linux/fs/nfs/nfs4proc.c
<<
>>
Prefs
   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/nfs.h>
  47#include <linux/nfs4.h>
  48#include <linux/nfs_fs.h>
  49#include <linux/nfs_page.h>
  50#include <linux/nfs_mount.h>
  51#include <linux/namei.h>
  52#include <linux/mount.h>
  53#include <linux/module.h>
  54#include <linux/nfs_idmap.h>
  55#include <linux/sunrpc/bc_xprt.h>
  56#include <linux/xattr.h>
  57#include <linux/utsname.h>
  58#include <linux/freezer.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#include "netns.h"
  67
  68#define NFSDBG_FACILITY         NFSDBG_PROC
  69
  70#define NFS4_POLL_RETRY_MIN     (HZ/10)
  71#define NFS4_POLL_RETRY_MAX     (15*HZ)
  72
  73#define NFS4_MAX_LOOP_ON_RECOVER (10)
  74
  75struct nfs4_opendata;
  76static int _nfs4_proc_open(struct nfs4_opendata *data);
  77static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  78static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  79static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  80static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  81static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *);
  82static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  83static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  84                            struct nfs_fattr *fattr, struct iattr *sattr,
  85                            struct nfs4_state *state);
  86#ifdef CONFIG_NFS_V4_1
  87static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
  88static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
  89#endif
  90/* Prevent leaks of NFSv4 errors into userland */
  91static int nfs4_map_errors(int err)
  92{
  93        if (err >= -1000)
  94                return err;
  95        switch (err) {
  96        case -NFS4ERR_RESOURCE:
  97                return -EREMOTEIO;
  98        case -NFS4ERR_WRONGSEC:
  99                return -EPERM;
 100        case -NFS4ERR_BADOWNER:
 101        case -NFS4ERR_BADNAME:
 102                return -EINVAL;
 103        case -NFS4ERR_SHARE_DENIED:
 104                return -EACCES;
 105        case -NFS4ERR_MINOR_VERS_MISMATCH:
 106                return -EPROTONOSUPPORT;
 107        case -NFS4ERR_ACCESS:
 108                return -EACCES;
 109        default:
 110                dprintk("%s could not handle NFSv4 error %d\n",
 111                                __func__, -err);
 112                break;
 113        }
 114        return -EIO;
 115}
 116
 117/*
 118 * This is our standard bitmap for GETATTR requests.
 119 */
 120const u32 nfs4_fattr_bitmap[3] = {
 121        FATTR4_WORD0_TYPE
 122        | FATTR4_WORD0_CHANGE
 123        | FATTR4_WORD0_SIZE
 124        | FATTR4_WORD0_FSID
 125        | FATTR4_WORD0_FILEID,
 126        FATTR4_WORD1_MODE
 127        | FATTR4_WORD1_NUMLINKS
 128        | FATTR4_WORD1_OWNER
 129        | FATTR4_WORD1_OWNER_GROUP
 130        | FATTR4_WORD1_RAWDEV
 131        | FATTR4_WORD1_SPACE_USED
 132        | FATTR4_WORD1_TIME_ACCESS
 133        | FATTR4_WORD1_TIME_METADATA
 134        | FATTR4_WORD1_TIME_MODIFY
 135};
 136
 137static const u32 nfs4_pnfs_open_bitmap[3] = {
 138        FATTR4_WORD0_TYPE
 139        | FATTR4_WORD0_CHANGE
 140        | FATTR4_WORD0_SIZE
 141        | FATTR4_WORD0_FSID
 142        | FATTR4_WORD0_FILEID,
 143        FATTR4_WORD1_MODE
 144        | FATTR4_WORD1_NUMLINKS
 145        | FATTR4_WORD1_OWNER
 146        | FATTR4_WORD1_OWNER_GROUP
 147        | FATTR4_WORD1_RAWDEV
 148        | FATTR4_WORD1_SPACE_USED
 149        | FATTR4_WORD1_TIME_ACCESS
 150        | FATTR4_WORD1_TIME_METADATA
 151        | FATTR4_WORD1_TIME_MODIFY,
 152        FATTR4_WORD2_MDSTHRESHOLD
 153};
 154
 155static const u32 nfs4_open_noattr_bitmap[3] = {
 156        FATTR4_WORD0_TYPE
 157        | FATTR4_WORD0_CHANGE
 158        | FATTR4_WORD0_FILEID,
 159};
 160
 161const u32 nfs4_statfs_bitmap[2] = {
 162        FATTR4_WORD0_FILES_AVAIL
 163        | FATTR4_WORD0_FILES_FREE
 164        | FATTR4_WORD0_FILES_TOTAL,
 165        FATTR4_WORD1_SPACE_AVAIL
 166        | FATTR4_WORD1_SPACE_FREE
 167        | FATTR4_WORD1_SPACE_TOTAL
 168};
 169
 170const u32 nfs4_pathconf_bitmap[2] = {
 171        FATTR4_WORD0_MAXLINK
 172        | FATTR4_WORD0_MAXNAME,
 173        0
 174};
 175
 176const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 177                        | FATTR4_WORD0_MAXREAD
 178                        | FATTR4_WORD0_MAXWRITE
 179                        | FATTR4_WORD0_LEASE_TIME,
 180                        FATTR4_WORD1_TIME_DELTA
 181                        | FATTR4_WORD1_FS_LAYOUT_TYPES,
 182                        FATTR4_WORD2_LAYOUT_BLKSIZE
 183};
 184
 185const u32 nfs4_fs_locations_bitmap[2] = {
 186        FATTR4_WORD0_TYPE
 187        | FATTR4_WORD0_CHANGE
 188        | FATTR4_WORD0_SIZE
 189        | FATTR4_WORD0_FSID
 190        | FATTR4_WORD0_FILEID
 191        | FATTR4_WORD0_FS_LOCATIONS,
 192        FATTR4_WORD1_MODE
 193        | FATTR4_WORD1_NUMLINKS
 194        | FATTR4_WORD1_OWNER
 195        | FATTR4_WORD1_OWNER_GROUP
 196        | FATTR4_WORD1_RAWDEV
 197        | FATTR4_WORD1_SPACE_USED
 198        | FATTR4_WORD1_TIME_ACCESS
 199        | FATTR4_WORD1_TIME_METADATA
 200        | FATTR4_WORD1_TIME_MODIFY
 201        | FATTR4_WORD1_MOUNTED_ON_FILEID
 202};
 203
 204static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 205                struct nfs4_readdir_arg *readdir)
 206{
 207        __be32 *start, *p;
 208
 209        BUG_ON(readdir->count < 80);
 210        if (cookie > 2) {
 211                readdir->cookie = cookie;
 212                memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 213                return;
 214        }
 215
 216        readdir->cookie = 0;
 217        memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 218        if (cookie == 2)
 219                return;
 220        
 221        /*
 222         * NFSv4 servers do not return entries for '.' and '..'
 223         * Therefore, we fake these entries here.  We let '.'
 224         * have cookie 0 and '..' have cookie 1.  Note that
 225         * when talking to the server, we always send cookie 0
 226         * instead of 1 or 2.
 227         */
 228        start = p = kmap_atomic(*readdir->pages);
 229        
 230        if (cookie == 0) {
 231                *p++ = xdr_one;                                  /* next */
 232                *p++ = xdr_zero;                   /* cookie, first word */
 233                *p++ = xdr_one;                   /* cookie, second word */
 234                *p++ = xdr_one;                             /* entry len */
 235                memcpy(p, ".\0\0\0", 4);                        /* entry */
 236                p++;
 237                *p++ = xdr_one;                         /* bitmap length */
 238                *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 239                *p++ = htonl(8);              /* attribute buffer length */
 240                p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 241        }
 242        
 243        *p++ = xdr_one;                                  /* next */
 244        *p++ = xdr_zero;                   /* cookie, first word */
 245        *p++ = xdr_two;                   /* cookie, second word */
 246        *p++ = xdr_two;                             /* entry len */
 247        memcpy(p, "..\0\0", 4);                         /* entry */
 248        p++;
 249        *p++ = xdr_one;                         /* bitmap length */
 250        *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 251        *p++ = htonl(8);              /* attribute buffer length */
 252        p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 253
 254        readdir->pgbase = (char *)p - (char *)start;
 255        readdir->count -= readdir->pgbase;
 256        kunmap_atomic(start);
 257}
 258
 259static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 260{
 261        int res;
 262
 263        might_sleep();
 264
 265        res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 266                        nfs_wait_bit_killable, TASK_KILLABLE);
 267        if (res)
 268                return res;
 269
 270        if (clp->cl_cons_state < 0)
 271                return clp->cl_cons_state;
 272        return 0;
 273}
 274
 275static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 276{
 277        int res = 0;
 278
 279        might_sleep();
 280
 281        if (*timeout <= 0)
 282                *timeout = NFS4_POLL_RETRY_MIN;
 283        if (*timeout > NFS4_POLL_RETRY_MAX)
 284                *timeout = NFS4_POLL_RETRY_MAX;
 285        freezable_schedule_timeout_killable(*timeout);
 286        if (fatal_signal_pending(current))
 287                res = -ERESTARTSYS;
 288        *timeout <<= 1;
 289        return res;
 290}
 291
 292/* This is the error handling routine for processes that are allowed
 293 * to sleep.
 294 */
 295static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 296{
 297        struct nfs_client *clp = server->nfs_client;
 298        struct nfs4_state *state = exception->state;
 299        struct inode *inode = exception->inode;
 300        int ret = errorcode;
 301
 302        exception->retry = 0;
 303        switch(errorcode) {
 304                case 0:
 305                        return 0;
 306                case -NFS4ERR_OPENMODE:
 307                        if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 308                                nfs4_inode_return_delegation(inode);
 309                                exception->retry = 1;
 310                                return 0;
 311                        }
 312                        if (state == NULL)
 313                                break;
 314                        nfs4_schedule_stateid_recovery(server, state);
 315                        goto wait_on_recovery;
 316                case -NFS4ERR_DELEG_REVOKED:
 317                case -NFS4ERR_ADMIN_REVOKED:
 318                case -NFS4ERR_BAD_STATEID:
 319                        if (state == NULL)
 320                                break;
 321                        nfs_remove_bad_delegation(state->inode);
 322                        nfs4_schedule_stateid_recovery(server, state);
 323                        goto wait_on_recovery;
 324                case -NFS4ERR_EXPIRED:
 325                        if (state != NULL)
 326                                nfs4_schedule_stateid_recovery(server, state);
 327                case -NFS4ERR_STALE_STATEID:
 328                case -NFS4ERR_STALE_CLIENTID:
 329                        nfs4_schedule_lease_recovery(clp);
 330                        goto wait_on_recovery;
 331#if defined(CONFIG_NFS_V4_1)
 332                case -NFS4ERR_BADSESSION:
 333                case -NFS4ERR_BADSLOT:
 334                case -NFS4ERR_BAD_HIGH_SLOT:
 335                case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 336                case -NFS4ERR_DEADSESSION:
 337                case -NFS4ERR_SEQ_FALSE_RETRY:
 338                case -NFS4ERR_SEQ_MISORDERED:
 339                        dprintk("%s ERROR: %d Reset session\n", __func__,
 340                                errorcode);
 341                        nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 342                        goto wait_on_recovery;
 343#endif /* defined(CONFIG_NFS_V4_1) */
 344                case -NFS4ERR_FILE_OPEN:
 345                        if (exception->timeout > HZ) {
 346                                /* We have retried a decent amount, time to
 347                                 * fail
 348                                 */
 349                                ret = -EBUSY;
 350                                break;
 351                        }
 352                case -NFS4ERR_GRACE:
 353                case -NFS4ERR_DELAY:
 354                case -EKEYEXPIRED:
 355                        ret = nfs4_delay(server->client, &exception->timeout);
 356                        if (ret != 0)
 357                                break;
 358                case -NFS4ERR_RETRY_UNCACHED_REP:
 359                case -NFS4ERR_OLD_STATEID:
 360                        exception->retry = 1;
 361                        break;
 362                case -NFS4ERR_BADOWNER:
 363                        /* The following works around a Linux server bug! */
 364                case -NFS4ERR_BADNAME:
 365                        if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 366                                server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 367                                exception->retry = 1;
 368                                printk(KERN_WARNING "NFS: v4 server %s "
 369                                                "does not accept raw "
 370                                                "uid/gids. "
 371                                                "Reenabling the idmapper.\n",
 372                                                server->nfs_client->cl_hostname);
 373                        }
 374        }
 375        /* We failed to handle the error */
 376        return nfs4_map_errors(ret);
 377wait_on_recovery:
 378        ret = nfs4_wait_clnt_recover(clp);
 379        if (ret == 0)
 380                exception->retry = 1;
 381        return ret;
 382}
 383
 384
 385static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 386{
 387        spin_lock(&clp->cl_lock);
 388        if (time_before(clp->cl_last_renewal,timestamp))
 389                clp->cl_last_renewal = timestamp;
 390        spin_unlock(&clp->cl_lock);
 391}
 392
 393static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 394{
 395        do_renew_lease(server->nfs_client, timestamp);
 396}
 397
 398#if defined(CONFIG_NFS_V4_1)
 399
 400/*
 401 * nfs4_free_slot - free a slot and efficiently update slot table.
 402 *
 403 * freeing a slot is trivially done by clearing its respective bit
 404 * in the bitmap.
 405 * If the freed slotid equals highest_used_slotid we want to update it
 406 * so that the server would be able to size down the slot table if needed,
 407 * otherwise we know that the highest_used_slotid is still in use.
 408 * When updating highest_used_slotid there may be "holes" in the bitmap
 409 * so we need to scan down from highest_used_slotid to 0 looking for the now
 410 * highest slotid in use.
 411 * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
 412 *
 413 * Must be called while holding tbl->slot_tbl_lock
 414 */
 415static void
 416nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
 417{
 418        BUG_ON(slotid >= NFS4_MAX_SLOT_TABLE);
 419        /* clear used bit in bitmap */
 420        __clear_bit(slotid, tbl->used_slots);
 421
 422        /* update highest_used_slotid when it is freed */
 423        if (slotid == tbl->highest_used_slotid) {
 424                slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 425                if (slotid < tbl->max_slots)
 426                        tbl->highest_used_slotid = slotid;
 427                else
 428                        tbl->highest_used_slotid = NFS4_NO_SLOT;
 429        }
 430        dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
 431                slotid, tbl->highest_used_slotid);
 432}
 433
 434bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
 435{
 436        rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 437        return true;
 438}
 439
 440/*
 441 * Signal state manager thread if session fore channel is drained
 442 */
 443static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 444{
 445        if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 446                rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
 447                                nfs4_set_task_privileged, NULL);
 448                return;
 449        }
 450
 451        if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 452                return;
 453
 454        dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 455        complete(&ses->fc_slot_table.complete);
 456}
 457
 458/*
 459 * Signal state manager thread if session back channel is drained
 460 */
 461void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 462{
 463        if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 464            ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 465                return;
 466        dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 467        complete(&ses->bc_slot_table.complete);
 468}
 469
 470static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 471{
 472        struct nfs4_slot_table *tbl;
 473
 474        tbl = &res->sr_session->fc_slot_table;
 475        if (!res->sr_slot) {
 476                /* just wake up the next guy waiting since
 477                 * we may have not consumed a slot after all */
 478                dprintk("%s: No slot\n", __func__);
 479                return;
 480        }
 481
 482        spin_lock(&tbl->slot_tbl_lock);
 483        nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 484        nfs4_check_drain_fc_complete(res->sr_session);
 485        spin_unlock(&tbl->slot_tbl_lock);
 486        res->sr_slot = NULL;
 487}
 488
 489static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 490{
 491        unsigned long timestamp;
 492        struct nfs_client *clp;
 493
 494        /*
 495         * sr_status remains 1 if an RPC level error occurred. The server
 496         * may or may not have processed the sequence operation..
 497         * Proceed as if the server received and processed the sequence
 498         * operation.
 499         */
 500        if (res->sr_status == 1)
 501                res->sr_status = NFS_OK;
 502
 503        /* don't increment the sequence number if the task wasn't sent */
 504        if (!RPC_WAS_SENT(task))
 505                goto out;
 506
 507        /* Check the SEQUENCE operation status */
 508        switch (res->sr_status) {
 509        case 0:
 510                /* Update the slot's sequence and clientid lease timer */
 511                ++res->sr_slot->seq_nr;
 512                timestamp = res->sr_renewal_time;
 513                clp = res->sr_session->clp;
 514                do_renew_lease(clp, timestamp);
 515                /* Check sequence flags */
 516                if (res->sr_status_flags != 0)
 517                        nfs4_schedule_lease_recovery(clp);
 518                break;
 519        case -NFS4ERR_DELAY:
 520                /* The server detected a resend of the RPC call and
 521                 * returned NFS4ERR_DELAY as per Section 2.10.6.2
 522                 * of RFC5661.
 523                 */
 524                dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 525                        __func__,
 526                        res->sr_slot - res->sr_session->fc_slot_table.slots,
 527                        res->sr_slot->seq_nr);
 528                goto out_retry;
 529        default:
 530                /* Just update the slot sequence no. */
 531                ++res->sr_slot->seq_nr;
 532        }
 533out:
 534        /* The session may be reset by one of the error handlers. */
 535        dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 536        nfs41_sequence_free_slot(res);
 537        return 1;
 538out_retry:
 539        if (!rpc_restart_call(task))
 540                goto out;
 541        rpc_delay(task, NFS4_POLL_RETRY_MAX);
 542        return 0;
 543}
 544
 545static int nfs4_sequence_done(struct rpc_task *task,
 546                               struct nfs4_sequence_res *res)
 547{
 548        if (res->sr_session == NULL)
 549                return 1;
 550        return nfs41_sequence_done(task, res);
 551}
 552
 553/*
 554 * nfs4_find_slot - efficiently look for a free slot
 555 *
 556 * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 557 * If found, we mark the slot as used, update the highest_used_slotid,
 558 * and respectively set up the sequence operation args.
 559 * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
 560 *
 561 * Note: must be called with under the slot_tbl_lock.
 562 */
 563static u32
 564nfs4_find_slot(struct nfs4_slot_table *tbl)
 565{
 566        u32 slotid;
 567        u32 ret_id = NFS4_NO_SLOT;
 568
 569        dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
 570                __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 571                tbl->max_slots);
 572        slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 573        if (slotid >= tbl->max_slots)
 574                goto out;
 575        __set_bit(slotid, tbl->used_slots);
 576        if (slotid > tbl->highest_used_slotid ||
 577                        tbl->highest_used_slotid == NFS4_NO_SLOT)
 578                tbl->highest_used_slotid = slotid;
 579        ret_id = slotid;
 580out:
 581        dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 582                __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 583        return ret_id;
 584}
 585
 586static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 587                struct nfs4_sequence_res *res, int cache_reply)
 588{
 589        args->sa_session = NULL;
 590        args->sa_cache_this = 0;
 591        if (cache_reply)
 592                args->sa_cache_this = 1;
 593        res->sr_session = NULL;
 594        res->sr_slot = NULL;
 595}
 596
 597int nfs41_setup_sequence(struct nfs4_session *session,
 598                                struct nfs4_sequence_args *args,
 599                                struct nfs4_sequence_res *res,
 600                                struct rpc_task *task)
 601{
 602        struct nfs4_slot *slot;
 603        struct nfs4_slot_table *tbl;
 604        u32 slotid;
 605
 606        dprintk("--> %s\n", __func__);
 607        /* slot already allocated? */
 608        if (res->sr_slot != NULL)
 609                return 0;
 610
 611        tbl = &session->fc_slot_table;
 612
 613        spin_lock(&tbl->slot_tbl_lock);
 614        if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 615            !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 616                /* The state manager will wait until the slot table is empty */
 617                rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 618                spin_unlock(&tbl->slot_tbl_lock);
 619                dprintk("%s session is draining\n", __func__);
 620                return -EAGAIN;
 621        }
 622
 623        if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 624            !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 625                rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 626                spin_unlock(&tbl->slot_tbl_lock);
 627                dprintk("%s enforce FIFO order\n", __func__);
 628                return -EAGAIN;
 629        }
 630
 631        slotid = nfs4_find_slot(tbl);
 632        if (slotid == NFS4_NO_SLOT) {
 633                rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 634                spin_unlock(&tbl->slot_tbl_lock);
 635                dprintk("<-- %s: no free slots\n", __func__);
 636                return -EAGAIN;
 637        }
 638        spin_unlock(&tbl->slot_tbl_lock);
 639
 640        rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 641        slot = tbl->slots + slotid;
 642        args->sa_session = session;
 643        args->sa_slotid = slotid;
 644
 645        dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 646
 647        res->sr_session = session;
 648        res->sr_slot = slot;
 649        res->sr_renewal_time = jiffies;
 650        res->sr_status_flags = 0;
 651        /*
 652         * sr_status is only set in decode_sequence, and so will remain
 653         * set to 1 if an rpc level failure occurs.
 654         */
 655        res->sr_status = 1;
 656        return 0;
 657}
 658EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 659
 660int nfs4_setup_sequence(const struct nfs_server *server,
 661                        struct nfs4_sequence_args *args,
 662                        struct nfs4_sequence_res *res,
 663                        struct rpc_task *task)
 664{
 665        struct nfs4_session *session = nfs4_get_session(server);
 666        int ret = 0;
 667
 668        if (session == NULL)
 669                goto out;
 670
 671        dprintk("--> %s clp %p session %p sr_slot %td\n",
 672                __func__, session->clp, session, res->sr_slot ?
 673                        res->sr_slot - session->fc_slot_table.slots : -1);
 674
 675        ret = nfs41_setup_sequence(session, args, res, task);
 676out:
 677        dprintk("<-- %s status=%d\n", __func__, ret);
 678        return ret;
 679}
 680
 681struct nfs41_call_sync_data {
 682        const struct nfs_server *seq_server;
 683        struct nfs4_sequence_args *seq_args;
 684        struct nfs4_sequence_res *seq_res;
 685};
 686
 687static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 688{
 689        struct nfs41_call_sync_data *data = calldata;
 690
 691        dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 692
 693        if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 694                                data->seq_res, task))
 695                return;
 696        rpc_call_start(task);
 697}
 698
 699static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 700{
 701        rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 702        nfs41_call_sync_prepare(task, calldata);
 703}
 704
 705static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 706{
 707        struct nfs41_call_sync_data *data = calldata;
 708
 709        nfs41_sequence_done(task, data->seq_res);
 710}
 711
 712static const struct rpc_call_ops nfs41_call_sync_ops = {
 713        .rpc_call_prepare = nfs41_call_sync_prepare,
 714        .rpc_call_done = nfs41_call_sync_done,
 715};
 716
 717static const struct rpc_call_ops nfs41_call_priv_sync_ops = {
 718        .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 719        .rpc_call_done = nfs41_call_sync_done,
 720};
 721
 722static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 723                                   struct nfs_server *server,
 724                                   struct rpc_message *msg,
 725                                   struct nfs4_sequence_args *args,
 726                                   struct nfs4_sequence_res *res,
 727                                   int privileged)
 728{
 729        int ret;
 730        struct rpc_task *task;
 731        struct nfs41_call_sync_data data = {
 732                .seq_server = server,
 733                .seq_args = args,
 734                .seq_res = res,
 735        };
 736        struct rpc_task_setup task_setup = {
 737                .rpc_client = clnt,
 738                .rpc_message = msg,
 739                .callback_ops = &nfs41_call_sync_ops,
 740                .callback_data = &data
 741        };
 742
 743        if (privileged)
 744                task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 745        task = rpc_run_task(&task_setup);
 746        if (IS_ERR(task))
 747                ret = PTR_ERR(task);
 748        else {
 749                ret = task->tk_status;
 750                rpc_put_task(task);
 751        }
 752        return ret;
 753}
 754
 755int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 756                            struct nfs_server *server,
 757                            struct rpc_message *msg,
 758                            struct nfs4_sequence_args *args,
 759                            struct nfs4_sequence_res *res,
 760                            int cache_reply)
 761{
 762        nfs41_init_sequence(args, res, cache_reply);
 763        return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
 764}
 765
 766#else
 767static inline
 768void nfs41_init_sequence(struct nfs4_sequence_args *args,
 769                struct nfs4_sequence_res *res, int cache_reply)
 770{
 771}
 772
 773static int nfs4_sequence_done(struct rpc_task *task,
 774                               struct nfs4_sequence_res *res)
 775{
 776        return 1;
 777}
 778#endif /* CONFIG_NFS_V4_1 */
 779
 780int _nfs4_call_sync(struct rpc_clnt *clnt,
 781                    struct nfs_server *server,
 782                    struct rpc_message *msg,
 783                    struct nfs4_sequence_args *args,
 784                    struct nfs4_sequence_res *res,
 785                    int cache_reply)
 786{
 787        nfs41_init_sequence(args, res, cache_reply);
 788        return rpc_call_sync(clnt, msg, 0);
 789}
 790
 791static inline
 792int nfs4_call_sync(struct rpc_clnt *clnt,
 793                   struct nfs_server *server,
 794                   struct rpc_message *msg,
 795                   struct nfs4_sequence_args *args,
 796                   struct nfs4_sequence_res *res,
 797                   int cache_reply)
 798{
 799        return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 800                                                args, res, cache_reply);
 801}
 802
 803static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 804{
 805        struct nfs_inode *nfsi = NFS_I(dir);
 806
 807        spin_lock(&dir->i_lock);
 808        nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 809        if (!cinfo->atomic || cinfo->before != dir->i_version)
 810                nfs_force_lookup_revalidate(dir);
 811        dir->i_version = cinfo->after;
 812        spin_unlock(&dir->i_lock);
 813}
 814
 815struct nfs4_opendata {
 816        struct kref kref;
 817        struct nfs_openargs o_arg;
 818        struct nfs_openres o_res;
 819        struct nfs_open_confirmargs c_arg;
 820        struct nfs_open_confirmres c_res;
 821        struct nfs4_string owner_name;
 822        struct nfs4_string group_name;
 823        struct nfs_fattr f_attr;
 824        struct dentry *dir;
 825        struct dentry *dentry;
 826        struct nfs4_state_owner *owner;
 827        struct nfs4_state *state;
 828        struct iattr attrs;
 829        unsigned long timestamp;
 830        unsigned int rpc_done : 1;
 831        int rpc_status;
 832        int cancelled;
 833};
 834
 835
 836static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 837{
 838        p->o_res.f_attr = &p->f_attr;
 839        p->o_res.seqid = p->o_arg.seqid;
 840        p->c_res.seqid = p->c_arg.seqid;
 841        p->o_res.server = p->o_arg.server;
 842        p->o_res.access_request = p->o_arg.access;
 843        nfs_fattr_init(&p->f_attr);
 844        nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 845}
 846
 847static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 848                struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 849                const struct iattr *attrs,
 850                gfp_t gfp_mask)
 851{
 852        struct dentry *parent = dget_parent(dentry);
 853        struct inode *dir = parent->d_inode;
 854        struct nfs_server *server = NFS_SERVER(dir);
 855        struct nfs4_opendata *p;
 856
 857        p = kzalloc(sizeof(*p), gfp_mask);
 858        if (p == NULL)
 859                goto err;
 860        p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 861        if (p->o_arg.seqid == NULL)
 862                goto err_free;
 863        nfs_sb_active(dentry->d_sb);
 864        p->dentry = dget(dentry);
 865        p->dir = parent;
 866        p->owner = sp;
 867        atomic_inc(&sp->so_count);
 868        p->o_arg.fh = NFS_FH(dir);
 869        p->o_arg.open_flags = flags;
 870        p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 871        /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
 872         * will return permission denied for all bits until close */
 873        if (!(flags & O_EXCL)) {
 874                /* ask server to check for all possible rights as results
 875                 * are cached */
 876                p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
 877                                  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
 878        }
 879        p->o_arg.clientid = server->nfs_client->cl_clientid;
 880        p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
 881        p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 882        p->o_arg.name = &dentry->d_name;
 883        p->o_arg.server = server;
 884        p->o_arg.bitmask = server->attr_bitmask;
 885        p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
 886        p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 887        if (attrs != NULL && attrs->ia_valid != 0) {
 888                __be32 verf[2];
 889
 890                p->o_arg.u.attrs = &p->attrs;
 891                memcpy(&p->attrs, attrs, sizeof(p->attrs));
 892
 893                verf[0] = jiffies;
 894                verf[1] = current->pid;
 895                memcpy(p->o_arg.u.verifier.data, verf,
 896                                sizeof(p->o_arg.u.verifier.data));
 897        }
 898        p->c_arg.fh = &p->o_res.fh;
 899        p->c_arg.stateid = &p->o_res.stateid;
 900        p->c_arg.seqid = p->o_arg.seqid;
 901        nfs4_init_opendata_res(p);
 902        kref_init(&p->kref);
 903        return p;
 904err_free:
 905        kfree(p);
 906err:
 907        dput(parent);
 908        return NULL;
 909}
 910
 911static void nfs4_opendata_free(struct kref *kref)
 912{
 913        struct nfs4_opendata *p = container_of(kref,
 914                        struct nfs4_opendata, kref);
 915        struct super_block *sb = p->dentry->d_sb;
 916
 917        nfs_free_seqid(p->o_arg.seqid);
 918        if (p->state != NULL)
 919                nfs4_put_open_state(p->state);
 920        nfs4_put_state_owner(p->owner);
 921        dput(p->dir);
 922        dput(p->dentry);
 923        nfs_sb_deactive(sb);
 924        nfs_fattr_free_names(&p->f_attr);
 925        kfree(p);
 926}
 927
 928static void nfs4_opendata_put(struct nfs4_opendata *p)
 929{
 930        if (p != NULL)
 931                kref_put(&p->kref, nfs4_opendata_free);
 932}
 933
 934static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 935{
 936        int ret;
 937
 938        ret = rpc_wait_for_completion_task(task);
 939        return ret;
 940}
 941
 942static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 943{
 944        int ret = 0;
 945
 946        if (open_mode & (O_EXCL|O_TRUNC))
 947                goto out;
 948        switch (mode & (FMODE_READ|FMODE_WRITE)) {
 949                case FMODE_READ:
 950                        ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 951                                && state->n_rdonly != 0;
 952                        break;
 953                case FMODE_WRITE:
 954                        ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 955                                && state->n_wronly != 0;
 956                        break;
 957                case FMODE_READ|FMODE_WRITE:
 958                        ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 959                                && state->n_rdwr != 0;
 960        }
 961out:
 962        return ret;
 963}
 964
 965static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 966{
 967        if (delegation == NULL)
 968                return 0;
 969        if ((delegation->type & fmode) != fmode)
 970                return 0;
 971        if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 972                return 0;
 973        nfs_mark_delegation_referenced(delegation);
 974        return 1;
 975}
 976
 977static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 978{
 979        switch (fmode) {
 980                case FMODE_WRITE:
 981                        state->n_wronly++;
 982                        break;
 983                case FMODE_READ:
 984                        state->n_rdonly++;
 985                        break;
 986                case FMODE_READ|FMODE_WRITE:
 987                        state->n_rdwr++;
 988        }
 989        nfs4_state_set_mode_locked(state, state->state | fmode);
 990}
 991
 992static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 993{
 994        if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 995                nfs4_stateid_copy(&state->stateid, stateid);
 996        nfs4_stateid_copy(&state->open_stateid, stateid);
 997        switch (fmode) {
 998                case FMODE_READ:
 999                        set_bit(NFS_O_RDONLY_STATE, &state->flags);
1000                        break;
1001                case FMODE_WRITE:
1002                        set_bit(NFS_O_WRONLY_STATE, &state->flags);
1003                        break;
1004                case FMODE_READ|FMODE_WRITE:
1005                        set_bit(NFS_O_RDWR_STATE, &state->flags);
1006        }
1007}
1008
1009static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1010{
1011        write_seqlock(&state->seqlock);
1012        nfs_set_open_stateid_locked(state, stateid, fmode);
1013        write_sequnlock(&state->seqlock);
1014}
1015
1016static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1017{
1018        /*
1019         * Protect the call to nfs4_state_set_mode_locked and
1020         * serialise the stateid update
1021         */
1022        write_seqlock(&state->seqlock);
1023        if (deleg_stateid != NULL) {
1024                nfs4_stateid_copy(&state->stateid, deleg_stateid);
1025                set_bit(NFS_DELEGATED_STATE, &state->flags);
1026        }
1027        if (open_stateid != NULL)
1028                nfs_set_open_stateid_locked(state, open_stateid, fmode);
1029        write_sequnlock(&state->seqlock);
1030        spin_lock(&state->owner->so_lock);
1031        update_open_stateflags(state, fmode);
1032        spin_unlock(&state->owner->so_lock);
1033}
1034
1035static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1036{
1037        struct nfs_inode *nfsi = NFS_I(state->inode);
1038        struct nfs_delegation *deleg_cur;
1039        int ret = 0;
1040
1041        fmode &= (FMODE_READ|FMODE_WRITE);
1042
1043        rcu_read_lock();
1044        deleg_cur = rcu_dereference(nfsi->delegation);
1045        if (deleg_cur == NULL)
1046                goto no_delegation;
1047
1048        spin_lock(&deleg_cur->lock);
1049        if (nfsi->delegation != deleg_cur ||
1050            (deleg_cur->type & fmode) != fmode)
1051                goto no_delegation_unlock;
1052
1053        if (delegation == NULL)
1054                delegation = &deleg_cur->stateid;
1055        else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1056                goto no_delegation_unlock;
1057
1058        nfs_mark_delegation_referenced(deleg_cur);
1059        __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1060        ret = 1;
1061no_delegation_unlock:
1062        spin_unlock(&deleg_cur->lock);
1063no_delegation:
1064        rcu_read_unlock();
1065
1066        if (!ret && open_stateid != NULL) {
1067                __update_open_stateid(state, open_stateid, NULL, fmode);
1068                ret = 1;
1069        }
1070
1071        return ret;
1072}
1073
1074
1075static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1076{
1077        struct nfs_delegation *delegation;
1078
1079        rcu_read_lock();
1080        delegation = rcu_dereference(NFS_I(inode)->delegation);
1081        if (delegation == NULL || (delegation->type & fmode) == fmode) {
1082                rcu_read_unlock();
1083                return;
1084        }
1085        rcu_read_unlock();
1086        nfs4_inode_return_delegation(inode);
1087}
1088
1089static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1090{
1091        struct nfs4_state *state = opendata->state;
1092        struct nfs_inode *nfsi = NFS_I(state->inode);
1093        struct nfs_delegation *delegation;
1094        int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1095        fmode_t fmode = opendata->o_arg.fmode;
1096        nfs4_stateid stateid;
1097        int ret = -EAGAIN;
1098
1099        for (;;) {
1100                if (can_open_cached(state, fmode, open_mode)) {
1101                        spin_lock(&state->owner->so_lock);
1102                        if (can_open_cached(state, fmode, open_mode)) {
1103                                update_open_stateflags(state, fmode);
1104                                spin_unlock(&state->owner->so_lock);
1105                                goto out_return_state;
1106                        }
1107                        spin_unlock(&state->owner->so_lock);
1108                }
1109                rcu_read_lock();
1110                delegation = rcu_dereference(nfsi->delegation);
1111                if (!can_open_delegated(delegation, fmode)) {
1112                        rcu_read_unlock();
1113                        break;
1114                }
1115                /* Save the delegation */
1116                nfs4_stateid_copy(&stateid, &delegation->stateid);
1117                rcu_read_unlock();
1118                ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1119                if (ret != 0)
1120                        goto out;
1121                ret = -EAGAIN;
1122
1123                /* Try to update the stateid using the delegation */
1124                if (update_open_stateid(state, NULL, &stateid, fmode))
1125                        goto out_return_state;
1126        }
1127out:
1128        return ERR_PTR(ret);
1129out_return_state:
1130        atomic_inc(&state->count);
1131        return state;
1132}
1133
1134static void
1135nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1136{
1137        struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1138        struct nfs_delegation *delegation;
1139        int delegation_flags = 0;
1140
1141        rcu_read_lock();
1142        delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1143        if (delegation)
1144                delegation_flags = delegation->flags;
1145        rcu_read_unlock();
1146        if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1147                pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1148                                   "returning a delegation for "
1149                                   "OPEN(CLAIM_DELEGATE_CUR)\n",
1150                                   clp->cl_hostname);
1151        } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1152                nfs_inode_set_delegation(state->inode,
1153                                         data->owner->so_cred,
1154                                         &data->o_res);
1155        else
1156                nfs_inode_reclaim_delegation(state->inode,
1157                                             data->owner->so_cred,
1158                                             &data->o_res);
1159}
1160
1161/*
1162 * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1163 * and update the nfs4_state.
1164 */
1165static struct nfs4_state *
1166_nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1167{
1168        struct inode *inode = data->state->inode;
1169        struct nfs4_state *state = data->state;
1170        int ret;
1171
1172        if (!data->rpc_done) {
1173                ret = data->rpc_status;
1174                goto err;
1175        }
1176
1177        ret = -ESTALE;
1178        if (!(data->f_attr.valid & NFS_ATTR_FATTR_TYPE) ||
1179            !(data->f_attr.valid & NFS_ATTR_FATTR_FILEID) ||
1180            !(data->f_attr.valid & NFS_ATTR_FATTR_CHANGE))
1181                goto err;
1182
1183        ret = -ENOMEM;
1184        state = nfs4_get_open_state(inode, data->owner);
1185        if (state == NULL)
1186                goto err;
1187
1188        ret = nfs_refresh_inode(inode, &data->f_attr);
1189        if (ret)
1190                goto err;
1191
1192        if (data->o_res.delegation_type != 0)
1193                nfs4_opendata_check_deleg(data, state);
1194        update_open_stateid(state, &data->o_res.stateid, NULL,
1195                            data->o_arg.fmode);
1196
1197        return state;
1198err:
1199        return ERR_PTR(ret);
1200
1201}
1202
1203static struct nfs4_state *
1204_nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1205{
1206        struct inode *inode;
1207        struct nfs4_state *state = NULL;
1208        int ret;
1209
1210        if (!data->rpc_done) {
1211                state = nfs4_try_open_cached(data);
1212                goto out;
1213        }
1214
1215        ret = -EAGAIN;
1216        if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1217                goto err;
1218        inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1219        ret = PTR_ERR(inode);
1220        if (IS_ERR(inode))
1221                goto err;
1222        ret = -ENOMEM;
1223        state = nfs4_get_open_state(inode, data->owner);
1224        if (state == NULL)
1225                goto err_put_inode;
1226        if (data->o_res.delegation_type != 0)
1227                nfs4_opendata_check_deleg(data, state);
1228        update_open_stateid(state, &data->o_res.stateid, NULL,
1229                        data->o_arg.fmode);
1230        iput(inode);
1231out:
1232        return state;
1233err_put_inode:
1234        iput(inode);
1235err:
1236        return ERR_PTR(ret);
1237}
1238
1239static struct nfs4_state *
1240nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1241{
1242        if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1243                return _nfs4_opendata_reclaim_to_nfs4_state(data);
1244        return _nfs4_opendata_to_nfs4_state(data);
1245}
1246
1247static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1248{
1249        struct nfs_inode *nfsi = NFS_I(state->inode);
1250        struct nfs_open_context *ctx;
1251
1252        spin_lock(&state->inode->i_lock);
1253        list_for_each_entry(ctx, &nfsi->open_files, list) {
1254                if (ctx->state != state)
1255                        continue;
1256                get_nfs_open_context(ctx);
1257                spin_unlock(&state->inode->i_lock);
1258                return ctx;
1259        }
1260        spin_unlock(&state->inode->i_lock);
1261        return ERR_PTR(-ENOENT);
1262}
1263
1264static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1265{
1266        struct nfs4_opendata *opendata;
1267
1268        opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1269        if (opendata == NULL)
1270                return ERR_PTR(-ENOMEM);
1271        opendata->state = state;
1272        atomic_inc(&state->count);
1273        return opendata;
1274}
1275
1276static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1277{
1278        struct nfs4_state *newstate;
1279        int ret;
1280
1281        opendata->o_arg.open_flags = 0;
1282        opendata->o_arg.fmode = fmode;
1283        memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1284        memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1285        nfs4_init_opendata_res(opendata);
1286        ret = _nfs4_recover_proc_open(opendata);
1287        if (ret != 0)
1288                return ret; 
1289        newstate = nfs4_opendata_to_nfs4_state(opendata);
1290        if (IS_ERR(newstate))
1291                return PTR_ERR(newstate);
1292        nfs4_close_state(newstate, fmode);
1293        *res = newstate;
1294        return 0;
1295}
1296
1297static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1298{
1299        struct nfs4_state *newstate;
1300        int ret;
1301
1302        /* memory barrier prior to reading state->n_* */
1303        clear_bit(NFS_DELEGATED_STATE, &state->flags);
1304        smp_rmb();
1305        if (state->n_rdwr != 0) {
1306                clear_bit(NFS_O_RDWR_STATE, &state->flags);
1307                ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1308                if (ret != 0)
1309                        return ret;
1310                if (newstate != state)
1311                        return -ESTALE;
1312        }
1313        if (state->n_wronly != 0) {
1314                clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1315                ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1316                if (ret != 0)
1317                        return ret;
1318                if (newstate != state)
1319                        return -ESTALE;
1320        }
1321        if (state->n_rdonly != 0) {
1322                clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1323                ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1324                if (ret != 0)
1325                        return ret;
1326                if (newstate != state)
1327                        return -ESTALE;
1328        }
1329        /*
1330         * We may have performed cached opens for all three recoveries.
1331         * Check if we need to update the current stateid.
1332         */
1333        if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1334            !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1335                write_seqlock(&state->seqlock);
1336                if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1337                        nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1338                write_sequnlock(&state->seqlock);
1339        }
1340        return 0;
1341}
1342
1343/*
1344 * OPEN_RECLAIM:
1345 *      reclaim state on the server after a reboot.
1346 */
1347static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1348{
1349        struct nfs_delegation *delegation;
1350        struct nfs4_opendata *opendata;
1351        fmode_t delegation_type = 0;
1352        int status;
1353
1354        opendata = nfs4_open_recoverdata_alloc(ctx, state);
1355        if (IS_ERR(opendata))
1356                return PTR_ERR(opendata);
1357        opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1358        opendata->o_arg.fh = NFS_FH(state->inode);
1359        rcu_read_lock();
1360        delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1361        if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1362                delegation_type = delegation->type;
1363        rcu_read_unlock();
1364        opendata->o_arg.u.delegation_type = delegation_type;
1365        status = nfs4_open_recover(opendata, state);
1366        nfs4_opendata_put(opendata);
1367        return status;
1368}
1369
1370static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1371{
1372        struct nfs_server *server = NFS_SERVER(state->inode);
1373        struct nfs4_exception exception = { };
1374        int err;
1375        do {
1376                err = _nfs4_do_open_reclaim(ctx, state);
1377                if (err != -NFS4ERR_DELAY)
1378                        break;
1379                nfs4_handle_exception(server, err, &exception);
1380        } while (exception.retry);
1381        return err;
1382}
1383
1384static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1385{
1386        struct nfs_open_context *ctx;
1387        int ret;
1388
1389        ctx = nfs4_state_find_open_context(state);
1390        if (IS_ERR(ctx))
1391                return PTR_ERR(ctx);
1392        ret = nfs4_do_open_reclaim(ctx, state);
1393        put_nfs_open_context(ctx);
1394        return ret;
1395}
1396
1397static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1398{
1399        struct nfs4_opendata *opendata;
1400        int ret;
1401
1402        opendata = nfs4_open_recoverdata_alloc(ctx, state);
1403        if (IS_ERR(opendata))
1404                return PTR_ERR(opendata);
1405        opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1406        nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1407        ret = nfs4_open_recover(opendata, state);
1408        nfs4_opendata_put(opendata);
1409        return ret;
1410}
1411
1412int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1413{
1414        struct nfs4_exception exception = { };
1415        struct nfs_server *server = NFS_SERVER(state->inode);
1416        int err;
1417        do {
1418                err = _nfs4_open_delegation_recall(ctx, state, stateid);
1419                switch (err) {
1420                        case 0:
1421                        case -ENOENT:
1422                        case -ESTALE:
1423                                goto out;
1424                        case -NFS4ERR_BADSESSION:
1425                        case -NFS4ERR_BADSLOT:
1426                        case -NFS4ERR_BAD_HIGH_SLOT:
1427                        case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1428                        case -NFS4ERR_DEADSESSION:
1429                                nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1430                                goto out;
1431                        case -NFS4ERR_STALE_CLIENTID:
1432                        case -NFS4ERR_STALE_STATEID:
1433                        case -NFS4ERR_EXPIRED:
1434                                /* Don't recall a delegation if it was lost */
1435                                nfs4_schedule_lease_recovery(server->nfs_client);
1436                                goto out;
1437                        case -ERESTARTSYS:
1438                                /*
1439                                 * The show must go on: exit, but mark the
1440                                 * stateid as needing recovery.
1441                                 */
1442                        case -NFS4ERR_DELEG_REVOKED:
1443                        case -NFS4ERR_ADMIN_REVOKED:
1444                        case -NFS4ERR_BAD_STATEID:
1445                                nfs_inode_find_state_and_recover(state->inode,
1446                                                stateid);
1447                                nfs4_schedule_stateid_recovery(server, state);
1448                        case -EKEYEXPIRED:
1449                                /*
1450                                 * User RPCSEC_GSS context has expired.
1451                                 * We cannot recover this stateid now, so
1452                                 * skip it and allow recovery thread to
1453                                 * proceed.
1454                                 */
1455                        case -ENOMEM:
1456                                err = 0;
1457                                goto out;
1458                }
1459                err = nfs4_handle_exception(server, err, &exception);
1460        } while (exception.retry);
1461out:
1462        return err;
1463}
1464
1465static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1466{
1467        struct nfs4_opendata *data = calldata;
1468
1469        data->rpc_status = task->tk_status;
1470        if (data->rpc_status == 0) {
1471                nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1472                nfs_confirm_seqid(&data->owner->so_seqid, 0);
1473                renew_lease(data->o_res.server, data->timestamp);
1474                data->rpc_done = 1;
1475        }
1476}
1477
1478static void nfs4_open_confirm_release(void *calldata)
1479{
1480        struct nfs4_opendata *data = calldata;
1481        struct nfs4_state *state = NULL;
1482
1483        /* If this request hasn't been cancelled, do nothing */
1484        if (data->cancelled == 0)
1485                goto out_free;
1486        /* In case of error, no cleanup! */
1487        if (!data->rpc_done)
1488                goto out_free;
1489        state = nfs4_opendata_to_nfs4_state(data);
1490        if (!IS_ERR(state))
1491                nfs4_close_state(state, data->o_arg.fmode);
1492out_free:
1493        nfs4_opendata_put(data);
1494}
1495
1496static const struct rpc_call_ops nfs4_open_confirm_ops = {
1497        .rpc_call_done = nfs4_open_confirm_done,
1498        .rpc_release = nfs4_open_confirm_release,
1499};
1500
1501/*
1502 * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1503 */
1504static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1505{
1506        struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1507        struct rpc_task *task;
1508        struct  rpc_message msg = {
1509                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1510                .rpc_argp = &data->c_arg,
1511                .rpc_resp = &data->c_res,
1512                .rpc_cred = data->owner->so_cred,
1513        };
1514        struct rpc_task_setup task_setup_data = {
1515                .rpc_client = server->client,
1516                .rpc_message = &msg,
1517                .callback_ops = &nfs4_open_confirm_ops,
1518                .callback_data = data,
1519                .workqueue = nfsiod_workqueue,
1520                .flags = RPC_TASK_ASYNC,
1521        };
1522        int status;
1523
1524        kref_get(&data->kref);
1525        data->rpc_done = 0;
1526        data->rpc_status = 0;
1527        data->timestamp = jiffies;
1528        task = rpc_run_task(&task_setup_data);
1529        if (IS_ERR(task))
1530                return PTR_ERR(task);
1531        status = nfs4_wait_for_completion_rpc_task(task);
1532        if (status != 0) {
1533                data->cancelled = 1;
1534                smp_wmb();
1535        } else
1536                status = data->rpc_status;
1537        rpc_put_task(task);
1538        return status;
1539}
1540
1541static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1542{
1543        struct nfs4_opendata *data = calldata;
1544        struct nfs4_state_owner *sp = data->owner;
1545
1546        if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1547                return;
1548        /*
1549         * Check if we still need to send an OPEN call, or if we can use
1550         * a delegation instead.
1551         */
1552        if (data->state != NULL) {
1553                struct nfs_delegation *delegation;
1554
1555                if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1556                        goto out_no_action;
1557                rcu_read_lock();
1558                delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1559                if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1560                    can_open_delegated(delegation, data->o_arg.fmode))
1561                        goto unlock_no_action;
1562                rcu_read_unlock();
1563        }
1564        /* Update client id. */
1565        data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1566        if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1567                task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1568                data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1569                nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1570        }
1571        data->timestamp = jiffies;
1572        if (nfs4_setup_sequence(data->o_arg.server,
1573                                &data->o_arg.seq_args,
1574                                &data->o_res.seq_res,
1575                                task) != 0)
1576                nfs_release_seqid(data->o_arg.seqid);
1577        else
1578                rpc_call_start(task);
1579        return;
1580unlock_no_action:
1581        rcu_read_unlock();
1582out_no_action:
1583        task->tk_action = NULL;
1584
1585}
1586
1587static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1588{
1589        rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1590        nfs4_open_prepare(task, calldata);
1591}
1592
1593static void nfs4_open_done(struct rpc_task *task, void *calldata)
1594{
1595        struct nfs4_opendata *data = calldata;
1596
1597        data->rpc_status = task->tk_status;
1598
1599        if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1600                return;
1601
1602        if (task->tk_status == 0) {
1603                if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1604                        switch (data->o_res.f_attr->mode & S_IFMT) {
1605                        case S_IFREG:
1606                                break;
1607                        case S_IFLNK:
1608                                data->rpc_status = -ELOOP;
1609                                break;
1610                        case S_IFDIR:
1611                                data->rpc_status = -EISDIR;
1612                                break;
1613                        default:
1614                                data->rpc_status = -ENOTDIR;
1615                        }
1616                }
1617                renew_lease(data->o_res.server, data->timestamp);
1618                if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1619                        nfs_confirm_seqid(&data->owner->so_seqid, 0);
1620        }
1621        data->rpc_done = 1;
1622}
1623
1624static void nfs4_open_release(void *calldata)
1625{
1626        struct nfs4_opendata *data = calldata;
1627        struct nfs4_state *state = NULL;
1628
1629        /* If this request hasn't been cancelled, do nothing */
1630        if (data->cancelled == 0)
1631                goto out_free;
1632        /* In case of error, no cleanup! */
1633        if (data->rpc_status != 0 || !data->rpc_done)
1634                goto out_free;
1635        /* In case we need an open_confirm, no cleanup! */
1636        if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1637                goto out_free;
1638        state = nfs4_opendata_to_nfs4_state(data);
1639        if (!IS_ERR(state))
1640                nfs4_close_state(state, data->o_arg.fmode);
1641out_free:
1642        nfs4_opendata_put(data);
1643}
1644
1645static const struct rpc_call_ops nfs4_open_ops = {
1646        .rpc_call_prepare = nfs4_open_prepare,
1647        .rpc_call_done = nfs4_open_done,
1648        .rpc_release = nfs4_open_release,
1649};
1650
1651static const struct rpc_call_ops nfs4_recover_open_ops = {
1652        .rpc_call_prepare = nfs4_recover_open_prepare,
1653        .rpc_call_done = nfs4_open_done,
1654        .rpc_release = nfs4_open_release,
1655};
1656
1657static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1658{
1659        struct inode *dir = data->dir->d_inode;
1660        struct nfs_server *server = NFS_SERVER(dir);
1661        struct nfs_openargs *o_arg = &data->o_arg;
1662        struct nfs_openres *o_res = &data->o_res;
1663        struct rpc_task *task;
1664        struct rpc_message msg = {
1665                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1666                .rpc_argp = o_arg,
1667                .rpc_resp = o_res,
1668                .rpc_cred = data->owner->so_cred,
1669        };
1670        struct rpc_task_setup task_setup_data = {
1671                .rpc_client = server->client,
1672                .rpc_message = &msg,
1673                .callback_ops = &nfs4_open_ops,
1674                .callback_data = data,
1675                .workqueue = nfsiod_workqueue,
1676                .flags = RPC_TASK_ASYNC,
1677        };
1678        int status;
1679
1680        nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1681        kref_get(&data->kref);
1682        data->rpc_done = 0;
1683        data->rpc_status = 0;
1684        data->cancelled = 0;
1685        if (isrecover)
1686                task_setup_data.callback_ops = &nfs4_recover_open_ops;
1687        task = rpc_run_task(&task_setup_data);
1688        if (IS_ERR(task))
1689                return PTR_ERR(task);
1690        status = nfs4_wait_for_completion_rpc_task(task);
1691        if (status != 0) {
1692                data->cancelled = 1;
1693                smp_wmb();
1694        } else
1695                status = data->rpc_status;
1696        rpc_put_task(task);
1697
1698        return status;
1699}
1700
1701static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1702{
1703        struct inode *dir = data->dir->d_inode;
1704        struct nfs_openres *o_res = &data->o_res;
1705        int status;
1706
1707        status = nfs4_run_open_task(data, 1);
1708        if (status != 0 || !data->rpc_done)
1709                return status;
1710
1711        nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1712
1713        if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1714                status = _nfs4_proc_open_confirm(data);
1715                if (status != 0)
1716                        return status;
1717        }
1718
1719        return status;
1720}
1721
1722static int nfs4_opendata_access(struct rpc_cred *cred,
1723                                struct nfs4_opendata *opendata,
1724                                struct nfs4_state *state, fmode_t fmode)
1725{
1726        struct nfs_access_entry cache;
1727        u32 mask;
1728
1729        /* access call failed or for some reason the server doesn't
1730         * support any access modes -- defer access call until later */
1731        if (opendata->o_res.access_supported == 0)
1732                return 0;
1733
1734        mask = 0;
1735        /* don't check MAY_WRITE - a newly created file may not have
1736         * write mode bits, but POSIX allows the creating process to write */
1737        if (fmode & FMODE_READ)
1738                mask |= MAY_READ;
1739        if (fmode & FMODE_EXEC)
1740                mask |= MAY_EXEC;
1741
1742        cache.cred = cred;
1743        cache.jiffies = jiffies;
1744        nfs_access_set_mask(&cache, opendata->o_res.access_result);
1745        nfs_access_add_cache(state->inode, &cache);
1746
1747        if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1748                return 0;
1749
1750        /* even though OPEN succeeded, access is denied. Close the file */
1751        nfs4_close_state(state, fmode);
1752        return -EACCES;
1753}
1754
1755/*
1756 * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1757 */
1758static int _nfs4_proc_open(struct nfs4_opendata *data)
1759{
1760        struct inode *dir = data->dir->d_inode;
1761        struct nfs_server *server = NFS_SERVER(dir);
1762        struct nfs_openargs *o_arg = &data->o_arg;
1763        struct nfs_openres *o_res = &data->o_res;
1764        int status;
1765
1766        status = nfs4_run_open_task(data, 0);
1767        if (!data->rpc_done)
1768                return status;
1769        if (status != 0) {
1770                if (status == -NFS4ERR_BADNAME &&
1771                                !(o_arg->open_flags & O_CREAT))
1772                        return -ENOENT;
1773                return status;
1774        }
1775
1776        nfs_fattr_map_and_free_names(server, &data->f_attr);
1777
1778        if (o_arg->open_flags & O_CREAT)
1779                update_changeattr(dir, &o_res->cinfo);
1780        if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1781                server->caps &= ~NFS_CAP_POSIX_LOCK;
1782        if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1783                status = _nfs4_proc_open_confirm(data);
1784                if (status != 0)
1785                        return status;
1786        }
1787        if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1788                _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1789        return 0;
1790}
1791
1792static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1793{
1794        unsigned int loop;
1795        int ret;
1796
1797        for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1798                ret = nfs4_wait_clnt_recover(clp);
1799                if (ret != 0)
1800                        break;
1801                if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1802                    !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1803                        break;
1804                nfs4_schedule_state_manager(clp);
1805                ret = -EIO;
1806        }
1807        return ret;
1808}
1809
1810static int nfs4_recover_expired_lease(struct nfs_server *server)
1811{
1812        return nfs4_client_recover_expired_lease(server->nfs_client);
1813}
1814
1815/*
1816 * OPEN_EXPIRED:
1817 *      reclaim state on the server after a network partition.
1818 *      Assumes caller holds the appropriate lock
1819 */
1820static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1821{
1822        struct nfs4_opendata *opendata;
1823        int ret;
1824
1825        opendata = nfs4_open_recoverdata_alloc(ctx, state);
1826        if (IS_ERR(opendata))
1827                return PTR_ERR(opendata);
1828        ret = nfs4_open_recover(opendata, state);
1829        if (ret == -ESTALE)
1830                d_drop(ctx->dentry);
1831        nfs4_opendata_put(opendata);
1832        return ret;
1833}
1834
1835static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1836{
1837        struct nfs_server *server = NFS_SERVER(state->inode);
1838        struct nfs4_exception exception = { };
1839        int err;
1840
1841        do {
1842                err = _nfs4_open_expired(ctx, state);
1843                switch (err) {
1844                default:
1845                        goto out;
1846                case -NFS4ERR_GRACE:
1847                case -NFS4ERR_DELAY:
1848                        nfs4_handle_exception(server, err, &exception);
1849                        err = 0;
1850                }
1851        } while (exception.retry);
1852out:
1853        return err;
1854}
1855
1856static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1857{
1858        struct nfs_open_context *ctx;
1859        int ret;
1860
1861        ctx = nfs4_state_find_open_context(state);
1862        if (IS_ERR(ctx))
1863                return PTR_ERR(ctx);
1864        ret = nfs4_do_open_expired(ctx, state);
1865        put_nfs_open_context(ctx);
1866        return ret;
1867}
1868
1869#if defined(CONFIG_NFS_V4_1)
1870static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1871{
1872        struct nfs_server *server = NFS_SERVER(state->inode);
1873        nfs4_stateid *stateid = &state->stateid;
1874        int status;
1875
1876        /* If a state reset has been done, test_stateid is unneeded */
1877        if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1878                return;
1879
1880        status = nfs41_test_stateid(server, stateid);
1881        if (status != NFS_OK) {
1882                /* Free the stateid unless the server explicitly
1883                 * informs us the stateid is unrecognized. */
1884                if (status != -NFS4ERR_BAD_STATEID)
1885                        nfs41_free_stateid(server, stateid);
1886                nfs_remove_bad_delegation(state->inode);
1887
1888                write_seqlock(&state->seqlock);
1889                nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1890                write_sequnlock(&state->seqlock);
1891                clear_bit(NFS_DELEGATED_STATE, &state->flags);
1892        }
1893}
1894
1895/**
1896 * nfs41_check_open_stateid - possibly free an open stateid
1897 *
1898 * @state: NFSv4 state for an inode
1899 *
1900 * Returns NFS_OK if recovery for this stateid is now finished.
1901 * Otherwise a negative NFS4ERR value is returned.
1902 */
1903static int nfs41_check_open_stateid(struct nfs4_state *state)
1904{
1905        struct nfs_server *server = NFS_SERVER(state->inode);
1906        nfs4_stateid *stateid = &state->open_stateid;
1907        int status;
1908
1909        /* If a state reset has been done, test_stateid is unneeded */
1910        if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1911            (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1912            (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1913                return -NFS4ERR_BAD_STATEID;
1914
1915        status = nfs41_test_stateid(server, stateid);
1916        if (status != NFS_OK) {
1917                /* Free the stateid unless the server explicitly
1918                 * informs us the stateid is unrecognized. */
1919                if (status != -NFS4ERR_BAD_STATEID)
1920                        nfs41_free_stateid(server, stateid);
1921
1922                clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1923                clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1924                clear_bit(NFS_O_RDWR_STATE, &state->flags);
1925        }
1926        return status;
1927}
1928
1929static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1930{
1931        int status;
1932
1933        nfs41_clear_delegation_stateid(state);
1934        status = nfs41_check_open_stateid(state);
1935        if (status != NFS_OK)
1936                status = nfs4_open_expired(sp, state);
1937        return status;
1938}
1939#endif
1940
1941/*
1942 * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1943 * fields corresponding to attributes that were used to store the verifier.
1944 * Make sure we clobber those fields in the later setattr call
1945 */
1946static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1947{
1948        if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1949            !(sattr->ia_valid & ATTR_ATIME_SET))
1950                sattr->ia_valid |= ATTR_ATIME;
1951
1952        if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1953            !(sattr->ia_valid & ATTR_MTIME_SET))
1954                sattr->ia_valid |= ATTR_MTIME;
1955}
1956
1957/*
1958 * Returns a referenced nfs4_state
1959 */
1960static int _nfs4_do_open(struct inode *dir,
1961                        struct dentry *dentry,
1962                        fmode_t fmode,
1963                        int flags,
1964                        struct iattr *sattr,
1965                        struct rpc_cred *cred,
1966                        struct nfs4_state **res,
1967                        struct nfs4_threshold **ctx_th)
1968{
1969        struct nfs4_state_owner  *sp;
1970        struct nfs4_state     *state = NULL;
1971        struct nfs_server       *server = NFS_SERVER(dir);
1972        struct nfs4_opendata *opendata;
1973        int status;
1974
1975        /* Protect against reboot recovery conflicts */
1976        status = -ENOMEM;
1977        sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1978        if (sp == NULL) {
1979                dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1980                goto out_err;
1981        }
1982        status = nfs4_recover_expired_lease(server);
1983        if (status != 0)
1984                goto err_put_state_owner;
1985        if (dentry->d_inode != NULL)
1986                nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1987        status = -ENOMEM;
1988        opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1989        if (opendata == NULL)
1990                goto err_put_state_owner;
1991
1992        if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1993                opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1994                if (!opendata->f_attr.mdsthreshold)
1995                        goto err_opendata_put;
1996                opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
1997        }
1998        if (dentry->d_inode != NULL)
1999                opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2000
2001        status = _nfs4_proc_open(opendata);
2002        if (status != 0)
2003                goto err_opendata_put;
2004
2005        state = nfs4_opendata_to_nfs4_state(opendata);
2006        status = PTR_ERR(state);
2007        if (IS_ERR(state))
2008                goto err_opendata_put;
2009        if (server->caps & NFS_CAP_POSIX_LOCK)
2010                set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2011
2012        status = nfs4_opendata_access(cred, opendata, state, fmode);
2013        if (status != 0)
2014                goto err_opendata_put;
2015
2016        if (opendata->o_arg.open_flags & O_EXCL) {
2017                nfs4_exclusive_attrset(opendata, sattr);
2018
2019                nfs_fattr_init(opendata->o_res.f_attr);
2020                status = nfs4_do_setattr(state->inode, cred,
2021                                opendata->o_res.f_attr, sattr,
2022                                state);
2023                if (status == 0)
2024                        nfs_setattr_update_inode(state->inode, sattr);
2025                nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2026        }
2027
2028        if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2029                *ctx_th = opendata->f_attr.mdsthreshold;
2030        else
2031                kfree(opendata->f_attr.mdsthreshold);
2032        opendata->f_attr.mdsthreshold = NULL;
2033
2034        nfs4_opendata_put(opendata);
2035        nfs4_put_state_owner(sp);
2036        *res = state;
2037        return 0;
2038err_opendata_put:
2039        kfree(opendata->f_attr.mdsthreshold);
2040        nfs4_opendata_put(opendata);
2041err_put_state_owner:
2042        nfs4_put_state_owner(sp);
2043out_err:
2044        *res = NULL;
2045        return status;
2046}
2047
2048
2049static struct nfs4_state *nfs4_do_open(struct inode *dir,
2050                                        struct dentry *dentry,
2051                                        fmode_t fmode,
2052                                        int flags,
2053                                        struct iattr *sattr,
2054                                        struct rpc_cred *cred,
2055                                        struct nfs4_threshold **ctx_th)
2056{
2057        struct nfs4_exception exception = { };
2058        struct nfs4_state *res;
2059        int status;
2060
2061        fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
2062        do {
2063                status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
2064                                       &res, ctx_th);
2065                if (status == 0)
2066                        break;
2067                /* NOTE: BAD_SEQID means the server and client disagree about the
2068                 * book-keeping w.r.t. state-changing operations
2069                 * (OPEN/CLOSE/LOCK/LOCKU...)
2070                 * It is actually a sign of a bug on the client or on the server.
2071                 *
2072                 * If we receive a BAD_SEQID error in the particular case of
2073                 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2074                 * have unhashed the old state_owner for us, and that we can
2075                 * therefore safely retry using a new one. We should still warn
2076                 * the user though...
2077                 */
2078                if (status == -NFS4ERR_BAD_SEQID) {
2079                        pr_warn_ratelimited("NFS: v4 server %s "
2080                                        " returned a bad sequence-id error!\n",
2081                                        NFS_SERVER(dir)->nfs_client->cl_hostname);
2082                        exception.retry = 1;
2083                        continue;
2084                }
2085                /*
2086                 * BAD_STATEID on OPEN means that the server cancelled our
2087                 * state before it received the OPEN_CONFIRM.
2088                 * Recover by retrying the request as per the discussion
2089                 * on Page 181 of RFC3530.
2090                 */
2091                if (status == -NFS4ERR_BAD_STATEID) {
2092                        exception.retry = 1;
2093                        continue;
2094                }
2095                if (status == -EAGAIN) {
2096                        /* We must have found a delegation */
2097                        exception.retry = 1;
2098                        continue;
2099                }
2100                res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
2101                                        status, &exception));
2102        } while (exception.retry);
2103        return res;
2104}
2105
2106static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2107                            struct nfs_fattr *fattr, struct iattr *sattr,
2108                            struct nfs4_state *state)
2109{
2110        struct nfs_server *server = NFS_SERVER(inode);
2111        struct nfs_setattrargs  arg = {
2112                .fh             = NFS_FH(inode),
2113                .iap            = sattr,
2114                .server         = server,
2115                .bitmask = server->attr_bitmask,
2116        };
2117        struct nfs_setattrres  res = {
2118                .fattr          = fattr,
2119                .server         = server,
2120        };
2121        struct rpc_message msg = {
2122                .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2123                .rpc_argp       = &arg,
2124                .rpc_resp       = &res,
2125                .rpc_cred       = cred,
2126        };
2127        unsigned long timestamp = jiffies;
2128        int status;
2129
2130        nfs_fattr_init(fattr);
2131
2132        if (state != NULL) {
2133                struct nfs_lockowner lockowner = {
2134                        .l_owner = current->files,
2135                        .l_pid = current->tgid,
2136                };
2137                nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2138                                &lockowner);
2139        } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
2140                                FMODE_WRITE)) {
2141                /* Use that stateid */
2142        } else
2143                nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2144
2145        status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2146        if (status == 0 && state != NULL)
2147                renew_lease(server, timestamp);
2148        return status;
2149}
2150
2151static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2152                           struct nfs_fattr *fattr, struct iattr *sattr,
2153                           struct nfs4_state *state)
2154{
2155        struct nfs_server *server = NFS_SERVER(inode);
2156        struct nfs4_exception exception = {
2157                .state = state,
2158                .inode = inode,
2159        };
2160        int err;
2161        do {
2162                err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2163                switch (err) {
2164                case -NFS4ERR_OPENMODE:
2165                        if (state && !(state->state & FMODE_WRITE)) {
2166                                err = -EBADF;
2167                                if (sattr->ia_valid & ATTR_OPEN)
2168                                        err = -EACCES;
2169                                goto out;
2170                        }
2171                }
2172                err = nfs4_handle_exception(server, err, &exception);
2173        } while (exception.retry);
2174out:
2175        return err;
2176}
2177
2178struct nfs4_closedata {
2179        struct inode *inode;
2180        struct nfs4_state *state;
2181        struct nfs_closeargs arg;
2182        struct nfs_closeres res;
2183        struct nfs_fattr fattr;
2184        unsigned long timestamp;
2185        bool roc;
2186        u32 roc_barrier;
2187};
2188
2189static void nfs4_free_closedata(void *data)
2190{
2191        struct nfs4_closedata *calldata = data;
2192        struct nfs4_state_owner *sp = calldata->state->owner;
2193        struct super_block *sb = calldata->state->inode->i_sb;
2194
2195        if (calldata->roc)
2196                pnfs_roc_release(calldata->state->inode);
2197        nfs4_put_open_state(calldata->state);
2198        nfs_free_seqid(calldata->arg.seqid);
2199        nfs4_put_state_owner(sp);
2200        nfs_sb_deactive_async(sb);
2201        kfree(calldata);
2202}
2203
2204static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2205                fmode_t fmode)
2206{
2207        spin_lock(&state->owner->so_lock);
2208        if (!(fmode & FMODE_READ))
2209                clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2210        if (!(fmode & FMODE_WRITE))
2211                clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2212        clear_bit(NFS_O_RDWR_STATE, &state->flags);
2213        spin_unlock(&state->owner->so_lock);
2214}
2215
2216static void nfs4_close_done(struct rpc_task *task, void *data)
2217{
2218        struct nfs4_closedata *calldata = data;
2219        struct nfs4_state *state = calldata->state;
2220        struct nfs_server *server = NFS_SERVER(calldata->inode);
2221
2222        dprintk("%s: begin!\n", __func__);
2223        if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2224                return;
2225        /* hmm. we are done with the inode, and in the process of freeing
2226         * the state_owner. we keep this around to process errors
2227         */
2228        switch (task->tk_status) {
2229                case 0:
2230                        if (calldata->roc)
2231                                pnfs_roc_set_barrier(state->inode,
2232                                                     calldata->roc_barrier);
2233                        nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2234                        renew_lease(server, calldata->timestamp);
2235                        nfs4_close_clear_stateid_flags(state,
2236                                        calldata->arg.fmode);
2237                        break;
2238                case -NFS4ERR_STALE_STATEID:
2239                case -NFS4ERR_OLD_STATEID:
2240                case -NFS4ERR_BAD_STATEID:
2241                case -NFS4ERR_EXPIRED:
2242                        if (calldata->arg.fmode == 0)
2243                                break;
2244                default:
2245                        if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2246                                rpc_restart_call_prepare(task);
2247        }
2248        nfs_release_seqid(calldata->arg.seqid);
2249        nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2250        dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2251}
2252
2253static void nfs4_close_prepare(struct rpc_task *task, void *data)
2254{
2255        struct nfs4_closedata *calldata = data;
2256        struct nfs4_state *state = calldata->state;
2257        struct inode *inode = calldata->inode;
2258        int call_close = 0;
2259
2260        dprintk("%s: begin!\n", __func__);
2261        if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2262                return;
2263
2264        task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2265        calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2266        spin_lock(&state->owner->so_lock);
2267        /* Calculate the change in open mode */
2268        if (state->n_rdwr == 0) {
2269                if (state->n_rdonly == 0) {
2270                        call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2271                        call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2272                        calldata->arg.fmode &= ~FMODE_READ;
2273                }
2274                if (state->n_wronly == 0) {
2275                        call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2276                        call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2277                        calldata->arg.fmode &= ~FMODE_WRITE;
2278                }
2279        }
2280        spin_unlock(&state->owner->so_lock);
2281
2282        if (!call_close) {
2283                /* Note: exit _without_ calling nfs4_close_done */
2284                task->tk_action = NULL;
2285                goto out;
2286        }
2287
2288        if (calldata->arg.fmode == 0) {
2289                task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2290                if (calldata->roc &&
2291                    pnfs_roc_drain(inode, &calldata->roc_barrier, task))
2292                        goto out;
2293        }
2294
2295        nfs_fattr_init(calldata->res.fattr);
2296        calldata->timestamp = jiffies;
2297        if (nfs4_setup_sequence(NFS_SERVER(inode),
2298                                &calldata->arg.seq_args,
2299                                &calldata->res.seq_res,
2300                                task) != 0)
2301                nfs_release_seqid(calldata->arg.seqid);
2302        else
2303                rpc_call_start(task);
2304out:
2305        dprintk("%s: done!\n", __func__);
2306}
2307
2308static const struct rpc_call_ops nfs4_close_ops = {
2309        .rpc_call_prepare = nfs4_close_prepare,
2310        .rpc_call_done = nfs4_close_done,
2311        .rpc_release = nfs4_free_closedata,
2312};
2313
2314/* 
2315 * It is possible for data to be read/written from a mem-mapped file 
2316 * after the sys_close call (which hits the vfs layer as a flush).
2317 * This means that we can't safely call nfsv4 close on a file until 
2318 * the inode is cleared. This in turn means that we are not good
2319 * NFSv4 citizens - we do not indicate to the server to update the file's 
2320 * share state even when we are done with one of the three share 
2321 * stateid's in the inode.
2322 *
2323 * NOTE: Caller must be holding the sp->so_owner semaphore!
2324 */
2325int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2326{
2327        struct nfs_server *server = NFS_SERVER(state->inode);
2328        struct nfs4_closedata *calldata;
2329        struct nfs4_state_owner *sp = state->owner;
2330        struct rpc_task *task;
2331        struct rpc_message msg = {
2332                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2333                .rpc_cred = state->owner->so_cred,
2334        };
2335        struct rpc_task_setup task_setup_data = {
2336                .rpc_client = server->client,
2337                .rpc_message = &msg,
2338                .callback_ops = &nfs4_close_ops,
2339                .workqueue = nfsiod_workqueue,
2340                .flags = RPC_TASK_ASYNC,
2341        };
2342        int status = -ENOMEM;
2343
2344        calldata = kzalloc(sizeof(*calldata), gfp_mask);
2345        if (calldata == NULL)
2346                goto out;
2347        nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2348        calldata->inode = state->inode;
2349        calldata->state = state;
2350        calldata->arg.fh = NFS_FH(state->inode);
2351        calldata->arg.stateid = &state->open_stateid;
2352        /* Serialization for the sequence id */
2353        calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2354        if (calldata->arg.seqid == NULL)
2355                goto out_free_calldata;
2356        calldata->arg.fmode = 0;
2357        calldata->arg.bitmask = server->cache_consistency_bitmask;
2358        calldata->res.fattr = &calldata->fattr;
2359        calldata->res.seqid = calldata->arg.seqid;
2360        calldata->res.server = server;
2361        calldata->roc = pnfs_roc(state->inode);
2362        nfs_sb_active(calldata->inode->i_sb);
2363
2364        msg.rpc_argp = &calldata->arg;
2365        msg.rpc_resp = &calldata->res;
2366        task_setup_data.callback_data = calldata;
2367        task = rpc_run_task(&task_setup_data);
2368        if (IS_ERR(task))
2369                return PTR_ERR(task);
2370        status = 0;
2371        if (wait)
2372                status = rpc_wait_for_completion_task(task);
2373        rpc_put_task(task);
2374        return status;
2375out_free_calldata:
2376        kfree(calldata);
2377out:
2378        nfs4_put_open_state(state);
2379        nfs4_put_state_owner(sp);
2380        return status;
2381}
2382
2383static struct inode *
2384nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2385{
2386        struct nfs4_state *state;
2387
2388        /* Protect against concurrent sillydeletes */
2389        state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2390                             ctx->cred, &ctx->mdsthreshold);
2391        if (IS_ERR(state))
2392                return ERR_CAST(state);
2393        ctx->state = state;
2394        return igrab(state->inode);
2395}
2396
2397static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2398{
2399        if (ctx->state == NULL)
2400                return;
2401        if (is_sync)
2402                nfs4_close_sync(ctx->state, ctx->mode);
2403        else
2404                nfs4_close_state(ctx->state, ctx->mode);
2405}
2406
2407static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2408{
2409        struct nfs4_server_caps_arg args = {
2410                .fhandle = fhandle,
2411        };
2412        struct nfs4_server_caps_res res = {};
2413        struct rpc_message msg = {
2414                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2415                .rpc_argp = &args,
2416                .rpc_resp = &res,
2417        };
2418        int status;
2419
2420        status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2421        if (status == 0) {
2422                memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2423                server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2424                                NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2425                                NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2426                                NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2427                                NFS_CAP_CTIME|NFS_CAP_MTIME);
2428                if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2429                        server->caps |= NFS_CAP_ACLS;
2430                if (res.has_links != 0)
2431                        server->caps |= NFS_CAP_HARDLINKS;
2432                if (res.has_symlinks != 0)
2433                        server->caps |= NFS_CAP_SYMLINKS;
2434                if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2435                        server->caps |= NFS_CAP_FILEID;
2436                if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2437                        server->caps |= NFS_CAP_MODE;
2438                if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2439                        server->caps |= NFS_CAP_NLINK;
2440                if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2441                        server->caps |= NFS_CAP_OWNER;
2442                if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2443                        server->caps |= NFS_CAP_OWNER_GROUP;
2444                if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2445                        server->caps |= NFS_CAP_ATIME;
2446                if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2447                        server->caps |= NFS_CAP_CTIME;
2448                if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2449                        server->caps |= NFS_CAP_MTIME;
2450
2451                memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2452                server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2453                server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2454                server->acl_bitmask = res.acl_bitmask;
2455                server->fh_expire_type = res.fh_expire_type;
2456        }
2457
2458        return status;
2459}
2460
2461int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2462{
2463        struct nfs4_exception exception = { };
2464        int err;
2465        do {
2466                err = nfs4_handle_exception(server,
2467                                _nfs4_server_capabilities(server, fhandle),
2468                                &exception);
2469        } while (exception.retry);
2470        return err;
2471}
2472
2473static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2474                struct nfs_fsinfo *info)
2475{
2476        struct nfs4_lookup_root_arg args = {
2477                .bitmask = nfs4_fattr_bitmap,
2478        };
2479        struct nfs4_lookup_res res = {
2480                .server = server,
2481                .fattr = info->fattr,
2482                .fh = fhandle,
2483        };
2484        struct rpc_message msg = {
2485                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2486                .rpc_argp = &args,
2487                .rpc_resp = &res,
2488        };
2489
2490        nfs_fattr_init(info->fattr);
2491        return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2492}
2493
2494static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2495                struct nfs_fsinfo *info)
2496{
2497        struct nfs4_exception exception = { };
2498        int err;
2499        do {
2500                err = _nfs4_lookup_root(server, fhandle, info);
2501                switch (err) {
2502                case 0:
2503                case -NFS4ERR_WRONGSEC:
2504                        goto out;
2505                default:
2506                        err = nfs4_handle_exception(server, err, &exception);
2507                }
2508        } while (exception.retry);
2509out:
2510        return err;
2511}
2512
2513static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2514                                struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2515{
2516        struct rpc_auth *auth;
2517        int ret;
2518
2519        auth = rpcauth_create(flavor, server->client);
2520        if (IS_ERR(auth)) {
2521                ret = -EIO;
2522                goto out;
2523        }
2524        ret = nfs4_lookup_root(server, fhandle, info);
2525out:
2526        return ret;
2527}
2528
2529static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2530                              struct nfs_fsinfo *info)
2531{
2532        int i, len, status = 0;
2533        rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2534
2535        len = rpcauth_list_flavors(flav_array, ARRAY_SIZE(flav_array));
2536        BUG_ON(len < 0);
2537
2538        for (i = 0; i < len; i++) {
2539                /* AUTH_UNIX is the default flavor if none was specified,
2540                 * thus has already been tried. */
2541                if (flav_array[i] == RPC_AUTH_UNIX)
2542                        continue;
2543
2544                status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2545                if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2546                        continue;
2547                break;
2548        }
2549        /*
2550         * -EACCESS could mean that the user doesn't have correct permissions
2551         * to access the mount.  It could also mean that we tried to mount
2552         * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2553         * existing mount programs don't handle -EACCES very well so it should
2554         * be mapped to -EPERM instead.
2555         */
2556        if (status == -EACCES)
2557                status = -EPERM;
2558        return status;
2559}
2560
2561/*
2562 * get the file handle for the "/" directory on the server
2563 */
2564int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2565                         struct nfs_fsinfo *info)
2566{
2567        int minor_version = server->nfs_client->cl_minorversion;
2568        int status = nfs4_lookup_root(server, fhandle, info);
2569        if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2570                /*
2571                 * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2572                 * by nfs4_map_errors() as this function exits.
2573                 */
2574                status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2575        if (status == 0)
2576                status = nfs4_server_capabilities(server, fhandle);
2577        if (status == 0)
2578                status = nfs4_do_fsinfo(server, fhandle, info);
2579        return nfs4_map_errors(status);
2580}
2581
2582static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2583                              struct nfs_fsinfo *info)
2584{
2585        int error;
2586        struct nfs_fattr *fattr = info->fattr;
2587
2588        error = nfs4_server_capabilities(server, mntfh);
2589        if (error < 0) {
2590                dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2591                return error;
2592        }
2593
2594        error = nfs4_proc_getattr(server, mntfh, fattr);
2595        if (error < 0) {
2596                dprintk("nfs4_get_root: getattr error = %d\n", -error);
2597                return error;
2598        }
2599
2600        if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2601            !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2602                memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2603
2604        return error;
2605}
2606
2607/*
2608 * Get locations and (maybe) other attributes of a referral.
2609 * Note that we'll actually follow the referral later when
2610 * we detect fsid mismatch in inode revalidation
2611 */
2612static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2613                             const struct qstr *name, struct nfs_fattr *fattr,
2614                             struct nfs_fh *fhandle)
2615{
2616        int status = -ENOMEM;
2617        struct page *page = NULL;
2618        struct nfs4_fs_locations *locations = NULL;
2619
2620        page = alloc_page(GFP_KERNEL);
2621        if (page == NULL)
2622                goto out;
2623        locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2624        if (locations == NULL)
2625                goto out;
2626
2627        status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2628        if (status != 0)
2629                goto out;
2630        /* Make sure server returned a different fsid for the referral */
2631        if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2632                dprintk("%s: server did not return a different fsid for"
2633                        " a referral at %s\n", __func__, name->name);
2634                status = -EIO;
2635                goto out;
2636        }
2637        /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2638        nfs_fixup_referral_attributes(&locations->fattr);
2639
2640        /* replace the lookup nfs_fattr with the locations nfs_fattr */
2641        memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2642        memset(fhandle, 0, sizeof(struct nfs_fh));
2643out:
2644        if (page)
2645                __free_page(page);
2646        kfree(locations);
2647        return status;
2648}
2649
2650static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2651{
2652        struct nfs4_getattr_arg args = {
2653                .fh = fhandle,
2654                .bitmask = server->attr_bitmask,
2655        };
2656        struct nfs4_getattr_res res = {
2657                .fattr = fattr,
2658                .server = server,
2659        };
2660        struct rpc_message msg = {
2661                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2662                .rpc_argp = &args,
2663                .rpc_resp = &res,
2664        };
2665        
2666        nfs_fattr_init(fattr);
2667        return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2668}
2669
2670static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2671{
2672        struct nfs4_exception exception = { };
2673        int err;
2674        do {
2675                err = nfs4_handle_exception(server,
2676                                _nfs4_proc_getattr(server, fhandle, fattr),
2677                                &exception);
2678        } while (exception.retry);
2679        return err;
2680}
2681
2682/* 
2683 * The file is not closed if it is opened due to the a request to change
2684 * the size of the file. The open call will not be needed once the
2685 * VFS layer lookup-intents are implemented.
2686 *
2687 * Close is called when the inode is destroyed.
2688 * If we haven't opened the file for O_WRONLY, we
2689 * need to in the size_change case to obtain a stateid.
2690 *
2691 * Got race?
2692 * Because OPEN is always done by name in nfsv4, it is
2693 * possible that we opened a different file by the same
2694 * name.  We can recognize this race condition, but we
2695 * can't do anything about it besides returning an error.
2696 *
2697 * This will be fixed with VFS changes (lookup-intent).
2698 */
2699static int
2700nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2701                  struct iattr *sattr)
2702{
2703        struct inode *inode = dentry->d_inode;
2704        struct rpc_cred *cred = NULL;
2705        struct nfs4_state *state = NULL;
2706        int status;
2707
2708        if (pnfs_ld_layoutret_on_setattr(inode))
2709                pnfs_return_layout(inode);
2710
2711        nfs_fattr_init(fattr);
2712        
2713        /* Deal with open(O_TRUNC) */
2714        if (sattr->ia_valid & ATTR_OPEN)
2715                sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2716
2717        /* Optimization: if the end result is no change, don't RPC */
2718        if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2719                return 0;
2720
2721        /* Search for an existing open(O_WRITE) file */
2722        if (sattr->ia_valid & ATTR_FILE) {
2723                struct nfs_open_context *ctx;
2724
2725                ctx = nfs_file_open_context(sattr->ia_file);
2726                if (ctx) {
2727                        cred = ctx->cred;
2728                        state = ctx->state;
2729                }
2730        }
2731
2732        status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2733        if (status == 0)
2734                nfs_setattr_update_inode(inode, sattr);
2735        return status;
2736}
2737
2738static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2739                const struct qstr *name, struct nfs_fh *fhandle,
2740                struct nfs_fattr *fattr)
2741{
2742        struct nfs_server *server = NFS_SERVER(dir);
2743        int                    status;
2744        struct nfs4_lookup_arg args = {
2745                .bitmask = server->attr_bitmask,
2746                .dir_fh = NFS_FH(dir),
2747                .name = name,
2748        };
2749        struct nfs4_lookup_res res = {
2750                .server = server,
2751                .fattr = fattr,
2752                .fh = fhandle,
2753        };
2754        struct rpc_message msg = {
2755                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2756                .rpc_argp = &args,
2757                .rpc_resp = &res,
2758        };
2759
2760        nfs_fattr_init(fattr);
2761
2762        dprintk("NFS call  lookup %s\n", name->name);
2763        status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2764        dprintk("NFS reply lookup: %d\n", status);
2765        return status;
2766}
2767
2768static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2769{
2770        fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2771                NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2772        fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2773        fattr->nlink = 2;
2774}
2775
2776static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2777                                   struct qstr *name, struct nfs_fh *fhandle,
2778                                   struct nfs_fattr *fattr)
2779{
2780        struct nfs4_exception exception = { };
2781        struct rpc_clnt *client = *clnt;
2782        int err;
2783        do {
2784                err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2785                switch (err) {
2786                case -NFS4ERR_BADNAME:
2787                        err = -ENOENT;
2788                        goto out;
2789                case -NFS4ERR_MOVED:
2790                        err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2791                        goto out;
2792                case -NFS4ERR_WRONGSEC:
2793                        err = -EPERM;
2794                        if (client != *clnt)
2795                                goto out;
2796
2797                        client = nfs4_create_sec_client(client, dir, name);
2798                        if (IS_ERR(client))
2799                                return PTR_ERR(client);
2800
2801                        exception.retry = 1;
2802                        break;
2803                default:
2804                        err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2805                }
2806        } while (exception.retry);
2807
2808out:
2809        if (err == 0)
2810                *clnt = client;
2811        else if (client != *clnt)
2812                rpc_shutdown_client(client);
2813
2814        return err;
2815}
2816
2817static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2818                            struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2819{
2820        int status;
2821        struct rpc_clnt *client = NFS_CLIENT(dir);
2822
2823        status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2824        if (client != NFS_CLIENT(dir)) {
2825                rpc_shutdown_client(client);
2826                nfs_fixup_secinfo_attributes(fattr);
2827        }
2828        return status;
2829}
2830
2831struct rpc_clnt *
2832nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2833                            struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2834{
2835        int status;
2836        struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2837
2838        status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2839        if (status < 0) {
2840                rpc_shutdown_client(client);
2841                return ERR_PTR(status);
2842        }
2843        return client;
2844}
2845
2846static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2847{
2848        struct nfs_server *server = NFS_SERVER(inode);
2849        struct nfs4_accessargs args = {
2850                .fh = NFS_FH(inode),
2851                .bitmask = server->cache_consistency_bitmask,
2852        };
2853        struct nfs4_accessres res = {
2854                .server = server,
2855        };
2856        struct rpc_message msg = {
2857                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2858                .rpc_argp = &args,
2859                .rpc_resp = &res,
2860                .rpc_cred = entry->cred,
2861        };
2862        int mode = entry->mask;
2863        int status;
2864
2865        /*
2866         * Determine which access bits we want to ask for...
2867         */
2868        if (mode & MAY_READ)
2869                args.access |= NFS4_ACCESS_READ;
2870        if (S_ISDIR(inode->i_mode)) {
2871                if (mode & MAY_WRITE)
2872                        args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2873                if (mode & MAY_EXEC)
2874                        args.access |= NFS4_ACCESS_LOOKUP;
2875        } else {
2876                if (mode & MAY_WRITE)
2877                        args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2878                if (mode & MAY_EXEC)
2879                        args.access |= NFS4_ACCESS_EXECUTE;
2880        }
2881
2882        res.fattr = nfs_alloc_fattr();
2883        if (res.fattr == NULL)
2884                return -ENOMEM;
2885
2886        status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2887        if (!status) {
2888                nfs_access_set_mask(entry, res.access);
2889                nfs_refresh_inode(inode, res.fattr);
2890        }
2891        nfs_free_fattr(res.fattr);
2892        return status;
2893}
2894
2895static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2896{
2897        struct nfs4_exception exception = { };
2898        int err;
2899        do {
2900                err = nfs4_handle_exception(NFS_SERVER(inode),
2901                                _nfs4_proc_access(inode, entry),
2902                                &exception);
2903        } while (exception.retry);
2904        return err;
2905}
2906
2907/*
2908 * TODO: For the time being, we don't try to get any attributes
2909 * along with any of the zero-copy operations READ, READDIR,
2910 * READLINK, WRITE.
2911 *
2912 * In the case of the first three, we want to put the GETATTR
2913 * after the read-type operation -- this is because it is hard
2914 * to predict the length of a GETATTR response in v4, and thus
2915 * align the READ data correctly.  This means that the GETATTR
2916 * may end up partially falling into the page cache, and we should
2917 * shift it into the 'tail' of the xdr_buf before processing.
2918 * To do this efficiently, we need to know the total length
2919 * of data received, which doesn't seem to be available outside
2920 * of the RPC layer.
2921 *
2922 * In the case of WRITE, we also want to put the GETATTR after
2923 * the operation -- in this case because we want to make sure
2924 * we get the post-operation mtime and size.
2925 *
2926 * Both of these changes to the XDR layer would in fact be quite
2927 * minor, but I decided to leave them for a subsequent patch.
2928 */
2929static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2930                unsigned int pgbase, unsigned int pglen)
2931{
2932        struct nfs4_readlink args = {
2933                .fh       = NFS_FH(inode),
2934                .pgbase   = pgbase,
2935                .pglen    = pglen,
2936                .pages    = &page,
2937        };
2938        struct nfs4_readlink_res res;
2939        struct rpc_message msg = {
2940                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2941                .rpc_argp = &args,
2942                .rpc_resp = &res,
2943        };
2944
2945        return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2946}
2947
2948static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2949                unsigned int pgbase, unsigned int pglen)
2950{
2951        struct nfs4_exception exception = { };
2952        int err;
2953        do {
2954                err = nfs4_handle_exception(NFS_SERVER(inode),
2955                                _nfs4_proc_readlink(inode, page, pgbase, pglen),
2956                                &exception);
2957        } while (exception.retry);
2958        return err;
2959}
2960
2961/*
2962 * This is just for mknod.  open(O_CREAT) will always do ->open_context().
2963 */
2964static int
2965nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2966                 int flags)
2967{
2968        struct nfs_open_context *ctx;
2969        struct nfs4_state *state;
2970        int status = 0;
2971
2972        ctx = alloc_nfs_open_context(dentry, FMODE_READ);
2973        if (IS_ERR(ctx))
2974                return PTR_ERR(ctx);
2975
2976        sattr->ia_mode &= ~current_umask();
2977        state = nfs4_do_open(dir, dentry, ctx->mode,
2978                        flags, sattr, ctx->cred,
2979                        &ctx->mdsthreshold);
2980        d_drop(dentry);
2981        if (IS_ERR(state)) {
2982                status = PTR_ERR(state);
2983                goto out;
2984        }
2985        d_add(dentry, igrab(state->inode));
2986        nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2987        ctx->state = state;
2988out:
2989        put_nfs_open_context(ctx);
2990        return status;
2991}
2992
2993static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2994{
2995        struct nfs_server *server = NFS_SERVER(dir);
2996        struct nfs_removeargs args = {
2997                .fh = NFS_FH(dir),
2998                .name = *name,
2999        };
3000        struct nfs_removeres res = {
3001                .server = server,
3002        };
3003        struct rpc_message msg = {
3004                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3005                .rpc_argp = &args,
3006                .rpc_resp = &res,
3007        };
3008        int status;
3009
3010        status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3011        if (status == 0)
3012                update_changeattr(dir, &res.cinfo);
3013        return status;
3014}
3015
3016static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3017{
3018        struct nfs4_exception exception = { };
3019        int err;
3020        do {
3021                err = nfs4_handle_exception(NFS_SERVER(dir),
3022                                _nfs4_proc_remove(dir, name),
3023                                &exception);
3024        } while (exception.retry);
3025        return err;
3026}
3027
3028static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3029{
3030        struct nfs_server *server = NFS_SERVER(dir);
3031        struct nfs_removeargs *args = msg->rpc_argp;
3032        struct nfs_removeres *res = msg->rpc_resp;
3033
3034        res->server = server;
3035        msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3036        nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3037}
3038
3039static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3040{
3041        if (nfs4_setup_sequence(NFS_SERVER(data->dir),
3042                                &data->args.seq_args,
3043                                &data->res.seq_res,
3044                                task))
3045                return;
3046        rpc_call_start(task);
3047}
3048
3049static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3050{
3051        struct nfs_removeres *res = task->tk_msg.rpc_resp;
3052
3053        if (!nfs4_sequence_done(task, &res->seq_res))
3054                return 0;
3055        if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3056                return 0;
3057        update_changeattr(dir, &res->cinfo);
3058        return 1;
3059}
3060
3061static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3062{
3063        struct nfs_server *server = NFS_SERVER(dir);
3064        struct nfs_renameargs *arg = msg->rpc_argp;
3065        struct nfs_renameres *res = msg->rpc_resp;
3066
3067        msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3068        res->server = server;
3069        nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3070}
3071
3072static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3073{
3074        if (nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3075                                &data->args.seq_args,
3076                                &data->res.seq_res,
3077                                task))
3078                return;
3079        rpc_call_start(task);
3080}
3081
3082static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3083                                 struct inode *new_dir)
3084{
3085        struct nfs_renameres *res = task->tk_msg.rpc_resp;
3086
3087        if (!nfs4_sequence_done(task, &res->seq_res))
3088                return 0;
3089        if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3090                return 0;
3091
3092        update_changeattr(old_dir, &res->old_cinfo);
3093        update_changeattr(new_dir, &res->new_cinfo);
3094        return 1;
3095}
3096
3097static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3098                struct inode *new_dir, struct qstr *new_name)
3099{
3100        struct nfs_server *server = NFS_SERVER(old_dir);
3101        struct nfs_renameargs arg = {
3102                .old_dir = NFS_FH(old_dir),
3103                .new_dir = NFS_FH(new_dir),
3104                .old_name = old_name,
3105                .new_name = new_name,
3106        };
3107        struct nfs_renameres res = {
3108                .server = server,
3109        };
3110        struct rpc_message msg = {
3111                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3112                .rpc_argp = &arg,
3113                .rpc_resp = &res,
3114        };
3115        int status = -ENOMEM;
3116        
3117        status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3118        if (!status) {
3119                update_changeattr(old_dir, &res.old_cinfo);
3120                update_changeattr(new_dir, &res.new_cinfo);
3121        }
3122        return status;
3123}
3124
3125static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3126                struct inode *new_dir, struct qstr *new_name)
3127{
3128        struct nfs4_exception exception = { };
3129        int err;
3130        do {
3131                err = nfs4_handle_exception(NFS_SERVER(old_dir),
3132                                _nfs4_proc_rename(old_dir, old_name,
3133                                        new_dir, new_name),
3134                                &exception);
3135        } while (exception.retry);
3136        return err;
3137}
3138
3139static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3140{
3141        struct nfs_server *server = NFS_SERVER(inode);
3142        struct nfs4_link_arg arg = {
3143                .fh     = NFS_FH(inode),
3144                .dir_fh = NFS_FH(dir),
3145                .name   = name,
3146                .bitmask = server->attr_bitmask,
3147        };
3148        struct nfs4_link_res res = {
3149                .server = server,
3150        };
3151        struct rpc_message msg = {
3152                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3153                .rpc_argp = &arg,
3154                .rpc_resp = &res,
3155        };
3156        int status = -ENOMEM;
3157
3158        res.fattr = nfs_alloc_fattr();
3159        if (res.fattr == NULL)
3160                goto out;
3161
3162        status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3163        if (!status) {
3164                update_changeattr(dir, &res.cinfo);
3165                nfs_post_op_update_inode(inode, res.fattr);
3166        }
3167out:
3168        nfs_free_fattr(res.fattr);
3169        return status;
3170}
3171
3172static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3173{
3174        struct nfs4_exception exception = { };
3175        int err;
3176        do {
3177                err = nfs4_handle_exception(NFS_SERVER(inode),
3178                                _nfs4_proc_link(inode, dir, name),
3179                                &exception);
3180        } while (exception.retry);
3181        return err;
3182}
3183
3184struct nfs4_createdata {
3185        struct rpc_message msg;
3186        struct nfs4_create_arg arg;
3187        struct nfs4_create_res res;
3188        struct nfs_fh fh;
3189        struct nfs_fattr fattr;
3190};
3191
3192static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3193                struct qstr *name, struct iattr *sattr, u32 ftype)
3194{
3195        struct nfs4_createdata *data;
3196
3197        data = kzalloc(sizeof(*data), GFP_KERNEL);
3198        if (data != NULL) {
3199                struct nfs_server *server = NFS_SERVER(dir);
3200
3201                data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3202                data->msg.rpc_argp = &data->arg;
3203                data->msg.rpc_resp = &data->res;
3204                data->arg.dir_fh = NFS_FH(dir);
3205                data->arg.server = server;
3206                data->arg.name = name;
3207                data->arg.attrs = sattr;
3208                data->arg.ftype = ftype;
3209                data->arg.bitmask = server->attr_bitmask;
3210                data->res.server = server;
3211                data->res.fh = &data->fh;
3212                data->res.fattr = &data->fattr;
3213                nfs_fattr_init(data->res.fattr);
3214        }
3215        return data;
3216}
3217
3218static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3219{
3220        int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3221                                    &data->arg.seq_args, &data->res.seq_res, 1);
3222        if (status == 0) {
3223                update_changeattr(dir, &data->res.dir_cinfo);
3224                status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3225        }
3226        return status;
3227}
3228
3229static void nfs4_free_createdata(struct nfs4_createdata *data)
3230{
3231        kfree(data);
3232}
3233
3234static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3235                struct page *page, unsigned int len, struct iattr *sattr)
3236{
3237        struct nfs4_createdata *data;
3238        int status = -ENAMETOOLONG;
3239
3240        if (len > NFS4_MAXPATHLEN)
3241                goto out;
3242
3243        status = -ENOMEM;
3244        data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3245        if (data == NULL)
3246                goto out;
3247
3248        data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3249        data->arg.u.symlink.pages = &page;
3250        data->arg.u.symlink.len = len;
3251        
3252        status = nfs4_do_create(dir, dentry, data);
3253
3254        nfs4_free_createdata(data);
3255out:
3256        return status;
3257}
3258
3259static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3260                struct page *page, unsigned int len, struct iattr *sattr)
3261{
3262        struct nfs4_exception exception = { };
3263        int err;
3264        do {
3265                err = nfs4_handle_exception(NFS_SERVER(dir),
3266                                _nfs4_proc_symlink(dir, dentry, page,
3267                                                        len, sattr),
3268                                &exception);
3269        } while (exception.retry);
3270        return err;
3271}
3272
3273static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3274                struct iattr *sattr)
3275{
3276        struct nfs4_createdata *data;
3277        int status = -ENOMEM;
3278
3279        data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3280        if (data == NULL)
3281                goto out;
3282
3283        status = nfs4_do_create(dir, dentry, data);
3284
3285        nfs4_free_createdata(data);
3286out:
3287        return status;
3288}
3289
3290static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3291                struct iattr *sattr)
3292{
3293        struct nfs4_exception exception = { };
3294        int err;
3295
3296        sattr->ia_mode &= ~current_umask();
3297        do {
3298                err = nfs4_handle_exception(NFS_SERVER(dir),
3299                                _nfs4_proc_mkdir(dir, dentry, sattr),
3300                                &exception);
3301        } while (exception.retry);
3302        return err;
3303}
3304
3305static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3306                u64 cookie, struct page **pages, unsigned int count, int plus)
3307{
3308        struct inode            *dir = dentry->d_inode;
3309        struct nfs4_readdir_arg args = {
3310                .fh = NFS_FH(dir),
3311                .pages = pages,
3312                .pgbase = 0,
3313                .count = count,
3314                .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3315                .plus = plus,
3316        };
3317        struct nfs4_readdir_res res;
3318        struct rpc_message msg = {
3319                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3320                .rpc_argp = &args,
3321                .rpc_resp = &res,
3322                .rpc_cred = cred,
3323        };
3324        int                     status;
3325
3326        dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3327                        dentry->d_parent->d_name.name,
3328                        dentry->d_name.name,
3329                        (unsigned long long)cookie);
3330        nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3331        res.pgbase = args.pgbase;
3332        status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3333        if (status >= 0) {
3334                memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3335                status += args.pgbase;
3336        }
3337
3338        nfs_invalidate_atime(dir);
3339
3340        dprintk("%s: returns %d\n", __func__, status);
3341        return status;
3342}
3343
3344static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3345                u64 cookie, struct page **pages, unsigned int count, int plus)
3346{
3347        struct nfs4_exception exception = { };
3348        int err;
3349        do {
3350                err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3351                                _nfs4_proc_readdir(dentry, cred, cookie,
3352                                        pages, count, plus),
3353                                &exception);
3354        } while (exception.retry);
3355        return err;
3356}
3357
3358static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3359                struct iattr *sattr, dev_t rdev)
3360{
3361        struct nfs4_createdata *data;
3362        int mode = sattr->ia_mode;
3363        int status = -ENOMEM;
3364
3365        BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3366        BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3367
3368        data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3369        if (data == NULL)
3370                goto out;
3371
3372        if (S_ISFIFO(mode))
3373                data->arg.ftype = NF4FIFO;
3374        else if (S_ISBLK(mode)) {
3375                data->arg.ftype = NF4BLK;
3376                data->arg.u.device.specdata1 = MAJOR(rdev);
3377                data->arg.u.device.specdata2 = MINOR(rdev);
3378        }
3379        else if (S_ISCHR(mode)) {
3380                data->arg.ftype = NF4CHR;
3381                data->arg.u.device.specdata1 = MAJOR(rdev);
3382                data->arg.u.device.specdata2 = MINOR(rdev);
3383        }
3384        
3385        status = nfs4_do_create(dir, dentry, data);
3386
3387        nfs4_free_createdata(data);
3388out:
3389        return status;
3390}
3391
3392static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3393                struct iattr *sattr, dev_t rdev)
3394{
3395        struct nfs4_exception exception = { };
3396        int err;
3397
3398        sattr->ia_mode &= ~current_umask();
3399        do {
3400                err = nfs4_handle_exception(NFS_SERVER(dir),
3401                                _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3402                                &exception);
3403        } while (exception.retry);
3404        return err;
3405}
3406
3407static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3408                 struct nfs_fsstat *fsstat)
3409{
3410        struct nfs4_statfs_arg args = {
3411                .fh = fhandle,
3412                .bitmask = server->attr_bitmask,
3413        };
3414        struct nfs4_statfs_res res = {
3415                .fsstat = fsstat,
3416        };
3417        struct rpc_message msg = {
3418                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3419                .rpc_argp = &args,
3420                .rpc_resp = &res,
3421        };
3422
3423        nfs_fattr_init(fsstat->fattr);
3424        return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3425}
3426
3427static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3428{
3429        struct nfs4_exception exception = { };
3430        int err;
3431        do {
3432                err = nfs4_handle_exception(server,
3433                                _nfs4_proc_statfs(server, fhandle, fsstat),
3434                                &exception);
3435        } while (exception.retry);
3436        return err;
3437}
3438
3439static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3440                struct nfs_fsinfo *fsinfo)
3441{
3442        struct nfs4_fsinfo_arg args = {
3443                .fh = fhandle,
3444                .bitmask = server->attr_bitmask,
3445        };
3446        struct nfs4_fsinfo_res res = {
3447                .fsinfo = fsinfo,
3448        };
3449        struct rpc_message msg = {
3450                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3451                .rpc_argp = &args,
3452                .rpc_resp = &res,
3453        };
3454
3455        return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3456}
3457
3458static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3459{
3460        struct nfs4_exception exception = { };
3461        int err;
3462
3463        do {
3464                err = nfs4_handle_exception(server,
3465                                _nfs4_do_fsinfo(server, fhandle, fsinfo),
3466                                &exception);
3467        } while (exception.retry);
3468        return err;
3469}
3470
3471static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3472{
3473        int error;
3474
3475        nfs_fattr_init(fsinfo->fattr);
3476        error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3477        if (error == 0) {
3478                /* block layout checks this! */
3479                server->pnfs_blksize = fsinfo->blksize;
3480                set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3481        }
3482
3483        return error;
3484}
3485
3486static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3487                struct nfs_pathconf *pathconf)
3488{
3489        struct nfs4_pathconf_arg args = {
3490                .fh = fhandle,
3491                .bitmask = server->attr_bitmask,
3492        };
3493        struct nfs4_pathconf_res res = {
3494                .pathconf = pathconf,
3495        };
3496        struct rpc_message msg = {
3497                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3498                .rpc_argp = &args,
3499                .rpc_resp = &res,
3500        };
3501
3502        /* None of the pathconf attributes are mandatory to implement */
3503        if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3504                memset(pathconf, 0, sizeof(*pathconf));
3505                return 0;
3506        }
3507
3508        nfs_fattr_init(pathconf->fattr);
3509        return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3510}
3511
3512static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3513                struct nfs_pathconf *pathconf)
3514{
3515        struct nfs4_exception exception = { };
3516        int err;
3517
3518        do {
3519                err = nfs4_handle_exception(server,
3520                                _nfs4_proc_pathconf(server, fhandle, pathconf),
3521                                &exception);
3522        } while (exception.retry);
3523        return err;
3524}
3525
3526void __nfs4_read_done_cb(struct nfs_read_data *data)
3527{
3528        nfs_invalidate_atime(data->header->inode);
3529}
3530
3531static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3532{
3533        struct nfs_server *server = NFS_SERVER(data->header->inode);
3534
3535        if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3536                rpc_restart_call_prepare(task);
3537                return -EAGAIN;
3538        }
3539
3540        __nfs4_read_done_cb(data);
3541        if (task->tk_status > 0)
3542                renew_lease(server, data->timestamp);
3543        return 0;
3544}
3545
3546static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3547{
3548
3549        dprintk("--> %s\n", __func__);
3550
3551        if (!nfs4_sequence_done(task, &data->res.seq_res))
3552                return -EAGAIN;
3553
3554        return data->read_done_cb ? data->read_done_cb(task, data) :
3555                                    nfs4_read_done_cb(task, data);
3556}
3557
3558static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3559{
3560        data->timestamp   = jiffies;
3561        data->read_done_cb = nfs4_read_done_cb;
3562        msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3563        nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3564}
3565
3566static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3567{
3568        if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3569                                &data->args.seq_args,
3570                                &data->res.seq_res,
3571                                task))
3572                return;
3573        rpc_call_start(task);
3574}
3575
3576static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3577{
3578        struct inode *inode = data->header->inode;
3579        
3580        if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3581                rpc_restart_call_prepare(task);
3582                return -EAGAIN;
3583        }
3584        if (task->tk_status >= 0) {
3585                renew_lease(NFS_SERVER(inode), data->timestamp);
3586                nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3587        }
3588        return 0;
3589}
3590
3591static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3592{
3593        if (!nfs4_sequence_done(task, &data->res.seq_res))
3594                return -EAGAIN;
3595        return data->write_done_cb ? data->write_done_cb(task, data) :
3596                nfs4_write_done_cb(task, data);
3597}
3598
3599static
3600bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3601{
3602        const struct nfs_pgio_header *hdr = data->header;
3603
3604        /* Don't request attributes for pNFS or O_DIRECT writes */
3605        if (data->ds_clp != NULL || hdr->dreq != NULL)
3606                return false;
3607        /* Otherwise, request attributes if and only if we don't hold
3608         * a delegation
3609         */
3610        return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3611}
3612
3613static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3614{
3615        struct nfs_server *server = NFS_SERVER(data->header->inode);
3616
3617        if (!nfs4_write_need_cache_consistency_data(data)) {
3618                data->args.bitmask = NULL;
3619                data->res.fattr = NULL;
3620        } else
3621                data->args.bitmask = server->cache_consistency_bitmask;
3622
3623        if (!data->write_done_cb)
3624                data->write_done_cb = nfs4_write_done_cb;
3625        data->res.server = server;
3626        data->timestamp   = jiffies;
3627
3628        msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3629        nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3630}
3631
3632static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3633{
3634        if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3635                                &data->args.seq_args,
3636                                &data->res.seq_res,
3637                                task))
3638                return;
3639        rpc_call_start(task);
3640}
3641
3642static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3643{
3644        if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3645                                &data->args.seq_args,
3646                                &data->res.seq_res,
3647                                task))
3648                return;
3649        rpc_call_start(task);
3650}
3651
3652static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3653{
3654        struct inode *inode = data->inode;
3655
3656        if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3657                rpc_restart_call_prepare(task);
3658                return -EAGAIN;
3659        }
3660        return 0;
3661}
3662
3663static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3664{
3665        if (!nfs4_sequence_done(task, &data->res.seq_res))
3666                return -EAGAIN;
3667        return data->commit_done_cb(task, data);
3668}
3669
3670static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3671{
3672        struct nfs_server *server = NFS_SERVER(data->inode);
3673
3674        if (data->commit_done_cb == NULL)
3675                data->commit_done_cb = nfs4_commit_done_cb;
3676        data->res.server = server;
3677        msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3678        nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3679}
3680
3681struct nfs4_renewdata {
3682        struct nfs_client       *client;
3683        unsigned long           timestamp;
3684};
3685
3686/*
3687 * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3688 * standalone procedure for queueing an asynchronous RENEW.
3689 */
3690static void nfs4_renew_release(void *calldata)
3691{
3692        struct nfs4_renewdata *data = calldata;
3693        struct nfs_client *clp = data->client;
3694
3695        if (atomic_read(&clp->cl_count) > 1)
3696                nfs4_schedule_state_renewal(clp);
3697        nfs_put_client(clp);
3698        kfree(data);
3699}
3700
3701static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3702{
3703        struct nfs4_renewdata *data = calldata;
3704        struct nfs_client *clp = data->client;
3705        unsigned long timestamp = data->timestamp;
3706
3707        if (task->tk_status < 0) {
3708                /* Unless we're shutting down, schedule state recovery! */
3709                if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3710                        return;
3711                if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3712                        nfs4_schedule_lease_recovery(clp);
3713                        return;
3714                }
3715                nfs4_schedule_path_down_recovery(clp);
3716        }
3717        do_renew_lease(clp, timestamp);
3718}
3719
3720static const struct rpc_call_ops nfs4_renew_ops = {
3721        .rpc_call_done = nfs4_renew_done,
3722        .rpc_release = nfs4_renew_release,
3723};
3724
3725static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3726{
3727        struct rpc_message msg = {
3728                .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3729                .rpc_argp       = clp,
3730                .rpc_cred       = cred,
3731        };
3732        struct nfs4_renewdata *data;
3733
3734        if (renew_flags == 0)
3735                return 0;
3736        if (!atomic_inc_not_zero(&clp->cl_count))
3737                return -EIO;
3738        data = kmalloc(sizeof(*data), GFP_NOFS);
3739        if (data == NULL)
3740                return -ENOMEM;
3741        data->client = clp;
3742        data->timestamp = jiffies;
3743        return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3744                        &nfs4_renew_ops, data);
3745}
3746
3747static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3748{
3749        struct rpc_message msg = {
3750                .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3751                .rpc_argp       = clp,
3752                .rpc_cred       = cred,
3753        };
3754        unsigned long now = jiffies;
3755        int status;
3756
3757        status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3758        if (status < 0)
3759                return status;
3760        do_renew_lease(clp, now);
3761        return 0;
3762}
3763
3764static inline int nfs4_server_supports_acls(struct nfs_server *server)
3765{
3766        return (server->caps & NFS_CAP_ACLS)
3767                && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3768                && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3769}
3770
3771/* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3772 * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3773 * the stack.
3774 */
3775#define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3776
3777static int buf_to_pages_noslab(const void *buf, size_t buflen,
3778                struct page **pages, unsigned int *pgbase)
3779{
3780        struct page *newpage, **spages;
3781        int rc = 0;
3782        size_t len;
3783        spages = pages;
3784
3785        do {
3786                len = min_t(size_t, PAGE_SIZE, buflen);
3787                newpage = alloc_page(GFP_KERNEL);
3788
3789                if (newpage == NULL)
3790                        goto unwind;
3791                memcpy(page_address(newpage), buf, len);
3792                buf += len;
3793                buflen -= len;
3794                *pages++ = newpage;
3795                rc++;
3796        } while (buflen != 0);
3797
3798        return rc;
3799
3800unwind:
3801        for(; rc > 0; rc--)
3802                __free_page(spages[rc-1]);
3803        return -ENOMEM;
3804}
3805
3806struct nfs4_cached_acl {
3807        int cached;
3808        size_t len;
3809        char data[0];
3810};
3811
3812static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3813{
3814        struct nfs_inode *nfsi = NFS_I(inode);
3815
3816        spin_lock(&inode->i_lock);
3817        kfree(nfsi->nfs4_acl);
3818        nfsi->nfs4_acl = acl;
3819        spin_unlock(&inode->i_lock);
3820}
3821
3822static void nfs4_zap_acl_attr(struct inode *inode)
3823{
3824        nfs4_set_cached_acl(inode, NULL);
3825}
3826
3827static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3828{
3829        struct nfs_inode *nfsi = NFS_I(inode);
3830        struct nfs4_cached_acl *acl;
3831        int ret = -ENOENT;
3832
3833        spin_lock(&inode->i_lock);
3834        acl = nfsi->nfs4_acl;
3835        if (acl == NULL)
3836                goto out;
3837        if (buf == NULL) /* user is just asking for length */
3838                goto out_len;
3839        if (acl->cached == 0)
3840                goto out;
3841        ret = -ERANGE; /* see getxattr(2) man page */
3842        if (acl->len > buflen)
3843                goto out;
3844        memcpy(buf, acl->data, acl->len);
3845out_len:
3846        ret = acl->len;
3847out:
3848        spin_unlock(&inode->i_lock);
3849        return ret;
3850}
3851
3852static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3853{
3854        struct nfs4_cached_acl *acl;
3855        size_t buflen = sizeof(*acl) + acl_len;
3856
3857        if (buflen <= PAGE_SIZE) {
3858                acl = kmalloc(buflen, GFP_KERNEL);
3859                if (acl == NULL)
3860                        goto out;
3861                acl->cached = 1;
3862                _copy_from_pages(acl->data, pages, pgbase, acl_len);
3863        } else {
3864                acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3865                if (acl == NULL)
3866                        goto out;
3867                acl->cached = 0;
3868        }
3869        acl->len = acl_len;
3870out:
3871        nfs4_set_cached_acl(inode, acl);
3872}
3873
3874/*
3875 * The getxattr API returns the required buffer length when called with a
3876 * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3877 * the required buf.  On a NULL buf, we send a page of data to the server
3878 * guessing that the ACL request can be serviced by a page. If so, we cache
3879 * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3880 * the cache. If not so, we throw away the page, and cache the required
3881 * length. The next getxattr call will then produce another round trip to
3882 * the server, this time with the input buf of the required size.
3883 */
3884static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3885{
3886        struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3887        struct nfs_getaclargs args = {
3888                .fh = NFS_FH(inode),
3889                .acl_pages = pages,
3890                .acl_len = buflen,
3891        };
3892        struct nfs_getaclres res = {
3893                .acl_len = buflen,
3894        };
3895        struct rpc_message msg = {
3896                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3897                .rpc_argp = &args,
3898                .rpc_resp = &res,
3899        };
3900        unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3901        int ret = -ENOMEM, i;
3902
3903        /* As long as we're doing a round trip to the server anyway,
3904         * let's be prepared for a page of acl data. */
3905        if (npages == 0)
3906                npages = 1;
3907        if (npages > ARRAY_SIZE(pages))
3908                return -ERANGE;
3909
3910        for (i = 0; i < npages; i++) {
3911                pages[i] = alloc_page(GFP_KERNEL);
3912                if (!pages[i])
3913                        goto out_free;
3914        }
3915
3916        /* for decoding across pages */
3917        res.acl_scratch = alloc_page(GFP_KERNEL);
3918        if (!res.acl_scratch)
3919                goto out_free;
3920
3921        args.acl_len = npages * PAGE_SIZE;
3922        args.acl_pgbase = 0;
3923
3924        dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3925                __func__, buf, buflen, npages, args.acl_len);
3926        ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3927                             &msg, &args.seq_args, &res.seq_res, 0);
3928        if (ret)
3929                goto out_free;
3930
3931        /* Handle the case where the passed-in buffer is too short */
3932        if (res.acl_flags & NFS4_ACL_TRUNC) {
3933                /* Did the user only issue a request for the acl length? */
3934                if (buf == NULL)
3935                        goto out_ok;
3936                ret = -ERANGE;
3937                goto out_free;
3938        }
3939        nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
3940        if (buf)
3941                _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
3942out_ok:
3943        ret = res.acl_len;
3944out_free:
3945        for (i = 0; i < npages; i++)
3946                if (pages[i])
3947                        __free_page(pages[i]);
3948        if (res.acl_scratch)
3949                __free_page(res.acl_scratch);
3950        return ret;
3951}
3952
3953static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3954{
3955        struct nfs4_exception exception = { };
3956        ssize_t ret;
3957        do {
3958                ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3959                if (ret >= 0)
3960                        break;
3961                ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3962        } while (exception.retry);
3963        return ret;
3964}
3965
3966static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3967{
3968        struct nfs_server *server = NFS_SERVER(inode);
3969        int ret;
3970
3971        if (!nfs4_server_supports_acls(server))
3972                return -EOPNOTSUPP;
3973        ret = nfs_revalidate_inode(server, inode);
3974        if (ret < 0)
3975                return ret;
3976        if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3977                nfs_zap_acl_cache(inode);
3978        ret = nfs4_read_cached_acl(inode, buf, buflen);
3979        if (ret != -ENOENT)
3980                /* -ENOENT is returned if there is no ACL or if there is an ACL
3981                 * but no cached acl data, just the acl length */
3982                return ret;
3983        return nfs4_get_acl_uncached(inode, buf, buflen);
3984}
3985
3986static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3987{
3988        struct nfs_server *server = NFS_SERVER(inode);
3989        struct page *pages[NFS4ACL_MAXPAGES];
3990        struct nfs_setaclargs arg = {
3991                .fh             = NFS_FH(inode),
3992                .acl_pages      = pages,
3993                .acl_len        = buflen,
3994        };
3995        struct nfs_setaclres res;
3996        struct rpc_message msg = {
3997                .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3998                .rpc_argp       = &arg,
3999                .rpc_resp       = &res,
4000        };
4001        unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4002        int ret, i;
4003
4004        if (!nfs4_server_supports_acls(server))
4005                return -EOPNOTSUPP;
4006        if (npages > ARRAY_SIZE(pages))
4007                return -ERANGE;
4008        i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4009        if (i < 0)
4010                return i;
4011        nfs4_inode_return_delegation(inode);
4012        ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4013
4014        /*
4015         * Free each page after tx, so the only ref left is
4016         * held by the network stack
4017         */
4018        for (; i > 0; i--)
4019                put_page(pages[i-1]);
4020
4021        /*
4022         * Acl update can result in inode attribute update.
4023         * so mark the attribute cache invalid.
4024         */
4025        spin_lock(&inode->i_lock);
4026        NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4027        spin_unlock(&inode->i_lock);
4028        nfs_access_zap_cache(inode);
4029        nfs_zap_acl_cache(inode);
4030        return ret;
4031}
4032
4033static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4034{
4035        struct nfs4_exception exception = { };
4036        int err;
4037        do {
4038                err = nfs4_handle_exception(NFS_SERVER(inode),
4039                                __nfs4_proc_set_acl(inode, buf, buflen),
4040                                &exception);
4041        } while (exception.retry);
4042        return err;
4043}
4044
4045static int
4046nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4047{
4048        struct nfs_client *clp = server->nfs_client;
4049
4050        if (task->tk_status >= 0)
4051                return 0;
4052        switch(task->tk_status) {
4053                case -NFS4ERR_DELEG_REVOKED:
4054                case -NFS4ERR_ADMIN_REVOKED:
4055                case -NFS4ERR_BAD_STATEID:
4056                        if (state == NULL)
4057                                break;
4058                        nfs_remove_bad_delegation(state->inode);
4059                case -NFS4ERR_OPENMODE:
4060                        if (state == NULL)
4061                                break;
4062                        nfs4_schedule_stateid_recovery(server, state);
4063                        goto wait_on_recovery;
4064                case -NFS4ERR_EXPIRED:
4065                        if (state != NULL)
4066                                nfs4_schedule_stateid_recovery(server, state);
4067                case -NFS4ERR_STALE_STATEID:
4068                case -NFS4ERR_STALE_CLIENTID:
4069                        nfs4_schedule_lease_recovery(clp);
4070                        goto wait_on_recovery;
4071#if defined(CONFIG_NFS_V4_1)
4072                case -NFS4ERR_BADSESSION:
4073                case -NFS4ERR_BADSLOT:
4074                case -NFS4ERR_BAD_HIGH_SLOT:
4075                case -NFS4ERR_DEADSESSION:
4076                case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4077                case -NFS4ERR_SEQ_FALSE_RETRY:
4078                case -NFS4ERR_SEQ_MISORDERED:
4079                        dprintk("%s ERROR %d, Reset session\n", __func__,
4080                                task->tk_status);
4081                        nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4082                        task->tk_status = 0;
4083                        return -EAGAIN;
4084#endif /* CONFIG_NFS_V4_1 */
4085                case -NFS4ERR_DELAY:
4086                        nfs_inc_server_stats(server, NFSIOS_DELAY);
4087                case -NFS4ERR_GRACE:
4088                case -EKEYEXPIRED:
4089                        rpc_delay(task, NFS4_POLL_RETRY_MAX);
4090                        task->tk_status = 0;
4091                        return -EAGAIN;
4092                case -NFS4ERR_RETRY_UNCACHED_REP:
4093                case -NFS4ERR_OLD_STATEID:
4094                        task->tk_status = 0;
4095                        return -EAGAIN;
4096        }
4097        task->tk_status = nfs4_map_errors(task->tk_status);
4098        return 0;
4099wait_on_recovery:
4100        rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4101        if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4102                rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4103        task->tk_status = 0;
4104        return -EAGAIN;
4105}
4106
4107static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4108                                    nfs4_verifier *bootverf)
4109{
4110        __be32 verf[2];
4111
4112        if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4113                /* An impossible timestamp guarantees this value
4114                 * will never match a generated boot time. */
4115                verf[0] = 0;
4116                verf[1] = (__be32)(NSEC_PER_SEC + 1);
4117        } else {
4118                struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4119                verf[0] = (__be32)nn->boot_time.tv_sec;
4120                verf[1] = (__be32)nn->boot_time.tv_nsec;
4121        }
4122        memcpy(bootverf->data, verf, sizeof(bootverf->data));
4123}
4124
4125static unsigned int
4126nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4127                                   char *buf, size_t len)
4128{
4129        unsigned int result;
4130
4131        rcu_read_lock();
4132        result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4133                                clp->cl_ipaddr,
4134                                rpc_peeraddr2str(clp->cl_rpcclient,
4135                                                        RPC_DISPLAY_ADDR),
4136                                rpc_peeraddr2str(clp->cl_rpcclient,
4137                                                        RPC_DISPLAY_PROTO));
4138        rcu_read_unlock();
4139        return result;
4140}
4141
4142static unsigned int
4143nfs4_init_uniform_client_string(const struct nfs_client *clp,
4144                                char *buf, size_t len)
4145{
4146        char *nodename = clp->cl_rpcclient->cl_nodename;
4147
4148        if (nfs4_client_id_uniquifier[0] != '\0')
4149                nodename = nfs4_client_id_uniquifier;
4150        return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4151                                clp->rpc_ops->version, clp->cl_minorversion,
4152                                nodename);
4153}
4154
4155/**
4156 * nfs4_proc_setclientid - Negotiate client ID
4157 * @clp: state data structure
4158 * @program: RPC program for NFSv4 callback service
4159 * @port: IP port number for NFS4 callback service
4160 * @cred: RPC credential to use for this call
4161 * @res: where to place the result
4162 *
4163 * Returns zero, a negative errno, or a negative NFS4ERR status code.
4164 */
4165int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4166                unsigned short port, struct rpc_cred *cred,
4167                struct nfs4_setclientid_res *res)
4168{
4169        nfs4_verifier sc_verifier;
4170        struct nfs4_setclientid setclientid = {
4171                .sc_verifier = &sc_verifier,
4172                .sc_prog = program,
4173                .sc_cb_ident = clp->cl_cb_ident,
4174        };
4175        struct rpc_message msg = {
4176                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4177                .rpc_argp = &setclientid,
4178                .rpc_resp = res,
4179                .rpc_cred = cred,
4180        };
4181        int status;
4182
4183        /* nfs_client_id4 */
4184        nfs4_init_boot_verifier(clp, &sc_verifier);
4185        if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4186                setclientid.sc_name_len =
4187                                nfs4_init_uniform_client_string(clp,
4188                                                setclientid.sc_name,
4189                                                sizeof(setclientid.sc_name));
4190        else
4191                setclientid.sc_name_len =
4192                                nfs4_init_nonuniform_client_string(clp,
4193                                                setclientid.sc_name,
4194                                                sizeof(setclientid.sc_name));
4195        /* cb_client4 */
4196        rcu_read_lock();
4197        setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4198                                sizeof(setclientid.sc_netid),
4199                                rpc_peeraddr2str(clp->cl_rpcclient,
4200                                                        RPC_DISPLAY_NETID));
4201        rcu_read_unlock();
4202        setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4203                                sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4204                                clp->cl_ipaddr, port >> 8, port & 255);
4205
4206        dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4207                clp->cl_rpcclient->cl_auth->au_ops->au_name,
4208                setclientid.sc_name_len, setclientid.sc_name);
4209        status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4210        dprintk("NFS reply setclientid: %d\n", status);
4211        return status;
4212}
4213
4214/**
4215 * nfs4_proc_setclientid_confirm - Confirm client ID
4216 * @clp: state data structure
4217 * @res: result of a previous SETCLIENTID
4218 * @cred: RPC credential to use for this call
4219 *
4220 * Returns zero, a negative errno, or a negative NFS4ERR status code.
4221 */
4222int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4223                struct nfs4_setclientid_res *arg,
4224                struct rpc_cred *cred)
4225{
4226        struct nfs_fsinfo fsinfo;
4227        struct rpc_message msg = {
4228                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4229                .rpc_argp = arg,
4230                .rpc_resp = &fsinfo,
4231                .rpc_cred = cred,
4232        };
4233        unsigned long now;
4234        int status;
4235
4236        dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4237                clp->cl_rpcclient->cl_auth->au_ops->au_name,
4238                clp->cl_clientid);
4239        now = jiffies;
4240        status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4241        if (status == 0) {
4242                spin_lock(&clp->cl_lock);
4243                clp->cl_lease_time = fsinfo.lease_time * HZ;
4244                clp->cl_last_renewal = now;
4245                spin_unlock(&clp->cl_lock);
4246        }
4247        dprintk("NFS reply setclientid_confirm: %d\n", status);
4248        return status;
4249}
4250
4251struct nfs4_delegreturndata {
4252        struct nfs4_delegreturnargs args;
4253        struct nfs4_delegreturnres res;
4254        struct nfs_fh fh;
4255        nfs4_stateid stateid;
4256        unsigned long timestamp;
4257        struct nfs_fattr fattr;
4258        int rpc_status;
4259};
4260
4261static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4262{
4263        struct nfs4_delegreturndata *data = calldata;
4264
4265        if (!nfs4_sequence_done(task, &data->res.seq_res))
4266                return;
4267
4268        switch (task->tk_status) {
4269        case -NFS4ERR_STALE_STATEID:
4270        case -NFS4ERR_EXPIRED:
4271        case 0:
4272                renew_lease(data->res.server, data->timestamp);
4273                break;
4274        default:
4275                if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4276                                -EAGAIN) {
4277                        rpc_restart_call_prepare(task);
4278                        return;
4279                }
4280        }
4281        data->rpc_status = task->tk_status;
4282}
4283
4284static void nfs4_delegreturn_release(void *calldata)
4285{
4286        kfree(calldata);
4287}
4288
4289#if defined(CONFIG_NFS_V4_1)
4290static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4291{
4292        struct nfs4_delegreturndata *d_data;
4293
4294        d_data = (struct nfs4_delegreturndata *)data;
4295
4296        if (nfs4_setup_sequence(d_data->res.server,
4297                                &d_data->args.seq_args,
4298                                &d_data->res.seq_res, task))
4299                return;
4300        rpc_call_start(task);
4301}
4302#endif /* CONFIG_NFS_V4_1 */
4303
4304static const struct rpc_call_ops nfs4_delegreturn_ops = {
4305#if defined(CONFIG_NFS_V4_1)
4306        .rpc_call_prepare = nfs4_delegreturn_prepare,
4307#endif /* CONFIG_NFS_V4_1 */
4308        .rpc_call_done = nfs4_delegreturn_done,
4309        .rpc_release = nfs4_delegreturn_release,
4310};
4311
4312static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4313{
4314        struct nfs4_delegreturndata *data;
4315        struct nfs_server *server = NFS_SERVER(inode);
4316        struct rpc_task *task;
4317        struct rpc_message msg = {
4318                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4319                .rpc_cred = cred,
4320        };
4321        struct rpc_task_setup task_setup_data = {
4322                .rpc_client = server->client,
4323                .rpc_message = &msg,
4324                .callback_ops = &nfs4_delegreturn_ops,
4325                .flags = RPC_TASK_ASYNC,
4326        };
4327        int status = 0;
4328
4329        data = kzalloc(sizeof(*data), GFP_NOFS);
4330        if (data == NULL)
4331                return -ENOMEM;
4332        nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4333        data->args.fhandle = &data->fh;
4334        data->args.stateid = &data->stateid;
4335        data->args.bitmask = server->cache_consistency_bitmask;
4336        nfs_copy_fh(&data->fh, NFS_FH(inode));
4337        nfs4_stateid_copy(&data->stateid, stateid);
4338        data->res.fattr = &data->fattr;
4339        data->res.server = server;
4340        nfs_fattr_init(data->res.fattr);
4341        data->timestamp = jiffies;
4342        data->rpc_status = 0;
4343
4344        task_setup_data.callback_data = data;
4345        msg.rpc_argp = &data->args;
4346        msg.rpc_resp = &data->res;
4347        task = rpc_run_task(&task_setup_data);
4348        if (IS_ERR(task))
4349                return PTR_ERR(task);
4350        if (!issync)
4351                goto out;
4352        status = nfs4_wait_for_completion_rpc_task(task);
4353        if (status != 0)
4354                goto out;
4355        status = data->rpc_status;
4356        if (status == 0)
4357                nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4358        else
4359                nfs_refresh_inode(inode, &data->fattr);
4360out:
4361        rpc_put_task(task);
4362        return status;
4363}
4364
4365int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4366{
4367        struct nfs_server *server = NFS_SERVER(inode);
4368        struct nfs4_exception exception = { };
4369        int err;
4370        do {
4371                err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4372                switch (err) {
4373                        case -NFS4ERR_STALE_STATEID:
4374                        case -NFS4ERR_EXPIRED:
4375                        case 0:
4376                                return 0;
4377                }
4378                err = nfs4_handle_exception(server, err, &exception);
4379        } while (exception.retry);
4380        return err;
4381}
4382
4383#define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4384#define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4385
4386/* 
4387 * sleep, with exponential backoff, and retry the LOCK operation. 
4388 */
4389static unsigned long
4390nfs4_set_lock_task_retry(unsigned long timeout)
4391{
4392        freezable_schedule_timeout_killable(timeout);
4393        timeout <<= 1;
4394        if (timeout > NFS4_LOCK_MAXTIMEOUT)
4395                return NFS4_LOCK_MAXTIMEOUT;
4396        return timeout;
4397}
4398
4399static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4400{
4401        struct inode *inode = state->inode;
4402        struct nfs_server *server = NFS_SERVER(inode);
4403        struct nfs_client *clp = server->nfs_client;
4404        struct nfs_lockt_args arg = {
4405                .fh = NFS_FH(inode),
4406                .fl = request,
4407        };
4408        struct nfs_lockt_res res = {
4409                .denied = request,
4410        };
4411        struct rpc_message msg = {
4412                .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4413                .rpc_argp       = &arg,
4414                .rpc_resp       = &res,
4415                .rpc_cred       = state->owner->so_cred,
4416        };
4417        struct nfs4_lock_state *lsp;
4418        int status;
4419
4420        arg.lock_owner.clientid = clp->cl_clientid;
4421        status = nfs4_set_lock_state(state, request);
4422        if (status != 0)
4423                goto out;
4424        lsp = request->fl_u.nfs4_fl.owner;
4425        arg.lock_owner.id = lsp->ls_seqid.owner_id;
4426        arg.lock_owner.s_dev = server->s_dev;
4427        status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4428        switch (status) {
4429                case 0:
4430                        request->fl_type = F_UNLCK;
4431                        break;
4432                case -NFS4ERR_DENIED:
4433                        status = 0;
4434        }
4435        request->fl_ops->fl_release_private(request);
4436out:
4437        return status;
4438}
4439
4440static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4441{
4442        struct nfs4_exception exception = { };
4443        int err;
4444
4445        do {
4446                err = nfs4_handle_exception(NFS_SERVER(state->inode),
4447                                _nfs4_proc_getlk(state, cmd, request),
4448                                &exception);
4449        } while (exception.retry);
4450        return err;
4451}
4452
4453static int do_vfs_lock(struct file *file, struct file_lock *fl)
4454{
4455        int res = 0;
4456        switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4457                case FL_POSIX:
4458                        res = posix_lock_file_wait(file, fl);
4459                        break;
4460                case FL_FLOCK:
4461                        res = flock_lock_file_wait(file, fl);
4462                        break;
4463                default:
4464                        BUG();
4465        }
4466        return res;
4467}
4468
4469struct nfs4_unlockdata {
4470        struct nfs_locku_args arg;
4471        struct nfs_locku_res res;
4472        struct nfs4_lock_state *lsp;
4473        struct nfs_open_context *ctx;
4474        struct file_lock fl;
4475        const struct nfs_server *server;
4476        unsigned long timestamp;
4477};
4478
4479static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4480                struct nfs_open_context *ctx,
4481                struct nfs4_lock_state *lsp,
4482                struct nfs_seqid *seqid)
4483{
4484        struct nfs4_unlockdata *p;
4485        struct inode *inode = lsp->ls_state->inode;
4486
4487        p = kzalloc(sizeof(*p), GFP_NOFS);
4488        if (p == NULL)
4489                return NULL;
4490        p->arg.fh = NFS_FH(inode);
4491        p->arg.fl = &p->fl;
4492        p->arg.seqid = seqid;
4493        p->res.seqid = seqid;
4494        p->arg.stateid = &lsp->ls_stateid;
4495        p->lsp = lsp;
4496        atomic_inc(&lsp->ls_count);
4497        /* Ensure we don't close file until we're done freeing locks! */
4498        p->ctx = get_nfs_open_context(ctx);
4499        memcpy(&p->fl, fl, sizeof(p->fl));
4500        p->server = NFS_SERVER(inode);
4501        return p;
4502}
4503
4504static void nfs4_locku_release_calldata(void *data)
4505{
4506        struct nfs4_unlockdata *calldata = data;
4507        nfs_free_seqid(calldata->arg.seqid);
4508        nfs4_put_lock_state(calldata->lsp);
4509        put_nfs_open_context(calldata->ctx);
4510        kfree(calldata);
4511}
4512
4513static void nfs4_locku_done(struct rpc_task *task, void *data)
4514{
4515        struct nfs4_unlockdata *calldata = data;
4516
4517        if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4518                return;
4519        switch (task->tk_status) {
4520                case 0:
4521                        nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4522                                        &calldata->res.stateid);
4523                        renew_lease(calldata->server, calldata->timestamp);
4524                        break;
4525                case -NFS4ERR_BAD_STATEID:
4526                case -NFS4ERR_OLD_STATEID:
4527                case -NFS4ERR_STALE_STATEID:
4528                case -NFS4ERR_EXPIRED:
4529                        break;
4530                default:
4531                        if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4532                                rpc_restart_call_prepare(task);
4533        }
4534        nfs_release_seqid(calldata->arg.seqid);
4535}
4536
4537static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4538{
4539        struct nfs4_unlockdata *calldata = data;
4540
4541        if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4542                return;
4543        if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4544                /* Note: exit _without_ running nfs4_locku_done */
4545                task->tk_action = NULL;
4546                return;
4547        }
4548        calldata->timestamp = jiffies;
4549        if (nfs4_setup_sequence(calldata->server,
4550                                &calldata->arg.seq_args,
4551                                &calldata->res.seq_res,
4552                                task) != 0)
4553                nfs_release_seqid(calldata->arg.seqid);
4554        else
4555                rpc_call_start(task);
4556}
4557
4558static const struct rpc_call_ops nfs4_locku_ops = {
4559        .rpc_call_prepare = nfs4_locku_prepare,
4560        .rpc_call_done = nfs4_locku_done,
4561        .rpc_release = nfs4_locku_release_calldata,
4562};
4563
4564static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4565                struct nfs_open_context *ctx,
4566                struct nfs4_lock_state *lsp,
4567                struct nfs_seqid *seqid)
4568{
4569        struct nfs4_unlockdata *data;
4570        struct rpc_message msg = {
4571                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4572                .rpc_cred = ctx->cred,
4573        };
4574        struct rpc_task_setup task_setup_data = {
4575                .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4576                .rpc_message = &msg,
4577                .callback_ops = &nfs4_locku_ops,
4578                .workqueue = nfsiod_workqueue,
4579                .flags = RPC_TASK_ASYNC,
4580        };
4581
4582        /* Ensure this is an unlock - when canceling a lock, the
4583         * canceled lock is passed in, and it won't be an unlock.
4584         */
4585        fl->fl_type = F_UNLCK;
4586
4587        data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4588        if (data == NULL) {
4589                nfs_free_seqid(seqid);
4590                return ERR_PTR(-ENOMEM);
4591        }
4592
4593        nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4594        msg.rpc_argp = &data->arg;
4595        msg.rpc_resp = &data->res;
4596        task_setup_data.callback_data = data;
4597        return rpc_run_task(&task_setup_data);
4598}
4599
4600static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4601{
4602        struct nfs_inode *nfsi = NFS_I(state->inode);
4603        struct nfs_seqid *seqid;
4604        struct nfs4_lock_state *lsp;
4605        struct rpc_task *task;
4606        int status = 0;
4607        unsigned char fl_flags = request->fl_flags;
4608
4609        status = nfs4_set_lock_state(state, request);
4610        /* Unlock _before_ we do the RPC call */
4611        request->fl_flags |= FL_EXISTS;
4612        down_read(&nfsi->rwsem);
4613        if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4614                up_read(&nfsi->rwsem);
4615                goto out;
4616        }
4617        up_read(&nfsi->rwsem);
4618        if (status != 0)
4619                goto out;
4620        /* Is this a delegated lock? */
4621        if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4622                goto out;
4623        lsp = request->fl_u.nfs4_fl.owner;
4624        seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4625        status = -ENOMEM;
4626        if (seqid == NULL)
4627                goto out;
4628        task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4629        status = PTR_ERR(task);
4630        if (IS_ERR(task))
4631                goto out;
4632        status = nfs4_wait_for_completion_rpc_task(task);
4633        rpc_put_task(task);
4634out:
4635        request->fl_flags = fl_flags;
4636        return status;
4637}
4638
4639struct nfs4_lockdata {
4640        struct nfs_lock_args arg;
4641        struct nfs_lock_res res;
4642        struct nfs4_lock_state *lsp;
4643        struct nfs_open_context *ctx;
4644        struct file_lock fl;
4645        unsigned long timestamp;
4646        int rpc_status;
4647        int cancelled;
4648        struct nfs_server *server;
4649};
4650
4651static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4652                struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4653                gfp_t gfp_mask)
4654{
4655        struct nfs4_lockdata *p;
4656        struct inode *inode = lsp->ls_state->inode;
4657        struct nfs_server *server = NFS_SERVER(inode);
4658
4659        p = kzalloc(sizeof(*p), gfp_mask);
4660        if (p == NULL)
4661                return NULL;
4662
4663        p->arg.fh = NFS_FH(inode);
4664        p->arg.fl = &p->fl;
4665        p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4666        if (p->arg.open_seqid == NULL)
4667                goto out_free;
4668        p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4669        if (p->arg.lock_seqid == NULL)
4670                goto out_free_seqid;
4671        p->arg.lock_stateid = &lsp->ls_stateid;
4672        p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4673        p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4674        p->arg.lock_owner.s_dev = server->s_dev;
4675        p->res.lock_seqid = p->arg.lock_seqid;
4676        p->lsp = lsp;
4677        p->server = server;
4678        atomic_inc(&lsp->ls_count);
4679        p->ctx = get_nfs_open_context(ctx);
4680        memcpy(&p->fl, fl, sizeof(p->fl));
4681        return p;
4682out_free_seqid:
4683        nfs_free_seqid(p->arg.open_seqid);
4684out_free:
4685        kfree(p);
4686        return NULL;
4687}
4688
4689static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4690{
4691        struct nfs4_lockdata *data = calldata;
4692        struct nfs4_state *state = data->lsp->ls_state;
4693
4694        dprintk("%s: begin!\n", __func__);
4695        if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4696                return;
4697        /* Do we need to do an open_to_lock_owner? */
4698        if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4699                if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4700                        goto out_release_lock_seqid;
4701                data->arg.open_stateid = &state->stateid;
4702                data->arg.new_lock_owner = 1;
4703                data->res.open_seqid = data->arg.open_seqid;
4704        } else
4705                data->arg.new_lock_owner = 0;
4706        data->timestamp = jiffies;
4707        if (nfs4_setup_sequence(data->server,
4708                                &data->arg.seq_args,
4709                                &data->res.seq_res,
4710                                task) == 0) {
4711                rpc_call_start(task);
4712                return;
4713        }
4714        nfs_release_seqid(data->arg.open_seqid);
4715out_release_lock_seqid:
4716        nfs_release_seqid(data->arg.lock_seqid);
4717        dprintk("%s: done!, ret = %d\n", __func__, task->tk_status);
4718}
4719
4720static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4721{
4722        rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4723        nfs4_lock_prepare(task, calldata);
4724}
4725
4726static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4727{
4728        struct nfs4_lockdata *data = calldata;
4729
4730        dprintk("%s: begin!\n", __func__);
4731
4732        if (!nfs4_sequence_done(task, &data->res.seq_res))
4733                return;
4734
4735        data->rpc_status = task->tk_status;
4736        if (data->arg.new_lock_owner != 0) {
4737                if (data->rpc_status == 0)
4738                        nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4739                else
4740                        goto out;
4741        }
4742        if (data->rpc_status == 0) {
4743                nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4744                set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4745                renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4746        }
4747out:
4748        dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4749}
4750
4751static void nfs4_lock_release(void *calldata)
4752{
4753        struct nfs4_lockdata *data = calldata;
4754
4755        dprintk("%s: begin!\n", __func__);
4756        nfs_free_seqid(data->arg.open_seqid);
4757        if (data->cancelled != 0) {
4758                struct rpc_task *task;
4759                task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4760                                data->arg.lock_seqid);
4761                if (!IS_ERR(task))
4762                        rpc_put_task_async(task);
4763                dprintk("%s: cancelling lock!\n", __func__);
4764        } else
4765                nfs_free_seqid(data->arg.lock_seqid);
4766        nfs4_put_lock_state(data->lsp);
4767        put_nfs_open_context(data->ctx);
4768        kfree(data);
4769        dprintk("%s: done!\n", __func__);
4770}
4771
4772static const struct rpc_call_ops nfs4_lock_ops = {
4773        .rpc_call_prepare = nfs4_lock_prepare,
4774        .rpc_call_done = nfs4_lock_done,
4775        .rpc_release = nfs4_lock_release,
4776};
4777
4778static const struct rpc_call_ops nfs4_recover_lock_ops = {
4779        .rpc_call_prepare = nfs4_recover_lock_prepare,
4780        .rpc_call_done = nfs4_lock_done,
4781        .rpc_release = nfs4_lock_release,
4782};
4783
4784static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4785{
4786        switch (error) {
4787        case -NFS4ERR_ADMIN_REVOKED:
4788        case -NFS4ERR_BAD_STATEID:
4789                lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4790                if (new_lock_owner != 0 ||
4791                   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4792                        nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4793                break;
4794        case -NFS4ERR_STALE_STATEID:
4795                lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4796        case -NFS4ERR_EXPIRED:
4797                nfs4_schedule_lease_recovery(server->nfs_client);
4798        };
4799}
4800
4801static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4802{
4803        struct nfs4_lockdata *data;
4804        struct rpc_task *task;
4805        struct rpc_message msg = {
4806                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4807                .rpc_cred = state->owner->so_cred,
4808        };
4809        struct rpc_task_setup task_setup_data = {
4810                .rpc_client = NFS_CLIENT(state->inode),
4811                .rpc_message = &msg,
4812                .callback_ops = &nfs4_lock_ops,
4813                .workqueue = nfsiod_workqueue,
4814                .flags = RPC_TASK_ASYNC,
4815        };
4816        int ret;
4817
4818        dprintk("%s: begin!\n", __func__);
4819        data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4820                        fl->fl_u.nfs4_fl.owner,
4821                        recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4822        if (data == NULL)
4823                return -ENOMEM;
4824        if (IS_SETLKW(cmd))
4825                data->arg.block = 1;
4826        if (recovery_type > NFS_LOCK_NEW) {
4827                if (recovery_type == NFS_LOCK_RECLAIM)
4828                        data->arg.reclaim = NFS_LOCK_RECLAIM;
4829                task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4830        }
4831        nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4832        msg.rpc_argp = &data->arg;
4833        msg.rpc_resp = &data->res;
4834        task_setup_data.callback_data = data;
4835        task = rpc_run_task(&task_setup_data);
4836        if (IS_ERR(task))
4837                return PTR_ERR(task);
4838        ret = nfs4_wait_for_completion_rpc_task(task);
4839        if (ret == 0) {
4840                ret = data->rpc_status;
4841                if (ret)
4842                        nfs4_handle_setlk_error(data->server, data->lsp,
4843                                        data->arg.new_lock_owner, ret);
4844        } else
4845                data->cancelled = 1;
4846        rpc_put_task(task);
4847        dprintk("%s: done, ret = %d!\n", __func__, ret);
4848        return ret;
4849}
4850
4851static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4852{
4853        struct nfs_server *server = NFS_SERVER(state->inode);
4854        struct nfs4_exception exception = {
4855                .inode = state->inode,
4856        };
4857        int err;
4858
4859        do {
4860                /* Cache the lock if possible... */
4861                if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4862                        return 0;
4863                err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4864                if (err != -NFS4ERR_DELAY)
4865                        break;
4866                nfs4_handle_exception(server, err, &exception);
4867        } while (exception.retry);
4868        return err;
4869}
4870
4871static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4872{
4873        struct nfs_server *server = NFS_SERVER(state->inode);
4874        struct nfs4_exception exception = {
4875                .inode = state->inode,
4876        };
4877        int err;
4878
4879        err = nfs4_set_lock_state(state, request);
4880        if (err != 0)
4881                return err;
4882        do {
4883                if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4884                        return 0;
4885                err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4886                switch (err) {
4887                default:
4888                        goto out;
4889                case -NFS4ERR_GRACE:
4890                case -NFS4ERR_DELAY:
4891                        nfs4_handle_exception(server, err, &exception);
4892                        err = 0;
4893                }
4894        } while (exception.retry);
4895out:
4896        return err;
4897}
4898
4899#if defined(CONFIG_NFS_V4_1)
4900/**
4901 * nfs41_check_expired_locks - possibly free a lock stateid
4902 *
4903 * @state: NFSv4 state for an inode
4904 *
4905 * Returns NFS_OK if recovery for this stateid is now finished.
4906 * Otherwise a negative NFS4ERR value is returned.
4907 */
4908static int nfs41_check_expired_locks(struct nfs4_state *state)
4909{
4910        int status, ret = -NFS4ERR_BAD_STATEID;
4911        struct nfs4_lock_state *lsp;
4912        struct nfs_server *server = NFS_SERVER(state->inode);
4913
4914        list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4915                if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
4916                        status = nfs41_test_stateid(server, &lsp->ls_stateid);
4917                        if (status != NFS_OK) {
4918                                /* Free the stateid unless the server
4919                                 * informs us the stateid is unrecognized. */
4920                                if (status != -NFS4ERR_BAD_STATEID)
4921                                        nfs41_free_stateid(server,
4922                                                        &lsp->ls_stateid);
4923                                clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
4924                                ret = status;
4925                        }
4926                }
4927        };
4928
4929        return ret;
4930}
4931
4932static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4933{
4934        int status = NFS_OK;
4935
4936        if (test_bit(LK_STATE_IN_USE, &state->flags))
4937                status = nfs41_check_expired_locks(state);
4938        if (status != NFS_OK)
4939                status = nfs4_lock_expired(state, request);
4940        return status;
4941}
4942#endif
4943
4944static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4945{
4946        struct nfs_inode *nfsi = NFS_I(state->inode);
4947        unsigned char fl_flags = request->fl_flags;
4948        int status = -ENOLCK;
4949
4950        if ((fl_flags & FL_POSIX) &&
4951                        !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4952                goto out;
4953        /* Is this a delegated open? */
4954        status = nfs4_set_lock_state(state, request);
4955        if (status != 0)
4956                goto out;
4957        request->fl_flags |= FL_ACCESS;
4958        status = do_vfs_lock(request->fl_file, request);
4959        if (status < 0)
4960                goto out;
4961        down_read(&nfsi->rwsem);
4962        if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4963                /* Yes: cache locks! */
4964                /* ...but avoid races with delegation recall... */
4965                request->fl_flags = fl_flags & ~FL_SLEEP;
4966                status = do_vfs_lock(request->fl_file, request);
4967                goto out_unlock;
4968        }
4969        status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4970        if (status != 0)
4971                goto out_unlock;
4972        /* Note: we always want to sleep here! */
4973        request->fl_flags = fl_flags | FL_SLEEP;
4974        if (do_vfs_lock(request->fl_file, request) < 0)
4975                printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4976                        "manager!\n", __func__);
4977out_unlock:
4978        up_read(&nfsi->rwsem);
4979out:
4980        request->fl_flags = fl_flags;
4981        return status;
4982}
4983
4984static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4985{
4986        struct nfs4_exception exception = {
4987                .state = state,
4988                .inode = state->inode,
4989        };
4990        int err;
4991
4992        do {
4993                err = _nfs4_proc_setlk(state, cmd, request);
4994                if (err == -NFS4ERR_DENIED)
4995                        err = -EAGAIN;
4996                err = nfs4_handle_exception(NFS_SERVER(state->inode),
4997                                err, &exception);
4998        } while (exception.retry);
4999        return err;
5000}
5001
5002static int
5003nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5004{
5005        struct nfs_open_context *ctx;
5006        struct nfs4_state *state;
5007        unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5008        int status;
5009
5010        /* verify open state */
5011        ctx = nfs_file_open_context(filp);
5012        state = ctx->state;
5013
5014        if (request->fl_start < 0 || request->fl_end < 0)
5015                return -EINVAL;
5016
5017        if (IS_GETLK(cmd)) {
5018                if (state != NULL)
5019                        return nfs4_proc_getlk(state, F_GETLK, request);
5020                return 0;
5021        }
5022
5023        if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5024                return -EINVAL;
5025
5026        if (request->fl_type == F_UNLCK) {
5027                if (state != NULL)
5028                        return nfs4_proc_unlck(state, cmd, request);
5029                return 0;
5030        }
5031
5032        if (state == NULL)
5033                return -ENOLCK;
5034        /*
5035         * Don't rely on the VFS having checked the file open mode,
5036         * since it won't do this for flock() locks.
5037         */
5038        switch (request->fl_type) {
5039        case F_RDLCK:
5040                if (!(filp->