);
 840 err:
 841         dput(parent);
 842         return NULL;
 843 }
 844
 845 static void nfs4_opendata_free(struct kref *kref)
 846 {
 847         struct nfs4_opendata *p = container_of(kref,
 848                         struct nfs4_opendata, kref);
 849         struct super_block *sb = p->dentry->d_sb;
 850
 851         nfs_free_seqid(p->o_arg.seqid);
 852         if (p->state != NULL)
 853                 nfs4_put_open_state(p->state);
 854         nfs4_put_state_owner(p->owner);
 855         dput(p->dir);
 856         dput(p->dentry);
 857         nfs_sb_deactive(sb);
 858         nfs_fattr_free_names(&p->f_attr);
 859         kfree(p);
 860 }
 861
 862 static void nfs4_opendata_put(struct nfs4_opendata *p)
 863 {
 864         if (p != NULL)
 865                 kref_put(&p->kref, nfs4_opendata_free);
 866 }
 867
 868 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 869 {
 870         int ret;
 871
 872         ret = rpc_wait_for_completion_task(task);
 873         return ret;
 874 }
 875
 876 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 877 {
 878         int ret = 0;
 879
 880         if (open_mode & O_EXCL)
 881                 goto out;
 882         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 883                 case FMODE_READ:
 884                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 885                                 && state->n_rdonly != 0;
 886                         break;
 887                 case FMODE_WRITE:
 888                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 889                                 && state->n_wronly != 0;
 890                         break;
 891                 case FMODE_READ|FMODE_WRITE:
 892                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 893                                 && state->n_rdwr != 0;
 894         }
 895 out:
 896         return ret;
 897 }
 898
 899 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 900 {
 901         if (delegation == NULL)
 902                 return 0;
 903         if ((delegation->type & fmode) != fmode)
 904                 return 0;
 905         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 906                 return 0;
 907         nfs_mark_delegation_referenced(delegation);
 908         return 1;
 909 }
 910
 911 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 912 {
 913         switch (fmode) {
 914                 case FMODE_WRITE:
 915                         state->n_wronly++;
 916                         break;
 917                 case FMODE_READ:
 918                         state->n_rdonly++;
 919                         break;
 920                 case FMODE_READ|FMODE_WRITE:
 921                         state->n_rdwr++;
 922         }
 923         nfs4_state_set_mode_locked(state, state->state | fmode);
 924 }
 925
 926 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 927 {
 928         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 929                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 930         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 931         switch (fmode) {
 932                 case FMODE_READ:
 933                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 934                         break;
 935                 case FMODE_WRITE:
 936                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 937                         break;
 938                 case FMODE_READ|FMODE_WRITE:
 939                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 940         }
 941 }
 942
 943 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 944 {
 945         write_seqlock(&state->seqlock);
 946         nfs_set_open_stateid_locked(state, stateid, fmode);
 947         write_sequnlock(&state->seqlock);
 948 }
 949
 950 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 951 {
 952         /*
 953          * Protect the call to nfs4_state_set_mode_locked and
 954          * serialise the stateid update
 955          */
 956         write_seqlock(&state->seqlock);
 957         if (deleg_stateid != NULL) {
 958                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 959                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 960         }
 961         if (open_stateid != NULL)
 962                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 963         write_sequnlock(&state->seqlock);
 964         spin_lock(&state->owner->so_lock);
 965         update_open_stateflags(state, fmode);
 966         spin_unlock(&state->owner->so_lock);
 967 }
 968
 969 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 970 {
 971         struct nfs_inode *nfsi = NFS_I(state->inode);
 972         struct nfs_delegation *deleg_cur;
 973         int ret = 0;
 974
 975         fmode &= (FMODE_READ|FMODE_WRITE);
 976
 977         rcu_read_lock();
 978         deleg_cur = rcu_dereference(nfsi->delegation);
 979         if (deleg_cur == NULL)
 980                 goto no_delegation;
 981
 982         spin_lock(&deleg_cur->lock);
 983         if (nfsi->delegation != deleg_cur ||
 984             (deleg_cur->type & fmode) != fmode)
 985                 goto no_delegation_unlock;
 986
 987         if (delegation == NULL)
 988                 delegation = &deleg_cur->stateid;
 989         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
 990                 goto no_delegation_unlock;
 991
 992         nfs_mark_delegation_referenced(deleg_cur);
 993         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
 994         ret = 1;
 995 no_delegation_unlock:
 996         spin_unlock(&deleg_cur->lock);
 997 no_delegation:
 998         rcu_read_unlock();
 999
1000         if (!ret && open_stateid != NULL) {
1001                 __update_open_stateid(state, open_stateid, NULL, fmode);
1002                 ret = 1;
1003         }
1004
1005         return ret;
1006 }
1007
1008
1009 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1010 {
1011         struct nfs_delegation *delegation;
1012
1013         rcu_read_lock();
1014         delegation = rcu_dereference(NFS_I(inode)->delegation);
1015         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1016                 rcu_read_unlock();
1017                 return;
1018         }
1019         rcu_read_unlock();
1020         nfs_inode_return_delegation(inode);
1021 }
1022
1023 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1024 {
1025         struct nfs4_state *state = opendata->state;
1026         struct nfs_inode *nfsi = NFS_I(state->inode);
1027         struct nfs_delegation *delegation;
1028         int open_mode = opendata->o_arg.open_flags & O_EXCL;
1029         fmode_t fmode = opendata->o_arg.fmode;
1030         nfs4_stateid stateid;
1031         int ret = -EAGAIN;
1032
1033         for (;;) {
1034                 if (can_open_cached(state, fmode, open_mode)) {
1035                         spin_lock(&state->owner->so_lock);
1036                         if (can_open_cached(state, fmode, open_mode)) {
1037                                 update_open_stateflags(state, fmode);
1038                                 spin_unlock(&state->owner->so_lock);
1039                                 goto out_return_state;
1040                         }
1041                         spin_unlock(&state->owner->so_lock);
1042                 }
1043                 rcu_read_lock();
1044                 delegation = rcu_dereference(nfsi->delegation);
1045                 if (!can_open_delegated(delegation, fmode)) {
1046                         rcu_read_unlock();
1047                         break;
1048                 }
1049                 /* Save the delegation */
1050                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1051                 rcu_read_unlock();
1052                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1053                 if (ret != 0)
1054                         goto out;
1055                 ret = -EAGAIN;
1056
1057                 /* Try to update the stateid using the delegation */
1058                 if (update_open_stateid(state, NULL, &stateid, fmode))
1059                         goto out_return_state;
1060         }
1061 out:
1062         return ERR_PTR(ret);
1063 out_return_state:
1064         atomic_inc(&state->count);
1065         return state;
1066 }
1067
1068 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1069 {
1070         struct inode *inode;
1071         struct nfs4_state *state = NULL;
1072         struct nfs_delegation *delegation;
1073         int ret;
1074
1075         if (!data->rpc_done) {
1076                 state = nfs4_try_open_cached(data);
1077                 goto out;
1078         }
1079
1080         ret = -EAGAIN;
1081         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1082                 goto err;
1083         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1084         ret = PTR_ERR(inode);
1085         if (IS_ERR(inode))
1086                 goto err;
1087         ret = -ENOMEM;
1088         state = nfs4_get_open_state(inode, data->owner);
1089         if (state == NULL)
1090                 goto err_put_inode;
1091         if (data->o_res.delegation_type != 0) {
1092                 int delegation_flags = 0;
1093
1094                 rcu_read_lock();
1095                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1096                 if (delegation)
1097                         delegation_flags = delegation->flags;
1098                 rcu_read_unlock();
1099                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1100                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1101                                         "returning a delegation for "
1102                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1103                                         NFS_CLIENT(inode)->cl_server);
1104                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1105                         nfs_inode_set_delegation(state->inode,
1106                                         data->owner->so_cred,
1107                                         &data->o_res);
1108                 else
1109                         nfs_inode_reclaim_delegation(state->inode,
1110                                         data->owner->so_cred,
1111                                         &data->o_res);
1112         }
1113
1114         update_open_stateid(state, &data->o_res.stateid, NULL,
1115                         data->o_arg.fmode);
1116         iput(inode);
1117 out:
1118         return state;
1119 err_put_inode:
1120         iput(inode);
1121 err:
1122         return ERR_PTR(ret);
1123 }
1124
1125 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1126 {
1127         struct nfs_inode *nfsi = NFS_I(state->inode);
1128         struct nfs_open_context *ctx;
1129
1130         spin_lock(&state->inode->i_lock);
1131         list_for_each_entry(ctx, &nfsi->open_files, list) {
1132                 if (ctx->state != state)
1133                         continue;
1134                 get_nfs_open_context(ctx);
1135                 spin_unlock(&state->inode->i_lock);
1136                 return ctx;
1137         }
1138         spin_unlock(&state->inode->i_lock);
1139         return ERR_PTR(-ENOENT);
1140 }
1141
1142 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1143 {
1144         struct nfs4_opendata *opendata;
1145
1146         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1147         if (opendata == NULL)
1148                 return ERR_PTR(-ENOMEM);
1149         opendata->state = state;
1150         atomic_inc(&state->count);
1151         return opendata;
1152 }
1153
1154 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1155 {
1156         struct nfs4_state *newstate;
1157         int ret;
1158
1159         opendata->o_arg.open_flags = 0;
1160         opendata->o_arg.fmode = fmode;
1161         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1162         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1163         nfs4_init_opendata_res(opendata);
1164         ret = _nfs4_recover_proc_open(opendata);
1165         if (ret != 0)
1166                 return ret;
1167         newstate = nfs4_opendata_to_nfs4_state(opendata);
1168         if (IS_ERR(newstate))
1169                 return PTR_ERR(newstate);
1170         nfs4_close_state(newstate, fmode);
1171         *res = newstate;
1172         return 0;
1173 }
1174
1175 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1176 {
1177         struct nfs4_state *newstate;
1178         int ret;
1179
1180         /* memory barrier prior to reading state->n_* */
1181         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1182         smp_rmb();
1183         if (state->n_rdwr != 0) {
1184                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1185                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1186                 if (ret != 0)
1187                         return ret;
1188                 if (newstate != state)
1189                         return -ESTALE;
1190         }
1191         if (state->n_wronly != 0) {
1192                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1193                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1194                 if (ret != 0)
1195                         return ret;
1196                 if (newstate != state)
1197                         return -ESTALE;
1198         }
1199         if (state->n_rdonly != 0) {
1200                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1201                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1202                 if (ret != 0)
1203                         return ret;
1204                 if (newstate != state)
1205                         return -ESTALE;
1206         }
1207         /*
1208          * We may have performed cached opens for all three recoveries.
1209          * Check if we need to update the current stateid.
1210          */
1211         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1212             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1213                 write_seqlock(&state->seqlock);
1214                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1215                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1216                 write_sequnlock(&state->seqlock);
1217         }
1218         return 0;
1219 }
1220
1221 /*
1222  * OPEN_RECLAIM:
1223  *      reclaim state on the server after a reboot.
1224  */
1225 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1226 {
1227         struct nfs_delegation *delegation;
1228         struct nfs4_opendata *opendata;
1229         fmode_t delegation_type = 0;
1230         int status;
1231
1232         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1233         if (IS_ERR(opendata))
1234                 return PTR_ERR(opendata);
1235         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1236         opendata->o_arg.fh = NFS_FH(state->inode);
1237         rcu_read_lock();
1238         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1239         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1240                 delegation_type = delegation->type;
1241         rcu_read_unlock();
1242         opendata->o_arg.u.delegation_type = delegation_type;
1243         status = nfs4_open_recover(opendata, state);
1244         nfs4_opendata_put(opendata);
1245         return status;
1246 }
1247
1248 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1249 {
1250         struct nfs_server *server = NFS_SERVER(state->inode);
1251         struct nfs4_exception exception = { };
1252         int err;
1253         do {
1254                 err = _nfs4_do_open_reclaim(ctx, state);
1255                 if (err != -NFS4ERR_DELAY)
1256                         break;
1257                 nfs4_handle_exception(server, err, &exception);
1258         } while (exception.retry);
1259         return err;
1260 }
1261
1262 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1263 {
1264         struct nfs_open_context *ctx;
1265         int ret;
1266
1267         ctx = nfs4_state_find_open_context(state);
1268         if (IS_ERR(ctx))
1269                 return PTR_ERR(ctx);
1270         ret = nfs4_do_open_reclaim(ctx, state);
1271         put_nfs_open_context(ctx);
1272         return ret;
1273 }
1274
1275 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1276 {
1277         struct nfs4_opendata *opendata;
1278         int ret;
1279
1280         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1281         if (IS_ERR(opendata))
1282                 return PTR_ERR(opendata);
1283         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1284         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1285                         sizeof(opendata->o_arg.u.delegation.data));
1286         ret = nfs4_open_recover(opendata, state);
1287         nfs4_opendata_put(opendata);
1288         return ret;
1289 }
1290
1291 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1292 {
1293         struct nfs4_exception exception = { };
1294         struct nfs_server *server = NFS_SERVER(state->inode);
1295         int err;
1296         do {
1297                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1298                 switch (err) {
1299                         case 0:
1300                         case -ENOENT:
1301                         case -ESTALE:
1302                                 goto out;
1303                         case -NFS4ERR_BADSESSION:
1304                         case -NFS4ERR_BADSLOT:
1305                         case -NFS4ERR_BAD_HIGH_SLOT:
1306                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1307                         case -NFS4ERR_DEADSESSION:
1308                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1309                                 goto out;
1310                         case -NFS4ERR_STALE_CLIENTID:
1311                         case -NFS4ERR_STALE_STATEID:
1312                         case -NFS4ERR_EXPIRED:
1313                                 /* Don't recall a delegation if it was lost */
1314                                 nfs4_schedule_lease_recovery(server->nfs_client);
1315                                 goto out;
1316                         case -ERESTARTSYS:
1317                                 /*
1318                                  * The show must go on: exit, but mark the
1319                                  * stateid as needing recovery.
1320                                  */
1321                         case -NFS4ERR_ADMIN_REVOKED:
1322                         case -NFS4ERR_BAD_STATEID:
1323                                 nfs4_schedule_stateid_recovery(server, state);
1324                         case -EKEYEXPIRED:
1325                                 /*
1326                                  * User RPCSEC_GSS context has expired.
1327                                  * We cannot recover this stateid now, so
1328                                  * skip it and allow recovery thread to
1329                                  * proceed.
1330                                  */
1331                         case -ENOMEM:
1332                                 err = 0;
1333                                 goto out;
1334                 }
1335                 err = nfs4_handle_exception(server, err, &exception);
1336         } while (exception.retry);
1337 out:
1338         return err;
1339 }
1340
1341 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1342 {
1343         struct nfs4_opendata *data = calldata;
1344
1345         data->rpc_status = task->tk_status;
1346         if (data->rpc_status == 0) {
1347                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1348                                 sizeof(data->o_res.stateid.data));
1349                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1350                 renew_lease(data->o_res.server, data->timestamp);
1351                 data->rpc_done = 1;
1352         }
1353 }
1354
1355 static void nfs4_open_confirm_release(void *calldata)
1356 {
1357         struct nfs4_opendata *data = calldata;
1358         struct nfs4_state *state = NULL;
1359
1360         /* If this request hasn't been cancelled, do nothing */
1361         if (data->cancelled == 0)
1362                 goto out_free;
1363         /* In case of error, no cleanup! */
1364         if (!data->rpc_done)
1365                 goto out_free;
1366         state = nfs4_opendata_to_nfs4_state(data);
1367         if (!IS_ERR(state))
1368                 nfs4_close_state(state, data->o_arg.fmode);
1369 out_free:
1370         nfs4_opendata_put(data);
1371 }
1372
1373 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1374         .rpc_call_done = nfs4_open_confirm_done,
1375         .rpc_release = nfs4_open_confirm_release,
1376 };
1377
1378 /*
1379  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1380  */
1381 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1382 {
1383         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1384         struct rpc_task *task;
1385         struct  rpc_message msg = {
1386                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1387                 .rpc_argp = &data->c_arg,
1388                 .rpc_resp = &data->c_res,
1389                 .rpc_cred = data->owner->so_cred,
1390         };
1391         struct rpc_task_setup task_setup_data = {
1392                 .rpc_client = server->client,
1393                 .rpc_message = &msg,
1394                 .callback_ops = &nfs4_open_confirm_ops,
1395                 .callback_data = data,
1396                 .workqueue = nfsiod_workqueue,
1397                 .flags = RPC_TASK_ASYNC,
1398         };
1399         int status;
1400
1401         kref_get(&data->kref);
1402         data->rpc_done = 0;
1403         data->rpc_status = 0;
1404         data->timestamp = jiffies;
1405         task = rpc_run_task(&task_setup_data);
1406         if (IS_ERR(task))
1407                 return PTR_ERR(task);
1408         status = nfs4_wait_for_completion_rpc_task(task);
1409         if (status != 0) {
1410                 data->cancelled = 1;
1411                 smp_wmb();
1412         } else
1413                 status = data->rpc_status;
1414         rpc_put_task(task);
1415         return status;
1416 }
1417
1418 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1419 {
1420         struct nfs4_opendata *data = calldata;
1421         struct nfs4_state_owner *sp = data->owner;
1422
1423         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1424                 return;
1425         /*
1426          * Check if we still need to send an OPEN call, or if we can use
1427          * a delegation instead.
1428          */
1429         if (data->state != NULL) {
1430                 struct nfs_delegation *delegation;
1431
1432                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1433                         goto out_no_action;
1434                 rcu_read_lock();
1435                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1436                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1437                     can_open_delegated(delegation, data->o_arg.fmode))
1438                         goto unlock_no_action;
1439                 rcu_read_unlock();
1440         }
1441         /* Update sequence id. */
1442         data->o_arg.id = sp->so_owner_id.id;
1443         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1444         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1445                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1446                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1447         }
1448         data->timestamp = jiffies;
1449         if (nfs4_setup_sequence(data->o_arg.server,
1450                                 &data->o_arg.seq_args,
1451                                 &data->o_res.seq_res, 1, task))
1452                 return;
1453         rpc_call_start(task);
1454         return;
1455 unlock_no_action:
1456         rcu_read_unlock();
1457 out_no_action:
1458         task->tk_action = NULL;
1459
1460 }
1461
1462 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1463 {
1464         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1465         nfs4_open_prepare(task, calldata);
1466 }
1467
1468 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1469 {
1470         struct nfs4_opendata *data = calldata;
1471
1472         data->rpc_status = task->tk_status;
1473
1474         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1475                 return;
1476
1477         if (task->tk_status == 0) {
1478                 switch (data->o_res.f_attr->mode & S_IFMT) {
1479                         case S_IFREG:
1480                                 break;
1481                         case S_IFLNK:
1482                                 data->rpc_status = -ELOOP;
1483                                 break;
1484                         case S_IFDIR:
1485                                 data->rpc_status = -EISDIR;
1486                                 break;
1487                         default:
1488                                 data->rpc_status = -ENOTDIR;
1489                 }
1490                 renew_lease(data->o_res.server, data->timestamp);
1491                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1492                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1493         }
1494         data->rpc_done = 1;
1495 }
1496
1497 static void nfs4_open_release(void *calldata)
1498 {
1499         struct nfs4_opendata *data = calldata;
1500         struct nfs4_state *state = NULL;
1501
1502         /* If this request hasn't been cancelled, do nothing */
1503         if (data->cancelled == 0)
1504                 goto out_free;
1505         /* In case of error, no cleanup! */
1506         if (data->rpc_status != 0 || !data->rpc_done)
1507                 goto out_free;
1508         /* In case we need an open_confirm, no cleanup! */
1509         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1510                 goto out_free;
1511         state = nfs4_opendata_to_nfs4_state(data);
1512         if (!IS_ERR(state))
1513                 nfs4_close_state(state, data->o_arg.fmode);
1514 out_free:
1515         nfs4_opendata_put(data);
1516 }
1517
1518 static const struct rpc_call_ops nfs4_open_ops = {
1519         .rpc_call_prepare = nfs4_open_prepare,
1520         .rpc_call_done = nfs4_open_done,
1521         .rpc_release = nfs4_open_release,
1522 };
1523
1524 static const struct rpc_call_ops nfs4_recover_open_ops = {
1525         .rpc_call_prepare = nfs4_recover_open_prepare,
1526         .rpc_call_done = nfs4_open_done,
1527         .rpc_release = nfs4_open_release,
1528 };
1529
1530 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1531 {
1532         struct inode *dir = data->dir->d_inode;
1533         struct nfs_server *server = NFS_SERVER(dir);
1534         struct nfs_openargs *o_arg = &data->o_arg;
1535         struct nfs_openres *o_res = &data->o_res;
1536         struct rpc_task *task;
1537         struct rpc_message msg = {
1538                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1539                 .rpc_argp = o_arg,
1540                 .rpc_resp = o_res,
1541                 .rpc_cred = data->owner->so_cred,
1542         };
1543         struct rpc_task_setup task_setup_data = {
1544                 .rpc_client = server->client,
1545                 .rpc_message = &msg,
1546                 .callback_ops = &nfs4_open_ops,
1547                 .callback_data = data,
1548                 .workqueue = nfsiod_workqueue,
1549                 .flags = RPC_TASK_ASYNC,
1550         };
1551         int status;
1552
1553         kref_get(&data->kref);
1554         data->rpc_done = 0;
1555         data->rpc_status = 0;
1556         data->cancelled = 0;
1557         if (isrecover)
1558                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1559         task = rpc_run_task(&task_setup_data);
1560         if (IS_ERR(task))
1561                 return PTR_ERR(task);
1562         status = nfs4_wait_for_completion_rpc_task(task);
1563         if (status != 0) {
1564                 data->cancelled = 1;
1565                 smp_wmb();
1566         } else
1567                 status = data->rpc_status;
1568         rpc_put_task(task);
1569
1570         return status;
1571 }
1572
1573 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1574 {
1575         struct inode *dir = data->dir->d_inode;
1576         struct nfs_openres *o_res = &data->o_res;
1577         int status;
1578
1579         status = nfs4_run_open_task(data, 1);
1580         if (status != 0 || !data->rpc_done)
1581                 return status;
1582
1583         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1584
1585         nfs_refresh_inode(dir, o_res->dir_attr);
1586
1587         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1588                 status = _nfs4_proc_open_confirm(data);
1589                 if (status != 0)
1590                         return status;
1591         }
1592
1593         return status;
1594 }
1595
1596 /*
1597  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1598  */
1599 static int _nfs4_proc_open(struct nfs4_opendata *data)
1600 {
1601         struct inode *dir = data->dir->d_inode;
1602         struct nfs_server *server = NFS_SERVER(dir);
1603         struct nfs_openargs *o_arg = &data->o_arg;
1604         struct nfs_openres *o_res = &data->o_res;
1605         int status;
1606
1607         status = nfs4_run_open_task(data, 0);
1608         if (!data->rpc_done)
1609                 return status;
1610         if (status != 0) {
1611                 if (status == -NFS4ERR_BADNAME &&
1612                                 !(o_arg->open_flags & O_CREAT))
1613                         return -ENOENT;
1614                 return status;
1615         }
1616
1617         nfs_fattr_map_and_free_names(server, &data->f_attr);
1618
1619         if (o_arg->open_flags & O_CREAT) {
1620                 update_changeattr(dir, &o_res->cinfo);
1621                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1622         } else
1623                 nfs_refresh_inode(dir, o_res->dir_attr);
1624         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1625                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1626         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1627                 status = _nfs4_proc_open_confirm(data);
1628                 if (status != 0)
1629                         return status;
1630         }
1631         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1632                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1633         return 0;
1634 }
1635
1636 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1637 {
1638         unsigned int loop;
1639         int ret;
1640
1641         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1642                 ret = nfs4_wait_clnt_recover(clp);
1643                 if (ret != 0)
1644                         break;
1645                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1646                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1647                         break;
1648                 nfs4_schedule_state_manager(clp);
1649                 ret = -EIO;
1650         }
1651         return ret;
1652 }
1653
1654 static int nfs4_recover_expired_lease(struct nfs_server *server)
1655 {
1656         return nfs4_client_recover_expired_lease(server->nfs_client);
1657 }
1658
1659 /*
1660  * OPEN_EXPIRED:
1661  *      reclaim state on the server after a network partition.
1662  *      Assumes caller holds the appropriate lock
1663  */
1664 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1665 {
1666         struct nfs4_opendata *opendata;
1667         int ret;
1668
1669         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1670         if (IS_ERR(opendata))
1671                 return PTR_ERR(opendata);
1672         ret = nfs4_open_recover(opendata, state);
1673         if (ret == -ESTALE)
1674                 d_drop(ctx->dentry);
1675         nfs4_opendata_put(opendata);
1676         return ret;
1677 }
1678
1679 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1680 {
1681         struct nfs_server *server = NFS_SERVER(state->inode);
1682         struct nfs4_exception exception = { };
1683         int err;
1684
1685         do {
1686                 err = _nfs4_open_expired(ctx, state);
1687                 switch (err) {
1688                 default:
1689                         goto out;
1690                 case -NFS4ERR_GRACE:
1691                 case -NFS4ERR_DELAY:
1692                         nfs4_handle_exception(server, err, &exception);
1693                         err = 0;
1694                 }
1695         } while (exception.retry);
1696 out:
1697         return err;
1698 }
1699
1700 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1701 {
1702         struct nfs_open_context *ctx;
1703         int ret;
1704
1705         ctx = nfs4_state_find_open_context(state);
1706         if (IS_ERR(ctx))
1707                 return PTR_ERR(ctx);
1708         ret = nfs4_do_open_expired(ctx, state);
1709         put_nfs_open_context(ctx);
1710         return ret;
1711 }
1712
1713 #if defined(CONFIG_NFS_V4_1)
1714 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1715 {
1716         int status;
1717         struct nfs_server *server = NFS_SERVER(state->inode);
1718
1719         status = nfs41_test_stateid(server, state);
1720         if (status == NFS_OK)
1721                 return 0;
1722         nfs41_free_stateid(server, state);
1723         return nfs4_open_expired(sp, state);
1724 }
1725 #endif
1726
1727 /*
1728  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1729  * fields corresponding to attributes that were used to store the verifier.
1730  * Make sure we clobber those fields in the later setattr call
1731  */
1732 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1733 {
1734         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1735             !(sattr->ia_valid & ATTR_ATIME_SET))
1736                 sattr->ia_valid |= ATTR_ATIME;
1737
1738         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1739             !(sattr->ia_valid & ATTR_MTIME_SET))
1740                 sattr->ia_valid |= ATTR_MTIME;
1741 }
1742
1743 /*
1744  * Returns a referenced nfs4_state
1745  */
1746 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)
1747 {
1748         struct nfs4_state_owner  *sp;
1749         struct nfs4_state     *state = NULL;
1750         struct nfs_server       *server = NFS_SERVER(dir);
1751         struct nfs4_opendata *opendata;
1752         int status;
1753
1754         /* Protect against reboot recovery conflicts */
1755         status = -ENOMEM;
1756         if (!(sp = nfs4_get_state_owner(server, cred))) {
1757                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1758                 goto out_err;
1759         }
1760         status = nfs4_recover_expired_lease(server);
1761         if (status != 0)
1762                 goto err_put_state_owner;
1763         if (dentry->d_inode != NULL)
1764                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1765         status = -ENOMEM;
1766         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1767         if (opendata == NULL)
1768                 goto err_put_state_owner;
1769
1770         if (dentry->d_inode != NULL)
1771                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1772
1773         status = _nfs4_proc_open(opendata);
1774         if (status != 0)
1775                 goto err_opendata_put;
1776
1777         state = nfs4_opendata_to_nfs4_state(opendata);
1778         status = PTR_ERR(state);
1779         if (IS_ERR(state))
1780                 goto err_opendata_put;
1781         if (server->caps & NFS_CAP_POSIX_LOCK)
1782                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1783
1784         if (opendata->o_arg.open_flags & O_EXCL) {
1785                 nfs4_exclusive_attrset(opendata, sattr);
1786
1787                 nfs_fattr_init(opendata->o_res.f_attr);
1788                 status = nfs4_do_setattr(state->inode, cred,
1789                                 opendata->o_res.f_attr, sattr,
1790                                 state);
1791                 if (status == 0)
1792                         nfs_setattr_update_inode(state->inode, sattr);
1793                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1794         }
1795         nfs4_opendata_put(opendata);
1796         nfs4_put_state_owner(sp);
1797         *res = state;
1798         return 0;
1799 err_opendata_put:
1800         nfs4_opendata_put(opendata);
1801 err_put_state_owner:
1802         nfs4_put_state_owner(sp);
1803 out_err:
1804         *res = NULL;
1805         return status;
1806 }
1807
1808
1809 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)
1810 {
1811         struct nfs4_exception exception = { };
1812         struct nfs4_state *res;
1813         int status;
1814
1815         do {
1816                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred, &res);
1817                 if (status == 0)
1818                         break;
1819                 /* NOTE: BAD_SEQID means the server and client disagree about the
1820                  * book-keeping w.r.t. state-changing operations
1821                  * (OPEN/CLOSE/LOCK/LOCKU...)
1822                  * It is actually a sign of a bug on the client or on the server.
1823                  *
1824                  * If we receive a BAD_SEQID error in the particular case of
1825                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1826                  * have unhashed the old state_owner for us, and that we can
1827                  * therefore safely retry using a new one. We should still warn
1828                  * the user though...
1829                  */
1830                 if (status == -NFS4ERR_BAD_SEQID) {
1831                         printk(KERN_WARNING "NFS: v4 server %s "
1832                                         " returned a bad sequence-id error!\n",
1833                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1834                         exception.retry = 1;
1835                         continue;
1836                 }
1837                 /*
1838                  * BAD_STATEID on OPEN means that the server cancelled our
1839                  * state before it received the OPEN_CONFIRM.
1840                  * Recover by retrying the request as per the discussion
1841                  * on Page 181 of RFC3530.
1842                  */
1843                 if (status == -NFS4ERR_BAD_STATEID) {
1844                         exception.retry = 1;
1845                         continue;
1846                 }
1847                 if (status == -EAGAIN) {
1848                         /* We must have found a delegation */
1849                         exception.retry = 1;
1850                         continue;
1851                 }
1852                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1853                                         status, &exception));
1854         } while (exception.retry);
1855         return res;
1856 }
1857
1858 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1859                             struct nfs_fattr *fattr, struct iattr *sattr,
1860                             struct nfs4_state *state)
1861 {
1862         struct nfs_server *server = NFS_SERVER(inode);
1863         struct nfs_setattrargs  arg = {
1864                 .fh             = NFS_FH(inode),
1865                 .iap            = sattr,
1866                 .server         = server,
1867                 .bitmask = server->attr_bitmask,
1868         };
1869         struct nfs_setattrres  res = {
1870                 .fattr          = fattr,
1871                 .server         = server,
1872         };
1873         struct rpc_message msg = {
1874                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1875                 .rpc_argp       = &arg,
1876                 .rpc_resp       = &res,
1877                 .rpc_cred       = cred,
1878         };
1879         unsigned long timestamp = jiffies;
1880         int status;
1881
1882         nfs_fattr_init(fattr);
1883
1884         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1885                 /* Use that stateid */
1886         } else if (state != NULL) {
1887                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1888         } else
1889                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1890
1891         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1892         if (status == 0 && state != NULL)
1893                 renew_lease(server, timestamp);
1894         return status;
1895 }
1896
1897 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1898                            struct nfs_fattr *fattr, struct iattr *sattr,
1899                            struct nfs4_state *state)
1900 {
1901         struct nfs_server *server = NFS_SERVER(inode);
1902         struct nfs4_exception exception = { };
1903         int err;
1904         do {
1905                 err = nfs4_handle_exception(server,
1906                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1907                                 &exception);
1908         } while (exception.retry);
1909         return err;
1910 }
1911
1912 struct nfs4_closedata {
1913         struct inode *inode;
1914         struct nfs4_state *state;
1915         struct nfs_closeargs arg;
1916         struct nfs_closeres res;
1917         struct nfs_fattr fattr;
1918         unsigned long timestamp;
1919         bool roc;
1920         u32 roc_barrier;
1921 };
1922
1923 static void nfs4_free_closedata(void *data)
1924 {
1925         struct nfs4_closedata *calldata = data;
1926         struct nfs4_state_owner *sp = calldata->state->owner;
1927         struct super_block *sb = calldata->state->inode->i_sb;
1928
1929         if (calldata->roc)
1930                 pnfs_roc_release(calldata->state->inode);
1931         nfs4_put_open_state(calldata->state);
1932         nfs_free_seqid(calldata->arg.seqid);
1933         nfs4_put_state_owner(sp);
1934         nfs_sb_deactive(sb);
1935         kfree(calldata);
1936 }
1937
1938 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1939                 fmode_t fmode)
1940 {
1941         spin_lock(&state->owner->so_lock);
1942         if (!(fmode & FMODE_READ))
1943                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1944         if (!(fmode & FMODE_WRITE))
1945                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1946         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1947         spin_unlock(&state->owner->so_lock);
1948 }
1949
1950 static void nfs4_close_done(struct rpc_task *task, void *data)
1951 {
1952         struct nfs4_closedata *calldata = data;
1953         struct nfs4_state *state = calldata->state;
1954         struct nfs_server *server = NFS_SERVER(calldata->inode);
1955
1956         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1957                 return;
1958         /* hmm. we are done with the inode, and in the process of freeing
1959          * the state_owner. we keep this around to process errors
1960          */
1961         switch (task->tk_status) {
1962                 case 0:
1963                         if (calldata->roc)
1964                                 pnfs_roc_set_barrier(state->inode,
1965                                                      calldata->roc_barrier);
1966                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1967                         renew_lease(server, calldata->timestamp);
1968                         nfs4_close_clear_stateid_flags(state,
1969                                         calldata->arg.fmode);
1970                         break;
1971                 case -NFS4ERR_STALE_STATEID:
1972                 case -NFS4ERR_OLD_STATEID:
1973                 case -NFS4ERR_BAD_STATEID:
1974                 case -NFS4ERR_EXPIRED:
1975                         if (calldata->arg.fmode == 0)
1976                                 break;
1977                 default:
1978                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1979                                 rpc_restart_call_prepare(task);
1980         }
1981         nfs_release_seqid(calldata->arg.seqid);
1982         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1983 }
1984
1985 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1986 {
1987         struct nfs4_closedata *calldata = data;
1988         struct nfs4_state *state = calldata->state;
1989         int call_close = 0;
1990
1991         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1992                 return;
1993
1994         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1995         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1996         spin_lock(&state->owner->so_lock);
1997         /* Calculate the change in open mode */
1998         if (state->n_rdwr == 0) {
1999                 if (state->n_rdonly == 0) {
2000                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2001                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2002                         calldata->arg.fmode &= ~FMODE_READ;
2003                 }
2004                 if (state->n_wronly == 0) {
2005                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2006                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2007                         calldata->arg.fmode &= ~FMODE_WRITE;
2008                 }
2009         }
2010         spin_unlock(&state->owner->so_lock);
2011
2012         if (!call_close) {
2013                 /* Note: exit _without_ calling nfs4_close_done */
2014                 task->tk_action = NULL;
2015                 return;
2016         }
2017
2018         if (calldata->arg.fmode == 0) {
2019                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2020                 if (calldata->roc &&
2021                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2022                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2023                                      task, NULL);
2024                         return;
2025                 }
2026         }
2027
2028         nfs_fattr_init(calldata->res.fattr);
2029         calldata->timestamp = jiffies;
2030         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2031                                 &calldata->arg.seq_args, &calldata->res.seq_res,
2032                                 1, task))
2033                 return;
2034         rpc_call_start(task);
2035 }
2036
2037 static const struct rpc_call_ops nfs4_close_ops = {
2038         .rpc_call_prepare = nfs4_close_prepare,
2039         .rpc_call_done = nfs4_close_done,
2040         .rpc_release = nfs4_free_closedata,
2041 };
2042
2043 /*
2044  * It is possible for data to be read/written from a mem-mapped file
2045  * after the sys_close call (which hits the vfs layer as a flush).
2046  * This means that we can't safely call nfsv4 close on a file until
2047  * the inode is cleared. This in turn means that we are not good
2048  * NFSv4 citizens - we do not indicate to the server to update the file's
2049  * share state even when we are done with one of the three share
2050  * stateid's in the inode.
2051  *
2052  * NOTE: Caller must be holding the sp->so_owner semaphore!
2053  */
2054 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2055 {
2056         struct nfs_server *server = NFS_SERVER(state->inode);
2057         struct nfs4_closedata *calldata;
2058         struct nfs4_state_owner *sp = state->owner;
2059         struct rpc_task *task;
2060         struct rpc_message msg = {
2061                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2062                 .rpc_cred = state->owner->so_cred,
2063         };
2064         struct rpc_task_setup task_setup_data = {
2065                 .rpc_client = server->client,
2066                 .rpc_message = &msg,
2067                 .callback_ops = &nfs4_close_ops,
2068                 .workqueue = nfsiod_workqueue,
2069                 .flags = RPC_TASK_ASYNC,
2070         };
2071         int status = -ENOMEM;
2072
2073         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2074         if (calldata == NULL)
2075                 goto out;
2076         calldata->inode = state->inode;
2077         calldata->state = state;
2078         calldata->arg.fh = NFS_FH(state->inode);
2079         calldata->arg.stateid = &state->open_stateid;
2080         /* Serialization for the sequence id */
2081         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2082         if (calldata->arg.seqid == NULL)
2083                 goto out_free_calldata;
2084         calldata->arg.fmode = 0;
2085         calldata->arg.bitmask = server->cache_consistency_bitmask;
2086         calldata->res.fattr = &calldata->fattr;
2087         calldata->res.seqid = calldata->arg.seqid;
2088         calldata->res.server = server;
2089         calldata->roc = roc;
2090         nfs_sb_active(calldata->inode->i_sb);
2091
2092         msg.rpc_argp = &calldata->arg;
2093         msg.rpc_resp = &calldata->res;
2094         task_setup_data.callback_data = calldata;
2095         task = rpc_run_task(&task_setup_data);
2096         if (IS_ERR(task))
2097                 return PTR_ERR(task);
2098         status = 0;
2099         if (wait)
2100                 status = rpc_wait_for_completion_task(task);
2101         rpc_put_task(task);
2102         return status;
2103 out_free_calldata:
2104         kfree(calldata);
2105 out:
2106         if (roc)
2107                 pnfs_roc_release(state->inode);
2108         nfs4_put_open_state(state);
2109         nfs4_put_state_owner(sp);
2110         return status;
2111 }
2112
2113 static struct inode *
2114 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2115 {
2116         struct nfs4_state *state;
2117
2118         /* Protect against concurrent sillydeletes */
2119         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2120         if (IS_ERR(state))
2121                 return ERR_CAST(state);
2122         ctx->state = state;
2123         return igrab(state->inode);
2124 }
2125
2126 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2127 {
2128         if (ctx->state == NULL)
2129                 return;
2130         if (is_sync)
2131                 nfs4_close_sync(ctx->state, ctx->mode);
2132         else
2133                 nfs4_close_state(ctx->state, ctx->mode);
2134 }
2135
2136 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2137 {
2138         struct nfs4_server_caps_arg args = {
2139                 .fhandle = fhandle,
2140         };
2141         struct nfs4_server_caps_res res = {};
2142         struct rpc_message msg = {
2143                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2144                 .rpc_argp = &args,
2145                 .rpc_resp = &res,
2146         };
2147         int status;
2148
2149         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2150         if (status == 0) {
2151                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2152                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2153                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2154                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2155                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2156                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2157                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2158                         server->caps |= NFS_CAP_ACLS;
2159                 if (res.has_links != 0)
2160                         server->caps |= NFS_CAP_HARDLINKS;
2161                 if (res.has_symlinks != 0)
2162                         server->caps |= NFS_CAP_SYMLINKS;
2163                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2164                         server->caps |= NFS_CAP_FILEID;
2165                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2166                         server->caps |= NFS_CAP_MODE;
2167                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2168                         server->caps |= NFS_CAP_NLINK;
2169                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2170                         server->caps |= NFS_CAP_OWNER;
2171                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2172                         server->caps |= NFS_CAP_OWNER_GROUP;
2173                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2174                         server->caps |= NFS_CAP_ATIME;
2175                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2176                         server->caps |= NFS_CAP_CTIME;
2177                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2178                         server->caps |= NFS_CAP_MTIME;
2179
2180                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2181                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2182                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2183                 server->acl_bitmask = res.acl_bitmask;
2184         }
2185
2186         return status;
2187 }
2188
2189 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2190 {
2191         struct nfs4_exception exception = { };
2192         int err;
2193         do {
2194                 err = nfs4_handle_exception(server,
2195                                 _nfs4_server_capabilities(server, fhandle),
2196                                 &exception);
2197         } while (exception.retry);
2198         return err;
2199 }
2200
2201 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2202                 struct nfs_fsinfo *info)
2203 {
2204         struct nfs4_lookup_root_arg args = {
2205                 .bitmask = nfs4_fattr_bitmap,
2206         };
2207         struct nfs4_lookup_res res = {
2208                 .server = server,
2209                 .fattr = info->fattr,
2210                 .fh = fhandle,
2211         };
2212         struct rpc_message msg = {
2213                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2214                 .rpc_argp = &args,
2215                 .rpc_resp = &res,
2216         };
2217
2218         nfs_fattr_init(info->fattr);
2219         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2220 }
2221
2222 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2223                 struct nfs_fsinfo *info)
2224 {
2225         struct nfs4_exception exception = { };
2226         int err;
2227         do {
2228                 err = _nfs4_lookup_root(server, fhandle, info);
2229                 switch (err) {
2230                 case 0:
2231                 case -NFS4ERR_WRONGSEC:
2232                         break;
2233                 default:
2234                         err = nfs4_handle_exception(server, err, &exception);
2235                 }
2236         } while (exception.retry);
2237         return err;
2238 }
2239
2240 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2241                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2242 {
2243         struct rpc_auth *auth;
2244         int ret;
2245
2246         auth = rpcauth_create(flavor, server->client);
2247         if (!auth) {
2248                 ret = -EIO;
2249                 goto out;
2250         }
2251         ret = nfs4_lookup_root(server, fhandle, info);
2252 out:
2253         return ret;
2254 }
2255
2256 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2257                               struct nfs_fsinfo *info)
2258 {
2259         int i, len, status = 0;
2260         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2261
2262         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2263         flav_array[len] = RPC_AUTH_NULL;
2264         len += 1;
2265
2266         for (i = 0; i < len; i++) {
2267                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2268                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2269                         continue;
2270                 break;
2271         }
2272         /*
2273          * -EACCESS could mean that the user doesn't have correct permissions
2274          * to access the mount.  It could also mean that we tried to mount
2275          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2276          * existing mount programs don't handle -EACCES very well so it should
2277          * be mapped to -EPERM instead.
2278          */
2279         if (status == -EACCES)
2280                 status = -EPERM;
2281         return status;
2282 }
2283
2284 /*
2285  * get the file handle for the "/" directory on the server
2286  */
2287 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2288                               struct nfs_fsinfo *info)
2289 {
2290         int minor_version = server->nfs_client->cl_minorversion;
2291         int status = nfs4_lookup_root(server, fhandle, info);
2292         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2293                 /*
2294                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2295                  * by nfs4_map_errors() as this function exits.
2296                  */
2297                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2298         if (status == 0)
2299                 status = nfs4_server_capabilities(server, fhandle);
2300         if (status == 0)
2301                 status = nfs4_do_fsinfo(server, fhandle, info);
2302         return nfs4_map_errors(status);
2303 }
2304
2305 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
2306 /*
2307  * Get locations and (maybe) other attributes of a referral.
2308  * Note that we'll actually follow the referral later when
2309  * we detect fsid mismatch in inode revalidation
2310  */
2311 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2312                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2313 {
2314         int status = -ENOMEM;
2315         struct page *page = NULL;
2316         struct nfs4_fs_locations *locations = NULL;
2317
2318         page = alloc_page(GFP_KERNEL);
2319         if (page == NULL)
2320                 goto out;
2321         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2322         if (locations == NULL)
2323                 goto out;
2324
2325         status = nfs4_proc_fs_locations(dir, name, locations, page);
2326         if (status != 0)
2327                 goto out;
2328         /* Make sure server returned a different fsid for the referral */
2329         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2330                 dprintk("%s: server did not return a different fsid for"
2331                         " a referral at %s\n", __func__, name->name);
2332                 status = -EIO;
2333                 goto out;
2334         }
2335         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2336         nfs_fixup_referral_attributes(&locations->fattr);
2337
2338         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2339         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2340         memset(fhandle, 0, sizeof(struct nfs_fh));
2341 out:
2342         if (page)
2343                 __free_page(page);
2344         kfree(locations);
2345         return status;
2346 }
2347
2348 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2349 {
2350         struct nfs4_getattr_arg args = {
2351                 .fh = fhandle,
2352                 .bitmask = server->attr_bitmask,
2353         };
2354         struct nfs4_getattr_res res = {
2355                 .fattr = fattr,
2356                 .server = server,
2357         };
2358         struct rpc_message msg = {
2359                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2360                 .rpc_argp = &args,
2361                 .rpc_resp = &res,
2362         };
2363
2364         nfs_fattr_init(fattr);
2365         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2366 }
2367
2368 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2369 {
2370         struct nfs4_exception exception = { };
2371         int err;
2372         do {
2373                 err = nfs4_handle_exception(server,
2374                                 _nfs4_proc_getattr(server, fhandle, fattr),
2375                                 &exception);
2376         } while (exception.retry);
2377         return err;
2378 }
2379
2380 /*
2381  * The file is not closed if it is opened due to the a request to change
2382  * the size of the file. The open call will not be needed once the
2383  * VFS layer lookup-intents are implemented.
2384  *
2385  * Close is called when the inode is destroyed.
2386  * If we haven't opened the file for O_WRONLY, we
2387  * need to in the size_change case to obtain a stateid.
2388  *
2389  * Got race?
2390  * Because OPEN is always done by name in nfsv4, it is
2391  * possible that we opened a different file by the same
2392  * name.  We can recognize this race condition, but we
2393  * can't do anything about it besides returning an error.
2394  *
2395  * This will be fixed with VFS changes (lookup-intent).
2396  */
2397 static int
2398 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2399                   struct iattr *sattr)
2400 {
2401         struct inode *inode = dentry->d_inode;
2402         struct rpc_cred *cred = NULL;
2403         struct nfs4_state *state = NULL;
2404         int status;
2405
2406         if (pnfs_ld_layoutret_on_setattr(inode))
2407                 pnfs_return_layout(inode);
2408
2409         nfs_fattr_init(fattr);
2410
2411         /* Search for an existing open(O_WRITE) file */
2412         if (sattr->ia_valid & ATTR_FILE) {
2413                 struct nfs_open_context *ctx;
2414
2415                 ctx = nfs_file_open_context(sattr->ia_file);
2416                 if (ctx) {
2417                         cred = ctx->cred;
2418                         state = ctx->state;
2419                 }
2420         }
2421
2422         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2423         if (status == 0)
2424                 nfs_setattr_update_inode(inode, sattr);
2425         return status;
2426 }
2427
2428 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2429                 const struct qstr *name, struct nfs_fh *fhandle,
2430                 struct nfs_fattr *fattr)
2431 {
2432         struct nfs_server *server = NFS_SERVER(dir);
2433         int                    status;
2434         struct nfs4_lookup_arg args = {
2435                 .bitmask = server->attr_bitmask,
2436                 .dir_fh = NFS_FH(dir),
2437                 .name = name,
2438         };
2439         struct nfs4_lookup_res res = {
2440                 .server = server,
2441                 .fattr = fattr,
2442                 .fh = fhandle,
2443         };
2444         struct rpc_message msg = {
2445                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2446                 .rpc_argp = &args,
2447                 .rpc_resp = &res,
2448         };
2449
2450         nfs_fattr_init(fattr);
2451
2452         dprintk("NFS call  lookup %s\n", name->name);
2453         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2454         dprintk("NFS reply lookup: %d\n", status);
2455         return status;
2456 }
2457
2458 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2459 {
2460         memset(fh, 0, sizeof(struct nfs_fh));
2461         fattr->fsid.major = 1;
2462         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2463                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2464         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2465         fattr->nlink = 2;
2466 }
2467
2468 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2469                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2470 {
2471         struct nfs4_exception exception = { };
2472         int err;
2473         do {
2474                 int status;
2475
2476                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2477                 switch (status) {
2478                 case -NFS4ERR_BADNAME:
2479                         return -ENOENT;
2480                 case -NFS4ERR_MOVED:
2481                         return nfs4_get_referral(dir, name, fattr, fhandle);
2482                 case -NFS4ERR_WRONGSEC:
2483                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2484                 }
2485                 err = nfs4_handle_exception(NFS_SERVER(dir),
2486                                 status, &exception);
2487         } while (exception.retry);
2488         return err;
2489 }
2490
2491 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2492 {
2493         struct nfs_server *server = NFS_SERVER(inode);
2494         struct nfs4_accessargs args = {
2495                 .fh = NFS_FH(inode),
2496                 .bitmask = server->attr_bitmask,
2497         };
2498         struct nfs4_accessres res = {
2499                 .server = server,
2500         };
2501         struct rpc_message msg = {
2502                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2503                 .rpc_argp = &args,
2504                 .rpc_resp = &res,
2505                 .rpc_cred = entry->cred,
2506         };
2507         int mode = entry->mask;
2508         int status;
2509
2510         /*
2511          * Determine which access bits we want to ask for...
2512          */
2513         if (mode & MAY_READ)
2514                 args.access |= NFS4_ACCESS_READ;
2515         if (S_ISDIR(inode->i_mode)) {
2516                 if (mode & MAY_WRITE)
2517                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2518                 if (mode & MAY_EXEC)
2519                         args.access |= NFS4_ACCESS_LOOKUP;
2520         } else {
2521                 if (mode & MAY_WRITE)
2522                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2523                 if (mode & MAY_EXEC)
2524                         args.access |= NFS4_ACCESS_EXECUTE;
2525         }
2526
2527         res.fattr = nfs_alloc_fattr();
2528         if (res.fattr == NULL)
2529                 return -ENOMEM;
2530
2531         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2532         if (!status) {
2533                 entry->mask = 0;
2534                 if (res.access & NFS4_ACCESS_READ)
2535                         entry->mask |= MAY_READ;
2536                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2537                         entry->mask |= MAY_WRITE;
2538                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2539                         entry->mask |= MAY_EXEC;
2540                 nfs_refresh_inode(inode, res.fattr);
2541         }
2542         nfs_free_fattr(res.fattr);
2543         return status;
2544 }
2545
2546 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2547 {
2548         struct nfs4_exception exception = { };
2549         int err;
2550         do {
2551                 err = nfs4_handle_exception(NFS_SERVER(inode),
2552                                 _nfs4_proc_access(inode, entry),
2553                                 &exception);
2554         } while (exception.retry);
2555         return err;
2556 }
2557
2558 /*
2559  * TODO: For the time being, we don't try to get any attributes
2560  * along with any of the zero-copy operations READ, READDIR,
2561  * READLINK, WRITE.
2562  *
2563  * In the case of the first three, we want to put the GETATTR
2564  * after the read-type operation -- this is because it is hard
2565  * to predict the length of a GETATTR response in v4, and thus
2566  * align the READ data correctly.  This means that the GETATTR
2567  * may end up partially falling into the page cache, and we should
2568  * shift it into the 'tail' of the xdr_buf before processing.
2569  * To do this efficiently, we need to know the total length
2570  * of data received, which doesn't seem to be available outside
2571  * of the RPC layer.
2572  *
2573  * In the case of WRITE, we also want to put the GETATTR after
2574  * the operation -- in this case because we want to make sure
2575  * we get the post-operation mtime and size.  This means that
2576  * we can't use xdr_encode_pages() as written: we need a variant
2577  * of it which would leave room in the 'tail' iovec.
2578  *
2579  * Both of these changes to the XDR layer would in fact be quite
2580  * minor, but I decided to leave them for a subsequent patch.
2581  */
2582 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2583                 unsigned int pgbase, unsigned int pglen)
2584 {
2585         struct nfs4_readlink args = {
2586                 .fh       = NFS_FH(inode),
2587                 .pgbase   = pgbase,
2588                 .pglen    = pglen,
2589                 .pages    = &page,
2590         };
2591         struct nfs4_readlink_res res;
2592         struct rpc_message msg = {
2593                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2594                 .rpc_argp = &args,
2595                 .rpc_resp = &res,
2596         };
2597
2598         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2599 }
2600
2601 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2602                 unsigned int pgbase, unsigned int pglen)
2603 {
2604         struct nfs4_exception exception = { };
2605         int err;
2606         do {
2607                 err = nfs4_handle_exception(NFS_SERVER(inode),
2608                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2609                                 &exception);
2610         } while (exception.retry);
2611         return err;
2612 }
2613
2614 /*
2615  * Got race?
2616  * We will need to arrange for the VFS layer to provide an atomic open.
2617  * Until then, this create/open method is prone to inefficiency and race
2618  * conditions due to the lookup, create, and open VFS calls from sys_open()
2619  * placed on the wire.
2620  *
2621  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2622  * The file will be opened again in the subsequent VFS open call
2623  * (nfs4_proc_file_open).
2624  *
2625  * The open for read will just hang around to be used by any process that
2626  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2627  */
2628
2629 static int
2630 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2631                  int flags, struct nfs_open_context *ctx)
2632 {
2633         struct dentry *de = dentry;
2634         struct nfs4_state *state;
2635         struct rpc_cred *cred = NULL;
2636         fmode_t fmode = 0;
2637         int status = 0;
2638
2639         if (ctx != NULL) {
2640                 cred = ctx->cred;
2641                 de = ctx->dentry;
2642                 fmode = ctx->mode;
2643         }
2644         sattr->ia_mode &= ~current_umask();
2645         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2646         d_drop(dentry);
2647         if (IS_ERR(state)) {
2648                 status = PTR_ERR(state);
2649                 goto out;
2650         }
2651         d_add(dentry, igrab(state->inode));
2652         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2653         if (ctx != NULL)
2654                 ctx->state = state;
2655         else
2656                 nfs4_close_sync(state, fmode);
2657 out:
2658         return status;
2659 }
2660
2661 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2662 {
2663         struct nfs_server *server = NFS_SERVER(dir);
2664         struct nfs_removeargs args = {
2665                 .fh = NFS_FH(dir),
2666                 .name.len = name->len,
2667                 .name.name = name->name,
2668                 .bitmask = server->attr_bitmask,
2669         };
2670         struct nfs_removeres res = {
2671                 .server = server,
2672         };
2673         struct rpc_message msg = {
2674                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2675                 .rpc_argp = &args,
2676                 .rpc_resp = &res,
2677         };
2678         int status = -ENOMEM;
2679
2680         res.dir_attr = nfs_alloc_fattr();
2681         if (res.dir_attr == NULL)
2682                 goto out;
2683
2684         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2685         if (status == 0) {
2686                 update_changeattr(dir, &res.cinfo);
2687                 nfs_post_op_update_inode(dir, res.dir_attr);
2688         }
2689         nfs_free_fattr(res.dir_attr);
2690 out:
2691         return status;
2692 }
2693
2694 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2695 {
2696         struct nfs4_exception exception = { };
2697         int err;
2698         do {
2699                 err = nfs4_handle_exception(NFS_SERVER(dir),
2700                                 _nfs4_proc_remove(dir, name),
2701                                 &exception);
2702         } while (exception.retry);
2703         return err;
2704 }
2705
2706 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2707 {
2708         struct nfs_server *server = NFS_SERVER(dir);
2709         struct nfs_removeargs *args = msg->rpc_argp;
2710         struct nfs_removeres *res = msg->rpc_resp;
2711
2712         args->bitmask = server->cache_consistency_bitmask;
2713         res->server = server;
2714         res->seq_res.sr_slot = NULL;
2715         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2716 }
2717
2718 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2719 {
2720         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2721
2722         if (!nfs4_sequence_done(task, &res->seq_res))
2723                 return 0;
2724         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2725                 return 0;
2726         update_changeattr(dir, &res->cinfo);
2727         nfs_post_op_update_inode(dir, res->dir_attr);
2728         return 1;
2729 }
2730
2731 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2732 {
2733         struct nfs_server *server = NFS_SERVER(dir);
2734         struct nfs_renameargs *arg = msg->rpc_argp;
2735         struct nfs_renameres *res = msg->rpc_resp;
2736
2737         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2738         arg->bitmask = server->attr_bitmask;
2739         res->server = server;
2740 }
2741
2742 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2743                                  struct inode *new_dir)
2744 {
2745         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2746
2747         if (!nfs4_sequence_done(task, &res->seq_res))
2748                 return 0;
2749         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2750                 return 0;
2751
2752         update_changeattr(old_dir, &res->old_cinfo);
2753         nfs_post_op_update_inode(old_dir, res->old_fattr);
2754         update_changeattr(new_dir, &res->new_cinfo);
2755         nfs_post_op_update_inode(new_dir, res->new_fattr);
2756         return 1;
2757 }
2758
2759 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2760                 struct inode *new_dir, struct qstr *new_name)
2761 {
2762         struct nfs_server *server = NFS_SERVER(old_dir);
2763         struct nfs_renameargs arg = {
2764                 .old_dir = NFS_FH(old_dir),
2765                 .new_dir = NFS_FH(new_dir),
2766                 .old_name = old_name,
2767                 .new_name = new_name,
2768                 .bitmask = server->attr_bitmask,
2769         };
2770         struct nfs_renameres res = {
2771                 .server = server,
2772         };
2773         struct rpc_message msg = {
2774                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2775                 .rpc_argp = &arg,
2776                 .rpc_resp = &res,
2777         };
2778         int status = -ENOMEM;
2779
2780         res.old_fattr = nfs_alloc_fattr();
2781         res.new_fattr = nfs_alloc_fattr();
2782         if (res.old_fattr == NULL || res.new_fattr == NULL)
2783                 goto out;
2784
2785         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2786         if (!status) {
2787                 update_changeattr(old_dir, &res.old_cinfo);
2788                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2789                 update_changeattr(new_dir, &res.new_cinfo);
2790                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2791         }
2792 out:
2793         nfs_free_fattr(res.new_fattr);
2794         nfs_free_fattr(res.old_fattr);
2795         return status;
2796 }
2797
2798 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2799                 struct inode *new_dir, struct qstr *new_name)
2800 {
2801         struct nfs4_exception exception = { };
2802         int err;
2803         do {
2804                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2805                                 _nfs4_proc_rename(old_dir, old_name,
2806                                         new_dir, new_name),
2807                                 &exception);
2808         } while (exception.retry);
2809         return err;
2810 }
2811
2812 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2813 {
2814         struct nfs_server *server = NFS_SERVER(inode);
2815         struct nfs4_link_arg arg = {
2816                 .fh     = NFS_FH(inode),
2817                 .dir_fh = NFS_FH(dir),
2818                 .name   = name,
2819                 .bitmask = server->attr_bitmask,
2820         };
2821         struct nfs4_link_res res = {
2822                 .server = server,
2823         };
2824         struct rpc_message msg = {
2825                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2826                 .rpc_argp = &arg,
2827                 .rpc_resp = &res,
2828         };
2829         int status = -ENOMEM;
2830
2831         res.fattr = nfs_alloc_fattr();
2832         res.dir_attr = nfs_alloc_fattr();
2833         if (res.fattr == NULL || res.dir_attr == NULL)
2834                 goto out;
2835
2836         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2837         if (!status) {
2838                 update_changeattr(dir, &res.cinfo);
2839                 nfs_post_op_update_inode(dir, res.dir_attr);
2840                 nfs_post_op_update_inode(inode, res.fattr);
2841         }
2842 out:
2843         nfs_free_fattr(res.dir_attr);
2844         nfs_free_fattr(res.fattr);
2845         return status;
2846 }
2847
2848 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2849 {
2850         struct nfs4_exception exception = { };
2851         int err;
2852         do {
2853                 err = nfs4_handle_exception(NFS_SERVER(inode),
2854                                 _nfs4_proc_link(inode, dir, name),
2855                                 &exception);
2856         } while (exception.retry);
2857         return err;
2858 }
2859
2860 struct nfs4_createdata {
2861         struct rpc_message msg;
2862         struct nfs4_create_arg arg;
2863         struct nfs4_create_res res;
2864         struct nfs_fh fh;
2865         struct nfs_fattr fattr;
2866         struct nfs_fattr dir_fattr;
2867 };
2868
2869 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2870                 struct qstr *name, struct iattr *sattr, u32 ftype)
2871 {
2872         struct nfs4_createdata *data;
2873
2874         data = kzalloc(sizeof(*data), GFP_KERNEL);
2875         if (data != NULL) {
2876                 struct nfs_server *server = NFS_SERVER(dir);
2877
2878                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2879                 data->msg.rpc_argp = &data->arg;
2880                 data->msg.rpc_resp = &data->res;
2881                 data->arg.dir_fh = NFS_FH(dir);
2882                 data->arg.server = server;
2883                 data->arg.name = name;
2884                 data->arg.attrs = sattr;
2885                 data->arg.ftype = ftype;
2886                 data->arg.bitmask = server->attr_bitmask;
2887                 data->res.server = server;
2888                 data->res.fh = &data->fh;
2889                 data->res.fattr = &data->fattr;
2890                 data->res.dir_fattr = &data->dir_fattr;
2891                 nfs_fattr_init(data->res.fattr);
2892                 nfs_fattr_init(data->res.dir_fattr);
2893         }
2894         return data;
2895 }
2896
2897 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2898 {
2899         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2900                                     &data->arg.seq_args, &data->res.seq_res, 1);
2901         if (status == 0) {
2902                 update_changeattr(dir, &data->res.dir_cinfo);
2903                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2904                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2905         }
2906         return status;
2907 }
2908
2909 static void nfs4_free_createdata(struct nfs4_createdata *data)
2910 {
2911         kfree(data);
2912 }
2913
2914 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2915                 struct page *page, unsigned int len, struct iattr *sattr)
2916 {
2917         struct nfs4_createdata *data;
2918         int status = -ENAMETOOLONG;
2919
2920         if (len > NFS4_MAXPATHLEN)
2921                 goto out;
2922
2923         status = -ENOMEM;
2924         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2925         if (data == NULL)
2926                 goto out;
2927
2928         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2929         data->arg.u.symlink.pages = &page;
2930         data->arg.u.symlink.len = len;
2931
2932         status = nfs4_do_create(dir, dentry, data);
2933
2934         nfs4_free_createdata(data);
2935 out:
2936         return status;
2937 }
2938
2939 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2940                 struct page *page, unsigned int len, struct iattr *sattr)
2941 {
2942         struct nfs4_exception exception = { };
2943         int err;
2944         do {
2945                 err = nfs4_handle_exception(NFS_SERVER(dir),
2946                                 _nfs4_proc_symlink(dir, dentry, page,
2947                                                         len, sattr),
2948                                 &exception);
2949         } while (exception.retry);
2950         return err;
2951 }
2952
2953 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2954                 struct iattr *sattr)
2955 {
2956         struct nfs4_createdata *data;
2957         int status = -ENOMEM;
2958
2959         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2960         if (data == NULL)
2961                 goto out;
2962
2963         status = nfs4_do_create(dir, dentry, data);
2964
2965         nfs4_free_createdata(data);
2966 out:
2967         return status;
2968 }
2969
2970 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2971                 struct iattr *sattr)
2972 {
2973         struct nfs4_exception exception = { };
2974         int err;
2975
2976         sattr->ia_mode &= ~current_umask();
2977         do {
2978                 err = nfs4_handle_exception(NFS_SERVER(dir),
2979                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2980                                 &exception);
2981         } while (exception.retry);
2982         return err;
2983 }
2984
2985 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2986                 u64 cookie, struct page **pages, unsigned int count, int plus)
2987 {
2988         struct inode            *dir = dentry->d_inode;
2989         struct nfs4_readdir_arg args = {
2990                 .fh = NFS_FH(dir),
2991                 .pages = pages,
2992                 .pgbase = 0,
2993                 .count = count,
2994                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2995                 .plus = plus,
2996         };
2997         struct nfs4_readdir_res res;
2998         struct rpc_message msg = {
2999                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3000                 .rpc_argp = &args,
3001                 .rpc_resp = &res,
3002                 .rpc_cred = cred,
3003         };
3004         int                     status;
3005
3006         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3007                         dentry->d_parent->d_name.name,
3008                         dentry->d_name.name,
3009                         (unsigned long long)cookie);
3010         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3011         res.pgbase = args.pgbase;
3012         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3013         if (status >= 0) {
3014                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3015                 status += args.pgbase;
3016         }
3017
3018         nfs_invalidate_atime(dir);
3019
3020         dprintk("%s: returns %d\n", __func__, status);
3021         return status;
3022 }
3023
3024 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3025                 u64 cookie, struct page **pages, unsigned int count, int plus)
3026 {
3027         struct nfs4_exception exception = { };
3028         int err;
3029         do {
3030                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3031                                 _nfs4_proc_readdir(dentry, cred, cookie,
3032                                         pages, count, plus),
3033                                 &exception);
3034         } while (exception.retry);
3035         return err;
3036 }
3037
3038 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3039                 struct iattr *sattr, dev_t rdev)
3040 {
3041         struct nfs4_createdata *data;
3042         int mode = sattr->ia_mode;
3043         int status = -ENOMEM;
3044
3045         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3046         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3047
3048         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3049         if (data == NULL)
3050                 goto out;
3051
3052         if (S_ISFIFO(mode))
3053                 data->arg.ftype = NF4FIFO;
3054         else if (S_ISBLK(mode)) {
3055                 data->arg.ftype = NF4BLK;
3056                 data->arg.u.device.specdata1 = MAJOR(rdev);
3057                 data->arg.u.device.specdata2 = MINOR(rdev);
3058         }
3059         else if (S_ISCHR(mode)) {
3060                 data->arg.ftype = NF4CHR;
3061                 data->arg.u.device.specdata1 = MAJOR(rdev);
3062                 data->arg.u.device.specdata2 = MINOR(rdev);
3063         }
3064
3065         status = nfs4_do_create(dir, dentry, data);
3066
3067         nfs4_free_createdata(data);
3068 out:
3069         return status;
3070 }
3071
3072 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3073                 struct iattr *sattr, dev_t rdev)
3074 {
3075         struct nfs4_exception exception = { };
3076         int err;
3077
3078         sattr->ia_mode &= ~current_umask();
3079         do {
3080                 err = nfs4_handle_exception(NFS_SERVER(dir),
3081                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3082                                 &exception);
3083         } while (exception.retry);
3084         return err;
3085 }
3086
3087 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3088                  struct nfs_fsstat *fsstat)
3089 {
3090         struct nfs4_statfs_arg args = {
3091                 .fh = fhandle,
3092                 .bitmask = server->attr_bitmask,
3093         };
3094         struct nfs4_statfs_res res = {
3095                 .fsstat = fsstat,
3096         };
3097         struct rpc_message msg = {
3098                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3099                 .rpc_argp = &args,
3100                 .rpc_resp = &res,
3101         };
3102
3103         nfs_fattr_init(fsstat->fattr);
3104         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3105 }
3106
3107 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3108 {
3109         struct nfs4_exception exception = { };
3110         int err;
3111         do {
3112                 err = nfs4_handle_exception(server,
3113                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3114                                 &exception);
3115         } while (exception.retry);
3116         return err;
3117 }
3118
3119 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3120                 struct nfs_fsinfo *fsinfo)
3121 {
3122         struct nfs4_fsinfo_arg args = {
3123                 .fh = fhandle,
3124                 .bitmask = server->attr_bitmask,
3125         };
3126         struct nfs4_fsinfo_res res = {
3127                 .fsinfo = fsinfo,
3128         };
3129         struct rpc_message msg = {
3130                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3131                 .rpc_argp = &args,
3132                 .rpc_resp = &res,
3133         };
3134
3135         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3136 }
3137
3138 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3139 {
3140         struct nfs4_exception exception = { };
3141         int err;
3142
3143         do {
3144                 err = nfs4_handle_exception(server,
3145                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3146                                 &exception);
3147         } while (exception.retry);
3148         return err;
3149 }
3150
3151 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3152 {
3153         nfs_fattr_init(fsinfo->fattr);
3154         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3155 }
3156
3157 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3158                 struct nfs_pathconf *pathconf)
3159 {
3160         struct nfs4_pathconf_arg args = {
3161                 .fh = fhandle,
3162                 .bitmask = server->attr_bitmask,
3163         };
3164         struct nfs4_pathconf_res res = {
3165                 .pathconf = pathconf,
3166         };
3167         struct rpc_message msg = {
3168                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3169                 .rpc_argp = &args,
3170                 .rpc_resp = &res,
3171         };
3172
3173         /* None of the pathconf attributes are mandatory to implement */
3174         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3175                 memset(pathconf, 0, sizeof(*pathconf));
3176                 return 0;
3177         }
3178
3179         nfs_fattr_init(pathconf->fattr);
3180         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3181 }
3182
3183 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3184                 struct nfs_pathconf *pathconf)
3185 {
3186         struct nfs4_exception exception = { };
3187         int err;
3188
3189         do {
3190                 err = nfs4_handle_exception(server,
3191                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3192                                 &exception);
3193         } while (exception.retry);
3194         return err;
3195 }
3196
3197 void __nfs4_read_done_cb(struct nfs_read_data *data)
3198 {
3199         nfs_invalidate_atime(data->inode);
3200 }
3201
3202 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3203 {
3204         struct nfs_server *server = NFS_SERVER(data->inode);
3205
3206         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3207                 rpc_restart_call_prepare(task);
3208                 return -EAGAIN;
3209         }
3210
3211         __nfs4_read_done_cb(data);
3212         if (task->tk_status > 0)
3213                 renew_lease(server, data->timestamp);
3214         return 0;
3215 }
3216
3217 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3218 {
3219
3220         dprintk("--> %s\n", __func__);
3221
3222         if (!nfs4_sequence_done(task, &data->res.seq_res))
3223                 return -EAGAIN;
3224
3225         return data->read_done_cb ? data->read_done_cb(task, data) :
3226                                     nfs4_read_done_cb(task, data);
3227 }
3228
3229 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3230 {
3231         data->timestamp   = jiffies;
3232         data->read_done_cb = nfs4_read_done_cb;
3233         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3234 }
3235
3236 /* Reset the the nfs_read_data to send the read to the MDS. */
3237 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3238 {
3239         dprintk("%s Reset task for i/o through\n", __func__);
3240         put_lseg(data->lseg);
3241         data->lseg = NULL;
3242         /* offsets will differ in the dense stripe case */
3243         data->args.offset = data->mds_offset;
3244         data->ds_clp = NULL;
3245         data->args.fh     = NFS_FH(data->inode);
3246         data->read_done_cb = nfs4_read_done_cb;
3247         task->tk_ops = data->mds_ops;
3248         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3249 }
3250 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3251
3252 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3253 {
3254         struct inode *inode = data->inode;
3255
3256         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3257                 rpc_restart_call_prepare(task);
3258                 return -EAGAIN;
3259         }
3260         if (task->tk_status >= 0) {
3261                 renew_lease(NFS_SERVER(inode), data->timestamp);
3262                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3263         }
3264         return 0;
3265 }
3266
3267 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3268 {
3269         if (!nfs4_sequence_done(task, &data->res.seq_res))
3270                 return -EAGAIN;
3271         return data->write_done_cb ? data->write_done_cb(task, data) :
3272                 nfs4_write_done_cb(task, data);
3273 }
3274
3275 /* Reset the the nfs_write_data to send the write to the MDS. */
3276 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3277 {
3278         dprintk("%s Reset task for i/o through\n", __func__);
3279         put_lseg(data->lseg);
3280         data->lseg          = NULL;
3281         data->ds_clp        = NULL;
3282         data->write_done_cb = nfs4_write_done_cb;
3283         data->args.fh       = NFS_FH(data->inode);
3284         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3285         data->args.offset   = data->mds_offset;
3286         data->res.fattr     = &data->fattr;
3287         task->tk_ops        = data->mds_ops;
3288         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3289 }
3290 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3291
3292 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3293 {
3294         struct nfs_server *server = NFS_SERVER(data->inode);
3295
3296         if (data->lseg) {
3297                 data->args.bitmask = NULL;
3298                 data->res.fattr = NULL;
3299         } else
3300                 data->args.bitmask = server->cache_consistency_bitmask;
3301         if (!data->write_done_cb)
3302                 data->write_done_cb = nfs4_write_done_cb;
3303         data->res.server = server;
3304         data->timestamp   = jiffies;
3305
3306         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3307 }
3308
3309 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3310 {
3311         struct inode *inode = data->inode;
3312
3313         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3314                 rpc_restart_call_prepare(task);
3315                 return -EAGAIN;
3316         }
3317         nfs_refresh_inode(inode, data->res.fattr);
3318         return 0;
3319 }
3320
3321 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3322 {
3323         if (!nfs4_sequence_done(task, &data->res.seq_res))
3324                 return -EAGAIN;
3325         return data->write_done_cb(task, data);
3326 }
3327
3328 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3329 {
3330         struct nfs_server *server = NFS_SERVER(data->inode);
3331
3332         if (data->lseg) {
3333                 data->args.bitmask = NULL;
3334                 data->res.fattr = NULL;
3335         } else
3336                 data->args.bitmask = server->cache_consistency_bitmask;
3337         if (!data->write_done_cb)
3338                 data->write_done_cb = nfs4_commit_done_cb;
3339         data->res.server = server;
3340         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3341 }
3342
3343 struct nfs4_renewdata {
3344         struct nfs_client       *client;
3345         unsigned long           timestamp;
3346 };
3347
3348 /*
3349  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3350  * standalone procedure for queueing an asynchronous RENEW.
3351  */
3352 static void nfs4_renew_release(void *calldata)
3353 {
3354         struct nfs4_renewdata *data = calldata;
3355         struct nfs_client *clp = data->client;
3356
3357         if (atomic_read(&clp->cl_count) > 1)
3358                 nfs4_schedule_state_renewal(clp);
3359         nfs_put_client(clp);
3360         kfree(data);
3361 }
3362
3363 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3364 {
3365         struct nfs4_renewdata *data = calldata;
3366         struct nfs_client *clp = data->client;
3367         unsigned long timestamp = data->timestamp;
3368
3369         if (task->tk_status < 0) {
3370                 /* Unless we're shutting down, schedule state recovery! */
3371                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3372                         return;
3373                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3374                         nfs4_schedule_lease_recovery(clp);
3375                         return;
3376                 }
3377                 nfs4_schedule_path_down_recovery(clp);
3378         }
3379         do_renew_lease(clp, timestamp);
3380 }
3381
3382 static const struct rpc_call_ops nfs4_renew_ops = {
3383         .rpc_call_done = nfs4_renew_done,
3384         .rpc_release = nfs4_renew_release,
3385 };
3386
3387 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3388 {
3389         struct rpc_message msg = {
3390                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3391                 .rpc_argp       = clp,
3392                 .rpc_cred       = cred,
3393         };
3394         struct nfs4_renewdata *data;
3395
3396         if (renew_flags == 0)
3397                 return 0;
3398         if (!atomic_inc_not_zero(&clp->cl_count))
3399                 return -EIO;
3400         data = kmalloc(sizeof(*data), GFP_NOFS);
3401         if (data == NULL)
3402                 return -ENOMEM;
3403         data->client = clp;
3404         data->timestamp = jiffies;
3405         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3406                         &nfs4_renew_ops, data);
3407 }
3408
3409 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3410 {
3411         struct rpc_message msg = {
3412                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3413                 .rpc_argp       = clp,
3414                 .rpc_cred       = cred,
3415         };
3416         unsigned long now = jiffies;
3417         int status;
3418
3419         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3420         if (status < 0)
3421                 return status;
3422         do_renew_lease(clp, now);
3423         return 0;
3424 }
3425
3426 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3427 {
3428         return (server->caps & NFS_CAP_ACLS)
3429                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3430                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3431 }
3432
3433 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3434  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3435  * the stack.
3436  */
3437 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3438
3439 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3440                 struct page **pages, unsigned int *pgbase)
3441 {
3442         struct page *newpage, **spages;
3443         int rc = 0;
3444         size_t len;
3445         spages = pages;
3446
3447         do {
3448                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3449                 newpage = alloc_page(GFP_KERNEL);
3450
3451                 if (newpage == NULL)
3452                         goto unwind;
3453                 memcpy(page_address(newpage), buf, len);
3454                 buf += len;
3455                 buflen -= len;
3456                 *pages++ = newpage;
3457                 rc++;
3458         } while (buflen != 0);
3459
3460         return rc;
3461
3462 unwind:
3463         for(; rc > 0; rc--)
3464                 __free_page(spages[rc-1]);
3465         return -ENOMEM;
3466 }
3467
3468 struct nfs4_cached_acl {
3469         int cached;
3470         size_t len;
3471         char data[0];
3472 };
3473
3474 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3475 {
3476         struct nfs_inode *nfsi = NFS_I(inode);
3477
3478         spin_lock(&inode->i_lock);
3479         kfree(nfsi->nfs4_acl);
3480         nfsi->nfs4_acl = acl;
3481         spin_unlock(&inode->i_lock);
3482 }
3483
3484 static void nfs4_zap_acl_attr(struct inode *inode)
3485 {
3486         nfs4_set_cached_acl(inode, NULL);
3487 }
3488
3489 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3490 {
3491         struct nfs_inode *nfsi = NFS_I(inode);
3492         struct nfs4_cached_acl *acl;
3493         int ret = -ENOENT;
3494
3495         spin_lock(&inode->i_lock);
3496         acl = nfsi->nfs4_acl;
3497         if (acl == NULL)
3498                 goto out;
3499         if (buf == NULL) /* user is just asking for length */
3500                 goto out_len;
3501         if (acl->cached == 0)
3502                 goto out;
3503         ret = -ERANGE; /* see getxattr(2) man page */
3504         if (acl->len > buflen)
3505                 goto out;
3506         memcpy(buf, acl->data, acl->len);
3507 out_len:
3508         ret = acl->len;
3509 out:
3510         spin_unlock(&inode->i_lock);
3511         return ret;
3512 }
3513
3514 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3515 {
3516         struct nfs4_cached_acl *acl;
3517
3518         if (buf && acl_len <= PAGE_SIZE) {
3519                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3520                 if (acl == NULL)
3521                         goto out;
3522                 acl->cached = 1;
3523                 memcpy(acl->data, buf, acl_len);
3524         } else {
3525                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3526                 if (acl == NULL)
3527                         goto out;
3528                 acl->cached = 0;
3529         }
3530         acl->len = acl_len;
3531 out:
3532         nfs4_set_cached_acl(inode, acl);
3533 }
3534
3535 /*
3536  * The getxattr API returns the required buffer length when called with a
3537  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3538  * the required buf.  On a NULL buf, we send a page of data to the server
3539  * guessing that the ACL request can be serviced by a page. If so, we cache
3540  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3541  * the cache. If not so, we throw away the page, and cache the required
3542  * length. The next getxattr call will then produce another round trip to
3543  * the server, this time with the input buf of the required size.
3544  */
3545 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3546 {
3547         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3548         struct nfs_getaclargs args = {
3549                 .fh = NFS_FH(inode),
3550                 .acl_pages = pages,
3551                 .acl_len = buflen,
3552         };
3553         struct nfs_getaclres res = {
3554                 .acl_len = buflen,
3555         };
3556         void *resp_buf;
3557         struct rpc_message msg = {
3558                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3559                 .rpc_argp = &args,
3560                 .rpc_resp = &res,
3561         };
3562         int ret = -ENOMEM, npages, i, acl_len = 0;
3563
3564         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3565         /* As long as we're doing a round trip to the server anyway,
3566          * let's be prepared for a page of acl data. */
3567         if (npages == 0)
3568                 npages = 1;
3569
3570         for (i = 0; i < npages; i++) {
3571                 pages[i] = alloc_page(GFP_KERNEL);
3572                 if (!pages[i])
3573                         goto out_free;
3574         }
3575         if (npages > 1) {
3576                 /* for decoding across pages */
3577                 args.acl_scratch = alloc_page(GFP_KERNEL);
3578                 if (!args.acl_scratch)
3579                         goto out_free;
3580         }
3581         args.acl_len = npages * PAGE_SIZE;
3582         args.acl_pgbase = 0;
3583         /* Let decode_getfacl know not to fail if the ACL data is larger than
3584          * the page we send as a guess */
3585         if (buf == NULL)
3586                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3587         resp_buf = page_address(pages[0]);
3588
3589         dprintk("%s  buf %p buflen %ld npages %d args.acl_len %ld\n",
3590                 __func__, buf, buflen, npages, args.acl_len);
3591         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3592                              &msg, &args.seq_args, &res.seq_res, 0);
3593         if (ret)
3594                 goto out_free;
3595
3596         acl_len = res.acl_len - res.acl_data_offset;
3597         if (acl_len > args.acl_len)
3598                 nfs4_write_cached_acl(inode, NULL, acl_len);
3599         else
3600                 nfs4_write_cached_acl(inode, resp_buf + res.acl_data_offset,
3601                                       acl_len);
3602         if (buf) {
3603                 ret = -ERANGE;
3604                 if (acl_len > buflen)
3605                         goto out_free;
3606                 _copy_from_pages(buf, pages, res.acl_data_offset,
3607                                 res.acl_len);
3608         }
3609         ret = acl_len;
3610 out_free:
3611         for (i = 0; i < npages; i++)
3612                 if (pages[i])
3613                         __free_page(pages[i]);
3614         if (args.acl_scratch)
3615                 __free_page(args.acl_scratch);
3616         return ret;
3617 }
3618
3619 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3620 {
3621         struct nfs4_exception exception = { };
3622         ssize_t ret;
3623         do {
3624                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3625                 if (ret >= 0)
3626                         break;
3627                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3628         } while (exception.retry);
3629         return ret;
3630 }
3631
3632 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3633 {
3634         struct nfs_server *server = NFS_SERVER(inode);
3635         int ret;
3636
3637         if (!nfs4_server_supports_acls(server))
3638                 return -EOPNOTSUPP;
3639         ret = nfs_revalidate_inode(server, inode);
3640         if (ret < 0)
3641                 return ret;
3642         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3643                 nfs_zap_acl_cache(inode);
3644         ret = nfs4_read_cached_acl(inode, buf, buflen);
3645         if (ret != -ENOENT)
3646                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3647                  * but no cached acl data, just the acl length */
3648                 return ret;
3649         return nfs4_get_acl_uncached(inode, buf, buflen);
3650 }
3651
3652 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3653 {
3654         struct nfs_server *server = NFS_SERVER(inode);
3655         struct page *pages[NFS4ACL_MAXPAGES];
3656         struct nfs_setaclargs arg = {
3657                 .fh             = NFS_FH(inode),
3658                 .acl_pages      = pages,
3659                 .acl_len        = buflen,
3660         };
3661         struct nfs_setaclres res;
3662         struct rpc_message msg = {
3663                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3664                 .rpc_argp       = &arg,
3665                 .rpc_resp       = &res,
3666         };
3667         int ret, i;
3668
3669         if (!nfs4_server_supports_acls(server))
3670                 return -EOPNOTSUPP;
3671         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3672         if (i < 0)
3673                 return i;
3674         nfs_inode_return_delegation(inode);
3675         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3676
3677         /*
3678          * Free each page after tx, so the only ref left is
3679          * held by the network stack
3680          */
3681         for (; i > 0; i--)
3682                 put_page(pages[i-1]);
3683
3684         /*
3685          * Acl update can result in inode attribute update.
3686          * so mark the attribute cache invalid.
3687          */
3688         spin_lock(&inode->i_lock);
3689         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3690         spin_unlock(&inode->i_lock);
3691         nfs_access_zap_cache(inode);
3692         nfs_zap_acl_cache(inode);
3693         return ret;
3694 }
3695
3696 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3697 {
3698         struct nfs4_exception exception = { };
3699         int err;
3700         do {
3701                 err = nfs4_handle_exception(NFS_SERVER(inode),
3702                                 __nfs4_proc_set_acl(inode, buf, buflen),
3703                                 &exception);
3704         } while (exception.retry);
3705         return err;
3706 }
3707
3708 static int
3709 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3710 {
3711         struct nfs_client *clp = server->nfs_client;
3712
3713         if (task->tk_status >= 0)
3714                 return 0;
3715         switch(task->tk_status) {
3716                 case -NFS4ERR_ADMIN_REVOKED:
3717                 case -NFS4ERR_BAD_STATEID:
3718                 case -NFS4ERR_OPENMODE:
3719                         if (state == NULL)
3720                                 break;
3721                         nfs4_schedule_stateid_recovery(server, state);
3722                         goto wait_on_recovery;
3723                 case -NFS4ERR_EXPIRED:
3724                         if (state != NULL)
3725                                 nfs4_schedule_stateid_recovery(server, state);
3726                 case -NFS4ERR_STALE_STATEID:
3727                 case -NFS4ERR_STALE_CLIENTID:
3728                         nfs4_schedule_lease_recovery(clp);
3729                         goto wait_on_recovery;
3730 #if defined(CONFIG_NFS_V4_1)
3731                 case -NFS4ERR_BADSESSION:
3732                 case -NFS4ERR_BADSLOT:
3733                 case -NFS4ERR_BAD_HIGH_SLOT:
3734                 case -NFS4ERR_DEADSESSION:
3735                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3736                 case -NFS4ERR_SEQ_FALSE_RETRY:
3737                 case -NFS4ERR_SEQ_MISORDERED:
3738                         dprintk("%s ERROR %d, Reset session\n", __func__,
3739                                 task->tk_status);
3740                         nfs4_schedule_session_recovery(clp->cl_session);
3741                         task->tk_status = 0;
3742                         return -EAGAIN;
3743 #endif /* CONFIG_NFS_V4_1 */
3744                 case -NFS4ERR_DELAY:
3745                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3746                 case -NFS4ERR_GRACE:
3747                 case -EKEYEXPIRED:
3748                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3749                         task->tk_status = 0;
3750                         return -EAGAIN;
3751                 case -NFS4ERR_RETRY_UNCACHED_REP:
3752                 case -NFS4ERR_OLD_STATEID:
3753                         task->tk_status = 0;
3754                         return -EAGAIN;
3755         }
3756         task->tk_status = nfs4_map_errors(task->tk_status);
3757         return 0;
3758 wait_on_recovery:
3759         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3760         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3761                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3762         task->tk_status = 0;
3763         return -EAGAIN;
3764 }
3765
3766 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3767                 unsigned short port, struct rpc_cred *cred,
3768                 struct nfs4_setclientid_res *res)
3769 {
3770         nfs4_verifier sc_verifier;
3771         struct nfs4_setclientid setclientid = {
3772                 .sc_verifier = &sc_verifier,
3773                 .sc_prog = program,
3774                 .sc_cb_ident = clp->cl_cb_ident,
3775         };
3776         struct rpc_message msg = {
3777                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3778                 .rpc_argp = &setclientid,
3779                 .rpc_resp = res,
3780                 .rpc_cred = cred,
3781         };
3782         __be32 *p;
3783         int loop = 0;
3784         int status;
3785
3786         p = (__be32*)sc_verifier.data;
3787         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3788         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3789
3790         for(;;) {
3791                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3792                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3793                                 clp->cl_ipaddr,
3794                                 rpc_peeraddr2str(clp->cl_rpcclient,
3795                                                         RPC_DISPLAY_ADDR),
3796                                 rpc_peeraddr2str(clp->cl_rpcclient,
3797                                                         RPC_DISPLAY_PROTO),
3798                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3799                                 clp->cl_id_uniquifier);
3800                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3801                                 sizeof(setclientid.sc_netid),
3802                                 rpc_peeraddr2str(clp->cl_rpcclient,
3803                                                         RPC_DISPLAY_NETID));
3804                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3805                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3806                                 clp->cl_ipaddr, port >> 8, port & 255);
3807
3808                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3809                 if (status != -NFS4ERR_CLID_INUSE)
3810                         break;
3811                 if (loop != 0) {
3812                         ++clp->cl_id_uniquifier;
3813                         break;
3814                 }
3815                 ++loop;
3816                 ssleep(clp->cl_lease_time / HZ + 1);
3817         }
3818         return status;
3819 }
3820
3821 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3822                 struct nfs4_setclientid_res *arg,
3823                 struct rpc_cred *cred)
3824 {
3825         struct nfs_fsinfo fsinfo;
3826         struct rpc_message msg = {
3827                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3828                 .rpc_argp = arg,
3829                 .rpc_resp = &fsinfo,
3830                 .rpc_cred = cred,
3831         };
3832         unsigned long now;
3833         int status;
3834
3835         now = jiffies;
3836         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3837         if (status == 0) {
3838                 spin_lock(&clp->cl_lock);
3839                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3840                 clp->cl_last_renewal = now;
3841                 spin_unlock(&clp->cl_lock);
3842         }
3843         return status;
3844 }
3845
3846 struct nfs4_delegreturndata {
3847         struct nfs4_delegreturnargs args;
3848         struct nfs4_delegreturnres res;
3849         struct nfs_fh fh;
3850         nfs4_stateid stateid;
3851         unsigned long timestamp;
3852         struct nfs_fattr fattr;
3853         int rpc_status;
3854 };
3855
3856 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3857 {
3858         struct nfs4_delegreturndata *data = calldata;
3859
3860         if (!nfs4_sequence_done(task, &data->res.seq_res))
3861                 return;
3862
3863         switch (task->tk_status) {
3864         case -NFS4ERR_STALE_STATEID:
3865         case -NFS4ERR_EXPIRED:
3866         case 0:
3867                 renew_lease(data->res.server, data->timestamp);
3868                 break;
3869         default:
3870                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3871                                 -EAGAIN) {
3872                         rpc_restart_call_prepare(task);
3873                         return;
3874                 }
3875         }
3876         data->rpc_status = task->tk_status;
3877 }
3878
3879 static void nfs4_delegreturn_release(void *calldata)
3880 {
3881         kfree(calldata);
3882 }
3883
3884 #if defined(CONFIG_NFS_V4_1)
3885 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3886 {
3887         struct nfs4_delegreturndata *d_data;
3888
3889         d_data = (struct nfs4_delegreturndata *)data;
3890
3891         if (nfs4_setup_sequence(d_data->res.server,
3892                                 &d_data->args.seq_args,
3893                                 &d_data->res.seq_res, 1, task))
3894                 return;
3895         rpc_call_start(task);
3896 }
3897 #endif /* CONFIG_NFS_V4_1 */
3898
3899 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3900 #if defined(CONFIG_NFS_V4_1)
3901         .rpc_call_prepare = nfs4_delegreturn_prepare,
3902 #endif /* CONFIG_NFS_V4_1 */
3903         .rpc_call_done = nfs4_delegreturn_done,
3904         .rpc_release = nfs4_delegreturn_release,
3905 };
3906
3907 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3908 {
3909         struct nfs4_delegreturndata *data;
3910         struct nfs_server *server = NFS_SERVER(inode);
3911         struct rpc_task *task;
3912         struct rpc_message msg = {
3913                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3914                 .rpc_cred = cred,
3915         };
3916         struct rpc_task_setup task_setup_data = {
3917                 .rpc_client = server->client,
3918                 .rpc_message = &msg,
3919                 .callback_ops = &nfs4_delegreturn_ops,
3920                 .flags = RPC_TASK_ASYNC,
3921         };
3922         int status = 0;
3923
3924         data = kzalloc(sizeof(*data), GFP_NOFS);
3925         if (data == NULL)
3926                 return -ENOMEM;
3927         data->args.fhandle = &data->fh;
3928         data->args.stateid = &data->stateid;
3929         data->args.bitmask = server->attr_bitmask;
3930         nfs_copy_fh(&data->fh, NFS_FH(inode));
3931         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3932         data->res.fattr = &data->fattr;
3933         data->res.server = server;
3934         nfs_fattr_init(data->res.fattr);
3935         data->timestamp = jiffies;
3936         data->rpc_status = 0;
3937
3938         task_setup_data.callback_data = data;
3939         msg.rpc_argp = &data->args;
3940         msg.rpc_resp = &data->res;
3941         task = rpc_run_task(&task_setup_data);
3942         if (IS_ERR(task))
3943                 return PTR_ERR(task);
3944         if (!issync)
3945                 goto out;
3946         status = nfs4_wait_for_completion_rpc_task(task);
3947         if (status != 0)
3948                 goto out;
3949         status = data->rpc_status;
3950         if (status != 0)
3951                 goto out;
3952         nfs_refresh_inode(inode, &data->fattr);
3953 out:
3954         rpc_put_task(task);
3955         return status;
3956 }
3957
3958 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3959 {
3960         struct nfs_server *server = NFS_SERVER(inode);
3961         struct nfs4_exception exception = { };
3962         int err;
3963         do {
3964                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3965                 switch (err) {
3966                         case -NFS4ERR_STALE_STATEID:
3967                         case -NFS4ERR_EXPIRED:
3968                         case 0:
3969                                 return 0;
3970                 }
3971                 err = nfs4_handle_exception(server, err, &exception);
3972         } while (exception.retry);
3973         return err;
3974 }
3975
3976 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3977 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3978
3979 /*
3980  * sleep, with exponential backoff, and retry the LOCK operation.
3981  */
3982 static unsigned long
3983 nfs4_set_lock_task_retry(unsigned long timeout)
3984 {
3985         schedule_timeout_killable(timeout);
3986         timeout <<= 1;
3987         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3988                 return NFS4_LOCK_MAXTIMEOUT;
3989         return timeout;
3990 }
3991
3992 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3993 {
3994         struct inode *inode = state->inode;
3995         struct nfs_server *server = NFS_SERVER(inode);
3996         struct nfs_client *clp = server->nfs_client;
3997         struct nfs_lockt_args arg = {
3998                 .fh = NFS_FH(inode),
3999                 .fl = request,
4000         };
4001         struct nfs_lockt_res res = {
4002                 .denied = request,
4003         };
4004         struct rpc_message msg = {
4005                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4006                 .rpc_argp       = &arg,
4007                 .rpc_resp       = &res,
4008                 .rpc_cred       = state->owner->so_cred,
4009         };
4010         struct nfs4_lock_state *lsp;
4011         int status;
4012
4013         arg.lock_owner.clientid = clp->cl_clientid;
4014         status = nfs4_set_lock_state(state, request);
4015         if (status != 0)
4016                 goto out;
4017         lsp = request->fl_u.nfs4_fl.owner;
4018         arg.lock_owner.id = lsp->ls_id.id;
4019         arg.lock_owner.s_dev = server->s_dev;
4020         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4021         switch (status) {
4022                 case 0:
4023                         request->fl_type = F_UNLCK;
4024                         break;
4025                 case -NFS4ERR_DENIED:
4026                         status = 0;
4027         }
4028         request->fl_ops->fl_release_private(request);
4029 out:
4030         return status;
4031 }
4032
4033 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4034 {
4035         struct nfs4_exception exception = { };
4036         int err;
4037
4038         do {
4039                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4040                                 _nfs4_proc_getlk(state, cmd, request),
4041                                 &exception);
4042         } while (exception.retry);
4043         return err;
4044 }
4045
4046 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4047 {
4048         int res = 0;
4049         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4050                 case FL_POSIX:
4051                         res = posix_lock_file_wait(file, fl);
4052                         break;
4053                 case FL_FLOCK:
4054                         res = flock_lock_file_wait(file, fl);
4055                         break;
4056                 default:
4057                         BUG();
4058         }
4059         return res;
4060 }
4061
4062 struct nfs4_unlockdata {
4063         struct nfs_locku_args arg;
4064         struct nfs_locku_res res;
4065         struct nfs4_lock_state *lsp;
4066         struct nfs_open_context *ctx;
4067         struct file_lock fl;
4068         const struct nfs_server *server;
4069         unsigned long timestamp;
4070 };
4071
4072 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4073                 struct nfs_open_context *ctx,
4074                 struct nfs4_lock_state *lsp,
4075                 struct nfs_seqid *seqid)
4076 {
4077         struct nfs4_unlockdata *p;
4078         struct inode *inode = lsp->ls_state->inode;
4079
4080         p = kzalloc(sizeof(*p), GFP_NOFS);
4081         if (p == NULL)
4082                 return NULL;
4083         p->arg.fh = NFS_FH(inode);
4084         p->arg.fl = &p->fl;
4085         p->arg.seqid = seqid;
4086         p->res.seqid = seqid;
4087         p->arg.stateid = &lsp->ls_stateid;
4088         p->lsp = lsp;
4089         atomic_inc(&lsp->ls_count);
4090         /* Ensure we don't close file until we're done freeing locks! */
4091         p->ctx = get_nfs_open_context(ctx);
4092         memcpy(&p->fl, fl, sizeof(p->fl));
4093         p->server = NFS_SERVER(inode);
4094         return p;
4095 }
4096
4097 static void nfs4_locku_release_calldata(void *data)
4098 {
4099         struct nfs4_unlockdata *calldata = data;
4100         nfs_free_seqid(calldata->arg.seqid);
4101         nfs4_put_lock_state(calldata->lsp);
4102         put_nfs_open_context(calldata->ctx);
4103         kfree(calldata);
4104 }
4105
4106 static void nfs4_locku_done(struct rpc_task *task, void *data)
4107 {
4108         struct nfs4_unlockdata *calldata = data;
4109
4110         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4111                 return;
4112         switch (task->tk_status) {
4113                 case 0:
4114                         memcpy(calldata->lsp->ls_stateid.data,
4115                                         calldata->res.stateid.data,
4116                                         sizeof(calldata->lsp->ls_stateid.data));
4117                         renew_lease(calldata->server, calldata->timestamp);
4118                         break;
4119                 case -NFS4ERR_BAD_STATEID:
4120                 case -NFS4ERR_OLD_STATEID:
4121                 case -NFS4ERR_STALE_STATEID:
4122                 case -NFS4ERR_EXPIRED:
4123                         break;
4124                 default:
4125                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4126                                 rpc_restart_call_prepare(task);
4127         }
4128 }
4129
4130 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4131 {
4132         struct nfs4_unlockdata *calldata = data;
4133
4134         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4135                 return;
4136         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4137                 /* Note: exit _without_ running nfs4_locku_done */
4138                 task->tk_action = NULL;
4139                 return;
4140         }
4141         calldata->timestamp = jiffies;
4142         if (nfs4_setup_sequence(calldata->server,
4143                                 &calldata->arg.seq_args,
4144                                 &calldata->res.seq_res, 1, task))
4145                 return;
4146         rpc_call_start(task);
4147 }
4148
4149 static const struct rpc_call_ops nfs4_locku_ops = {
4150         .rpc_call_prepare = nfs4_locku_prepare,
4151         .rpc_call_done = nfs4_locku_done,
4152         .rpc_release = nfs4_locku_release_calldata,
4153 };
4154
4155 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4156                 struct nfs_open_context *ctx,
4157                 struct nfs4_lock_state *lsp,
4158                 struct nfs_seqid *seqid)
4159 {
4160         struct nfs4_unlockdata *data;
4161         struct rpc_message msg = {
4162                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4163                 .rpc_cred = ctx->cred,
4164         };
4165         struct rpc_task_setup task_setup_data = {
4166                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4167                 .rpc_message = &msg,
4168                 .callback_ops = &nfs4_locku_ops,
4169                 .workqueue = nfsiod_workqueue,
4170                 .flags = RPC_TASK_ASYNC,
4171         };
4172
4173         /* Ensure this is an unlock - when canceling a lock, the
4174          * canceled lock is passed in, and it won't be an unlock.
4175          */
4176         fl->fl_type = F_UNLCK;
4177
4178         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4179         if (data == NULL) {
4180                 nfs_free_seqid(seqid);
4181                 return ERR_PTR(-ENOMEM);
4182         }
4183
4184         msg.rpc_argp = &data->arg;
4185         msg.rpc_resp = &data->res;
4186         task_setup_data.callback_data = data;
4187         return rpc_run_task(&task_setup_data);
4188 }
4189
4190 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4191 {
4192         struct nfs_inode *nfsi = NFS_I(state->inode);
4193         struct nfs_seqid *seqid;
4194         struct nfs4_lock_state *lsp;
4195         struct rpc_task *task;
4196         int status = 0;
4197         unsigned char fl_flags = request->fl_flags;
4198
4199         status = nfs4_set_lock_state(state, request);
4200         /* Unlock _before_ we do the RPC call */
4201         request->fl_flags |= FL_EXISTS;
4202         down_read(&nfsi->rwsem);
4203         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4204                 up_read(&nfsi->rwsem);
4205                 goto out;
4206         }
4207         up_read(&nfsi->rwsem);
4208         if (status != 0)
4209                 goto out;
4210         /* Is this a delegated lock? */
4211         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4212                 goto out;
4213         lsp = request->fl_u.nfs4_fl.owner;
4214         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4215         status = -ENOMEM;
4216         if (seqid == NULL)
4217                 goto out;
4218         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4219         status = PTR_ERR(task);
4220         if (IS_ERR(task))
4221                 goto out;
4222         status = nfs4_wait_for_completion_rpc_task(task);
4223         rpc_put_task(task);
4224 out:
4225         request->fl_flags = fl_flags;
4226         return status;
4227 }
4228
4229 struct nfs4_lockdata {
4230         struct nfs_lock_args arg;
4231         struct nfs_lock_res res;
4232         struct nfs4_lock_state *lsp;
4233         struct nfs_open_context *ctx;
4234         struct file_lock fl;
4235         unsigned long timestamp;
4236         int rpc_status;
4237         int cancelled;
4238         struct nfs_server *server;
4239 };
4240
4241 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4242                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4243                 gfp_t gfp_mask)
4244 {
4245         struct nfs4_lockdata *p;
4246         struct inode *inode = lsp->ls_state->inode;
4247         struct nfs_server *server = NFS_SERVER(inode);
4248
4249         p = kzalloc(sizeof(*p), gfp_mask);
4250         if (p == NULL)
4251                 return NULL;
4252
4253         p->arg.fh = NFS_FH(inode);
4254         p->arg.fl = &p->fl;
4255         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4256         if (p->arg.open_seqid == NULL)
4257                 goto out_free;
4258         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4259         if (p->arg.lock_seqid == NULL)
4260                 goto out_free_seqid;
4261         p->arg.lock_stateid = &lsp->ls_stateid;
4262         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4263         p->arg.lock_owner.id = lsp->ls_id.id;
4264         p->arg.lock_owner.s_dev = server->s_dev;
4265         p->res.lock_seqid = p->arg.lock_seqid;
4266         p->lsp = lsp;
4267         p->server = server;
4268         atomic_inc(&lsp->ls_count);
4269         p->ctx = get_nfs_open_context(ctx);
4270         memcpy(&p->fl, fl, sizeof(p->fl));
4271         return p;
4272 out_free_seqid:
4273         nfs_free_seqid(p->arg.open_seqid);
4274 out_free:
4275         kfree(p);
4276         return NULL;
4277 }
4278
4279 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4280 {
4281         struct nfs4_lockdata *data = calldata;
4282         struct nfs4_state *state = data->lsp->ls_state;
4283
4284         dprintk("%s: begin!\n", __func__);
4285         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4286                 return;
4287         /* Do we need to do an open_to_lock_owner? */
4288         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4289                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4290                         return;
4291                 data->arg.open_stateid = &state->stateid;
4292                 data->arg.new_lock_owner = 1;
4293                 data->res.open_seqid = data->arg.open_seqid;
4294         } else
4295                 data->arg.new_lock_owner = 0;
4296         data->timestamp = jiffies;
4297         if (nfs4_setup_sequence(data->server,
4298                                 &data->arg.seq_args,
4299                                 &data->res.seq_res, 1, task))
4300                 return;
4301         rpc_call_start(task);
4302         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4303 }
4304
4305 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4306 {
4307         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4308         nfs4_lock_prepare(task, calldata);
4309 }
4310
4311 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4312 {
4313         struct nfs4_lockdata *data = calldata;
4314
4315         dprintk("%s: begin!\n", __func__);
4316
4317         if (!nfs4_sequence_done(task, &data->res.seq_res))
4318                 return;
4319
4320         data->rpc_status = task->tk_status;
4321         if (data->arg.new_lock_owner != 0) {
4322                 if (data->rpc_status == 0)
4323                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4324                 else
4325                         goto out;
4326         }
4327         if (data->rpc_status == 0) {
4328                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4329                                         sizeof(data->lsp->ls_stateid.data));
4330                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4331                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4332         }
4333 out:
4334         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4335 }
4336
4337 static void nfs4_lock_release(void *calldata)
4338 {
4339         struct nfs4_lockdata *data = calldata;
4340
4341         dprintk("%s: begin!\n", __func__);
4342         nfs_free_seqid(data->arg.open_seqid);
4343         if (data->cancelled != 0) {
4344                 struct rpc_task *task;
4345                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4346                                 data->arg.lock_seqid);
4347                 if (!IS_ERR(task))
4348                         rpc_put_task_async(task);
4349                 dprintk("%s: cancelling lock!\n", __func__);
4350         } else
4351                 nfs_free_seqid(data->arg.lock_seqid);
4352         nfs4_put_lock_state(data->lsp);
4353         put_nfs_open_context(data->ctx);
4354         kfree(data);
4355         dprintk("%s: done!\n", __func__);
4356 }
4357
4358 static const struct rpc_call_ops nfs4_lock_ops = {
4359         .rpc_call_prepare = nfs4_lock_prepare,
4360         .rpc_call_done = nfs4_lock_done,
4361         .rpc_release = nfs4_lock_release,
4362 };
4363
4364 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4365         .rpc_call_prepare = nfs4_recover_lock_prepare,
4366         .rpc_call_done = nfs4_lock_done,
4367         .rpc_release = nfs4_lock_release,
4368 };
4369
4370 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4371 {
4372         switch (error) {
4373         case -NFS4ERR_ADMIN_REVOKED:
4374         case -NFS4ERR_BAD_STATEID:
4375                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4376                 if (new_lock_owner != 0 ||
4377                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4378                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4379                 break;
4380         case -NFS4ERR_STALE_STATEID:
4381                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4382         case -NFS4ERR_EXPIRED:
4383                 nfs4_schedule_lease_recovery(server->nfs_client);
4384         };
4385 }
4386
4387 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4388 {
4389         struct nfs4_lockdata *data;
4390         struct rpc_task *task;
4391         struct rpc_message msg = {
4392                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4393                 .rpc_cred = state->owner->so_cred,
4394         };
4395         struct rpc_task_setup task_setup_data = {
4396                 .rpc_client = NFS_CLIENT(state->inode),
4397                 .rpc_message = &msg,
4398                 .callback_ops = &nfs4_lock_ops,
4399                 .workqueue = nfsiod_workqueue,
4400                 .flags = RPC_TASK_ASYNC,
4401         };
4402         int ret;
4403
4404         dprintk("%s: begin!\n", __func__);
4405         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4406                         fl->fl_u.nfs4_fl.owner,
4407                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4408         if (data == NULL)
4409                 return -ENOMEM;
4410         if (IS_SETLKW(cmd))
4411                 data->arg.block = 1;
4412         if (recovery_type > NFS_LOCK_NEW) {
4413                 if (recovery_type == NFS_LOCK_RECLAIM)
4414                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4415                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4416         }
4417         msg.rpc_argp = &data->arg;
4418         msg.rpc_resp = &data->res;
4419         task_setup_data.callback_data = data;
4420         task = rpc_run_task(&task_setup_data);
4421         if (IS_ERR(task))
4422                 return PTR_ERR(task);
4423         ret = nfs4_wait_for_completion_rpc_task(task);
4424         if (ret == 0) {
4425                 ret = data->rpc_status;
4426                 if (ret)
4427                         nfs4_handle_setlk_error(data->server, data->lsp,
4428                                         data->arg.new_lock_owner, ret);
4429         } else
4430                 data->cancelled = 1;
4431         rpc_put_task(task);
4432         dprintk("%s: done, ret = %d!\n", __func__, ret);
4433         return ret;
4434 }
4435
4436 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4437 {
4438         struct nfs_server *server = NFS_SERVER(state->inode);
4439         struct nfs4_exception exception = { };
4440         int err;
4441
4442         do {
4443                 /* Cache the lock if possible... */
4444                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4445                         return 0;
4446                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4447                 if (err != -NFS4ERR_DELAY)
4448                         break;
4449                 nfs4_handle_exception(server, err, &exception);
4450         } while (exception.retry);
4451         return err;
4452 }
4453
4454 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4455 {
4456         struct nfs_server *server = NFS_SERVER(state->inode);
4457         struct nfs4_exception exception = { };
4458         int err;
4459
4460         err = nfs4_set_lock_state(state, request);
4461         if (err != 0)
4462                 return err;
4463         do {
4464                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4465                         return 0;
4466                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4467                 switch (err) {
4468                 default:
4469                         goto out;
4470                 case -NFS4ERR_GRACE:
4471                 case -NFS4ERR_DELAY:
4472                         nfs4_handle_exception(server, err, &exception);
4473                         err = 0;
4474                 }
4475         } while (exception.retry);
4476 out:
4477         return err;
4478 }
4479
4480 #if defined(CONFIG_NFS_V4_1)
4481 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4482 {
4483         int status;
4484         struct nfs_server *server = NFS_SERVER(state->inode);
4485
4486         status = nfs41_test_stateid(server, state);
4487         if (status == NFS_OK)
4488                 return 0;
4489         nfs41_free_stateid(server, state);
4490         return nfs4_lock_expired(state, request);
4491 }
4492 #endif
4493
4494 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4495 {
4496         struct nfs_inode *nfsi = NFS_I(state->inode);
4497         unsigned char fl_flags = request->fl_flags;
4498         int status = -ENOLCK;
4499
4500         if ((fl_flags & FL_POSIX) &&
4501                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4502                 goto out;
4503         /* Is this a delegated open? */
4504         status = nfs4_set_lock_state(state, request);
4505         if (status != 0)
4506                 goto out;
4507         request->fl_flags |= FL_ACCESS;
4508         status = do_vfs_lock(request->fl_file, request);
4509         if (status < 0)
4510                 goto out;
4511         down_read(&nfsi->rwsem);
4512         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4513                 /* Yes: cache locks! */
4514                 /* ...but avoid races with delegation recall... */
4515                 request->fl_flags = fl_flags & ~FL_SLEEP;
4516                 status = do_vfs_lock(request->fl_file, request);
4517                 goto out_unlock;
4518         }
4519         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4520         if (status != 0)
4521                 goto out_unlock;
4522         /* Note: we always want to sleep here! */
4523         request->fl_flags = fl_flags | FL_SLEEP;
4524         if (do_vfs_lock(request->fl_file, request) < 0)
4525                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4526 out_unlock:
4527         up_read(&nfsi->rwsem);
4528 out:
4529         request->fl_flags = fl_flags;
4530         return status;
4531 }
4532
4533 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4534 {
4535         struct nfs4_exception exception = { };
4536         int err;
4537
4538         do {
4539                 err = _nfs4_proc_setlk(state, cmd, request);
4540                 if (err == -NFS4ERR_DENIED)
4541                         err = -EAGAIN;
4542                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4543                                 err, &exception);
4544         } while (exception.retry);
4545         return err;
4546 }
4547
4548 static int
4549 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4550 {
4551         struct nfs_open_context *ctx;
4552         struct nfs4_state *state;
4553         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4554         int status;
4555
4556         /* verify open state */
4557         ctx = nfs_file_open_context(filp);
4558         state = ctx->state;
4559
4560         if (request->fl_start < 0 || request->fl_end < 0)
4561                 return -EINVAL;
4562
4563         if (IS_GETLK(cmd)) {
4564                 if (state != NULL)
4565                         return nfs4_proc_getlk(state, F_GETLK, request);
4566                 return 0;
4567         }
4568
4569         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4570                 return -EINVAL;
4571
4572         if (request->fl_type == F_UNLCK) {
4573                 if (state != NULL)
4574                         return nfs4_proc_unlck(state, cmd, request);
4575                 return 0;
4576         }
4577
4578         if (state == NULL)
4579                 return -ENOLCK;
4580         do {
4581                 status = nfs4_proc_setlk(state, cmd, request);
4582                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4583                         break;
4584                 timeout = nfs4_set_lock_task_retry(timeout);
4585                 status = -ERESTARTSYS;
4586                 if (signalled())
4587                         break;
4588         } while(status < 0);
4589         return status;
4590 }
4591
4592 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4593 {
4594         struct nfs_server *server = NFS_SERVER(state->inode);
4595         struct nfs4_exception exception = { };
4596         int err;
4597
4598         err = nfs4_set_lock_state(state, fl);
4599         if (err != 0)
4600                 goto out;
4601         do {
4602                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4603                 switch (err) {
4604                         default:
4605                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4606                                                 __func__, err);
4607                         case 0:
4608                         case -ESTALE:
4609                                 goto out;
4610                         case -NFS4ERR_EXPIRED:
4611                                 nfs4_schedule_stateid_recovery(server, state);
4612                         case -NFS4ERR_STALE_CLIENTID:
4613                         case -NFS4ERR_STALE_STATEID:
4614                                 nfs4_schedule_lease_recovery(server->nfs_client);
4615                                 goto out;
4616                         case -NFS4ERR_BADSESSION:
4617                         case -NFS4ERR_BADSLOT:
4618                         case -NFS4ERR_BAD_HIGH_SLOT:
4619                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4620                         case -NFS4ERR_DEADSESSION:
4621                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4622                                 goto out;
4623                         case -ERESTARTSYS:
4624                                 /*
4625                                  * The show must go on: exit, but mark the
4626                                  * stateid as needing recovery.
4627                                  */
4628                         case -NFS4ERR_ADMIN_REVOKED:
4629                         case -NFS4ERR_BAD_STATEID:
4630                         case -NFS4ERR_OPENMODE:
4631                                 nfs4_schedule_stateid_recovery(server, state);
4632                                 err = 0;
4633                                 goto out;
4634                         case -EKEYEXPIRED:
4635                                 /*
4636                                  * User RPCSEC_GSS context has expired.
4637                                  * We cannot recover this stateid now, so
4638                                  * skip it and allow recovery thread to
4639                                  * proceed.
4640                                  */
4641                                 err = 0;
4642                                 goto out;
4643                         case -ENOMEM:
4644                         case -NFS4ERR_DENIED:
4645                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4646                                 err = 0;
4647                                 goto out;
4648                         case -NFS4ERR_DELAY:
4649                                 break;
4650                 }
4651                 err = nfs4_handle_exception(server, err, &exception);
4652         } while (exception.retry);
4653 out:
4654         return err;
4655 }
4656
4657 static void nfs4_release_lockowner_release(void *calldata)
4658 {
4659         kfree(calldata);
4660 }
4661
4662 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4663         .rpc_release = nfs4_release_lockowner_release,
4664 };
4665
4666 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4667 {
4668         struct nfs_server *server = lsp->ls_state->owner->so_server;
4669         struct nfs_release_lockowner_args *args;
4670         struct rpc_message msg = {
4671                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4672         };
4673
4674         if (server->nfs_client->cl_mvops->minor_version != 0)
4675                 return;
4676         args = kmalloc(sizeof(*args), GFP_NOFS);
4677         if (!args)
4678                 return;
4679         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4680         args->lock_owner.id = lsp->ls_id.id;
4681         args->lock_owner.s_dev = server->s_dev;
4682         msg.rpc_argp = args;
4683         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4684 }
4685
4686 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4687
4688 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4689                                    const void *buf, size_t buflen,
4690                                    int flags, int type)
4691 {
4692         if (strcmp(key, "") != 0)
4693                 return -EINVAL;
4694
4695         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4696 }
4697
4698 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4699                                    void *buf, size_t buflen, int type)
4700 {
4701         if (strcmp(key, "") != 0)
4702                 return -EINVAL;
4703
4704         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4705 }
4706
4707 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4708                                        size_t list_len, const char *name,
4709                                        size_t name_len, int type)
4710 {
4711         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4712
4713         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4714                 return 0;
4715
4716         if (list && len <= list_len)
4717                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4718         return len;
4719 }
4720
4721 /*
4722  * nfs_fhget will use either the mounted_on_fileid or the fileid
4723  */
4724 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4725 {
4726         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4727                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4728               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4729               (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4730                 return;
4731
4732         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4733                 NFS_ATTR_FATTR_NLINK;
4734         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4735         fattr->nlink = 2;
4736 }
4737
4738 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4739                 struct nfs4_fs_locations *fs_locations, struct page *page)
4740 {
4741         struct nfs_server *server = NFS_SERVER(dir);
4742         u32 bitmask[2] = {
4743                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4744         };
4745         struct nfs4_fs_locations_arg args = {
4746                 .dir_fh = NFS_FH(dir),
4747                 .name = name,
4748                 .page = page,
4749                 .bitmask = bitmask,
4750         };
4751         struct nfs4_fs_locations_res res = {
4752                 .fs_locations = fs_locations,
4753         };
4754         struct rpc_message msg = {
4755                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4756                 .rpc_argp = &args,
4757                 .rpc_resp = &res,
4758         };
4759         int status;
4760
4761         dprintk("%s: start\n", __func__);
4762
4763         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4764          * is not supported */
4765         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4766                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4767         else
4768                 bitmask[0] |= FATTR4_WORD0_FILEID;
4769
4770         nfs_fattr_init(&fs_locations->fattr);
4771         fs_locations->server = server;
4772         fs_locations->nlocations = 0;
4773         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4774         dprintk("%s: returned status = %d\n", __func__, status);
4775         return status;
4776 }
4777
4778 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4779 {
4780         int status;
4781         struct nfs4_secinfo_arg args = {
4782                 .dir_fh = NFS_FH(dir),
4783                 .name   = name,
4784         };
4785         struct nfs4_secinfo_res res = {
4786                 .flavors     = flavors,
4787         };
4788         struct rpc_message msg = {
4789                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4790                 .rpc_argp = &args,
4791                 .rpc_resp = &res,
4792         };
4793
4794         dprintk("NFS call  secinfo %s\n", name->name);
4795         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4796         dprintk("NFS reply  secinfo: %d\n", status);
4797         return status;
4798 }
4799
4800 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4801 {
4802         struct nfs4_exception exception = { };
4803         int err;
4804         do {
4805                 err = nfs4_handle_exception(NFS_SERVER(dir),
4806                                 _nfs4_proc_secinfo(dir, name, flavors),
4807                                 &exception);
4808         } while (exception.retry);
4809         return err;
4810 }
4811
4812 #ifdef CONFIG_NFS_V4_1
4813 /*
4814  * Check the exchange flags returned by the server for invalid flags, having
4815  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4816  * DS flags set.
4817  */
4818 static int nfs4_check_cl_exchange_flags(u32 flags)
4819 {
4820         if (flags & ~EXCHGID4_FLAG_MASK_R)
4821                 goto out_inval;
4822         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4823             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4824                 goto out_inval;
4825         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4826                 goto out_inval;
4827         return NFS_OK;
4828 out_inval:
4829         return -NFS4ERR_INVAL;
4830 }
4831
4832 static bool
4833 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
4834 {
4835         if (a->server_scope_sz == b->server_scope_sz &&
4836             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
4837                 return true;
4838
4839         return false;
4840 }
4841
4842 /*
4843  * nfs4_proc_exchange_id()
4844  *
4845  * Since the clientid has expired, all compounds using sessions
4846  * associated with the stale clientid will be returning
4847  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4848  * be in some phase of session reset.
4849  */
4850 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4851 {
4852         nfs4_verifier verifier;
4853         struct nfs41_exchange_id_args args = {
4854                 .client = clp,
4855                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4856         };
4857         struct nfs41_exchange_id_res res = {
4858                 .client = clp,
4859         };
4860         int status;
4861         struct rpc_message msg = {
4862                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4863                 .rpc_argp = &args,
4864                 .rpc_resp = &res,
4865                 .rpc_cred = cred,
4866         };
4867         __be32 *p;
4868
4869         dprintk("--> %s\n", __func__);
4870         BUG_ON(clp == NULL);
4871
4872         p = (u32 *)verifier.data;
4873         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4874         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4875         args.verifier = &verifier;
4876
4877         args.id_len = scnprintf(args.id, sizeof(args.id),
4878                                 "%s/%s.%s/%u",
4879                                 clp->cl_ipaddr,
4880                                 init_utsname()->nodename,
4881                                 init_utsname()->domainname,
4882                                 clp->cl_rpcclient->cl_auth->au_flavor);
4883
4884         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
4885         if (unlikely(!res.server_scope))
4886                 return -ENOMEM;
4887
4888         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4889         if (!status)
4890                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4891
4892         if (!status) {
4893                 if (clp->server_scope &&
4894                     !nfs41_same_server_scope(clp->server_scope,
4895                                              res.server_scope)) {
4896                         dprintk("%s: server_scope mismatch detected\n",
4897                                 __func__);
4898                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
4899                         kfree(clp->server_scope);
4900                         clp->server_scope = NULL;
4901                 }
4902
4903                 if (!clp->server_scope)
4904                         clp->server_scope = res.server_scope;
4905                 else
4906                         kfree(res.server_scope);
4907         }
4908
4909         dprintk("<-- %s status= %d\n", __func__, status);
4910         return status;
4911 }
4912
4913 struct nfs4_get_lease_time_data {
4914         struct nfs4_get_lease_time_args *args;
4915         struct nfs4_get_lease_time_res *res;
4916         struct nfs_client *clp;
4917 };
4918
4919 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4920                                         void *calldata)
4921 {
4922         int ret;
4923         struct nfs4_get_lease_time_data *data =
4924                         (struct nfs4_get_lease_time_data *)calldata;
4925
4926         dprintk("--> %s\n", __func__);
4927         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4928         /* just setup sequence, do not trigger session recovery
4929            since we're invoked within one */
4930         ret = nfs41_setup_sequence(data->clp->cl_session,
4931                                    &data->args->la_seq_args,
4932                                    &data->res->lr_seq_res, 0, task);
4933
4934         BUG_ON(ret == -EAGAIN);
4935         rpc_call_start(task);
4936         dprintk("<-- %s\n", __func__);
4937 }
4938
4939 /*
4940  * Called from nfs4_state_manager thread for session setup, so don't recover
4941  * from sequence operation or clientid errors.
4942  */
4943 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4944 {
4945         struct nfs4_get_lease_time_data *data =
4946                         (struct nfs4_get_lease_time_data *)calldata;
4947
4948         dprintk("--> %s\n", __func__);
4949         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
4950                 return;
4951         switch (task->tk_status) {
4952         case -NFS4ERR_DELAY:
4953         case -NFS4ERR_GRACE:
4954                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4955                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4956                 task->tk_status = 0;
4957                 /* fall through */
4958         case -NFS4ERR_RETRY_UNCACHED_REP:
4959                 rpc_restart_call_prepare(task);
4960                 return;
4961         }
4962         dprintk("<-- %s\n", __func__);
4963 }
4964
4965 struct rpc_call_ops nfs4_get_lease_time_ops = {
4966         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4967         .rpc_call_done = nfs4_get_lease_time_done,
4968 };
4969
4970 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4971 {
4972         struct rpc_task *task;
4973         struct nfs4_get_lease_time_args args;
4974         struct nfs4_get_lease_time_res res = {
4975                 .lr_fsinfo = fsinfo,
4976         };
4977         struct nfs4_get_lease_time_data data = {
4978                 .args = &args,
4979                 .res = &res,
4980                 .clp = clp,
4981         };
4982         struct rpc_message msg = {
4983                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4984                 .rpc_argp = &args,
4985                 .rpc_resp = &res,
4986         };
4987         struct rpc_task_setup task_setup = {
4988                 .rpc_client = clp->cl_rpcclient,
4989                 .rpc_message = &msg,
4990                 .callback_ops = &nfs4_get_lease_time_ops,
4991                 .callback_data = &data,
4992                 .flags = RPC_TASK_TIMEOUT,
4993         };
4994         int status;
4995
4996         dprintk("--> %s\n", __func__);
4997         task = rpc_run_task(&task_setup);
4998
4999         if (IS_ERR(task))
5000                 status = PTR_ERR(task);
5001         else {
5002                 status = task->tk_status;
5003                 rpc_put_task(task);
5004         }
5005         dprintk("<-- %s return %d\n", __func__, status);
5006
5007         return status;
5008 }
5009
5010 /*
5011  * Reset a slot table
5012  */
5013 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5014                                  int ivalue)
5015 {
5016         struct nfs4_slot *new = NULL;
5017         int i;
5018         int ret = 0;
5019
5020         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5021                 max_reqs, tbl->max_slots);
5022
5023         /* Does the newly negotiated max_reqs match the existing slot table? */
5024         if (max_reqs != tbl->max_slots) {
5025                 ret = -ENOMEM;
5026                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
5027                               GFP_NOFS);
5028                 if (!new)
5029                         goto out;
5030                 ret = 0;
5031                 kfree(tbl->slots);
5032         }
5033         spin_lock(&tbl->slot_tbl_lock);
5034         if (new) {
5035                 tbl->slots = new;
5036                 tbl->max_slots = max_reqs;
5037         }
5038         for (i = 0; i < tbl->max_slots; ++i)
5039                 tbl->slots[i].seq_nr = ivalue;
5040         spin_unlock(&tbl->slot_tbl_lock);
5041         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5042                 tbl, tbl->slots, tbl->max_slots);
5043 out:
5044         dprintk("<-- %s: return %d\n", __func__, ret);
5045         return ret;
5046 }
5047
5048 /* Destroy the slot table */
5049 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5050 {
5051         if (session->fc_slot_table.slots != NULL) {
5052                 kfree(session->fc_slot_table.slots);
5053                 session->fc_slot_table.slots = NULL;
5054         }
5055         if (session->bc_slot_table.slots != NULL) {
5056                 kfree(session->bc_slot_table.slots);
5057                 session->bc_slot_table.slots = NULL;
5058         }
5059         return;
5060 }
5061
5062 /*
5063  * Initialize slot table
5064  */
5065 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
5066                 int max_slots, int ivalue)
5067 {
5068         struct nfs4_slot *slot;
5069         int ret = -ENOMEM;
5070
5071         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
5072
5073         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
5074
5075         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
5076         if (!slot)
5077                 goto out;
5078         ret = 0;
5079
5080         spin_lock(&tbl->slot_tbl_lock);
5081         tbl->max_slots = max_slots;
5082         tbl->slots = slot;
5083         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5084         spin_unlock(&tbl->slot_tbl_lock);
5085         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5086                 tbl, tbl->slots, tbl->max_slots);
5087 out:
5088         dprintk("<-- %s: return %d\n", __func__, ret);
5089         return ret;
5090 }
5091
5092 /*
5093  * Initialize or reset the forechannel and backchannel tables
5094  */
5095 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5096 {
5097         struct nfs4_slot_table *tbl;
5098         int status;
5099
5100         dprintk("--> %s\n", __func__);
5101         /* Fore channel */
5102         tbl = &ses->fc_slot_table;
5103         if (tbl->slots == NULL) {
5104                 status = nfs4_init_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5105                 if (status) /* -ENOMEM */
5106                         return status;
5107         } else {
5108                 status = nfs4_reset_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5109                 if (status)
5110                         return status;
5111         }
5112         /* Back channel */
5113         tbl = &ses->bc_slot_table;
5114         if (tbl->slots == NULL) {
5115                 status = nfs4_init_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5116                 if (status)
5117                         /* Fore and back channel share a connection so get
5118                          * both slot tables or neither */
5119                         nfs4_destroy_slot_tables(ses);
5120         } else
5121                 status = nfs4_reset_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5122         return status;
5123 }
5124
5125 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5126 {
5127         struct nfs4_session *session;
5128         struct nfs4_slot_table *tbl;
5129
5130         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5131         if (!session)
5132                 return NULL;
5133
5134         tbl = &session->fc_slot_table;
5135         tbl->highest_used_slotid = -1;
5136         spin_lock_init(&tbl->slot_tbl_lock);
5137         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5138         init_completion(&tbl->complete);
5139
5140         tbl = &session->bc_slot_table;
5141         tbl->highest_used_slotid = -1;
5142         spin_lock_init(&tbl->slot_tbl_lock);
5143         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5144         init_completion(&tbl->complete);
5145
5146         session->session_state = 1<<NFS4_SESSION_INITING;
5147
5148         session->clp = clp;
5149         return session;
5150 }
5151
5152 void nfs4_destroy_session(struct nfs4_session *session)
5153 {
5154         nfs4_proc_destroy_session(session);
5155         dprintk("%s Destroy backchannel for xprt %p\n",
5156                 __func__, session->clp->cl_rpcclient->cl_xprt);
5157         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
5158                                 NFS41_BC_MIN_CALLBACKS);
5159         nfs4_destroy_slot_tables(session);
5160         kfree(session);
5161 }
5162
5163 /*
5164  * Initialize the values to be used by the client in CREATE_SESSION
5165  * If nfs4_init_session set the fore channel request and response sizes,
5166  * use them.
5167  *
5168  * Set the back channel max_resp_sz_cached to zero to force the client to
5169  * always set csa_cachethis to FALSE because the current implementation
5170  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5171  */
5172 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5173 {
5174         struct nfs4_session *session = args->client->cl_session;
5175         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5176                      mxresp_sz = session->fc_attrs.max_resp_sz;
5177
5178         if (mxrqst_sz == 0)
5179                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5180         if (mxresp_sz == 0)
5181                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5182         /* Fore channel attributes */
5183         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5184         args->fc_attrs.max_resp_sz = mxresp_sz;
5185         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5186         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
5187
5188         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5189                 "max_ops=%u max_reqs=%u\n",
5190                 __func__,
5191                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5192                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5193
5194         /* Back channel attributes */
5195         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5196         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5197         args->bc_attrs.max_resp_sz_cached = 0;
5198         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5199         args->bc_attrs.max_reqs = 1;
5200
5201         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5202                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5203                 __func__,
5204                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5205                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5206                 args->bc_attrs.max_reqs);
5207 }
5208
5209 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5210 {
5211         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5212         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5213
5214         if (rcvd->max_resp_sz > sent->max_resp_sz)
5215                 return -EINVAL;
5216         /*
5217          * Our requested max_ops is the minimum we need; we're not
5218          * prepared to break up compounds into smaller pieces than that.
5219          * So, no point even trying to continue if the server won't
5220          * cooperate:
5221          */
5222         if (rcvd->max_ops < sent->max_ops)
5223                 return -EINVAL;
5224         if (rcvd->max_reqs == 0)
5225                 return -EINVAL;
5226         return 0;
5227 }
5228
5229 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5230 {
5231         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5232         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5233
5234         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5235                 return -EINVAL;
5236         if (rcvd->max_resp_sz < sent->max_resp_sz)
5237                 return -EINVAL;
5238         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5239                 return -EINVAL;
5240         /* These would render the backchannel useless: */
5241         if (rcvd->max_ops  == 0)
5242                 return -EINVAL;
5243         if (rcvd->max_reqs == 0)
5244                 return -EINVAL;
5245         return 0;
5246 }
5247
5248 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5249                                      struct nfs4_session *session)
5250 {
5251         int ret;
5252
5253         ret = nfs4_verify_fore_channel_attrs(args, session);
5254         if (ret)
5255                 return ret;
5256         return nfs4_verify_back_channel_attrs(args, session);
5257 }
5258
5259 static int _nfs4_proc_create_session(struct nfs_client *clp)
5260 {
5261         struct nfs4_session *session = clp->cl_session;
5262         struct nfs41_create_session_args args = {
5263                 .client = clp,
5264                 .cb_program = NFS4_CALLBACK,
5265         };
5266         struct nfs41_create_session_res res = {
5267                 .client = clp,
5268         };
5269         struct rpc_message msg = {
5270                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5271                 .rpc_argp = &args,
5272                 .rpc_resp = &res,
5273         };
5274         int status;
5275
5276         nfs4_init_channel_attrs(&args);
5277         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5278
5279         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5280
5281         if (!status)
5282                 /* Verify the session's negotiated channel_attrs values */
5283                 status = nfs4_verify_channel_attrs(&args, session);
5284         if (!status) {
5285                 /* Increment the clientid slot sequence id */
5286                 clp->cl_seqid++;
5287         }
5288
5289         return status;
5290 }
5291
5292 /*
5293  * Issues a CREATE_SESSION operation to the server.
5294  * It is the responsibility of the caller to verify the session is
5295  * expired before calling this routine.
5296  */
5297 int nfs4_proc_create_session(struct nfs_client *clp)
5298 {
5299         int status;
5300         unsigned *ptr;
5301         struct nfs4_session *session = clp->cl_session;
5302
5303         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5304
5305         status = _nfs4_proc_create_session(clp);
5306         if (status)
5307                 goto out;
5308
5309         /* Init or reset the session slot tables */
5310         status = nfs4_setup_session_slot_tables(session);
5311         dprintk("slot table setup returned %d\n", status);
5312         if (status)
5313                 goto out;
5314
5315         ptr = (unsigned *)&session->sess_id.data[0];
5316         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5317                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5318 out:
5319         dprintk("<-- %s\n", __func__);
5320         return status;
5321 }
5322
5323 /*
5324  * Issue the over-the-wire RPC DESTROY_SESSION.
5325  * The caller must serialize access to this routine.
5326  */
5327 int nfs4_proc_destroy_session(struct nfs4_session *session)
5328 {
5329         int status = 0;
5330         struct rpc_message msg;
5331
5332         dprintk("--> nfs4_proc_destroy_session\n");
5333
5334         /* session is still being setup */
5335         if (session->clp->cl_cons_state != NFS_CS_READY)
5336                 return status;
5337
5338         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5339         msg.rpc_argp = session;
5340         msg.rpc_resp = NULL;
5341         msg.rpc_cred = NULL;
5342         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5343
5344         if (status)
5345                 printk(KERN_WARNING
5346                         "Got error %d from the server on DESTROY_SESSION. "
5347                         "Session has been destroyed regardless...\n", status);
5348
5349         dprintk("<-- nfs4_proc_destroy_session\n");
5350         return status;
5351 }
5352
5353 int nfs4_init_session(struct nfs_server *server)
5354 {
5355         struct nfs_client *clp = server->nfs_client;
5356         struct nfs4_session *session;
5357         unsigned int rsize, wsize;
5358         int ret;
5359
5360         if (!nfs4_has_session(clp))
5361                 return 0;
5362
5363         session = clp->cl_session;
5364         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5365                 return 0;
5366
5367         rsize = server->rsize;
5368         if (rsize == 0)
5369                 rsize = NFS_MAX_FILE_IO_SIZE;
5370         wsize = server->wsize;
5371         if (wsize == 0)
5372                 wsize = NFS_MAX_FILE_IO_SIZE;
5373
5374         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5375         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5376
5377         ret = nfs4_recover_expired_lease(server);
5378         if (!ret)
5379                 ret = nfs4_check_client_ready(clp);
5380         return ret;
5381 }
5382
5383 int nfs4_init_ds_session(struct nfs_client *clp)
5384 {
5385         struct nfs4_session *session = clp->cl_session;
5386         int ret;
5387
5388         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5389                 return 0;
5390
5391         ret = nfs4_client_recover_expired_lease(clp);
5392         if (!ret)
5393                 /* Test for the DS role */
5394                 if (!is_ds_client(clp))
5395                         ret = -ENODEV;
5396         if (!ret)
5397                 ret = nfs4_check_client_ready(clp);
5398         return ret;
5399
5400 }
5401 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5402
5403
5404 /*
5405  * Renew the cl_session lease.
5406  */
5407 struct nfs4_sequence_data {
5408         struct nfs_client *clp;
5409         struct nfs4_sequence_args args;
5410         struct nfs4_sequence_res res;
5411 };
5412
5413 static void nfs41_sequence_release(void *data)
5414 {
5415         struct nfs4_sequence_data *calldata = data;
5416         struct nfs_client *clp = calldata->clp;
5417
5418         if (atomic_read(&clp->cl_count) > 1)
5419                 nfs4_schedule_state_renewal(clp);
5420         nfs_put_client(clp);
5421         kfree(calldata);
5422 }
5423
5424 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5425 {
5426         switch(task->tk_status) {
5427         case -NFS4ERR_DELAY:
5428                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5429                 return -EAGAIN;
5430         default:
5431                 nfs4_schedule_lease_recovery(clp);
5432         }
5433         return 0;
5434 }
5435
5436 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5437 {
5438         struct nfs4_sequence_data *calldata = data;
5439         struct nfs_client *clp = calldata->clp;
5440
5441         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5442                 return;
5443
5444         if (task->tk_status < 0) {
5445                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5446                 if (atomic_read(&clp->cl_count) == 1)
5447                         goto out;
5448
5449                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5450                         rpc_restart_call_prepare(task);
5451                         return;
5452                 }
5453         }
5454         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5455 out:
5456         dprintk("<-- %s\n", __func__);
5457 }
5458
5459 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5460 {
5461         struct nfs4_sequence_data *calldata = data;
5462         struct nfs_client *clp = calldata->clp;
5463         struct nfs4_sequence_args *args;
5464         struct nfs4_sequence_res *res;
5465
5466         args = task->tk_msg.rpc_argp;
5467         res = task->tk_msg.rpc_resp;
5468
5469         if (nfs41_setup_sequence(clp->cl_session, args, res, 0, task))
5470                 return;
5471         rpc_call_start(task);
5472 }
5473
5474 static const struct rpc_call_ops nfs41_sequence_ops = {
5475         .rpc_call_done = nfs41_sequence_call_done,
5476         .rpc_call_prepare = nfs41_sequence_prepare,
5477         .rpc_release = nfs41_sequence_release,
5478 };
5479
5480 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5481 {
5482         struct nfs4_sequence_data *calldata;
5483         struct rpc_message msg = {
5484                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5485                 .rpc_cred = cred,
5486         };
5487         struct rpc_task_setup task_setup_data = {
5488                 .rpc_client = clp->cl_rpcclient,
5489                 .rpc_message = &msg,
5490                 .callback_ops = &nfs41_sequence_ops,
5491                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5492         };
5493
5494         if (!atomic_inc_not_zero(&clp->cl_count))
5495                 return ERR_PTR(-EIO);
5496         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5497         if (calldata == NULL) {
5498                 nfs_put_client(clp);
5499                 return ERR_PTR(-ENOMEM);
5500         }
5501         msg.rpc_argp = &calldata->args;
5502         msg.rpc_resp = &calldata->res;
5503         calldata->clp = clp;
5504         task_setup_data.callback_data = calldata;
5505
5506         return rpc_run_task(&task_setup_data);
5507 }
5508
5509 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5510 {
5511         struct rpc_task *task;
5512         int ret = 0;
5513
5514         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5515                 return 0;
5516         task = _nfs41_proc_sequence(clp, cred);
5517         if (IS_ERR(task))
5518                 ret = PTR_ERR(task);
5519         else
5520                 rpc_put_task_async(task);
5521         dprintk("<-- %s status=%d\n", __func__, ret);
5522         return ret;
5523 }
5524
5525 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5526 {
5527         struct rpc_task *task;
5528         int ret;
5529
5530         task = _nfs41_proc_sequence(clp, cred);
5531         if (IS_ERR(task)) {
5532                 ret = PTR_ERR(task);
5533                 goto out;
5534         }
5535         ret = rpc_wait_for_completion_task(task);
5536         if (!ret) {
5537                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5538
5539                 if (task->tk_status == 0)
5540                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5541                 ret = task->tk_status;
5542         }
5543         rpc_put_task(task);
5544 out:
5545         dprintk("<-- %s status=%d\n", __func__, ret);
5546         return ret;
5547 }
5548
5549 struct nfs4_reclaim_complete_data {
5550         struct nfs_client *clp;
5551         struct nfs41_reclaim_complete_args arg;
5552         struct nfs41_reclaim_complete_res res;
5553 };
5554
5555 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5556 {
5557         struct nfs4_reclaim_complete_data *calldata = data;
5558
5559         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5560         if (nfs41_setup_sequence(calldata->clp->cl_session,
5561                                 &calldata->arg.seq_args,
5562                                 &calldata->res.seq_res, 0, task))
5563                 return;
5564
5565         rpc_call_start(task);
5566 }
5567
5568 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5569 {
5570         switch(task->tk_status) {
5571         case 0:
5572         case -NFS4ERR_COMPLETE_ALREADY:
5573         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5574                 break;
5575         case -NFS4ERR_DELAY:
5576                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5577                 /* fall through */
5578         case -NFS4ERR_RETRY_UNCACHED_REP:
5579                 return -EAGAIN;
5580         default:
5581                 nfs4_schedule_lease_recovery(clp);
5582         }
5583         return 0;
5584 }
5585
5586 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5587 {
5588         struct nfs4_reclaim_complete_data *calldata = data;
5589         struct nfs_client *clp = calldata->clp;
5590         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5591
5592         dprintk("--> %s\n", __func__);
5593         if (!nfs41_sequence_done(task, res))
5594                 return;
5595
5596         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5597                 rpc_restart_call_prepare(task);
5598                 return;
5599         }
5600         dprintk("<-- %s\n", __func__);
5601 }
5602
5603 static void nfs4_free_reclaim_complete_data(void *data)
5604 {
5605         struct nfs4_reclaim_complete_data *calldata = data;
5606
5607         kfree(calldata);
5608 }
5609
5610 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5611         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5612         .rpc_call_done = nfs4_reclaim_complete_done,
5613         .rpc_release = nfs4_free_reclaim_complete_data,
5614 };
5615
5616 /*
5617  * Issue a global reclaim complete.
5618  */
5619 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5620 {
5621         struct nfs4_reclaim_complete_data *calldata;
5622         struct rpc_task *task;
5623         struct rpc_message msg = {
5624                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5625         };
5626         struct rpc_task_setup task_setup_data = {
5627                 .rpc_client = clp->cl_rpcclient,
5628                 .rpc_message = &msg,
5629                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5630                 .flags = RPC_TASK_ASYNC,
5631         };
5632         int status = -ENOMEM;
5633
5634         dprintk("--> %s\n", __func__);
5635         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5636         if (calldata == NULL)
5637                 goto out;
5638         calldata->clp = clp;
5639         calldata->arg.one_fs = 0;
5640
5641         msg.rpc_argp = &calldata->arg;
5642         msg.rpc_resp = &calldata->res;
5643         task_setup_data.callback_data = calldata;
5644         task = rpc_run_task(&task_setup_data);
5645         if (IS_ERR(task)) {
5646                 status = PTR_ERR(task);
5647                 goto out;
5648         }
5649         status = nfs4_wait_for_completion_rpc_task(task);
5650         if (status == 0)
5651                 status = task->tk_status;
5652         rpc_put_task(task);
5653         return 0;
5654 out:
5655         dprintk("<-- %s status=%d\n", __func__, status);
5656         return status;
5657 }
5658
5659 static void
5660 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5661 {
5662         struct nfs4_layoutget *lgp = calldata;
5663         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5664
5665         dprintk("--> %s\n", __func__);
5666         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5667          * right now covering the LAYOUTGET we are about to send.
5668          * However, that is not so catastrophic, and there seems
5669          * to be no way to prevent it completely.
5670          */
5671         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5672                                 &lgp->res.seq_res, 0, task))
5673                 return;
5674         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5675                                           NFS_I(lgp->args.inode)->layout,
5676                                           lgp->args.ctx->state)) {
5677                 rpc_exit(task, NFS4_OK);
5678                 return;
5679         }
5680         rpc_call_start(task);
5681 }
5682
5683 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5684 {
5685         struct nfs4_layoutget *lgp = calldata;
5686         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5687
5688         dprintk("--> %s\n", __func__);
5689
5690         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5691                 return;
5692
5693         switch (task->tk_status) {
5694         case 0:
5695                 break;
5696         case -NFS4ERR_LAYOUTTRYLATER:
5697         case -NFS4ERR_RECALLCONFLICT:
5698                 task->tk_status = -NFS4ERR_DELAY;
5699                 /* Fall through */
5700         default:
5701                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5702                         rpc_restart_call_prepare(task);
5703                         return;
5704                 }
5705         }
5706         dprintk("<-- %s\n", __func__);
5707 }
5708
5709 static void nfs4_layoutget_release(void *calldata)
5710 {
5711         struct nfs4_layoutget *lgp = calldata;
5712
5713         dprintk("--> %s\n", __func__);
5714         put_nfs_open_context(lgp->args.ctx);
5715         kfree(calldata);
5716         dprintk("<-- %s\n", __func__);
5717 }
5718
5719 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5720         .rpc_call_prepare = nfs4_layoutget_prepare,
5721         .rpc_call_done = nfs4_layoutget_done,
5722         .rpc_release = nfs4_layoutget_release,
5723 };
5724
5725 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5726 {
5727         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5728         struct rpc_task *task;
5729         struct rpc_message msg = {
5730                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5731                 .rpc_argp = &lgp->args,
5732                 .rpc_resp = &lgp->res,
5733         };
5734         struct rpc_task_setup task_setup_data = {
5735                 .rpc_client = server->client,
5736                 .rpc_message = &msg,
5737                 .callback_ops = &nfs4_layoutget_call_ops,
5738                 .callback_data = lgp,
5739                 .flags = RPC_TASK_ASYNC,
5740         };
5741         int status = 0;
5742
5743         dprintk("--> %s\n", __func__);
5744
5745         lgp->res.layoutp = &lgp->args.layout;
5746         lgp->res.seq_res.sr_slot = NULL;
5747         task = rpc_run_task(&task_setup_data);
5748         if (IS_ERR(task))
5749                 return PTR_ERR(task);
5750         status = nfs4_wait_for_completion_rpc_task(task);
5751         if (status == 0)
5752                 status = task->tk_status;
5753         if (status == 0)
5754                 status = pnfs_layout_process(lgp);
5755         rpc_put_task(task);
5756         dprintk("<-- %s status=%d\n", __func__, status);
5757         return status;
5758 }
5759
5760 static void
5761 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5762 {
5763         struct nfs4_layoutreturn *lrp = calldata;
5764
5765         dprintk("--> %s\n", __func__);
5766         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5767                                 &lrp->res.seq_res, 0, task))
5768                 return;
5769         rpc_call_start(task);
5770 }
5771
5772 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5773 {
5774         struct nfs4_layoutreturn *lrp = calldata;
5775         struct nfs_server *server;
5776         struct pnfs_layout_hdr *lo = lrp->args.layout;
5777
5778         dprintk("--> %s\n", __func__);
5779
5780         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5781                 return;
5782
5783         server = NFS_SERVER(lrp->args.inode);
5784         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5785                 rpc_restart_call_prepare(task);
5786                 return;
5787         }
5788         spin_lock(&lo->plh_inode->i_lock);
5789         if (task->tk_status == 0) {
5790                 if (lrp->res.lrs_present) {
5791                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5792                 } else
5793                         BUG_ON(!list_empty(&lo->plh_segs));
5794         }
5795         lo->plh_block_lgets--;
5796         spin_unlock(&lo->plh_inode->i_lock);
5797         dprintk("<-- %s\n", __func__);
5798 }
5799
5800 static void nfs4_layoutreturn_release(void *calldata)
5801 {
5802         struct nfs4_layoutreturn *lrp = calldata;
5803
5804         dprintk("--> %s\n", __func__);
5805         put_layout_hdr(lrp->args.layout);
5806         kfree(calldata);
5807         dprintk("<-- %s\n", __func__);
5808 }
5809
5810 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
5811         .rpc_call_prepare = nfs4_layoutreturn_prepare,
5812         .rpc_call_done = nfs4_layoutreturn_done,
5813         .rpc_release = nfs4_layoutreturn_release,
5814 };
5815
5816 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
5817 {
5818         struct rpc_task *task;
5819         struct rpc_message msg = {
5820                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
5821                 .rpc_argp = &lrp->args,
5822                 .rpc_resp = &lrp->res,
5823         };
5824         struct rpc_task_setup task_setup_data = {
5825                 .rpc_client = lrp->clp->cl_rpcclient,
5826                 .rpc_message = &msg,
5827                 .callback_ops = &nfs4_layoutreturn_call_ops,
5828                 .callback_data = lrp,
5829         };
5830         int status;
5831
5832         dprintk("--> %s\n", __func__);
5833         task = rpc_run_task(&task_setup_data);
5834         if (IS_ERR(task))
5835                 return PTR_ERR(task);
5836         status = task->tk_status;
5837         dprintk("<-- %s status=%d\n", __func__, status);
5838         rpc_put_task(task);
5839         return status;
5840 }
5841
5842 /*
5843  * Retrieve the list of Data Server devices from the MDS.
5844  */
5845 static int _nfs4_getdevicelist(struct nfs_server *server,
5846                                     const struct nfs_fh *fh,
5847                                     struct pnfs_devicelist *devlist)
5848 {
5849         struct nfs4_getdevicelist_args args = {
5850                 .fh = fh,
5851                 .layoutclass = server->pnfs_curr_ld->id,
5852         };
5853         struct nfs4_getdevicelist_res res = {
5854                 .devlist = devlist,
5855         };
5856         struct rpc_message msg = {
5857                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
5858                 .rpc_argp = &args,
5859                 .rpc_resp = &res,
5860         };
5861         int status;
5862
5863         dprintk("--> %s\n", __func__);
5864         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5865                                 &res.seq_res, 0);
5866         dprintk("<-- %s status=%d\n", __func__, status);
5867         return status;
5868 }
5869
5870 int nfs4_proc_getdevicelist(struct nfs_server *server,
5871                             const struct nfs_fh *fh,
5872                             struct pnfs_devicelist *devlist)
5873 {
5874         struct nfs4_exception exception = { };
5875         int err;
5876
5877         do {
5878                 err = nfs4_handle_exception(server,
5879                                 _nfs4_getdevicelist(server, fh, devlist),
5880                                 &exception);
5881         } while (exception.retry);
5882
5883         dprintk("%s: err=%d, num_devs=%u\n", __func__,
5884                 err, devlist->num_devs);
5885
5886         return err;
5887 }
5888 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
5889
5890 static int
5891 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5892 {
5893         struct nfs4_getdeviceinfo_args args = {
5894                 .pdev = pdev,
5895         };
5896         struct nfs4_getdeviceinfo_res res = {
5897                 .pdev = pdev,
5898         };
5899         struct rpc_message msg = {
5900                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5901                 .rpc_argp = &args,
5902                 .rpc_resp = &res,
5903         };
5904         int status;
5905
5906         dprintk("--> %s\n", __func__);
5907         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5908         dprintk("<-- %s status=%d\n", __func__, status);
5909
5910         return status;
5911 }
5912
5913 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5914 {
5915         struct nfs4_exception exception = { };
5916         int err;
5917
5918         do {
5919                 err = nfs4_handle_exception(server,
5920                                         _nfs4_proc_getdeviceinfo(server, pdev),
5921                                         &exception);
5922         } while (exception.retry);
5923         return err;
5924 }
5925 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5926
5927 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
5928 {
5929         struct nfs4_layoutcommit_data *data = calldata;
5930         struct nfs_server *server = NFS_SERVER(data->args.inode);
5931
5932         if (nfs4_setup_sequence(server, &data->args.seq_args,
5933                                 &data->res.seq_res, 1, task))
5934                 return;
5935         rpc_call_start(task);
5936 }
5937
5938 static void
5939 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
5940 {
5941         struct nfs4_layoutcommit_data *data = calldata;
5942         struct nfs_server *server = NFS_SERVER(data->args.inode);
5943
5944         if (!nfs4_sequence_done(task, &data->res.seq_res))
5945                 return;
5946
5947         switch (task->tk_status) { /* Just ignore these failures */
5948         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
5949         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
5950         case NFS4ERR_BADLAYOUT:     /* no layout */
5951         case NFS4ERR_GRACE:         /* loca_recalim always false */
5952                 task->tk_status = 0;
5953         }
5954
5955         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5956                 rpc_restart_call_prepare(task);
5957                 return;
5958         }
5959
5960         if (task->tk_status == 0)
5961                 nfs_post_op_update_inode_force_wcc(data->args.inode,
5962                                                    data->res.fattr);
5963 }
5964
5965 static void nfs4_layoutcommit_release(void *calldata)
5966 {
5967         struct nfs4_layoutcommit_data *data = calldata;
5968         struct pnfs_layout_segment *lseg, *tmp;
5969         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
5970
5971         pnfs_cleanup_layoutcommit(data);
5972         /* Matched by references in pnfs_set_layoutcommit */
5973         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
5974                 list_del_init(&lseg->pls_lc_list);
5975                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
5976                                        &lseg->pls_flags))
5977                         put_lseg(lseg);
5978         }
5979
5980         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
5981         smp_mb__after_clear_bit();
5982         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
5983
5984         put_rpccred(data->cred);
5985         kfree(data);
5986 }
5987
5988 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
5989         .rpc_call_prepare = nfs4_layoutcommit_prepare,
5990         .rpc_call_done = nfs4_layoutcommit_done,
5991         .rpc_release = nfs4_layoutcommit_release,
5992 };
5993
5994 int
5995 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
5996 {
5997         struct rpc_message msg = {
5998                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
5999                 .rpc_argp = &data->args,
6000                 .rpc_resp = &data->res,
6001                 .rpc_cred = data->cred,
6002         };
6003         struct rpc_task_setup task_setup_data = {
6004                 .task = &data->task,
6005                 .rpc_client = NFS_CLIENT(data->args.inode),
6006                 .rpc_message = &msg,
6007                 .callback_ops = &nfs4_layoutcommit_ops,
6008                 .callback_data = data,
6009                 .flags = RPC_TASK_ASYNC,
6010         };
6011         struct rpc_task *task;
6012         int status = 0;
6013
6014         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6015                 "lbw: %llu inode %lu\n",
6016                 data->task.tk_pid, sync,
6017                 data->args.lastbytewritten,
6018                 data->args.inode->i_ino);
6019
6020         task = rpc_run_task(&task_setup_data);
6021         if (IS_ERR(task))
6022                 return PTR_ERR(task);
6023         if (sync == false)
6024                 goto out;
6025         status = nfs4_wait_for_completion_rpc_task(task);
6026         if (status != 0)
6027                 goto out;
6028         status = task->tk_status;
6029 out:
6030         dprintk("%s: status %d\n", __func__, status);
6031         rpc_put_task(task);
6032         return status;
6033 }
6034
6035 static int
6036 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6037                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6038 {
6039         struct nfs41_secinfo_no_name_args args = {
6040                 .style = SECINFO_STYLE_CURRENT_FH,
6041         };
6042         struct nfs4_secinfo_res res = {
6043                 .flavors = flavors,
6044         };
6045         struct rpc_message msg = {
6046                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6047                 .rpc_argp = &args,
6048                 .rpc_resp = &res,
6049         };
6050         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6051 }
6052
6053 static int
6054 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6055                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6056 {
6057         struct nfs4_exception exception = { };
6058         int err;
6059         do {
6060                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6061                 switch (err) {
6062                 case 0:
6063                 case -NFS4ERR_WRONGSEC:
6064                 case -NFS4ERR_NOTSUPP:
6065                         break;
6066                 default:
6067                         err = nfs4_handle_exception(server, err, &exception);
6068                 }
6069         } while (exception.retry);
6070         return err;
6071 }
6072
6073 static int
6074 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6075                     struct nfs_fsinfo *info)
6076 {
6077         int err;
6078         struct page *page;
6079         rpc_authflavor_t flavor;
6080         struct nfs4_secinfo_flavors *flavors;
6081
6082         page = alloc_page(GFP_KERNEL);
6083         if (!page) {
6084                 err = -ENOMEM;
6085                 goto out;
6086         }
6087
6088         flavors = page_address(page);
6089         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6090
6091         /*
6092          * Fall back on "guess and check" method if
6093          * the server doesn't support SECINFO_NO_NAME
6094          */
6095         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6096                 err = nfs4_find_root_sec(server, fhandle, info);
6097                 goto out_freepage;
6098         }
6099         if (err)
6100                 goto out_freepage;
6101
6102         flavor = nfs_find_best_sec(flavors);
6103         if (err == 0)
6104                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6105
6106 out_freepage:
6107         put_page(page);
6108         if (err == -EACCES)
6109                 return -EPERM;
6110 out:
6111         return err;
6112 }
6113 static int _nfs41_test_stateid(struct nfs_server *server, struct nfs4_state *state)
6114 {
6115         int status;
6116         struct nfs41_test_stateid_args args = {
6117                 .stateid = &state->stateid,
6118         };
6119         struct nfs41_test_stateid_res res;
6120         struct rpc_message msg = {
6121                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6122                 .rpc_argp = &args,
6123                 .rpc_resp = &res,
6124         };
6125         args.seq_args.sa_session = res.seq_res.sr_session = NULL;
6126         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 0, 1);
6127         return status;
6128 }
6129
6130 static int nfs41_test_stateid(struct nfs_server *server, struct nfs4_state *state)
6131 {
6132         struct nfs4_exception exception = { };
6133         int err;
6134         do {
6135                 err = nfs4_handle_exception(server,
6136                                 _nfs41_test_stateid(server, state),
6137                                 &exception);
6138         } while (exception.retry);
6139         return err;
6140 }
6141
6142 static int _nfs4_free_stateid(struct nfs_server *server, struct nfs4_state *state)
6143 {
6144         int status;
6145         struct nfs41_free_stateid_args args = {
6146                 .stateid = &state->stateid,
6147         };
6148         struct nfs41_free_stateid_res res;
6149         struct rpc_message msg = {
6150                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6151                 .rpc_argp = &args,
6152                 .rpc_resp = &res,
6153         };
6154
6155         args.seq_args.sa_session = res.seq_res.sr_session = NULL;
6156         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 0, 1);
6157         return status;
6158 }
6159
6160 static int nfs41_free_stateid(struct nfs_server *server, struct nfs4_state *state)
6161 {
6162         struct nfs4_exception exception = { };
6163         int err;
6164         do {
6165                 err = nfs4_handle_exception(server,
6166                                 _nfs4_free_stateid(server, state),
6167                                 &exception);
6168         } while (exception.retry);
6169         return err;
6170 }
6171 #endif /* CONFIG_NFS_V4_1 */
6172
6173 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6174         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6175         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6176         .recover_open   = nfs4_open_reclaim,
6177         .recover_lock   = nfs4_lock_reclaim,
6178         .establish_clid = nfs4_init_clientid,
6179         .get_clid_cred  = nfs4_get_setclientid_cred,
6180 };
6181
6182 #if defined(CONFIG_NFS_V4_1)
6183 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6184         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6185         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6186         .recover_open   = nfs4_open_reclaim,
6187         .recover_lock   = nfs4_lock_reclaim,
6188         .establish_clid = nfs41_init_clientid,
6189         .get_clid_cred  = nfs4_get_exchange_id_cred,
6190         .reclaim_complete = nfs41_proc_reclaim_complete,
6191 };
6192 #endif /* CONFIG_NFS_V4_1 */
6193
6194 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6195         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6196         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6197         .recover_open   = nfs4_open_expired,
6198         .recover_lock   = nfs4_lock_expired,
6199         .establish_clid = nfs4_init_clientid,
6200         .get_clid_cred  = nfs4_get_setclientid_cred,
6201 };
6202
6203 #if defined(CONFIG_NFS_V4_1)
6204 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6205         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6206         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6207         .recover_open   = nfs41_open_expired,
6208         .recover_lock   = nfs41_lock_expired,
6209         .establish_clid = nfs41_init_clientid,
6210         .get_clid_cred  = nfs4_get_exchange_id_cred,
6211 };
6212 #endif /* CONFIG_NFS_V4_1 */
6213
6214 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6215         .sched_state_renewal = nfs4_proc_async_renew,
6216         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6217         .renew_lease = nfs4_proc_renew,
6218 };
6219
6220 #if defined(CONFIG_NFS_V4_1)
6221 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6222         .sched_state_renewal = nfs41_proc_async_sequence,
6223         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6224         .renew_lease = nfs4_proc_sequence,
6225 };
6226 #endif
6227
6228 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6229         .minor_version = 0,
6230         .call_sync = _nfs4_call_sync,
6231         .validate_stateid = nfs4_validate_delegation_stateid,
6232         .find_root_sec = nfs4_find_root_sec,
6233         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6234         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6235         .state_renewal_ops = &nfs40_state_renewal_ops,
6236 };
6237
6238 #if defined(CONFIG_NFS_V4_1)
6239 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6240         .minor_version = 1,
6241         .call_sync = _nfs4_call_sync_session,
6242         .validate_stateid = nfs41_validate_delegation_stateid,
6243         .find_root_sec = nfs41_find_root_sec,
6244         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6245         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6246         .state_renewal_ops = &nfs41_state_renewal_ops,
6247 };
6248 #endif
6249
6250 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6251         [0] = &nfs_v4_0_minor_ops,
6252 #if defined(CONFIG_NFS_V4_1)
6253         [1] = &nfs_v4_1_minor_ops,
6254 #endif
6255 };
6256
6257 static const struct inode_operations nfs4_file_inode_operations = {
6258         .permission     = nfs_permission,
6259         .getattr        = nfs_getattr,
6260         .setattr        = nfs_setattr,
6261         .getxattr       = generic_getxattr,
6262         .setxattr       = generic_setxattr,
6263         .listxattr      = generic_listxattr,
6264         .removexattr    = generic_removexattr,
6265 };
6266
6267 const struct nfs_rpc_ops nfs_v4_clientops = {
6268         .version        = 4,                    /* protocol version */
6269         .dentry_ops     = &nfs4_dentry_operations,
6270         .dir_inode_ops  = &nfs4_dir_inode_operations,
6271         .file_inode_ops = &nfs4_file_inode_operations,
6272         .file_ops       = &nfs4_file_operations,
6273         .getroot        = nfs4_proc_get_root,
6274         .getattr        = nfs4_proc_getattr,
6275         .setattr        = nfs4_proc_setattr,
6276         .lookup         = nfs4_proc_lookup,
6277         .access         = nfs4_proc_access,
6278         .readlink       = nfs4_proc_readlink,
6279         .create         = nfs4_proc_create,
6280         .remove         = nfs4_proc_remove,
6281         .unlink_setup   = nfs4_proc_unlink_setup,
6282         .unlink_done    = nfs4_proc_unlink_done,
6283         .rename         = nfs4_proc_rename,
6284         .rename_setup   = nfs4_proc_rename_setup,
6285         .rename_done    = nfs4_proc_rename_done,
6286         .link           = nfs4_proc_link,
6287         .symlink        = nfs4_proc_symlink,
6288         .mkdir          = nfs4_proc_mkdir,
6289         .rmdir          = nfs4_proc_remove,
6290         .readdir        = nfs4_proc_readdir,
6291         .mknod          = nfs4_proc_mknod,
6292         .statfs         = nfs4_proc_statfs,
6293         .fsinfo         = nfs4_proc_fsinfo,
6294         .pathconf       = nfs4_proc_pathconf,
6295         .set_capabilities = nfs4_server_capabilities,
6296         .decode_dirent  = nfs4_decode_dirent,
6297         .read_setup     = nfs4_proc_read_setup,
6298         .read_done      = nfs4_read_done,
6299         .write_setup    = nfs4_proc_write_setup,
6300         .write_done     = nfs4_write_done,
6301         .commit_setup   = nfs4_proc_commit_setup,
6302         .commit_done    = nfs4_commit_done,
6303         .lock           = nfs4_proc_lock,
6304         .clear_acl_cache = nfs4_zap_acl_attr,
6305         .close_context  = nfs4_close_context,
6306         .open_context   = nfs4_atomic_open,
6307         .init_client    = nfs4_init_client,
6308         .secinfo        = nfs4_proc_secinfo,
6309 };
6310
6311 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6312         .prefix = XATTR_NAME_NFSV4_ACL,
6313         .list   = nfs4_xattr_list_nfs4_acl,
6314         .get    = nfs4_xattr_get_nfs4_acl,
6315         .set    = nfs4_xattr_set_nfs4_acl,
6316 };
6317
6318 const struct xattr_handler *nfs4_xattr_handlers[] = {
6319         &nfs4_xattr_nfs4_acl_handler,
6320         NULL
6321 };
6322
6323 /*
6324  * Local variables:
6325  *  c-basic-offset: 8
6326  * End:
6327  */