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        default:
 108                dprintk("%s could not handle NFSv4 error %d\n",
 109                                __func__, -err);
 110                break;
 111        }
 112        return -EIO;
 113}
 114
 115/*
 116 * This is our standard bitmap for GETATTR requests.
 117 */
 118const u32 nfs4_fattr_bitmap[3] = {
 119        FATTR4_WORD0_TYPE
 120        | FATTR4_WORD0_CHANGE
 121        | FATTR4_WORD0_SIZE
 122        | FATTR4_WORD0_FSID
 123        | FATTR4_WORD0_FILEID,
 124        FATTR4_WORD1_MODE
 125        | FATTR4_WORD1_NUMLINKS
 126        | FATTR4_WORD1_OWNER
 127        | FATTR4_WORD1_OWNER_GROUP
 128        | FATTR4_WORD1_RAWDEV
 129        | FATTR4_WORD1_SPACE_USED
 130        | FATTR4_WORD1_TIME_ACCESS
 131        | FATTR4_WORD1_TIME_METADATA
 132        | FATTR4_WORD1_TIME_MODIFY
 133};
 134
 135static const u32 nfs4_pnfs_open_bitmap[3] = {
 136        FATTR4_WORD0_TYPE
 137        | FATTR4_WORD0_CHANGE
 138        | FATTR4_WORD0_SIZE
 139        | FATTR4_WORD0_FSID
 140        | FATTR4_WORD0_FILEID,
 141        FATTR4_WORD1_MODE
 142        | FATTR4_WORD1_NUMLINKS
 143        | FATTR4_WORD1_OWNER
 144        | FATTR4_WORD1_OWNER_GROUP
 145        | FATTR4_WORD1_RAWDEV
 146        | FATTR4_WORD1_SPACE_USED
 147        | FATTR4_WORD1_TIME_ACCESS
 148        | FATTR4_WORD1_TIME_METADATA
 149        | FATTR4_WORD1_TIME_MODIFY,
 150        FATTR4_WORD2_MDSTHRESHOLD
 151};
 152
 153const u32 nfs4_statfs_bitmap[2] = {
 154        FATTR4_WORD0_FILES_AVAIL
 155        | FATTR4_WORD0_FILES_FREE
 156        | FATTR4_WORD0_FILES_TOTAL,
 157        FATTR4_WORD1_SPACE_AVAIL
 158        | FATTR4_WORD1_SPACE_FREE
 159        | FATTR4_WORD1_SPACE_TOTAL
 160};
 161
 162const u32 nfs4_pathconf_bitmap[2] = {
 163        FATTR4_WORD0_MAXLINK
 164        | FATTR4_WORD0_MAXNAME,
 165        0
 166};
 167
 168const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 169                        | FATTR4_WORD0_MAXREAD
 170                        | FATTR4_WORD0_MAXWRITE
 171                        | FATTR4_WORD0_LEASE_TIME,
 172                        FATTR4_WORD1_TIME_DELTA
 173                        | FATTR4_WORD1_FS_LAYOUT_TYPES,
 174                        FATTR4_WORD2_LAYOUT_BLKSIZE
 175};
 176
 177const u32 nfs4_fs_locations_bitmap[2] = {
 178        FATTR4_WORD0_TYPE
 179        | FATTR4_WORD0_CHANGE
 180        | FATTR4_WORD0_SIZE
 181        | FATTR4_WORD0_FSID
 182        | FATTR4_WORD0_FILEID
 183        | FATTR4_WORD0_FS_LOCATIONS,
 184        FATTR4_WORD1_MODE
 185        | FATTR4_WORD1_NUMLINKS
 186        | FATTR4_WORD1_OWNER
 187        | FATTR4_WORD1_OWNER_GROUP
 188        | FATTR4_WORD1_RAWDEV
 189        | FATTR4_WORD1_SPACE_USED
 190        | FATTR4_WORD1_TIME_ACCESS
 191        | FATTR4_WORD1_TIME_METADATA
 192        | FATTR4_WORD1_TIME_MODIFY
 193        | FATTR4_WORD1_MOUNTED_ON_FILEID
 194};
 195
 196static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 197                struct nfs4_readdir_arg *readdir)
 198{
 199        __be32 *start, *p;
 200
 201        BUG_ON(readdir->count < 80);
 202        if (cookie > 2) {
 203                readdir->cookie = cookie;
 204                memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 205                return;
 206        }
 207
 208        readdir->cookie = 0;
 209        memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 210        if (cookie == 2)
 211                return;
 212        
 213        /*
 214         * NFSv4 servers do not return entries for '.' and '..'
 215         * Therefore, we fake these entries here.  We let '.'
 216         * have cookie 0 and '..' have cookie 1.  Note that
 217         * when talking to the server, we always send cookie 0
 218         * instead of 1 or 2.
 219         */
 220        start = p = kmap_atomic(*readdir->pages);
 221        
 222        if (cookie == 0) {
 223                *p++ = xdr_one;                                  /* next */
 224                *p++ = xdr_zero;                   /* cookie, first word */
 225                *p++ = xdr_one;                   /* cookie, second word */
 226                *p++ = xdr_one;                             /* entry len */
 227                memcpy(p, ".\0\0\0", 4);                        /* entry */
 228                p++;
 229                *p++ = xdr_one;                         /* bitmap length */
 230                *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 231                *p++ = htonl(8);              /* attribute buffer length */
 232                p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 233        }
 234        
 235        *p++ = xdr_one;                                  /* next */
 236        *p++ = xdr_zero;                   /* cookie, first word */
 237        *p++ = xdr_two;                   /* cookie, second word */
 238        *p++ = xdr_two;                             /* entry len */
 239        memcpy(p, "..\0\0", 4);                         /* entry */
 240        p++;
 241        *p++ = xdr_one;                         /* bitmap length */
 242        *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 243        *p++ = htonl(8);              /* attribute buffer length */
 244        p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 245
 246        readdir->pgbase = (char *)p - (char *)start;
 247        readdir->count -= readdir->pgbase;
 248        kunmap_atomic(start);
 249}
 250
 251static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 252{
 253        int res;
 254
 255        might_sleep();
 256
 257        res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 258                        nfs_wait_bit_killable, TASK_KILLABLE);
 259        if (res)
 260                return res;
 261
 262        if (clp->cl_cons_state < 0)
 263                return clp->cl_cons_state;
 264        return 0;
 265}
 266
 267static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 268{
 269        int res = 0;
 270
 271        might_sleep();
 272
 273        if (*timeout <= 0)
 274                *timeout = NFS4_POLL_RETRY_MIN;
 275        if (*timeout > NFS4_POLL_RETRY_MAX)
 276                *timeout = NFS4_POLL_RETRY_MAX;
 277        freezable_schedule_timeout_killable(*timeout);
 278        if (fatal_signal_pending(current))
 279                res = -ERESTARTSYS;
 280        *timeout <<= 1;
 281        return res;
 282}
 283
 284/* This is the error handling routine for processes that are allowed
 285 * to sleep.
 286 */
 287static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 288{
 289        struct nfs_client *clp = server->nfs_client;
 290        struct nfs4_state *state = exception->state;
 291        struct inode *inode = exception->inode;
 292        int ret = errorcode;
 293
 294        exception->retry = 0;
 295        switch(errorcode) {
 296                case 0:
 297                        return 0;
 298                case -NFS4ERR_OPENMODE:
 299                        if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 300                                nfs4_inode_return_delegation(inode);
 301                                exception->retry = 1;
 302                                return 0;
 303                        }
 304                        if (state == NULL)
 305                                break;
 306                        nfs4_schedule_stateid_recovery(server, state);
 307                        goto wait_on_recovery;
 308                case -NFS4ERR_DELEG_REVOKED:
 309                case -NFS4ERR_ADMIN_REVOKED:
 310                case -NFS4ERR_BAD_STATEID:
 311                        if (state == NULL)
 312                                break;
 313                        nfs_remove_bad_delegation(state->inode);
 314                        nfs4_schedule_stateid_recovery(server, state);
 315                        goto wait_on_recovery;
 316                case -NFS4ERR_EXPIRED:
 317                        if (state != NULL)
 318                                nfs4_schedule_stateid_recovery(server, state);
 319                case -NFS4ERR_STALE_STATEID:
 320                case -NFS4ERR_STALE_CLIENTID:
 321                        nfs4_schedule_lease_recovery(clp);
 322                        goto wait_on_recovery;
 323#if defined(CONFIG_NFS_V4_1)
 324                case -NFS4ERR_BADSESSION:
 325                case -NFS4ERR_BADSLOT:
 326                case -NFS4ERR_BAD_HIGH_SLOT:
 327                case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 328                case -NFS4ERR_DEADSESSION:
 329                case -NFS4ERR_SEQ_FALSE_RETRY:
 330                case -NFS4ERR_SEQ_MISORDERED:
 331                        dprintk("%s ERROR: %d Reset session\n", __func__,
 332                                errorcode);
 333                        nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 334                        exception->retry = 1;
 335                        break;
 336#endif /* defined(CONFIG_NFS_V4_1) */
 337                case -NFS4ERR_FILE_OPEN:
 338                        if (exception->timeout > HZ) {
 339                                /* We have retried a decent amount, time to
 340                                 * fail
 341                                 */
 342                                ret = -EBUSY;
 343                                break;
 344                        }
 345                case -NFS4ERR_GRACE:
 346                case -NFS4ERR_DELAY:
 347                case -EKEYEXPIRED:
 348                        ret = nfs4_delay(server->client, &exception->timeout);
 349                        if (ret != 0)
 350                                break;
 351                case -NFS4ERR_RETRY_UNCACHED_REP:
 352                case -NFS4ERR_OLD_STATEID:
 353                        exception->retry = 1;
 354                        break;
 355                case -NFS4ERR_BADOWNER:
 356                        /* The following works around a Linux server bug! */
 357                case -NFS4ERR_BADNAME:
 358                        if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 359                                server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 360                                exception->retry = 1;
 361                                printk(KERN_WARNING "NFS: v4 server %s "
 362                                                "does not accept raw "
 363                                                "uid/gids. "
 364                                                "Reenabling the idmapper.\n",
 365                                                server->nfs_client->cl_hostname);
 366                        }
 367        }
 368        /* We failed to handle the error */
 369        return nfs4_map_errors(ret);
 370wait_on_recovery:
 371        ret = nfs4_wait_clnt_recover(clp);
 372        if (ret == 0)
 373                exception->retry = 1;
 374        return ret;
 375}
 376
 377
 378static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 379{
 380        spin_lock(&clp->cl_lock);
 381        if (time_before(clp->cl_last_renewal,timestamp))
 382                clp->cl_last_renewal = timestamp;
 383        spin_unlock(&clp->cl_lock);
 384}
 385
 386static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 387{
 388        do_renew_lease(server->nfs_client, timestamp);
 389}
 390
 391#if defined(CONFIG_NFS_V4_1)
 392
 393/*
 394 * nfs4_free_slot - free a slot and efficiently update slot table.
 395 *
 396 * freeing a slot is trivially done by clearing its respective bit
 397 * in the bitmap.
 398 * If the freed slotid equals highest_used_slotid we want to update it
 399 * so that the server would be able to size down the slot table if needed,
 400 * otherwise we know that the highest_used_slotid is still in use.
 401 * When updating highest_used_slotid there may be "holes" in the bitmap
 402 * so we need to scan down from highest_used_slotid to 0 looking for the now
 403 * highest slotid in use.
 404 * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
 405 *
 406 * Must be called while holding tbl->slot_tbl_lock
 407 */
 408static void
 409nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
 410{
 411        BUG_ON(slotid >= NFS4_MAX_SLOT_TABLE);
 412        /* clear used bit in bitmap */
 413        __clear_bit(slotid, tbl->used_slots);
 414
 415        /* update highest_used_slotid when it is freed */
 416        if (slotid == tbl->highest_used_slotid) {
 417                slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 418                if (slotid < tbl->max_slots)
 419                        tbl->highest_used_slotid = slotid;
 420                else
 421                        tbl->highest_used_slotid = NFS4_NO_SLOT;
 422        }
 423        dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
 424                slotid, tbl->highest_used_slotid);
 425}
 426
 427bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
 428{
 429        rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 430        return true;
 431}
 432
 433/*
 434 * Signal state manager thread if session fore channel is drained
 435 */
 436static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 437{
 438        if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 439                rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
 440                                nfs4_set_task_privileged, NULL);
 441                return;
 442        }
 443
 444        if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 445                return;
 446
 447        dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 448        complete(&ses->fc_slot_table.complete);
 449}
 450
 451/*
 452 * Signal state manager thread if session back channel is drained
 453 */
 454void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 455{
 456        if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 457            ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 458                return;
 459        dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 460        complete(&ses->bc_slot_table.complete);
 461}
 462
 463static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 464{
 465        struct nfs4_slot_table *tbl;
 466
 467        tbl = &res->sr_session->fc_slot_table;
 468        if (!res->sr_slot) {
 469                /* just wake up the next guy waiting since
 470                 * we may have not consumed a slot after all */
 471                dprintk("%s: No slot\n", __func__);
 472                return;
 473        }
 474
 475        spin_lock(&tbl->slot_tbl_lock);
 476        nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 477        nfs4_check_drain_fc_complete(res->sr_session);
 478        spin_unlock(&tbl->slot_tbl_lock);
 479        res->sr_slot = NULL;
 480}
 481
 482static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 483{
 484        unsigned long timestamp;
 485        struct nfs_client *clp;
 486
 487        /*
 488         * sr_status remains 1 if an RPC level error occurred. The server
 489         * may or may not have processed the sequence operation..
 490         * Proceed as if the server received and processed the sequence
 491         * operation.
 492         */
 493        if (res->sr_status == 1)
 494                res->sr_status = NFS_OK;
 495
 496        /* don't increment the sequence number if the task wasn't sent */
 497        if (!RPC_WAS_SENT(task))
 498                goto out;
 499
 500        /* Check the SEQUENCE operation status */
 501        switch (res->sr_status) {
 502        case 0:
 503                /* Update the slot's sequence and clientid lease timer */
 504                ++res->sr_slot->seq_nr;
 505                timestamp = res->sr_renewal_time;
 506                clp = res->sr_session->clp;
 507                do_renew_lease(clp, timestamp);
 508                /* Check sequence flags */
 509                if (res->sr_status_flags != 0)
 510                        nfs4_schedule_lease_recovery(clp);
 511                break;
 512        case -NFS4ERR_DELAY:
 513                /* The server detected a resend of the RPC call and
 514                 * returned NFS4ERR_DELAY as per Section 2.10.6.2
 515                 * of RFC5661.
 516                 */
 517                dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 518                        __func__,
 519                        res->sr_slot - res->sr_session->fc_slot_table.slots,
 520                        res->sr_slot->seq_nr);
 521                goto out_retry;
 522        default:
 523                /* Just update the slot sequence no. */
 524                ++res->sr_slot->seq_nr;
 525        }
 526out:
 527        /* The session may be reset by one of the error handlers. */
 528        dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 529        nfs41_sequence_free_slot(res);
 530        return 1;
 531out_retry:
 532        if (!rpc_restart_call(task))
 533                goto out;
 534        rpc_delay(task, NFS4_POLL_RETRY_MAX);
 535        return 0;
 536}
 537
 538static int nfs4_sequence_done(struct rpc_task *task,
 539                               struct nfs4_sequence_res *res)
 540{
 541        if (res->sr_session == NULL)
 542                return 1;
 543        return nfs41_sequence_done(task, res);
 544}
 545
 546/*
 547 * nfs4_find_slot - efficiently look for a free slot
 548 *
 549 * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 550 * If found, we mark the slot as used, update the highest_used_slotid,
 551 * and respectively set up the sequence operation args.
 552 * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
 553 *
 554 * Note: must be called with under the slot_tbl_lock.
 555 */
 556static u32
 557nfs4_find_slot(struct nfs4_slot_table *tbl)
 558{
 559        u32 slotid;
 560        u32 ret_id = NFS4_NO_SLOT;
 561
 562        dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
 563                __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 564                tbl->max_slots);
 565        slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 566        if (slotid >= tbl->max_slots)
 567                goto out;
 568        __set_bit(slotid, tbl->used_slots);
 569        if (slotid > tbl->highest_used_slotid ||
 570                        tbl->highest_used_slotid == NFS4_NO_SLOT)
 571                tbl->highest_used_slotid = slotid;
 572        ret_id = slotid;
 573out:
 574        dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 575                __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 576        return ret_id;
 577}
 578
 579static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 580                struct nfs4_sequence_res *res, int cache_reply)
 581{
 582        args->sa_session = NULL;
 583        args->sa_cache_this = 0;
 584        if (cache_reply)
 585                args->sa_cache_this = 1;
 586        res->sr_session = NULL;
 587        res->sr_slot = NULL;
 588}
 589
 590int nfs41_setup_sequence(struct nfs4_session *session,
 591                                struct nfs4_sequence_args *args,
 592                                struct nfs4_sequence_res *res,
 593                                struct rpc_task *task)
 594{
 595        struct nfs4_slot *slot;
 596        struct nfs4_slot_table *tbl;
 597        u32 slotid;
 598
 599        dprintk("--> %s\n", __func__);
 600        /* slot already allocated? */
 601        if (res->sr_slot != NULL)
 602                return 0;
 603
 604        tbl = &session->fc_slot_table;
 605
 606        spin_lock(&tbl->slot_tbl_lock);
 607        if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 608            !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 609                /* The state manager will wait until the slot table is empty */
 610                rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 611                spin_unlock(&tbl->slot_tbl_lock);
 612                dprintk("%s session is draining\n", __func__);
 613                return -EAGAIN;
 614        }
 615
 616        if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 617            !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 618                rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 619                spin_unlock(&tbl->slot_tbl_lock);
 620                dprintk("%s enforce FIFO order\n", __func__);
 621                return -EAGAIN;
 622        }
 623
 624        slotid = nfs4_find_slot(tbl);
 625        if (slotid == NFS4_NO_SLOT) {
 626                rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 627                spin_unlock(&tbl->slot_tbl_lock);
 628                dprintk("<-- %s: no free slots\n", __func__);
 629                return -EAGAIN;
 630        }
 631        spin_unlock(&tbl->slot_tbl_lock);
 632
 633        rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 634        slot = tbl->slots + slotid;
 635        args->sa_session = session;
 636        args->sa_slotid = slotid;
 637
 638        dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 639
 640        res->sr_session = session;
 641        res->sr_slot = slot;
 642        res->sr_renewal_time = jiffies;
 643        res->sr_status_flags = 0;
 644        /*
 645         * sr_status is only set in decode_sequence, and so will remain
 646         * set to 1 if an rpc level failure occurs.
 647         */
 648        res->sr_status = 1;
 649        return 0;
 650}
 651EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 652
 653int nfs4_setup_sequence(const struct nfs_server *server,
 654                        struct nfs4_sequence_args *args,
 655                        struct nfs4_sequence_res *res,
 656                        struct rpc_task *task)
 657{
 658        struct nfs4_session *session = nfs4_get_session(server);
 659        int ret = 0;
 660
 661        if (session == NULL)
 662                goto out;
 663
 664        dprintk("--> %s clp %p session %p sr_slot %td\n",
 665                __func__, session->clp, session, res->sr_slot ?
 666                        res->sr_slot - session->fc_slot_table.slots : -1);
 667
 668        ret = nfs41_setup_sequence(session, args, res, task);
 669out:
 670        dprintk("<-- %s status=%d\n", __func__, ret);
 671        return ret;
 672}
 673
 674struct nfs41_call_sync_data {
 675        const struct nfs_server *seq_server;
 676        struct nfs4_sequence_args *seq_args;
 677        struct nfs4_sequence_res *seq_res;
 678};
 679
 680static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 681{
 682        struct nfs41_call_sync_data *data = calldata;
 683
 684        dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 685
 686        if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 687                                data->seq_res, task))
 688                return;
 689        rpc_call_start(task);
 690}
 691
 692static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 693{
 694        rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 695        nfs41_call_sync_prepare(task, calldata);
 696}
 697
 698static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 699{
 700        struct nfs41_call_sync_data *data = calldata;
 701
 702        nfs41_sequence_done(task, data->seq_res);
 703}
 704
 705static const struct rpc_call_ops nfs41_call_sync_ops = {
 706        .rpc_call_prepare = nfs41_call_sync_prepare,
 707        .rpc_call_done = nfs41_call_sync_done,
 708};
 709
 710static const struct rpc_call_ops nfs41_call_priv_sync_ops = {
 711        .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 712        .rpc_call_done = nfs41_call_sync_done,
 713};
 714
 715static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 716                                   struct nfs_server *server,
 717                                   struct rpc_message *msg,
 718                                   struct nfs4_sequence_args *args,
 719                                   struct nfs4_sequence_res *res,
 720                                   int privileged)
 721{
 722        int ret;
 723        struct rpc_task *task;
 724        struct nfs41_call_sync_data data = {
 725                .seq_server = server,
 726                .seq_args = args,
 727                .seq_res = res,
 728        };
 729        struct rpc_task_setup task_setup = {
 730                .rpc_client = clnt,
 731                .rpc_message = msg,
 732                .callback_ops = &nfs41_call_sync_ops,
 733                .callback_data = &data
 734        };
 735
 736        if (privileged)
 737                task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 738        task = rpc_run_task(&task_setup);
 739        if (IS_ERR(task))
 740                ret = PTR_ERR(task);
 741        else {
 742                ret = task->tk_status;
 743                rpc_put_task(task);
 744        }
 745        return ret;
 746}
 747
 748int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 749                            struct nfs_server *server,
 750                            struct rpc_message *msg,
 751                            struct nfs4_sequence_args *args,
 752                            struct nfs4_sequence_res *res,
 753                            int cache_reply)
 754{
 755        nfs41_init_sequence(args, res, cache_reply);
 756        return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
 757}
 758
 759#else
 760static inline
 761void nfs41_init_sequence(struct nfs4_sequence_args *args,
 762                struct nfs4_sequence_res *res, int cache_reply)
 763{
 764}
 765
 766static int nfs4_sequence_done(struct rpc_task *task,
 767                               struct nfs4_sequence_res *res)
 768{
 769        return 1;
 770}
 771#endif /* CONFIG_NFS_V4_1 */
 772
 773int _nfs4_call_sync(struct rpc_clnt *clnt,
 774                    struct nfs_server *server,
 775                    struct rpc_message *msg,
 776                    struct nfs4_sequence_args *args,
 777                    struct nfs4_sequence_res *res,
 778                    int cache_reply)
 779{
 780        nfs41_init_sequence(args, res, cache_reply);
 781        return rpc_call_sync(clnt, msg, 0);
 782}
 783
 784static inline
 785int nfs4_call_sync(struct rpc_clnt *clnt,
 786                   struct nfs_server *server,
 787                   struct rpc_message *msg,
 788                   struct nfs4_sequence_args *args,
 789                   struct nfs4_sequence_res *res,
 790                   int cache_reply)
 791{
 792        return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 793                                                args, res, cache_reply);
 794}
 795
 796static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 797{
 798        struct nfs_inode *nfsi = NFS_I(dir);
 799
 800        spin_lock(&dir->i_lock);
 801        nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 802        if (!cinfo->atomic || cinfo->before != dir->i_version)
 803                nfs_force_lookup_revalidate(dir);
 804        dir->i_version = cinfo->after;
 805        spin_unlock(&dir->i_lock);
 806}
 807
 808struct nfs4_opendata {
 809        struct kref kref;
 810        struct nfs_openargs o_arg;
 811        struct nfs_openres o_res;
 812        struct nfs_open_confirmargs c_arg;
 813        struct nfs_open_confirmres c_res;
 814        struct nfs4_string owner_name;
 815        struct nfs4_string group_name;
 816        struct nfs_fattr f_attr;
 817        struct dentry *dir;
 818        struct dentry *dentry;
 819        struct nfs4_state_owner *owner;
 820        struct nfs4_state *state;
 821        struct iattr attrs;
 822        unsigned long timestamp;
 823        unsigned int rpc_done : 1;
 824        int rpc_status;
 825        int cancelled;
 826};
 827
 828
 829static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 830{
 831        p->o_res.f_attr = &p->f_attr;
 832        p->o_res.seqid = p->o_arg.seqid;
 833        p->c_res.seqid = p->c_arg.seqid;
 834        p->o_res.server = p->o_arg.server;
 835        nfs_fattr_init(&p->f_attr);
 836        nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 837}
 838
 839static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 840                struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 841                const struct iattr *attrs,
 842                gfp_t gfp_mask)
 843{
 844        struct dentry *parent = dget_parent(dentry);
 845        struct inode *dir = parent->d_inode;
 846        struct nfs_server *server = NFS_SERVER(dir);
 847        struct nfs4_opendata *p;
 848
 849        p = kzalloc(sizeof(*p), gfp_mask);
 850        if (p == NULL)
 851                goto err;
 852        p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 853        if (p->o_arg.seqid == NULL)
 854                goto err_free;
 855        nfs_sb_active(dentry->d_sb);
 856        p->dentry = dget(dentry);
 857        p->dir = parent;
 858        p->owner = sp;
 859        atomic_inc(&sp->so_count);
 860        p->o_arg.fh = NFS_FH(dir);
 861        p->o_arg.open_flags = flags;
 862        p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 863        p->o_arg.clientid = server->nfs_client->cl_clientid;
 864        p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
 865        p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 866        p->o_arg.name = &dentry->d_name;
 867        p->o_arg.server = server;
 868        p->o_arg.bitmask = server->attr_bitmask;
 869        p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
 870        p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 871        if (attrs != NULL && attrs->ia_valid != 0) {
 872                __be32 verf[2];
 873
 874                p->o_arg.u.attrs = &p->attrs;
 875                memcpy(&p->attrs, attrs, sizeof(p->attrs));
 876
 877                verf[0] = jiffies;
 878                verf[1] = current->pid;
 879                memcpy(p->o_arg.u.verifier.data, verf,
 880                                sizeof(p->o_arg.u.verifier.data));
 881        }
 882        p->c_arg.fh = &p->o_res.fh;
 883        p->c_arg.stateid = &p->o_res.stateid;
 884        p->c_arg.seqid = p->o_arg.seqid;
 885        nfs4_init_opendata_res(p);
 886        kref_init(&p->kref);
 887        return p;
 888err_free:
 889        kfree(p);
 890err:
 891        dput(parent);
 892        return NULL;
 893}
 894
 895static void nfs4_opendata_free(struct kref *kref)
 896{
 897        struct nfs4_opendata *p = container_of(kref,
 898                        struct nfs4_opendata, kref);
 899        struct super_block *sb = p->dentry->d_sb;
 900
 901        nfs_free_seqid(p->o_arg.seqid);
 902        if (p->state != NULL)
 903                nfs4_put_open_state(p->state);
 904        nfs4_put_state_owner(p->owner);
 905        dput(p->dir);
 906        dput(p->dentry);
 907        nfs_sb_deactive(sb);
 908        nfs_fattr_free_names(&p->f_attr);
 909        kfree(p);
 910}
 911
 912static void nfs4_opendata_put(struct nfs4_opendata *p)
 913{
 914        if (p != NULL)
 915                kref_put(&p->kref, nfs4_opendata_free);
 916}
 917
 918static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 919{
 920        int ret;
 921
 922        ret = rpc_wait_for_completion_task(task);
 923        return ret;
 924}
 925
 926static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 927{
 928        int ret = 0;
 929
 930        if (open_mode & (O_EXCL|O_TRUNC))
 931                goto out;
 932        switch (mode & (FMODE_READ|FMODE_WRITE)) {
 933                case FMODE_READ:
 934                        ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 935                                && state->n_rdonly != 0;
 936                        break;
 937                case FMODE_WRITE:
 938                        ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 939                                && state->n_wronly != 0;
 940                        break;
 941                case FMODE_READ|FMODE_WRITE:
 942                        ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 943                                && state->n_rdwr != 0;
 944        }
 945out:
 946        return ret;
 947}
 948
 949static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 950{
 951        if (delegation == NULL)
 952                return 0;
 953        if ((delegation->type & fmode) != fmode)
 954                return 0;
 955        if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 956                return 0;
 957        nfs_mark_delegation_referenced(delegation);
 958        return 1;
 959}
 960
 961static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 962{
 963        switch (fmode) {
 964                case FMODE_WRITE:
 965                        state->n_wronly++;
 966                        break;
 967                case FMODE_READ:
 968                        state->n_rdonly++;
 969                        break;
 970                case FMODE_READ|FMODE_WRITE:
 971                        state->n_rdwr++;
 972        }
 973        nfs4_state_set_mode_locked(state, state->state | fmode);
 974}
 975
 976static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 977{
 978        if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 979                nfs4_stateid_copy(&state->stateid, stateid);
 980        nfs4_stateid_copy(&state->open_stateid, stateid);
 981        switch (fmode) {
 982                case FMODE_READ:
 983                        set_bit(NFS_O_RDONLY_STATE, &state->flags);
 984                        break;
 985                case FMODE_WRITE:
 986                        set_bit(NFS_O_WRONLY_STATE, &state->flags);
 987                        break;
 988                case FMODE_READ|FMODE_WRITE:
 989                        set_bit(NFS_O_RDWR_STATE, &state->flags);
 990        }
 991}
 992
 993static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 994{
 995        write_seqlock(&state->seqlock);
 996        nfs_set_open_stateid_locked(state, stateid, fmode);
 997        write_sequnlock(&state->seqlock);
 998}
 999
1000static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1001{
1002        /*
1003         * Protect the call to nfs4_state_set_mode_locked and
1004         * serialise the stateid update
1005         */
1006        write_seqlock(&state->seqlock);
1007        if (deleg_stateid != NULL) {
1008                nfs4_stateid_copy(&state->stateid, deleg_stateid);
1009                set_bit(NFS_DELEGATED_STATE, &state->flags);
1010        }
1011        if (open_stateid != NULL)
1012                nfs_set_open_stateid_locked(state, open_stateid, fmode);
1013        write_sequnlock(&state->seqlock);
1014        spin_lock(&state->owner->so_lock);
1015        update_open_stateflags(state, fmode);
1016        spin_unlock(&state->owner->so_lock);
1017}
1018
1019static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1020{
1021        struct nfs_inode *nfsi = NFS_I(state->inode);
1022        struct nfs_delegation *deleg_cur;
1023        int ret = 0;
1024
1025        fmode &= (FMODE_READ|FMODE_WRITE);
1026
1027        rcu_read_lock();
1028        deleg_cur = rcu_dereference(nfsi->delegation);
1029        if (deleg_cur == NULL)
1030                goto no_delegation;
1031
1032        spin_lock(&deleg_cur->lock);
1033        if (nfsi->delegation != deleg_cur ||
1034            (deleg_cur->type & fmode) != fmode)
1035                goto no_delegation_unlock;
1036
1037        if (delegation == NULL)
1038                delegation = &deleg_cur->stateid;
1039        else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1040                goto no_delegation_unlock;
1041
1042        nfs_mark_delegation_referenced(deleg_cur);
1043        __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1044        ret = 1;
1045no_delegation_unlock:
1046        spin_unlock(&deleg_cur->lock);
1047no_delegation:
1048        rcu_read_unlock();
1049
1050        if (!ret && open_stateid != NULL) {
1051                __update_open_stateid(state, open_stateid, NULL, fmode);
1052                ret = 1;
1053        }
1054
1055        return ret;
1056}
1057
1058
1059static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1060{
1061        struct nfs_delegation *delegation;
1062
1063        rcu_read_lock();
1064        delegation = rcu_dereference(NFS_I(inode)->delegation);
1065        if (delegation == NULL || (delegation->type & fmode) == fmode) {
1066                rcu_read_unlock();
1067                return;
1068        }
1069        rcu_read_unlock();
1070        nfs4_inode_return_delegation(inode);
1071}
1072
1073static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1074{
1075        struct nfs4_state *state = opendata->state;
1076        struct nfs_inode *nfsi = NFS_I(state->inode);
1077        struct nfs_delegation *delegation;
1078        int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1079        fmode_t fmode = opendata->o_arg.fmode;
1080        nfs4_stateid stateid;
1081        int ret = -EAGAIN;
1082
1083        for (;;) {
1084                if (can_open_cached(state, fmode, open_mode)) {
1085                        spin_lock(&state->owner->so_lock);
1086                        if (can_open_cached(state, fmode, open_mode)) {
1087                                update_open_stateflags(state, fmode);
1088                                spin_unlock(&state->owner->so_lock);
1089                                goto out_return_state;
1090                        }
1091                        spin_unlock(&state->owner->so_lock);
1092                }
1093                rcu_read_lock();
1094                delegation = rcu_dereference(nfsi->delegation);
1095                if (!can_open_delegated(delegation, fmode)) {
1096                        rcu_read_unlock();
1097                        break;
1098                }
1099                /* Save the delegation */
1100                nfs4_stateid_copy(&stateid, &delegation->stateid);
1101                rcu_read_unlock();
1102                ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1103                if (ret != 0)
1104                        goto out;
1105                ret = -EAGAIN;
1106
1107                /* Try to update the stateid using the delegation */
1108                if (update_open_stateid(state, NULL, &stateid, fmode))
1109                        goto out_return_state;
1110        }
1111out:
1112        return ERR_PTR(ret);
1113out_return_state:
1114        atomic_inc(&state->count);
1115        return state;
1116}
1117
1118static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1119{
1120        struct inode *inode;
1121        struct nfs4_state *state = NULL;
1122        struct nfs_delegation *delegation;
1123        int ret;
1124
1125        if (!data->rpc_done) {
1126                state = nfs4_try_open_cached(data);
1127                goto out;
1128        }
1129
1130        ret = -EAGAIN;
1131        if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1132                goto err;
1133        inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1134        ret = PTR_ERR(inode);
1135        if (IS_ERR(inode))
1136                goto err;
1137        ret = -ENOMEM;
1138        state = nfs4_get_open_state(inode, data->owner);
1139        if (state == NULL)
1140                goto err_put_inode;
1141        if (data->o_res.delegation_type != 0) {
1142                struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1143                int delegation_flags = 0;
1144
1145                rcu_read_lock();
1146                delegation = rcu_dereference(NFS_I(inode)->delegation);
1147                if (delegation)
1148                        delegation_flags = delegation->flags;
1149                rcu_read_unlock();
1150                if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1151                        pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1152                                        "returning a delegation for "
1153                                        "OPEN(CLAIM_DELEGATE_CUR)\n",
1154                                        clp->cl_hostname);
1155                } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1156                        nfs_inode_set_delegation(state->inode,
1157                                        data->owner->so_cred,
1158                                        &data->o_res);
1159                else
1160                        nfs_inode_reclaim_delegation(state->inode,
1161                                        data->owner->so_cred,
1162                                        &data->o_res);
1163        }
1164
1165        update_open_stateid(state, &data->o_res.stateid, NULL,
1166                        data->o_arg.fmode);
1167        iput(inode);
1168out:
1169        return state;
1170err_put_inode:
1171        iput(inode);
1172err:
1173        return ERR_PTR(ret);
1174}
1175
1176static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1177{
1178        struct nfs_inode *nfsi = NFS_I(state->inode);
1179        struct nfs_open_context *ctx;
1180
1181        spin_lock(&state->inode->i_lock);
1182        list_for_each_entry(ctx, &nfsi->open_files, list) {
1183                if (ctx->state != state)
1184                        continue;
1185                get_nfs_open_context(ctx);
1186                spin_unlock(&state->inode->i_lock);
1187                return ctx;
1188        }
1189        spin_unlock(&state->inode->i_lock);
1190        return ERR_PTR(-ENOENT);
1191}
1192
1193static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1194{
1195        struct nfs4_opendata *opendata;
1196
1197        opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1198        if (opendata == NULL)
1199                return ERR_PTR(-ENOMEM);
1200        opendata->state = state;
1201        atomic_inc(&state->count);
1202        return opendata;
1203}
1204
1205static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1206{
1207        struct nfs4_state *newstate;
1208        int ret;
1209
1210        opendata->o_arg.open_flags = 0;
1211        opendata->o_arg.fmode = fmode;
1212        memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1213        memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1214        nfs4_init_opendata_res(opendata);
1215        ret = _nfs4_recover_proc_open(opendata);
1216        if (ret != 0)
1217                return ret; 
1218        newstate = nfs4_opendata_to_nfs4_state(opendata);
1219        if (IS_ERR(newstate))
1220                return PTR_ERR(newstate);
1221        nfs4_close_state(newstate, fmode);
1222        *res = newstate;
1223        return 0;
1224}
1225
1226static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1227{
1228        struct nfs4_state *newstate;
1229        int ret;
1230
1231        /* memory barrier prior to reading state->n_* */
1232        clear_bit(NFS_DELEGATED_STATE, &state->flags);
1233        smp_rmb();
1234        if (state->n_rdwr != 0) {
1235                clear_bit(NFS_O_RDWR_STATE, &state->flags);
1236                ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1237                if (ret != 0)
1238                        return ret;
1239                if (newstate != state)
1240                        return -ESTALE;
1241        }
1242        if (state->n_wronly != 0) {
1243                clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1244                ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1245                if (ret != 0)
1246                        return ret;
1247                if (newstate != state)
1248                        return -ESTALE;
1249        }
1250        if (state->n_rdonly != 0) {
1251                clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1252                ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1253                if (ret != 0)
1254                        return ret;
1255                if (newstate != state)
1256                        return -ESTALE;
1257        }
1258        /*
1259         * We may have performed cached opens for all three recoveries.
1260         * Check if we need to update the current stateid.
1261         */
1262        if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1263            !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1264                write_seqlock(&state->seqlock);
1265                if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1266                        nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1267                write_sequnlock(&state->seqlock);
1268        }
1269        return 0;
1270}
1271
1272/*
1273 * OPEN_RECLAIM:
1274 *      reclaim state on the server after a reboot.
1275 */
1276static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1277{
1278        struct nfs_delegation *delegation;
1279        struct nfs4_opendata *opendata;
1280        fmode_t delegation_type = 0;
1281        int status;
1282
1283        opendata = nfs4_open_recoverdata_alloc(ctx, state);
1284        if (IS_ERR(opendata))
1285                return PTR_ERR(opendata);
1286        opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1287        opendata->o_arg.fh = NFS_FH(state->inode);
1288        rcu_read_lock();
1289        delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1290        if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1291                delegation_type = delegation->type;
1292        rcu_read_unlock();
1293        opendata->o_arg.u.delegation_type = delegation_type;
1294        status = nfs4_open_recover(opendata, state);
1295        nfs4_opendata_put(opendata);
1296        return status;
1297}
1298
1299static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1300{
1301        struct nfs_server *server = NFS_SERVER(state->inode);
1302        struct nfs4_exception exception = { };
1303        int err;
1304        do {
1305                err = _nfs4_do_open_reclaim(ctx, state);
1306                if (err != -NFS4ERR_DELAY)
1307                        break;
1308                nfs4_handle_exception(server, err, &exception);
1309        } while (exception.retry);
1310        return err;
1311}
1312
1313static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1314{
1315        struct nfs_open_context *ctx;
1316        int ret;
1317
1318        ctx = nfs4_state_find_open_context(state);
1319        if (IS_ERR(ctx))
1320                return PTR_ERR(ctx);
1321        ret = nfs4_do_open_reclaim(ctx, state);
1322        put_nfs_open_context(ctx);
1323        return ret;
1324}
1325
1326static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1327{
1328        struct nfs4_opendata *opendata;
1329        int ret;
1330
1331        opendata = nfs4_open_recoverdata_alloc(ctx, state);
1332        if (IS_ERR(opendata))
1333                return PTR_ERR(opendata);
1334        opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1335        nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1336        ret = nfs4_open_recover(opendata, state);
1337        nfs4_opendata_put(opendata);
1338        return ret;
1339}
1340
1341int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1342{
1343        struct nfs4_exception exception = { };
1344        struct nfs_server *server = NFS_SERVER(state->inode);
1345        int err;
1346        do {
1347                err = _nfs4_open_delegation_recall(ctx, state, stateid);
1348                switch (err) {
1349                        case 0:
1350                        case -ENOENT:
1351                        case -ESTALE:
1352                                goto out;
1353                        case -NFS4ERR_BADSESSION:
1354                        case -NFS4ERR_BADSLOT:
1355                        case -NFS4ERR_BAD_HIGH_SLOT:
1356                        case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1357                        case -NFS4ERR_DEADSESSION:
1358                                nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1359                                goto out;
1360                        case -NFS4ERR_STALE_CLIENTID:
1361                        case -NFS4ERR_STALE_STATEID:
1362                        case -NFS4ERR_EXPIRED:
1363                                /* Don't recall a delegation if it was lost */
1364                                nfs4_schedule_lease_recovery(server->nfs_client);
1365                                goto out;
1366                        case -ERESTARTSYS:
1367                                /*
1368                                 * The show must go on: exit, but mark the
1369                                 * stateid as needing recovery.
1370                                 */
1371                        case -NFS4ERR_DELEG_REVOKED:
1372                        case -NFS4ERR_ADMIN_REVOKED:
1373                        case -NFS4ERR_BAD_STATEID:
1374                                nfs_inode_find_state_and_recover(state->inode,
1375                                                stateid);
1376                                nfs4_schedule_stateid_recovery(server, state);
1377                        case -EKEYEXPIRED:
1378                                /*
1379                                 * User RPCSEC_GSS context has expired.
1380                                 * We cannot recover this stateid now, so
1381                                 * skip it and allow recovery thread to
1382                                 * proceed.
1383                                 */
1384                        case -ENOMEM:
1385                                err = 0;
1386                                goto out;
1387                }
1388                err = nfs4_handle_exception(server, err, &exception);
1389        } while (exception.retry);
1390out:
1391        return err;
1392}
1393
1394static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1395{
1396        struct nfs4_opendata *data = calldata;
1397
1398        data->rpc_status = task->tk_status;
1399        if (data->rpc_status == 0) {
1400                nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1401                nfs_confirm_seqid(&data->owner->so_seqid, 0);
1402                renew_lease(data->o_res.server, data->timestamp);
1403                data->rpc_done = 1;
1404        }
1405}
1406
1407static void nfs4_open_confirm_release(void *calldata)
1408{
1409        struct nfs4_opendata *data = calldata;
1410        struct nfs4_state *state = NULL;
1411
1412        /* If this request hasn't been cancelled, do nothing */
1413        if (data->cancelled == 0)
1414                goto out_free;
1415        /* In case of error, no cleanup! */
1416        if (!data->rpc_done)
1417                goto out_free;
1418        state = nfs4_opendata_to_nfs4_state(data);
1419        if (!IS_ERR(state))
1420                nfs4_close_state(state, data->o_arg.fmode);
1421out_free:
1422        nfs4_opendata_put(data);
1423}
1424
1425static const struct rpc_call_ops nfs4_open_confirm_ops = {
1426        .rpc_call_done = nfs4_open_confirm_done,
1427        .rpc_release = nfs4_open_confirm_release,
1428};
1429
1430/*
1431 * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1432 */
1433static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1434{
1435        struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1436        struct rpc_task *task;
1437        struct  rpc_message msg = {
1438                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1439                .rpc_argp = &data->c_arg,
1440                .rpc_resp = &data->c_res,
1441                .rpc_cred = data->owner->so_cred,
1442        };
1443        struct rpc_task_setup task_setup_data = {
1444                .rpc_client = server->client,
1445                .rpc_message = &msg,
1446                .callback_ops = &nfs4_open_confirm_ops,
1447                .callback_data = data,
1448                .workqueue = nfsiod_workqueue,
1449                .flags = RPC_TASK_ASYNC,
1450        };
1451        int status;
1452
1453        kref_get(&data->kref);
1454        data->rpc_done = 0;
1455        data->rpc_status = 0;
1456        data->timestamp = jiffies;
1457        task = rpc_run_task(&task_setup_data);
1458        if (IS_ERR(task))
1459                return PTR_ERR(task);
1460        status = nfs4_wait_for_completion_rpc_task(task);
1461        if (status != 0) {
1462                data->cancelled = 1;
1463                smp_wmb();
1464        } else
1465                status = data->rpc_status;
1466        rpc_put_task(task);
1467        return status;
1468}
1469
1470static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1471{
1472        struct nfs4_opendata *data = calldata;
1473        struct nfs4_state_owner *sp = data->owner;
1474
1475        if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1476                return;
1477        /*
1478         * Check if we still need to send an OPEN call, or if we can use
1479         * a delegation instead.
1480         */
1481        if (data->state != NULL) {
1482                struct nfs_delegation *delegation;
1483
1484                if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1485                        goto out_no_action;
1486                rcu_read_lock();
1487                delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1488                if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1489                    can_open_delegated(delegation, data->o_arg.fmode))
1490                        goto unlock_no_action;
1491                rcu_read_unlock();
1492        }
1493        /* Update client id. */
1494        data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1495        if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1496                task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1497                nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1498        }
1499        data->timestamp = jiffies;
1500        if (nfs4_setup_sequence(data->o_arg.server,
1501                                &data->o_arg.seq_args,
1502                                &data->o_res.seq_res, task))
1503                return;
1504        rpc_call_start(task);
1505        return;
1506unlock_no_action:
1507        rcu_read_unlock();
1508out_no_action:
1509        task->tk_action = NULL;
1510
1511}
1512
1513static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1514{
1515        rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1516        nfs4_open_prepare(task, calldata);
1517}
1518
1519static void nfs4_open_done(struct rpc_task *task, void *calldata)
1520{
1521        struct nfs4_opendata *data = calldata;
1522
1523        data->rpc_status = task->tk_status;
1524
1525        if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1526                return;
1527
1528        if (task->tk_status == 0) {
1529                switch (data->o_res.f_attr->mode & S_IFMT) {
1530                        case S_IFREG:
1531                                break;
1532                        case S_IFLNK:
1533                                data->rpc_status = -ELOOP;
1534                                break;
1535                        case S_IFDIR:
1536                                data->rpc_status = -EISDIR;
1537                                break;
1538                        default:
1539                                data->rpc_status = -ENOTDIR;
1540                }
1541                renew_lease(data->o_res.server, data->timestamp);
1542                if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1543                        nfs_confirm_seqid(&data->owner->so_seqid, 0);
1544        }
1545        data->rpc_done = 1;
1546}
1547
1548static void nfs4_open_release(void *calldata)
1549{
1550        struct nfs4_opendata *data = calldata;
1551        struct nfs4_state *state = NULL;
1552
1553        /* If this request hasn't been cancelled, do nothing */
1554        if (data->cancelled == 0)
1555                goto out_free;
1556        /* In case of error, no cleanup! */
1557        if (data->rpc_status != 0 || !data->rpc_done)
1558                goto out_free;
1559        /* In case we need an open_confirm, no cleanup! */
1560        if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1561                goto out_free;
1562        state = nfs4_opendata_to_nfs4_state(data);
1563        if (!IS_ERR(state))
1564                nfs4_close_state(state, data->o_arg.fmode);
1565out_free:
1566        nfs4_opendata_put(data);
1567}
1568
1569static const struct rpc_call_ops nfs4_open_ops = {
1570        .rpc_call_prepare = nfs4_open_prepare,
1571        .rpc_call_done = nfs4_open_done,
1572        .rpc_release = nfs4_open_release,
1573};
1574
1575static const struct rpc_call_ops nfs4_recover_open_ops = {
1576        .rpc_call_prepare = nfs4_recover_open_prepare,
1577        .rpc_call_done = nfs4_open_done,
1578        .rpc_release = nfs4_open_release,
1579};
1580
1581static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1582{
1583        struct inode *dir = data->dir->d_inode;
1584        struct nfs_server *server = NFS_SERVER(dir);
1585        struct nfs_openargs *o_arg = &data->o_arg;
1586        struct nfs_openres *o_res = &data->o_res;
1587        struct rpc_task *task;
1588        struct rpc_message msg = {
1589                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1590                .rpc_argp = o_arg,
1591                .rpc_resp = o_res,
1592                .rpc_cred = data->owner->so_cred,
1593        };
1594        struct rpc_task_setup task_setup_data = {
1595                .rpc_client = server->client,
1596                .rpc_message = &msg,
1597                .callback_ops = &nfs4_open_ops,
1598                .callback_data = data,
1599                .workqueue = nfsiod_workqueue,
1600                .flags = RPC_TASK_ASYNC,
1601        };
1602        int status;
1603
1604        nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1605        kref_get(&data->kref);
1606        data->rpc_done = 0;
1607        data->rpc_status = 0;
1608        data->cancelled = 0;
1609        if (isrecover)
1610                task_setup_data.callback_ops = &nfs4_recover_open_ops;
1611        task = rpc_run_task(&task_setup_data);
1612        if (IS_ERR(task))
1613                return PTR_ERR(task);
1614        status = nfs4_wait_for_completion_rpc_task(task);
1615        if (status != 0) {
1616                data->cancelled = 1;
1617                smp_wmb();
1618        } else
1619                status = data->rpc_status;
1620        rpc_put_task(task);
1621
1622        return status;
1623}
1624
1625static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1626{
1627        struct inode *dir = data->dir->d_inode;
1628        struct nfs_openres *o_res = &data->o_res;
1629        int status;
1630
1631        status = nfs4_run_open_task(data, 1);
1632        if (status != 0 || !data->rpc_done)
1633                return status;
1634
1635        nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1636
1637        if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1638                status = _nfs4_proc_open_confirm(data);
1639                if (status != 0)
1640                        return status;
1641        }
1642
1643        return status;
1644}
1645
1646/*
1647 * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1648 */
1649static int _nfs4_proc_open(struct nfs4_opendata *data)
1650{
1651        struct inode *dir = data->dir->d_inode;
1652        struct nfs_server *server = NFS_SERVER(dir);
1653        struct nfs_openargs *o_arg = &data->o_arg;
1654        struct nfs_openres *o_res = &data->o_res;
1655        int status;
1656
1657        status = nfs4_run_open_task(data, 0);
1658        if (!data->rpc_done)
1659                return status;
1660        if (status != 0) {
1661                if (status == -NFS4ERR_BADNAME &&
1662                                !(o_arg->open_flags & O_CREAT))
1663                        return -ENOENT;
1664                return status;
1665        }
1666
1667        nfs_fattr_map_and_free_names(server, &data->f_attr);
1668
1669        if (o_arg->open_flags & O_CREAT)
1670                update_changeattr(dir, &o_res->cinfo);
1671        if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1672                server->caps &= ~NFS_CAP_POSIX_LOCK;
1673        if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1674                status = _nfs4_proc_open_confirm(data);
1675                if (status != 0)
1676                        return status;
1677        }
1678        if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1679                _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1680        return 0;
1681}
1682
1683static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1684{
1685        unsigned int loop;
1686        int ret;
1687
1688        for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1689                ret = nfs4_wait_clnt_recover(clp);
1690                if (ret != 0)
1691                        break;
1692                if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1693                    !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1694                        break;
1695                nfs4_schedule_state_manager(clp);
1696                ret = -EIO;
1697        }
1698        return ret;
1699}
1700
1701static int nfs4_recover_expired_lease(struct nfs_server *server)
1702{
1703        return nfs4_client_recover_expired_lease(server->nfs_client);
1704}
1705
1706/*
1707 * OPEN_EXPIRED:
1708 *      reclaim state on the server after a network partition.
1709 *      Assumes caller holds the appropriate lock
1710 */
1711static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1712{
1713        struct nfs4_opendata *opendata;
1714        int ret;
1715
1716        opendata = nfs4_open_recoverdata_alloc(ctx, state);
1717        if (IS_ERR(opendata))
1718                return PTR_ERR(opendata);
1719        ret = nfs4_open_recover(opendata, state);
1720        if (ret == -ESTALE)
1721                d_drop(ctx->dentry);
1722        nfs4_opendata_put(opendata);
1723        return ret;
1724}
1725
1726static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1727{
1728        struct nfs_server *server = NFS_SERVER(state->inode);
1729        struct nfs4_exception exception = { };
1730        int err;
1731
1732        do {
1733                err = _nfs4_open_expired(ctx, state);
1734                switch (err) {
1735                default:
1736                        goto out;
1737                case -NFS4ERR_GRACE:
1738                case -NFS4ERR_DELAY:
1739                        nfs4_handle_exception(server, err, &exception);
1740                        err = 0;
1741                }
1742        } while (exception.retry);
1743out:
1744        return err;
1745}
1746
1747static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1748{
1749        struct nfs_open_context *ctx;
1750        int ret;
1751
1752        ctx = nfs4_state_find_open_context(state);
1753        if (IS_ERR(ctx))
1754                return PTR_ERR(ctx);
1755        ret = nfs4_do_open_expired(ctx, state);
1756        put_nfs_open_context(ctx);
1757        return ret;
1758}
1759
1760#if defined(CONFIG_NFS_V4_1)
1761static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1762{
1763        struct nfs_server *server = NFS_SERVER(state->inode);
1764        nfs4_stateid *stateid = &state->stateid;
1765        int status;
1766
1767        /* If a state reset has been done, test_stateid is unneeded */
1768        if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1769                return;
1770
1771        status = nfs41_test_stateid(server, stateid);
1772        if (status != NFS_OK) {
1773                /* Free the stateid unless the server explicitly
1774                 * informs us the stateid is unrecognized. */
1775                if (status != -NFS4ERR_BAD_STATEID)
1776                        nfs41_free_stateid(server, stateid);
1777
1778                clear_bit(NFS_DELEGATED_STATE, &state->flags);
1779        }
1780}
1781
1782/**
1783 * nfs41_check_open_stateid - possibly free an open stateid
1784 *
1785 * @state: NFSv4 state for an inode
1786 *
1787 * Returns NFS_OK if recovery for this stateid is now finished.
1788 * Otherwise a negative NFS4ERR value is returned.
1789 */
1790static int nfs41_check_open_stateid(struct nfs4_state *state)
1791{
1792        struct nfs_server *server = NFS_SERVER(state->inode);
1793        nfs4_stateid *stateid = &state->stateid;
1794        int status;
1795
1796        /* If a state reset has been done, test_stateid is unneeded */
1797        if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1798            (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1799            (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1800                return -NFS4ERR_BAD_STATEID;
1801
1802        status = nfs41_test_stateid(server, stateid);
1803        if (status != NFS_OK) {
1804                /* Free the stateid unless the server explicitly
1805                 * informs us the stateid is unrecognized. */
1806                if (status != -NFS4ERR_BAD_STATEID)
1807                        nfs41_free_stateid(server, stateid);
1808
1809                clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1810                clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1811                clear_bit(NFS_O_RDWR_STATE, &state->flags);
1812        }
1813        return status;
1814}
1815
1816static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1817{
1818        int status;
1819
1820        nfs41_clear_delegation_stateid(state);
1821        status = nfs41_check_open_stateid(state);
1822        if (status != NFS_OK)
1823                status = nfs4_open_expired(sp, state);
1824        return status;
1825}
1826#endif
1827
1828/*
1829 * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1830 * fields corresponding to attributes that were used to store the verifier.
1831 * Make sure we clobber those fields in the later setattr call
1832 */
1833static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1834{
1835        if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1836            !(sattr->ia_valid & ATTR_ATIME_SET))
1837                sattr->ia_valid |= ATTR_ATIME;
1838
1839        if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1840            !(sattr->ia_valid & ATTR_MTIME_SET))
1841                sattr->ia_valid |= ATTR_MTIME;
1842}
1843
1844/*
1845 * Returns a referenced nfs4_state
1846 */
1847static int _nfs4_do_open(struct inode *dir,
1848                        struct dentry *dentry,
1849                        fmode_t fmode,
1850                        int flags,
1851                        struct iattr *sattr,
1852                        struct rpc_cred *cred,
1853                        struct nfs4_state **res,
1854                        struct nfs4_threshold **ctx_th)
1855{
1856        struct nfs4_state_owner  *sp;
1857        struct nfs4_state     *state = NULL;
1858        struct nfs_server       *server = NFS_SERVER(dir);
1859        struct nfs4_opendata *opendata;
1860        int status;
1861
1862        /* Protect against reboot recovery conflicts */
1863        status = -ENOMEM;
1864        sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1865        if (sp == NULL) {
1866                dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1867                goto out_err;
1868        }
1869        status = nfs4_recover_expired_lease(server);
1870        if (status != 0)
1871                goto err_put_state_owner;
1872        if (dentry->d_inode != NULL)
1873                nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1874        status = -ENOMEM;
1875        opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1876        if (opendata == NULL)
1877                goto err_put_state_owner;
1878
1879        if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1880                opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1881                if (!opendata->f_attr.mdsthreshold)
1882                        goto err_opendata_put;
1883                opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
1884        }
1885        if (dentry->d_inode != NULL)
1886                opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1887
1888        status = _nfs4_proc_open(opendata);
1889        if (status != 0)
1890                goto err_opendata_put;
1891
1892        state = nfs4_opendata_to_nfs4_state(opendata);
1893        status = PTR_ERR(state);
1894        if (IS_ERR(state))
1895                goto err_opendata_put;
1896        if (server->caps & NFS_CAP_POSIX_LOCK)
1897                set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1898
1899        if (opendata->o_arg.open_flags & O_EXCL) {
1900                nfs4_exclusive_attrset(opendata, sattr);
1901
1902                nfs_fattr_init(opendata->o_res.f_attr);
1903                status = nfs4_do_setattr(state->inode, cred,
1904                                opendata->o_res.f_attr, sattr,
1905                                state);
1906                if (status == 0)
1907                        nfs_setattr_update_inode(state->inode, sattr);
1908                nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1909        }
1910
1911        if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
1912                *ctx_th = opendata->f_attr.mdsthreshold;
1913        else
1914                kfree(opendata->f_attr.mdsthreshold);
1915        opendata->f_attr.mdsthreshold = NULL;
1916
1917        nfs4_opendata_put(opendata);
1918        nfs4_put_state_owner(sp);
1919        *res = state;
1920        return 0;
1921err_opendata_put:
1922        kfree(opendata->f_attr.mdsthreshold);
1923        nfs4_opendata_put(opendata);
1924err_put_state_owner:
1925        nfs4_put_state_owner(sp);
1926out_err:
1927        *res = NULL;
1928        return status;
1929}
1930
1931
1932static struct nfs4_state *nfs4_do_open(struct inode *dir,
1933                                        struct dentry *dentry,
1934                                        fmode_t fmode,
1935                                        int flags,
1936                                        struct iattr *sattr,
1937                                        struct rpc_cred *cred,
1938                                        struct nfs4_threshold **ctx_th)
1939{
1940        struct nfs4_exception exception = { };
1941        struct nfs4_state *res;
1942        int status;
1943
1944        fmode &= FMODE_READ|FMODE_WRITE;
1945        do {
1946                status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
1947                                       &res, ctx_th);
1948                if (status == 0)
1949                        break;
1950                /* NOTE: BAD_SEQID means the server and client disagree about the
1951                 * book-keeping w.r.t. state-changing operations
1952                 * (OPEN/CLOSE/LOCK/LOCKU...)
1953                 * It is actually a sign of a bug on the client or on the server.
1954                 *
1955                 * If we receive a BAD_SEQID error in the particular case of
1956                 * doing an OPEN, we assume that nfs_increment_open_seqid() will
1957                 * have unhashed the old state_owner for us, and that we can
1958                 * therefore safely retry using a new one. We should still warn
1959                 * the user though...
1960                 */
1961                if (status == -NFS4ERR_BAD_SEQID) {
1962                        pr_warn_ratelimited("NFS: v4 server %s "
1963                                        " returned a bad sequence-id error!\n",
1964                                        NFS_SERVER(dir)->nfs_client->cl_hostname);
1965                        exception.retry = 1;
1966                        continue;
1967                }
1968                /*
1969                 * BAD_STATEID on OPEN means that the server cancelled our
1970                 * state before it received the OPEN_CONFIRM.
1971                 * Recover by retrying the request as per the discussion
1972                 * on Page 181 of RFC3530.
1973                 */
1974                if (status == -NFS4ERR_BAD_STATEID) {
1975                        exception.retry = 1;
1976                        continue;
1977                }
1978                if (status == -EAGAIN) {
1979                        /* We must have found a delegation */
1980                        exception.retry = 1;
1981                        continue;
1982                }
1983                res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1984                                        status, &exception));
1985        } while (exception.retry);
1986        return res;
1987}
1988
1989static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1990                            struct nfs_fattr *fattr, struct iattr *sattr,
1991                            struct nfs4_state *state)
1992{
1993        struct nfs_server *server = NFS_SERVER(inode);
1994        struct nfs_setattrargs  arg = {
1995                .fh             = NFS_FH(inode),
1996                .iap            = sattr,
1997                .server         = server,
1998                .bitmask = server->attr_bitmask,
1999        };
2000        struct nfs_setattrres  res = {
2001                .fattr          = fattr,
2002                .server         = server,
2003        };
2004        struct rpc_message msg = {
2005                .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2006                .rpc_argp       = &arg,
2007                .rpc_resp       = &res,
2008                .rpc_cred       = cred,
2009        };
2010        unsigned long timestamp = jiffies;
2011        int status;
2012
2013        nfs_fattr_init(fattr);
2014
2015        if (state != NULL) {
2016                nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2017                                current->files, current->tgid);
2018        } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
2019                                FMODE_WRITE)) {
2020                /* Use that stateid */
2021        } else
2022                nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2023
2024        status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2025        if (status == 0 && state != NULL)
2026                renew_lease(server, timestamp);
2027        return status;
2028}
2029
2030static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2031                           struct nfs_fattr *fattr, struct iattr *sattr,
2032                           struct nfs4_state *state)
2033{
2034        struct nfs_server *server = NFS_SERVER(inode);
2035        struct nfs4_exception exception = {
2036                .state = state,
2037                .inode = inode,
2038        };
2039        int err;
2040        do {
2041                err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2042                switch (err) {
2043                case -NFS4ERR_OPENMODE:
2044                        if (state && !(state->state & FMODE_WRITE)) {
2045                                err = -EBADF;
2046                                if (sattr->ia_valid & ATTR_OPEN)
2047                                        err = -EACCES;
2048                                goto out;
2049                        }
2050                }
2051                err = nfs4_handle_exception(server, err, &exception);
2052        } while (exception.retry);
2053out:
2054        return err;
2055}
2056
2057struct nfs4_closedata {
2058        struct inode *inode;
2059        struct nfs4_state *state;
2060        struct nfs_closeargs arg;
2061        struct nfs_closeres res;
2062        struct nfs_fattr fattr;
2063        unsigned long timestamp;
2064        bool roc;
2065        u32 roc_barrier;
2066};
2067
2068static void nfs4_free_closedata(void *data)
2069{
2070        struct nfs4_closedata *calldata = data;
2071        struct nfs4_state_owner *sp = calldata->state->owner;
2072        struct super_block *sb = calldata->state->inode->i_sb;
2073
2074        if (calldata->roc)
2075                pnfs_roc_release(calldata->state->inode);
2076        nfs4_put_open_state(calldata->state);
2077        nfs_free_seqid(calldata->arg.seqid);
2078        nfs4_put_state_owner(sp);
2079        nfs_sb_deactive(sb);
2080        kfree(calldata);
2081}
2082
2083static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2084                fmode_t fmode)
2085{
2086        spin_lock(&state->owner->so_lock);
2087        if (!(fmode & FMODE_READ))
2088                clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2089        if (!(fmode & FMODE_WRITE))
2090                clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2091        clear_bit(NFS_O_RDWR_STATE, &state->flags);
2092        spin_unlock(&state->owner->so_lock);
2093}
2094
2095static void nfs4_close_done(struct rpc_task *task, void *data)
2096{
2097        struct nfs4_closedata *calldata = data;
2098        struct nfs4_state *state = calldata->state;
2099        struct nfs_server *server = NFS_SERVER(calldata->inode);
2100
2101        dprintk("%s: begin!\n", __func__);
2102        if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2103                return;
2104        /* hmm. we are done with the inode, and in the process of freeing
2105         * the state_owner. we keep this around to process errors
2106         */
2107        switch (task->tk_status) {
2108                case 0:
2109                        if (calldata->roc)
2110                                pnfs_roc_set_barrier(state->inode,
2111                                                     calldata->roc_barrier);
2112                        nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2113                        renew_lease(server, calldata->timestamp);
2114                        nfs4_close_clear_stateid_flags(state,
2115                                        calldata->arg.fmode);
2116                        break;
2117                case -NFS4ERR_STALE_STATEID:
2118                case -NFS4ERR_OLD_STATEID:
2119                case -NFS4ERR_BAD_STATEID:
2120                case -NFS4ERR_EXPIRED:
2121                        if (calldata->arg.fmode == 0)
2122                                break;
2123                default:
2124                        if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2125                                rpc_restart_call_prepare(task);
2126        }
2127        nfs_release_seqid(calldata->arg.seqid);
2128        nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2129        dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2130}
2131
2132static void nfs4_close_prepare(struct rpc_task *task, void *data)
2133{
2134        struct nfs4_closedata *calldata = data;
2135        struct nfs4_state *state = calldata->state;
2136        int call_close = 0;
2137
2138        dprintk("%s: begin!\n", __func__);
2139        if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2140                return;
2141
2142        task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2143        calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2144        spin_lock(&state->owner->so_lock);
2145        /* Calculate the change in open mode */
2146        if (state->n_rdwr == 0) {
2147                if (state->n_rdonly == 0) {
2148                        call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2149                        call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2150                        calldata->arg.fmode &= ~FMODE_READ;
2151                }
2152                if (state->n_wronly == 0) {
2153                        call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2154                        call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2155                        calldata->arg.fmode &= ~FMODE_WRITE;
2156                }
2157        }
2158        spin_unlock(&state->owner->so_lock);
2159
2160        if (!call_close) {
2161                /* Note: exit _without_ calling nfs4_close_done */
2162                task->tk_action = NULL;
2163                goto out;
2164        }
2165
2166        if (calldata->arg.fmode == 0) {
2167                task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2168                if (calldata->roc &&
2169                    pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2170                        rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2171                                     task, NULL);
2172                        goto out;
2173                }
2174        }
2175
2176        nfs_fattr_init(calldata->res.fattr);
2177        calldata->timestamp = jiffies;
2178        if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2179                                &calldata->arg.seq_args,
2180                                &calldata->res.seq_res,
2181                                task))
2182                goto out;
2183        rpc_call_start(task);
2184out:
2185        dprintk("%s: done!\n", __func__);
2186}
2187
2188static const struct rpc_call_ops nfs4_close_ops = {
2189        .rpc_call_prepare = nfs4_close_prepare,
2190        .rpc_call_done = nfs4_close_done,
2191        .rpc_release = nfs4_free_closedata,
2192};
2193
2194/* 
2195 * It is possible for data to be read/written from a mem-mapped file 
2196 * after the sys_close call (which hits the vfs layer as a flush).
2197 * This means that we can't safely call nfsv4 close on a file until 
2198 * the inode is cleared. This in turn means that we are not good
2199 * NFSv4 citizens - we do not indicate to the server to update the file's 
2200 * share state even when we are done with one of the three share 
2201 * stateid's in the inode.
2202 *
2203 * NOTE: Caller must be holding the sp->so_owner semaphore!
2204 */
2205int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2206{
2207        struct nfs_server *server = NFS_SERVER(state->inode);
2208        struct nfs4_closedata *calldata;
2209        struct nfs4_state_owner *sp = state->owner;
2210        struct rpc_task *task;
2211        struct rpc_message msg = {
2212                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2213                .rpc_cred = state->owner->so_cred,
2214        };
2215        struct rpc_task_setup task_setup_data = {
2216                .rpc_client = server->client,
2217                .rpc_message = &msg,
2218                .callback_ops = &nfs4_close_ops,
2219                .workqueue = nfsiod_workqueue,
2220                .flags = RPC_TASK_ASYNC,
2221        };
2222        int status = -ENOMEM;
2223
2224        calldata = kzalloc(sizeof(*calldata), gfp_mask);
2225        if (calldata == NULL)
2226                goto out;
2227        nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2228        calldata->inode = state->inode;
2229        calldata->state = state;
2230        calldata->arg.fh = NFS_FH(state->inode);
2231        calldata->arg.stateid = &state->open_stateid;
2232        /* Serialization for the sequence id */
2233        calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2234        if (calldata->arg.seqid == NULL)
2235                goto out_free_calldata;
2236        calldata->arg.fmode = 0;
2237        calldata->arg.bitmask = server->cache_consistency_bitmask;
2238        calldata->res.fattr = &calldata->fattr;
2239        calldata->res.seqid = calldata->arg.seqid;
2240        calldata->res.server = server;
2241        calldata->roc = roc;
2242        nfs_sb_active(calldata->inode->i_sb);
2243
2244        msg.rpc_argp = &calldata->arg;
2245        msg.rpc_resp = &calldata->res;
2246        task_setup_data.callback_data = calldata;
2247        task = rpc_run_task(&task_setup_data);
2248        if (IS_ERR(task))
2249                return PTR_ERR(task);
2250        status = 0;
2251        if (wait)
2252                status = rpc_wait_for_completion_task(task);
2253        rpc_put_task(task);
2254        return status;
2255out_free_calldata:
2256        kfree(calldata);
2257out:
2258        if (roc)
2259                pnfs_roc_release(state->inode);
2260        nfs4_put_open_state(state);
2261        nfs4_put_state_owner(sp);
2262        return status;
2263}
2264
2265static struct inode *
2266nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2267{
2268        struct nfs4_state *state;
2269
2270        /* Protect against concurrent sillydeletes */
2271        state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2272                             ctx->cred, &ctx->mdsthreshold);
2273        if (IS_ERR(state))
2274                return ERR_CAST(state);
2275        ctx->state = state;
2276        return igrab(state->inode);
2277}
2278
2279static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2280{
2281        if (ctx->state == NULL)
2282                return;
2283        if (is_sync)
2284                nfs4_close_sync(ctx->state, ctx->mode);
2285        else
2286                nfs4_close_state(ctx->state, ctx->mode);
2287}
2288
2289static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2290{
2291        struct nfs4_server_caps_arg args = {
2292                .fhandle = fhandle,
2293        };
2294        struct nfs4_server_caps_res res = {};
2295        struct rpc_message msg = {
2296                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2297                .rpc_argp = &args,
2298                .rpc_resp = &res,
2299        };
2300        int status;
2301
2302        status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2303        if (status == 0) {
2304                memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2305                server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2306                                NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2307                                NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2308                                NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2309                                NFS_CAP_CTIME|NFS_CAP_MTIME);
2310                if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2311                        server->caps |= NFS_CAP_ACLS;
2312                if (res.has_links != 0)
2313                        server->caps |= NFS_CAP_HARDLINKS;
2314                if (res.has_symlinks != 0)
2315                        server->caps |= NFS_CAP_SYMLINKS;
2316                if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2317                        server->caps |= NFS_CAP_FILEID;
2318                if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2319                        server->caps |= NFS_CAP_MODE;
2320                if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2321                        server->caps |= NFS_CAP_NLINK;
2322                if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2323                        server->caps |= NFS_CAP_OWNER;
2324                if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2325                        server->caps |= NFS_CAP_OWNER_GROUP;
2326                if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2327                        server->caps |= NFS_CAP_ATIME;
2328                if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2329                        server->caps |= NFS_CAP_CTIME;
2330                if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2331                        server->caps |= NFS_CAP_MTIME;
2332
2333                memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2334                server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2335                server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2336                server->acl_bitmask = res.acl_bitmask;
2337                server->fh_expire_type = res.fh_expire_type;
2338        }
2339
2340        return status;
2341}
2342
2343int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2344{
2345        struct nfs4_exception exception = { };
2346        int err;
2347        do {
2348                err = nfs4_handle_exception(server,
2349                                _nfs4_server_capabilities(server, fhandle),
2350                                &exception);
2351        } while (exception.retry);
2352        return err;
2353}
2354
2355static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2356                struct nfs_fsinfo *info)
2357{
2358        struct nfs4_lookup_root_arg args = {
2359                .bitmask = nfs4_fattr_bitmap,
2360        };
2361        struct nfs4_lookup_res res = {
2362                .server = server,
2363                .fattr = info->fattr,
2364                .fh = fhandle,
2365        };
2366        struct rpc_message msg = {
2367                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2368                .rpc_argp = &args,
2369                .rpc_resp = &res,
2370        };
2371
2372        nfs_fattr_init(info->fattr);
2373        return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2374}
2375
2376static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2377                struct nfs_fsinfo *info)
2378{
2379        struct nfs4_exception exception = { };
2380        int err;
2381        do {
2382                err = _nfs4_lookup_root(server, fhandle, info);
2383                switch (err) {
2384                case 0:
2385                case -NFS4ERR_WRONGSEC:
2386                        goto out;
2387                default:
2388                        err = nfs4_handle_exception(server, err, &exception);
2389                }
2390        } while (exception.retry);
2391out:
2392        return err;
2393}
2394
2395static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2396                                struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2397{
2398        struct rpc_auth *auth;
2399        int ret;
2400
2401        auth = rpcauth_create(flavor, server->client);
2402        if (!auth) {
2403                ret = -EIO;
2404                goto out;
2405        }
2406        ret = nfs4_lookup_root(server, fhandle, info);
2407out:
2408        return ret;
2409}
2410
2411static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2412                              struct nfs_fsinfo *info)
2413{
2414        int i, len, status = 0;
2415        rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2416
2417        len = rpcauth_list_flavors(flav_array, ARRAY_SIZE(flav_array));
2418        BUG_ON(len < 0);
2419
2420        for (i = 0; i < len; i++) {
2421                /* AUTH_UNIX is the default flavor if none was specified,
2422                 * thus has already been tried. */
2423                if (flav_array[i] == RPC_AUTH_UNIX)
2424                        continue;
2425
2426                status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2427                if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2428                        continue;
2429                break;
2430        }
2431        /*
2432         * -EACCESS could mean that the user doesn't have correct permissions
2433         * to access the mount.  It could also mean that we tried to mount
2434         * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2435         * existing mount programs don't handle -EACCES very well so it should
2436         * be mapped to -EPERM instead.
2437         */
2438        if (status == -EACCES)
2439                status = -EPERM;
2440        return status;
2441}
2442
2443/*
2444 * get the file handle for the "/" directory on the server
2445 */
2446int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2447                         struct nfs_fsinfo *info)
2448{
2449        int minor_version = server->nfs_client->cl_minorversion;
2450        int status = nfs4_lookup_root(server, fhandle, info);
2451        if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2452                /*
2453                 * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2454                 * by nfs4_map_errors() as this function exits.
2455                 */
2456                status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2457        if (status == 0)
2458                status = nfs4_server_capabilities(server, fhandle);
2459        if (status == 0)
2460                status = nfs4_do_fsinfo(server, fhandle, info);
2461        return nfs4_map_errors(status);
2462}
2463
2464static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2465                              struct nfs_fsinfo *info)
2466{
2467        int error;
2468        struct nfs_fattr *fattr = info->fattr;
2469
2470        error = nfs4_server_capabilities(server, mntfh);
2471        if (error < 0) {
2472                dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2473                return error;
2474        }
2475
2476        error = nfs4_proc_getattr(server, mntfh, fattr);
2477        if (error < 0) {
2478                dprintk("nfs4_get_root: getattr error = %d\n", -error);
2479                return error;
2480        }
2481
2482        if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2483            !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2484                memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2485
2486        return error;
2487}
2488
2489/*
2490 * Get locations and (maybe) other attributes of a referral.
2491 * Note that we'll actually follow the referral later when
2492 * we detect fsid mismatch in inode revalidation
2493 */
2494static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2495                             const struct qstr *name, struct nfs_fattr *fattr,
2496                             struct nfs_fh *fhandle)
2497{
2498        int status = -ENOMEM;
2499        struct page *page = NULL;
2500        struct nfs4_fs_locations *locations = NULL;
2501
2502        page = alloc_page(GFP_KERNEL);
2503        if (page == NULL)
2504                goto out;
2505        locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2506        if (locations == NULL)
2507                goto out;
2508
2509        status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2510        if (status != 0)
2511                goto out;
2512        /* Make sure server returned a different fsid for the referral */
2513        if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2514                dprintk("%s: server did not return a different fsid for"
2515                        " a referral at %s\n", __func__, name->name);
2516                status = -EIO;
2517                goto out;
2518        }
2519        /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2520        nfs_fixup_referral_attributes(&locations->fattr);
2521
2522        /* replace the lookup nfs_fattr with the locations nfs_fattr */
2523        memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2524        memset(fhandle, 0, sizeof(struct nfs_fh));
2525out:
2526        if (page)
2527                __free_page(page);
2528        kfree(locations);
2529        return status;
2530}
2531
2532static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2533{
2534        struct nfs4_getattr_arg args = {
2535                .fh = fhandle,
2536                .bitmask = server->attr_bitmask,
2537        };
2538        struct nfs4_getattr_res res = {
2539                .fattr = fattr,
2540                .server = server,
2541        };
2542        struct rpc_message msg = {
2543                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2544                .rpc_argp = &args,
2545                .rpc_resp = &res,
2546        };
2547        
2548        nfs_fattr_init(fattr);
2549        return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2550}
2551
2552static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2553{
2554        struct nfs4_exception exception = { };
2555        int err;
2556        do {
2557                err = nfs4_handle_exception(server,
2558                                _nfs4_proc_getattr(server, fhandle, fattr),
2559                                &exception);
2560        } while (exception.retry);
2561        return err;
2562}
2563
2564/* 
2565 * The file is not closed if it is opened due to the a request to change
2566 * the size of the file. The open call will not be needed once the
2567 * VFS layer lookup-intents are implemented.
2568 *
2569 * Close is called when the inode is destroyed.
2570 * If we haven't opened the file for O_WRONLY, we
2571 * need to in the size_change case to obtain a stateid.
2572 *
2573 * Got race?
2574 * Because OPEN is always done by name in nfsv4, it is
2575 * possible that we opened a different file by the same
2576 * name.  We can recognize this race condition, but we
2577 * can't do anything about it besides returning an error.
2578 *
2579 * This will be fixed with VFS changes (lookup-intent).
2580 */
2581static int
2582nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2583                  struct iattr *sattr)
2584{
2585        struct inode *inode = dentry->d_inode;
2586        struct rpc_cred *cred = NULL;
2587        struct nfs4_state *state = NULL;
2588        int status;
2589
2590        if (pnfs_ld_layoutret_on_setattr(inode))
2591                pnfs_return_layout(inode);
2592
2593        nfs_fattr_init(fattr);
2594        
2595        /* Deal with open(O_TRUNC) */
2596        if (sattr->ia_valid & ATTR_OPEN)
2597                sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2598
2599        /* Optimization: if the end result is no change, don't RPC */
2600        if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2601                return 0;
2602
2603        /* Search for an existing open(O_WRITE) file */
2604        if (sattr->ia_valid & ATTR_FILE) {
2605                struct nfs_open_context *ctx;
2606
2607                ctx = nfs_file_open_context(sattr->ia_file);
2608                if (ctx) {
2609                        cred = ctx->cred;
2610                        state = ctx->state;
2611                }
2612        }
2613
2614        status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2615        if (status == 0)
2616                nfs_setattr_update_inode(inode, sattr);
2617        return status;
2618}
2619
2620static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2621                const struct qstr *name, struct nfs_fh *fhandle,
2622                struct nfs_fattr *fattr)
2623{
2624        struct nfs_server *server = NFS_SERVER(dir);
2625        int                    status;
2626        struct nfs4_lookup_arg args = {
2627                .bitmask = server->attr_bitmask,
2628                .dir_fh = NFS_FH(dir),
2629                .name = name,
2630        };
2631        struct nfs4_lookup_res res = {
2632                .server = server,
2633                .fattr = fattr,
2634                .fh = fhandle,
2635        };
2636        struct rpc_message msg = {
2637                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2638                .rpc_argp = &args,
2639                .rpc_resp = &res,
2640        };
2641
2642        nfs_fattr_init(fattr);
2643
2644        dprintk("NFS call  lookup %s\n", name->name);
2645        status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2646        dprintk("NFS reply lookup: %d\n", status);
2647        return status;
2648}
2649
2650static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2651{
2652        fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2653                NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2654        fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2655        fattr->nlink = 2;
2656}
2657
2658static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2659                                   struct qstr *name, struct nfs_fh *fhandle,
2660                                   struct nfs_fattr *fattr)
2661{
2662        struct nfs4_exception exception = { };
2663        struct rpc_clnt *client = *clnt;
2664        int err;
2665        do {
2666                err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2667                switch (err) {
2668                case -NFS4ERR_BADNAME:
2669                        err = -ENOENT;
2670                        goto out;
2671                case -NFS4ERR_MOVED:
2672                        err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2673                        goto out;
2674                case -NFS4ERR_WRONGSEC:
2675                        err = -EPERM;
2676                        if (client != *clnt)
2677                                goto out;
2678
2679                        client = nfs4_create_sec_client(client, dir, name);
2680                        if (IS_ERR(client))
2681                                return PTR_ERR(client);
2682
2683                        exception.retry = 1;
2684                        break;
2685                default:
2686                        err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2687                }
2688        } while (exception.retry);
2689
2690out:
2691        if (err == 0)
2692                *clnt = client;
2693        else if (client != *clnt)
2694                rpc_shutdown_client(client);
2695
2696        return err;
2697}
2698
2699static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2700                            struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2701{
2702        int status;
2703        struct rpc_clnt *client = NFS_CLIENT(dir);
2704
2705        status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2706        if (client != NFS_CLIENT(dir)) {
2707                rpc_shutdown_client(client);
2708                nfs_fixup_secinfo_attributes(fattr);
2709        }
2710        return status;
2711}
2712
2713struct rpc_clnt *
2714nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2715                            struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2716{
2717        int status;
2718        struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2719
2720        status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2721        if (status < 0) {
2722                rpc_shutdown_client(client);
2723                return ERR_PTR(status);
2724        }
2725        return client;
2726}
2727
2728static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2729{
2730        struct nfs_server *server = NFS_SERVER(inode);
2731        struct nfs4_accessargs args = {
2732                .fh = NFS_FH(inode),
2733                .bitmask = server->cache_consistency_bitmask,
2734        };
2735        struct nfs4_accessres res = {
2736                .server = server,
2737        };
2738        struct rpc_message msg = {
2739                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2740                .rpc_argp = &args,
2741                .rpc_resp = &res,
2742                .rpc_cred = entry->cred,
2743        };
2744        int mode = entry->mask;
2745        int status;
2746
2747        /*
2748         * Determine which access bits we want to ask for...
2749         */
2750        if (mode & MAY_READ)
2751                args.access |= NFS4_ACCESS_READ;
2752        if (S_ISDIR(inode->i_mode)) {
2753                if (mode & MAY_WRITE)
2754                        args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2755                if (mode & MAY_EXEC)
2756                        args.access |= NFS4_ACCESS_LOOKUP;
2757        } else {
2758                if (mode & MAY_WRITE)
2759                        args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2760                if (mode & MAY_EXEC)
2761                        args.access |= NFS4_ACCESS_EXECUTE;
2762        }
2763
2764        res.fattr = nfs_alloc_fattr();
2765        if (res.fattr == NULL)
2766                return -ENOMEM;
2767
2768        status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2769        if (!status) {
2770                entry->mask = 0;
2771                if (res.access & NFS4_ACCESS_READ)
2772                        entry->mask |= MAY_READ;
2773                if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2774                        entry->mask |= MAY_WRITE;
2775                if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2776                        entry->mask |= MAY_EXEC;
2777                nfs_refresh_inode(inode, res.fattr);
2778        }
2779        nfs_free_fattr(res.fattr);
2780        return status;
2781}
2782
2783static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2784{
2785        struct nfs4_exception exception = { };
2786        int err;
2787        do {
2788                err = nfs4_handle_exception(NFS_SERVER(inode),
2789                                _nfs4_proc_access(inode, entry),
2790                                &exception);
2791        } while (exception.retry);
2792        return err;
2793}
2794
2795/*
2796 * TODO: For the time being, we don't try to get any attributes
2797 * along with any of the zero-copy operations READ, READDIR,
2798 * READLINK, WRITE.
2799 *
2800 * In the case of the first three, we want to put the GETATTR
2801 * after the read-type operation -- this is because it is hard
2802 * to predict the length of a GETATTR response in v4, and thus
2803 * align the READ data correctly.  This means that the GETATTR
2804 * may end up partially falling into the page cache, and we should
2805 * shift it into the 'tail' of the xdr_buf before processing.
2806 * To do this efficiently, we need to know the total length
2807 * of data received, which doesn't seem to be available outside
2808 * of the RPC layer.
2809 *
2810 * In the case of WRITE, we also want to put the GETATTR after
2811 * the operation -- in this case because we want to make sure
2812 * we get the post-operation mtime and size.
2813 *
2814 * Both of these changes to the XDR layer would in fact be quite
2815 * minor, but I decided to leave them for a subsequent patch.
2816 */
2817static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2818                unsigned int pgbase, unsigned int pglen)
2819{
2820        struct nfs4_readlink args = {
2821                .fh       = NFS_FH(inode),
2822                .pgbase   = pgbase,
2823                .pglen    = pglen,
2824                .pages    = &page,
2825        };
2826        struct nfs4_readlink_res res;
2827        struct rpc_message msg = {
2828                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2829                .rpc_argp = &args,
2830                .rpc_resp = &res,
2831        };
2832
2833        return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2834}
2835
2836static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2837                unsigned int pgbase, unsigned int pglen)
2838{
2839        struct nfs4_exception exception = { };
2840        int err;
2841        do {
2842                err = nfs4_handle_exception(NFS_SERVER(inode),
2843                                _nfs4_proc_readlink(inode, page, pgbase, pglen),
2844                                &exception);
2845        } while (exception.retry);
2846        return err;
2847}
2848
2849/*
2850 * This is just for mknod.  open(O_CREAT) will always do ->open_context().
2851 */
2852static int
2853nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2854                 int flags)
2855{
2856        struct nfs_open_context *ctx;
2857        struct nfs4_state *state;
2858        int status = 0;
2859
2860        ctx = alloc_nfs_open_context(dentry, FMODE_READ);
2861        if (IS_ERR(ctx))
2862                return PTR_ERR(ctx);
2863
2864        sattr->ia_mode &= ~current_umask();
2865        state = nfs4_do_open(dir, dentry, ctx->mode,
2866                        flags, sattr, ctx->cred,
2867                        &ctx->mdsthreshold);
2868        d_drop(dentry);
2869        if (IS_ERR(state)) {
2870                status = PTR_ERR(state);
2871                goto out;
2872        }
2873        d_add(dentry, igrab(state->inode));
2874        nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2875        ctx->state = state;
2876out:
2877        put_nfs_open_context(ctx);
2878        return status;
2879}
2880
2881static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2882{
2883        struct nfs_server *server = NFS_SERVER(dir);
2884        struct nfs_removeargs args = {
2885                .fh = NFS_FH(dir),
2886                .name = *name,
2887        };
2888        struct nfs_removeres res = {
2889                .server = server,
2890        };
2891        struct rpc_message msg = {
2892                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2893                .rpc_argp = &args,
2894                .rpc_resp = &res,
2895        };
2896        int status;
2897
2898        status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2899        if (status == 0)
2900                update_changeattr(dir, &res.cinfo);
2901        return status;
2902}
2903
2904static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2905{
2906        struct nfs4_exception exception = { };
2907        int err;
2908        do {
2909                err = nfs4_handle_exception(NFS_SERVER(dir),
2910                                _nfs4_proc_remove(dir, name),
2911                                &exception);
2912        } while (exception.retry);
2913        return err;
2914}
2915
2916static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2917{
2918        struct nfs_server *server = NFS_SERVER(dir);
2919        struct nfs_removeargs *args = msg->rpc_argp;
2920        struct nfs_removeres *res = msg->rpc_resp;
2921
2922        res->server = server;
2923        msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2924        nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2925}
2926
2927static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
2928{
2929        if (nfs4_setup_sequence(NFS_SERVER(data->dir),
2930                                &data->args.seq_args,
2931                                &data->res.seq_res,
2932                                task))
2933                return;
2934        rpc_call_start(task);
2935}
2936
2937static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2938{
2939        struct nfs_removeres *res = task->tk_msg.rpc_resp;
2940
2941        if (!nfs4_sequence_done(task, &res->seq_res))
2942                return 0;
2943        if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2944                return 0;
2945        update_changeattr(dir, &res->cinfo);
2946        return 1;
2947}
2948
2949static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2950{
2951        struct nfs_server *server = NFS_SERVER(dir);
2952        struct nfs_renameargs *arg = msg->rpc_argp;
2953        struct nfs_renameres *res = msg->rpc_resp;
2954
2955        msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2956        res->server = server;
2957        nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2958}
2959
2960static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
2961{
2962        if (nfs4_setup_sequence(NFS_SERVER(data->old_dir),
2963                                &data->args.seq_args,
2964                                &data->res.seq_res,
2965                                task))
2966                return;
2967        rpc_call_start(task);
2968}
2969
2970static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2971                                 struct inode *new_dir)
2972{
2973        struct nfs_renameres *res = task->tk_msg.rpc_resp;
2974
2975        if (!nfs4_sequence_done(task, &res->seq_res))
2976                return 0;
2977        if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2978                return 0;
2979
2980        update_changeattr(old_dir, &res->old_cinfo);
2981        update_changeattr(new_dir, &res->new_cinfo);
2982        return 1;
2983}
2984
2985static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2986                struct inode *new_dir, struct qstr *new_name)
2987{
2988        struct nfs_server *server = NFS_SERVER(old_dir);
2989        struct nfs_renameargs arg = {
2990                .old_dir = NFS_FH(old_dir),
2991                .new_dir = NFS_FH(new_dir),
2992                .old_name = old_name,
2993                .new_name = new_name,
2994        };
2995        struct nfs_renameres res = {
2996                .server = server,
2997        };
2998        struct rpc_message msg = {
2999                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3000                .rpc_argp = &arg,
3001                .rpc_resp = &res,
3002        };
3003        int status = -ENOMEM;
3004        
3005        status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3006        if (!status) {
3007                update_changeattr(old_dir, &res.old_cinfo);
3008                update_changeattr(new_dir, &res.new_cinfo);
3009        }
3010        return status;
3011}
3012
3013static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3014                struct inode *new_dir, struct qstr *new_name)
3015{
3016        struct nfs4_exception exception = { };
3017        int err;
3018        do {
3019                err = nfs4_handle_exception(NFS_SERVER(old_dir),
3020                                _nfs4_proc_rename(old_dir, old_name,
3021                                        new_dir, new_name),
3022                                &exception);
3023        } while (exception.retry);
3024        return err;
3025}
3026
3027static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3028{
3029        struct nfs_server *server = NFS_SERVER(inode);
3030        struct nfs4_link_arg arg = {
3031                .fh     = NFS_FH(inode),
3032                .dir_fh = NFS_FH(dir),
3033                .name   = name,
3034                .bitmask = server->attr_bitmask,
3035        };
3036        struct nfs4_link_res res = {
3037                .server = server,
3038        };
3039        struct rpc_message msg = {
3040                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3041                .rpc_argp = &arg,
3042                .rpc_resp = &res,
3043        };
3044        int status = -ENOMEM;
3045
3046        res.fattr = nfs_alloc_fattr();
3047        if (res.fattr == NULL)
3048                goto out;
3049
3050        status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3051        if (!status) {
3052                update_changeattr(dir, &res.cinfo);
3053                nfs_post_op_update_inode(inode, res.fattr);
3054        }
3055out:
3056        nfs_free_fattr(res.fattr);
3057        return status;
3058}
3059
3060static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3061{
3062        struct nfs4_exception exception = { };
3063        int err;
3064        do {
3065                err = nfs4_handle_exception(NFS_SERVER(inode),
3066                                _nfs4_proc_link(inode, dir, name),
3067                                &exception);
3068        } while (exception.retry);
3069        return err;
3070}
3071
3072struct nfs4_createdata {
3073        struct rpc_message msg;
3074        struct nfs4_create_arg arg;
3075        struct nfs4_create_res res;
3076        struct nfs_fh fh;
3077        struct nfs_fattr fattr;
3078};
3079
3080static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3081                struct qstr *name, struct iattr *sattr, u32 ftype)
3082{
3083        struct nfs4_createdata *data;
3084
3085        data = kzalloc(sizeof(*data), GFP_KERNEL);
3086        if (data != NULL) {
3087                struct nfs_server *server = NFS_SERVER(dir);
3088
3089                data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3090                data->msg.rpc_argp = &data->arg;
3091                data->msg.rpc_resp = &data->res;
3092                data->arg.dir_fh = NFS_FH(dir);
3093                data->arg.server = server;
3094                data->arg.name = name;
3095                data->arg.attrs = sattr;
3096                data->arg.ftype = ftype;
3097                data->arg.bitmask = server->attr_bitmask;
3098                data->res.server = server;
3099                data->res.fh = &data->fh;
3100                data->res.fattr = &data->fattr;
3101                nfs_fattr_init(data->res.fattr);
3102        }
3103        return data;
3104}
3105
3106static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3107{
3108        int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3109                                    &data->arg.seq_args, &data->res.seq_res, 1);
3110        if (status == 0) {
3111                update_changeattr(dir, &data->res.dir_cinfo);
3112                status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3113        }
3114        return status;
3115}
3116
3117static void nfs4_free_createdata(struct nfs4_createdata *data)
3118{
3119        kfree(data);
3120}
3121
3122static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3123                struct page *page, unsigned int len, struct iattr *sattr)
3124{
3125        struct nfs4_createdata *data;
3126        int status = -ENAMETOOLONG;
3127
3128        if (len > NFS4_MAXPATHLEN)
3129                goto out;
3130
3131        status = -ENOMEM;
3132        data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3133        if (data == NULL)
3134                goto out;
3135
3136        data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3137        data->arg.u.symlink.pages = &page;
3138        data->arg.u.symlink.len = len;
3139        
3140        status = nfs4_do_create(dir, dentry, data);
3141
3142        nfs4_free_createdata(data);
3143out:
3144        return status;
3145}
3146
3147static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3148                struct page *page, unsigned int len, struct iattr *sattr)
3149{
3150        struct nfs4_exception exception = { };
3151        int err;
3152        do {
3153                err = nfs4_handle_exception(NFS_SERVER(dir),
3154                                _nfs4_proc_symlink(dir, dentry, page,
3155                                                        len, sattr),
3156                                &exception);
3157        } while (exception.retry);
3158        return err;
3159}
3160
3161static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3162                struct iattr *sattr)
3163{
3164        struct nfs4_createdata *data;
3165        int status = -ENOMEM;
3166
3167        data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3168        if (data == NULL)
3169                goto out;
3170
3171        status = nfs4_do_create(dir, dentry, data);
3172
3173        nfs4_free_createdata(data);
3174out:
3175        return status;
3176}
3177
3178static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3179                struct iattr *sattr)
3180{
3181        struct nfs4_exception exception = { };
3182        int err;
3183
3184        sattr->ia_mode &= ~current_umask();
3185        do {
3186                err = nfs4_handle_exception(NFS_SERVER(dir),
3187                                _nfs4_proc_mkdir(dir, dentry, sattr),
3188                                &exception);
3189        } while (exception.retry);
3190        return err;
3191}
3192
3193static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3194                u64 cookie, struct page **pages, unsigned int count, int plus)
3195{
3196        struct inode            *dir = dentry->d_inode;
3197        struct nfs4_readdir_arg args = {
3198                .fh = NFS_FH(dir),
3199                .pages = pages,
3200                .pgbase = 0,
3201                .count = count,
3202                .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3203                .plus = plus,
3204        };
3205        struct nfs4_readdir_res res;
3206        struct rpc_message msg = {
3207                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3208                .rpc_argp = &args,
3209                .rpc_resp = &res,
3210                .rpc_cred = cred,
3211        };
3212        int                     status;
3213
3214        dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3215                        dentry->d_parent->d_name.name,
3216                        dentry->d_name.name,
3217                        (unsigned long long)cookie);
3218        nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3219        res.pgbase = args.pgbase;
3220        status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3221        if (status >= 0) {
3222                memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3223                status += args.pgbase;
3224        }
3225
3226        nfs_invalidate_atime(dir);
3227
3228        dprintk("%s: returns %d\n", __func__, status);
3229        return status;
3230}
3231
3232static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3233                u64 cookie, struct page **pages, unsigned int count, int plus)
3234{
3235        struct nfs4_exception exception = { };
3236        int err;
3237        do {
3238                err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3239                                _nfs4_proc_readdir(dentry, cred, cookie,
3240                                        pages, count, plus),
3241                                &exception);
3242        } while (exception.retry);
3243        return err;
3244}
3245
3246static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3247                struct iattr *sattr, dev_t rdev)
3248{
3249        struct nfs4_createdata *data;
3250        int mode = sattr->ia_mode;
3251        int status = -ENOMEM;
3252
3253        BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3254        BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3255
3256        data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3257        if (data == NULL)
3258                goto out;
3259
3260        if (S_ISFIFO(mode))
3261                data->arg.ftype = NF4FIFO;
3262        else if (S_ISBLK(mode)) {
3263                data->arg.ftype = NF4BLK;
3264                data->arg.u.device.specdata1 = MAJOR(rdev);
3265                data->arg.u.device.specdata2 = MINOR(rdev);
3266        }
3267        else if (S_ISCHR(mode)) {
3268                data->arg.ftype = NF4CHR;
3269                data->arg.u.device.specdata1 = MAJOR(rdev);
3270                data->arg.u.device.specdata2 = MINOR(rdev);
3271        }
3272        
3273        status = nfs4_do_create(dir, dentry, data);
3274
3275        nfs4_free_createdata(data);
3276out:
3277        return status;
3278}
3279
3280static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3281                struct iattr *sattr, dev_t rdev)
3282{
3283        struct nfs4_exception exception = { };
3284        int err;
3285
3286        sattr->ia_mode &= ~current_umask();
3287        do {
3288                err = nfs4_handle_exception(NFS_SERVER(dir),
3289                                _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3290                                &exception);
3291        } while (exception.retry);
3292        return err;
3293}
3294
3295static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3296                 struct nfs_fsstat *fsstat)
3297{
3298        struct nfs4_statfs_arg args = {
3299                .fh = fhandle,
3300                .bitmask = server->attr_bitmask,
3301        };
3302        struct nfs4_statfs_res res = {
3303                .fsstat = fsstat,
3304        };
3305        struct rpc_message msg = {
3306                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3307                .rpc_argp = &args,
3308                .rpc_resp = &res,
3309        };
3310
3311        nfs_fattr_init(fsstat->fattr);
3312        return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3313}
3314
3315static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3316{
3317        struct nfs4_exception exception = { };
3318        int err;
3319        do {
3320                err = nfs4_handle_exception(server,
3321                                _nfs4_proc_statfs(server, fhandle, fsstat),
3322                                &exception);
3323        } while (exception.retry);
3324        return err;
3325}
3326
3327static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3328                struct nfs_fsinfo *fsinfo)
3329{
3330        struct nfs4_fsinfo_arg args = {
3331                .fh = fhandle,
3332                .bitmask = server->attr_bitmask,
3333        };
3334        struct nfs4_fsinfo_res res = {
3335                .fsinfo = fsinfo,
3336        };
3337        struct rpc_message msg = {
3338                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3339                .rpc_argp = &args,
3340                .rpc_resp = &res,
3341        };
3342
3343        return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3344}
3345
3346static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3347{
3348        struct nfs4_exception exception = { };
3349        int err;
3350
3351        do {
3352                err = nfs4_handle_exception(server,
3353                                _nfs4_do_fsinfo(server, fhandle, fsinfo),
3354                                &exception);
3355        } while (exception.retry);
3356        return err;
3357}
3358
3359static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3360{
3361        int error;
3362
3363        nfs_fattr_init(fsinfo->fattr);
3364        error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3365        if (error == 0)
3366                set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3367
3368        return error;
3369}
3370
3371static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3372                struct nfs_pathconf *pathconf)
3373{
3374        struct nfs4_pathconf_arg args = {
3375                .fh = fhandle,
3376                .bitmask = server->attr_bitmask,
3377        };
3378        struct nfs4_pathconf_res res = {
3379                .pathconf = pathconf,
3380        };
3381        struct rpc_message msg = {
3382                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3383                .rpc_argp = &args,
3384                .rpc_resp = &res,
3385        };
3386
3387        /* None of the pathconf attributes are mandatory to implement */
3388        if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3389                memset(pathconf, 0, sizeof(*pathconf));
3390                return 0;
3391        }
3392
3393        nfs_fattr_init(pathconf->fattr);
3394        return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3395}
3396
3397static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3398                struct nfs_pathconf *pathconf)
3399{
3400        struct nfs4_exception exception = { };
3401        int err;
3402
3403        do {
3404                err = nfs4_handle_exception(server,
3405                                _nfs4_proc_pathconf(server, fhandle, pathconf),
3406                                &exception);
3407        } while (exception.retry);
3408        return err;
3409}
3410
3411void __nfs4_read_done_cb(struct nfs_read_data *data)
3412{
3413        nfs_invalidate_atime(data->header->inode);
3414}
3415
3416static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3417{
3418        struct nfs_server *server = NFS_SERVER(data->header->inode);
3419
3420        if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3421                rpc_restart_call_prepare(task);
3422                return -EAGAIN;
3423        }
3424
3425        __nfs4_read_done_cb(data);
3426        if (task->tk_status > 0)
3427                renew_lease(server, data->timestamp);
3428        return 0;
3429}
3430
3431static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3432{
3433
3434        dprintk("--> %s\n", __func__);
3435
3436        if (!nfs4_sequence_done(task, &data->res.seq_res))
3437                return -EAGAIN;
3438
3439        return data->read_done_cb ? data->read_done_cb(task, data) :
3440                                    nfs4_read_done_cb(task, data);
3441}
3442
3443static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3444{
3445        data->timestamp   = jiffies;
3446        data->read_done_cb = nfs4_read_done_cb;
3447        msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3448        nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3449}
3450
3451static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3452{
3453        if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3454                                &data->args.seq_args,
3455                                &data->res.seq_res,
3456                                task))
3457                return;
3458        rpc_call_start(task);
3459}
3460
3461static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3462{
3463        struct inode *inode = data->header->inode;
3464        
3465        if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3466                rpc_restart_call_prepare(task);
3467                return -EAGAIN;
3468        }
3469        if (task->tk_status >= 0) {
3470                renew_lease(NFS_SERVER(inode), data->timestamp);
3471                nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3472        }
3473        return 0;
3474}
3475
3476static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3477{
3478        if (!nfs4_sequence_done(task, &data->res.seq_res))
3479                return -EAGAIN;
3480        return data->write_done_cb ? data->write_done_cb(task, data) :
3481                nfs4_write_done_cb(task, data);
3482}
3483
3484static
3485bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3486{
3487        const struct nfs_pgio_header *hdr = data->header;
3488
3489        /* Don't request attributes for pNFS or O_DIRECT writes */
3490        if (data->ds_clp != NULL || hdr->dreq != NULL)
3491                return false;
3492        /* Otherwise, request attributes if and only if we don't hold
3493         * a delegation
3494         */
3495        return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3496}
3497
3498static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3499{
3500        struct nfs_server *server = NFS_SERVER(data->header->inode);
3501
3502        if (!nfs4_write_need_cache_consistency_data(data)) {
3503                data->args.bitmask = NULL;
3504                data->res.fattr = NULL;
3505        } else
3506                data->args.bitmask = server->cache_consistency_bitmask;
3507
3508        if (!data->write_done_cb)
3509                data->write_done_cb = nfs4_write_done_cb;
3510        data->res.server = server;
3511        data->timestamp   = jiffies;
3512
3513        msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3514        nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3515}
3516
3517static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3518{
3519        if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3520                                &data->args.seq_args,
3521                                &data->res.seq_res,
3522                                task))
3523                return;
3524        rpc_call_start(task);
3525}
3526
3527static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3528{
3529        if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3530                                &data->args.seq_args,
3531                                &data->res.seq_res,
3532                                task))
3533                return;
3534        rpc_call_start(task);
3535}
3536
3537static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3538{
3539        struct inode *inode = data->inode;
3540
3541        if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3542                rpc_restart_call_prepare(task);
3543                return -EAGAIN;
3544        }
3545        return 0;
3546}
3547
3548static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3549{
3550        if (!nfs4_sequence_done(task, &data->res.seq_res))
3551                return -EAGAIN;
3552        return data->commit_done_cb(task, data);
3553}
3554
3555static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3556{
3557        struct nfs_server *server = NFS_SERVER(data->inode);
3558
3559        if (data->commit_done_cb == NULL)
3560                data->commit_done_cb = nfs4_commit_done_cb;
3561        data->res.server = server;
3562        msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3563        nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3564}
3565
3566struct nfs4_renewdata {
3567        struct nfs_client       *client;
3568        unsigned long           timestamp;
3569};
3570
3571/*
3572 * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3573 * standalone procedure for queueing an asynchronous RENEW.
3574 */
3575static void nfs4_renew_release(void *calldata)
3576{
3577        struct nfs4_renewdata *data = calldata;
3578        struct nfs_client *clp = data->client;
3579
3580        if (atomic_read(&clp->cl_count) > 1)
3581                nfs4_schedule_state_renewal(clp);
3582        nfs_put_client(clp);
3583        kfree(data);
3584}
3585
3586static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3587{
3588        struct nfs4_renewdata *data = calldata;
3589        struct nfs_client *clp = data->client;
3590        unsigned long timestamp = data->timestamp;
3591
3592        if (task->tk_status < 0) {
3593                /* Unless we're shutting down, schedule state recovery! */
3594                if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3595                        return;
3596                if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3597                        nfs4_schedule_lease_recovery(clp);
3598                        return;
3599                }
3600                nfs4_schedule_path_down_recovery(clp);
3601        }
3602        do_renew_lease(clp, timestamp);
3603}
3604
3605static const struct rpc_call_ops nfs4_renew_ops = {
3606        .rpc_call_done = nfs4_renew_done,
3607        .rpc_release = nfs4_renew_release,
3608};
3609
3610static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3611{
3612        struct rpc_message msg = {
3613                .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3614                .rpc_argp       = clp,
3615                .rpc_cred       = cred,
3616        };
3617        struct nfs4_renewdata *data;
3618
3619        if (renew_flags == 0)
3620                return 0;
3621        if (!atomic_inc_not_zero(&clp->cl_count))
3622                return -EIO;
3623        data = kmalloc(sizeof(*data), GFP_NOFS);
3624        if (data == NULL)
3625                return -ENOMEM;
3626        data->client = clp;
3627        data->timestamp = jiffies;
3628        return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3629                        &nfs4_renew_ops, data);
3630}
3631
3632static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3633{
3634        struct rpc_message msg = {
3635                .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3636                .rpc_argp       = clp,
3637                .rpc_cred       = cred,
3638        };
3639        unsigned long now = jiffies;
3640        int status;
3641
3642        status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3643        if (status < 0)
3644                return status;
3645        do_renew_lease(clp, now);
3646        return 0;
3647}
3648
3649static inline int nfs4_server_supports_acls(struct nfs_server *server)
3650{
3651        return (server->caps & NFS_CAP_ACLS)
3652                && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3653                && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3654}
3655
3656/* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3657 * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3658 * the stack.
3659 */
3660#define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3661
3662static int buf_to_pages_noslab(const void *buf, size_t buflen,
3663                struct page **pages, unsigned int *pgbase)
3664{
3665        struct page *newpage, **spages;
3666        int rc = 0;
3667        size_t len;
3668        spages = pages;
3669
3670        do {
3671                len = min_t(size_t, PAGE_SIZE, buflen);
3672                newpage = alloc_page(GFP_KERNEL);
3673
3674                if (newpage == NULL)
3675                        goto unwind;
3676                memcpy(page_address(newpage), buf, len);
3677                buf += len;
3678                buflen -= len;
3679                *pages++ = newpage;
3680                rc++;
3681        } while (buflen != 0);
3682
3683        return rc;
3684
3685unwind:
3686        for(; rc > 0; rc--)
3687                __free_page(spages[rc-1]);
3688        return -ENOMEM;
3689}
3690
3691struct nfs4_cached_acl {
3692        int cached;
3693        size_t len;
3694        char data[0];
3695};
3696
3697static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3698{
3699        struct nfs_inode *nfsi = NFS_I(inode);
3700
3701        spin_lock(&inode->i_lock);
3702        kfree(nfsi->nfs4_acl);
3703        nfsi->nfs4_acl = acl;
3704        spin_unlock(&inode->i_lock);
3705}
3706
3707static void nfs4_zap_acl_attr(struct inode *inode)
3708{
3709        nfs4_set_cached_acl(inode, NULL);
3710}
3711
3712static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3713{
3714        struct nfs_inode *nfsi = NFS_I(inode);
3715        struct nfs4_cached_acl *acl;
3716        int ret = -ENOENT;
3717
3718        spin_lock(&inode->i_lock);
3719        acl = nfsi->nfs4_acl;
3720        if (acl == NULL)
3721                goto out;
3722        if (buf == NULL) /* user is just asking for length */
3723                goto out_len;
3724        if (acl->cached == 0)
3725                goto out;
3726        ret = -ERANGE; /* see getxattr(2) man page */
3727        if (acl->len > buflen)
3728                goto out;
3729        memcpy(buf, acl->data, acl->len);
3730out_len:
3731        ret = acl->len;
3732out:
3733        spin_unlock(&inode->i_lock);
3734        return ret;
3735}
3736
3737static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3738{
3739        struct nfs4_cached_acl *acl;
3740        size_t buflen = sizeof(*acl) + acl_len;
3741
3742        if (buflen <= PAGE_SIZE) {
3743                acl = kmalloc(buflen, GFP_KERNEL);
3744                if (acl == NULL)
3745                        goto out;
3746                acl->cached = 1;
3747                _copy_from_pages(acl->data, pages, pgbase, acl_len);
3748        } else {
3749                acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3750                if (acl == NULL)
3751                        goto out;
3752                acl->cached = 0;
3753        }
3754        acl->len = acl_len;
3755out:
3756        nfs4_set_cached_acl(inode, acl);
3757}
3758
3759/*
3760 * The getxattr API returns the required buffer length when called with a
3761 * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3762 * the required buf.  On a NULL buf, we send a page of data to the server
3763 * guessing that the ACL request can be serviced by a page. If so, we cache
3764 * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3765 * the cache. If not so, we throw away the page, and cache the required
3766 * length. The next getxattr call will then produce another round trip to
3767 * the server, this time with the input buf of the required size.
3768 */
3769static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3770{
3771        struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3772        struct nfs_getaclargs args = {
3773                .fh = NFS_FH(inode),
3774                .acl_pages = pages,
3775                .acl_len = buflen,
3776        };
3777        struct nfs_getaclres res = {
3778                .acl_len = buflen,
3779        };
3780        struct rpc_message msg = {
3781                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3782                .rpc_argp = &args,
3783                .rpc_resp = &res,
3784        };
3785        unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3786        int ret = -ENOMEM, i;
3787
3788        /* As long as we're doing a round trip to the server anyway,
3789         * let's be prepared for a page of acl data. */
3790        if (npages == 0)
3791                npages = 1;
3792        if (npages > ARRAY_SIZE(pages))
3793                return -ERANGE;
3794
3795        for (i = 0; i < npages; i++) {
3796                pages[i] = alloc_page(GFP_KERNEL);
3797                if (!pages[i])
3798                        goto out_free;
3799        }
3800
3801        /* for decoding across pages */
3802        res.acl_scratch = alloc_page(GFP_KERNEL);
3803        if (!res.acl_scratch)
3804                goto out_free;
3805
3806        args.acl_len = npages * PAGE_SIZE;
3807        args.acl_pgbase = 0;
3808
3809        dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3810                __func__, buf, buflen, npages, args.acl_len);
3811        ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3812                             &msg, &args.seq_args, &res.seq_res, 0);
3813        if (ret)
3814                goto out_free;
3815
3816        /* Handle the case where the passed-in buffer is too short */
3817        if (res.acl_flags & NFS4_ACL_TRUNC) {
3818                /* Did the user only issue a request for the acl length? */
3819                if (buf == NULL)
3820                        goto out_ok;
3821                ret = -ERANGE;
3822                goto out_free;
3823        }
3824        nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
3825        if (buf)
3826                _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
3827out_ok:
3828        ret = res.acl_len;
3829out_free:
3830        for (i = 0; i < npages; i++)
3831                if (pages[i])
3832                        __free_page(pages[i]);
3833        if (res.acl_scratch)
3834                __free_page(res.acl_scratch);
3835        return ret;
3836}
3837
3838static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3839{
3840        struct nfs4_exception exception = { };
3841        ssize_t ret;
3842        do {
3843                ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3844                if (ret >= 0)
3845                        break;
3846                ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3847        } while (exception.retry);
3848        return ret;
3849}
3850
3851static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3852{
3853        struct nfs_server *server = NFS_SERVER(inode);
3854        int ret;
3855
3856        if (!nfs4_server_supports_acls(server))
3857                return -EOPNOTSUPP;
3858        ret = nfs_revalidate_inode(server, inode);
3859        if (ret < 0)
3860                return ret;
3861        if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3862                nfs_zap_acl_cache(inode);
3863        ret = nfs4_read_cached_acl(inode, buf, buflen);
3864        if (ret != -ENOENT)
3865                /* -ENOENT is returned if there is no ACL or if there is an ACL
3866                 * but no cached acl data, just the acl length */
3867                return ret;
3868        return nfs4_get_acl_uncached(inode, buf, buflen);
3869}
3870
3871static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3872{
3873        struct nfs_server *server = NFS_SERVER(inode);
3874        struct page *pages[NFS4ACL_MAXPAGES];
3875        struct nfs_setaclargs arg = {
3876                .fh             = NFS_FH(inode),
3877                .acl_pages      = pages,
3878                .acl_len        = buflen,
3879        };
3880        struct nfs_setaclres res;
3881        struct rpc_message msg = {
3882                .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3883                .rpc_argp       = &arg,
3884                .rpc_resp       = &res,
3885        };
3886        unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3887        int ret, i;
3888
3889        if (!nfs4_server_supports_acls(server))
3890                return -EOPNOTSUPP;
3891        if (npages > ARRAY_SIZE(pages))
3892                return -ERANGE;
3893        i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3894        if (i < 0)
3895                return i;
3896        nfs4_inode_return_delegation(inode);
3897        ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3898
3899        /*
3900         * Free each page after tx, so the only ref left is
3901         * held by the network stack
3902         */
3903        for (; i > 0; i--)
3904                put_page(pages[i-1]);
3905
3906        /*
3907         * Acl update can result in inode attribute update.
3908         * so mark the attribute cache invalid.
3909         */
3910        spin_lock(&inode->i_lock);
3911        NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3912        spin_unlock(&inode->i_lock);
3913        nfs_access_zap_cache(inode);
3914        nfs_zap_acl_cache(inode);
3915        return ret;
3916}
3917
3918static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3919{
3920        struct nfs4_exception exception = { };
3921        int err;
3922        do {
3923                err = nfs4_handle_exception(NFS_SERVER(inode),
3924                                __nfs4_proc_set_acl(inode, buf, buflen),
3925                                &exception);
3926        } while (exception.retry);
3927        return err;
3928}
3929
3930static int
3931nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3932{
3933        struct nfs_client *clp = server->nfs_client;
3934
3935        if (task->tk_status >= 0)
3936                return 0;
3937        switch(task->tk_status) {
3938                case -NFS4ERR_DELEG_REVOKED:
3939                case -NFS4ERR_ADMIN_REVOKED:
3940                case -NFS4ERR_BAD_STATEID:
3941                        if (state == NULL)
3942                                break;
3943                        nfs_remove_bad_delegation(state->inode);
3944                case -NFS4ERR_OPENMODE:
3945                        if (state == NULL)
3946                                break;
3947                        nfs4_schedule_stateid_recovery(server, state);
3948                        goto wait_on_recovery;
3949                case -NFS4ERR_EXPIRED:
3950                        if (state != NULL)
3951                                nfs4_schedule_stateid_recovery(server, state);
3952                case -NFS4ERR_STALE_STATEID:
3953                case -NFS4ERR_STALE_CLIENTID:
3954                        nfs4_schedule_lease_recovery(clp);
3955                        goto wait_on_recovery;
3956#if defined(CONFIG_NFS_V4_1)
3957                case -NFS4ERR_BADSESSION:
3958                case -NFS4ERR_BADSLOT:
3959                case -NFS4ERR_BAD_HIGH_SLOT:
3960                case -NFS4ERR_DEADSESSION:
3961                case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3962                case -NFS4ERR_SEQ_FALSE_RETRY:
3963                case -NFS4ERR_SEQ_MISORDERED:
3964                        dprintk("%s ERROR %d, Reset session\n", __func__,
3965                                task->tk_status);
3966                        nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
3967                        task->tk_status = 0;
3968                        return -EAGAIN;
3969#endif /* CONFIG_NFS_V4_1 */
3970                case -NFS4ERR_DELAY:
3971                        nfs_inc_server_stats(server, NFSIOS_DELAY);
3972                case -NFS4ERR_GRACE:
3973                case -EKEYEXPIRED:
3974                        rpc_delay(task, NFS4_POLL_RETRY_MAX);
3975                        task->tk_status = 0;
3976                        return -EAGAIN;
3977                case -NFS4ERR_RETRY_UNCACHED_REP:
3978                case -NFS4ERR_OLD_STATEID:
3979                        task->tk_status = 0;
3980                        return -EAGAIN;
3981        }
3982        task->tk_status = nfs4_map_errors(task->tk_status);
3983        return 0;
3984wait_on_recovery:
3985        rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3986        if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3987                rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3988        task->tk_status = 0;
3989        return -EAGAIN;
3990}
3991
3992static void nfs4_init_boot_verifier(const struct nfs_client *clp,
3993                                    nfs4_verifier *bootverf)
3994{
3995        __be32 verf[2];
3996
3997        if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
3998                /* An impossible timestamp guarantees this value
3999                 * will never match a generated boot time. */
4000                verf[0] = 0;
4001                verf[1] = (__be32)(NSEC_PER_SEC + 1);
4002        } else {
4003                struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4004                verf[0] = (__be32)nn->boot_time.tv_sec;
4005                verf[1] = (__be32)nn->boot_time.tv_nsec;
4006        }
4007        memcpy(bootverf->data, verf, sizeof(bootverf->data));
4008}
4009
4010/**
4011 * nfs4_proc_setclientid - Negotiate client ID
4012 * @clp: state data structure
4013 * @program: RPC program for NFSv4 callback service
4014 * @port: IP port number for NFS4 callback service
4015 * @cred: RPC credential to use for this call
4016 * @res: where to place the result
4017 *
4018 * Returns zero, a negative errno, or a negative NFS4ERR status code.
4019 */
4020int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4021                unsigned short port, struct rpc_cred *cred,
4022                struct nfs4_setclientid_res *res)
4023{
4024        nfs4_verifier sc_verifier;
4025        struct nfs4_setclientid setclientid = {
4026                .sc_verifier = &sc_verifier,
4027                .sc_prog = program,
4028                .sc_cb_ident = clp->cl_cb_ident,
4029        };
4030        struct rpc_message msg = {
4031                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4032                .rpc_argp = &setclientid,
4033                .rpc_resp = res,
4034                .rpc_cred = cred,
4035        };
4036        int status;
4037
4038        /* nfs_client_id4 */
4039        nfs4_init_boot_verifier(clp, &sc_verifier);
4040        rcu_read_lock();
4041        setclientid.sc_name_len = scnprintf(setclientid.sc_name,
4042                        sizeof(setclientid.sc_name), "%s/%s %s",
4043                        clp->cl_ipaddr,
4044                        rpc_peeraddr2str(clp->cl_rpcclient,
4045                                                RPC_DISPLAY_ADDR),
4046                        rpc_peeraddr2str(clp->cl_rpcclient,
4047                                                RPC_DISPLAY_PROTO));
4048        /* cb_client4 */
4049        setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4050                                sizeof(setclientid.sc_netid),
4051                                rpc_peeraddr2str(clp->cl_rpcclient,
4052                                                        RPC_DISPLAY_NETID));
4053        rcu_read_unlock();
4054        setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4055                                sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4056                                clp->cl_ipaddr, port >> 8, port & 255);
4057
4058        dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4059                clp->cl_rpcclient->cl_auth->au_ops->au_name,
4060                setclientid.sc_name_len, setclientid.sc_name);
4061        status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4062        dprintk("NFS reply setclientid: %d\n", status);
4063        return status;
4064}
4065
4066/**
4067 * nfs4_proc_setclientid_confirm - Confirm client ID
4068 * @clp: state data structure
4069 * @res: result of a previous SETCLIENTID
4070 * @cred: RPC credential to use for this call
4071 *
4072 * Returns zero, a negative errno, or a negative NFS4ERR status code.
4073 */
4074int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4075                struct nfs4_setclientid_res *arg,
4076                struct rpc_cred *cred)
4077{
4078        struct nfs_fsinfo fsinfo;
4079        struct rpc_message msg = {
4080                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4081                .rpc_argp = arg,
4082                .rpc_resp = &fsinfo,
4083                .rpc_cred = cred,
4084        };
4085        unsigned long now;
4086        int status;
4087
4088        dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4089                clp->cl_rpcclient->cl_auth->au_ops->au_name,
4090                clp->cl_clientid);
4091        now = jiffies;
4092        status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4093        if (status == 0) {
4094                spin_lock(&clp->cl_lock);
4095                clp->cl_lease_time = fsinfo.lease_time * HZ;
4096                clp->cl_last_renewal = now;
4097                spin_unlock(&clp->cl_lock);
4098        }
4099        dprintk("NFS reply setclientid_confirm: %d\n", status);
4100        return status;
4101}
4102
4103struct nfs4_delegreturndata {
4104        struct nfs4_delegreturnargs args;
4105        struct nfs4_delegreturnres res;
4106        struct nfs_fh fh;
4107        nfs4_stateid stateid;
4108        unsigned long timestamp;
4109        struct nfs_fattr fattr;
4110        int rpc_status;
4111};
4112
4113static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4114{
4115        struct nfs4_delegreturndata *data = calldata;
4116
4117        if (!nfs4_sequence_done(task, &data->res.seq_res))
4118                return;
4119
4120        switch (task->tk_status) {
4121        case -NFS4ERR_STALE_STATEID:
4122        case -NFS4ERR_EXPIRED:
4123        case 0:
4124                renew_lease(data->res.server, data->timestamp);
4125                break;
4126        default:
4127                if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4128                                -EAGAIN) {
4129                        rpc_restart_call_prepare(task);
4130                        return;
4131                }
4132        }
4133        data->rpc_status = task->tk_status;
4134}
4135
4136static void nfs4_delegreturn_release(void *calldata)
4137{
4138        kfree(calldata);
4139}
4140
4141#if defined(CONFIG_NFS_V4_1)
4142static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4143{
4144        struct nfs4_delegreturndata *d_data;
4145
4146        d_data = (struct nfs4_delegreturndata *)data;
4147
4148        if (nfs4_setup_sequence(d_data->res.server,
4149                                &d_data->args.seq_args,
4150                                &d_data->res.seq_res, task))
4151                return;
4152        rpc_call_start(task);
4153}
4154#endif /* CONFIG_NFS_V4_1 */
4155
4156static const struct rpc_call_ops nfs4_delegreturn_ops = {
4157#if defined(CONFIG_NFS_V4_1)
4158        .rpc_call_prepare = nfs4_delegreturn_prepare,
4159#endif /* CONFIG_NFS_V4_1 */
4160        .rpc_call_done = nfs4_delegreturn_done,
4161        .rpc_release = nfs4_delegreturn_release,
4162};
4163
4164static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4165{
4166        struct nfs4_delegreturndata *data;
4167        struct nfs_server *server = NFS_SERVER(inode);
4168        struct rpc_task *task;
4169        struct rpc_message msg = {
4170                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4171                .rpc_cred = cred,
4172        };
4173        struct rpc_task_setup task_setup_data = {
4174                .rpc_client = server->client,
4175                .rpc_message = &msg,
4176                .callback_ops = &nfs4_delegreturn_ops,
4177                .flags = RPC_TASK_ASYNC,
4178        };
4179        int status = 0;
4180
4181        data = kzalloc(sizeof(*data), GFP_NOFS);
4182        if (data == NULL)
4183                return -ENOMEM;
4184        nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4185        data->args.fhandle = &data->fh;
4186        data->args.stateid = &data->stateid;
4187        data->args.bitmask = server->cache_consistency_bitmask;
4188        nfs_copy_fh(&data->fh, NFS_FH(inode));
4189        nfs4_stateid_copy(&data->stateid, stateid);
4190        data->res.fattr = &data->fattr;
4191        data->res.server = server;
4192        nfs_fattr_init(data->res.fattr);
4193        data->timestamp = jiffies;
4194        data->rpc_status = 0;
4195
4196        task_setup_data.callback_data = data;
4197        msg.rpc_argp = &data->args;
4198        msg.rpc_resp = &data->res;
4199        task = rpc_run_task(&task_setup_data);
4200        if (IS_ERR(task))
4201                return PTR_ERR(task);
4202        if (!issync)
4203                goto out;
4204        status = nfs4_wait_for_completion_rpc_task(task);
4205        if (status != 0)
4206                goto out;
4207        status = data->rpc_status;
4208        if (status == 0)
4209                nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4210        else
4211                nfs_refresh_inode(inode, &data->fattr);
4212out:
4213        rpc_put_task(task);
4214        return status;
4215}
4216
4217int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4218{
4219        struct nfs_server *server = NFS_SERVER(inode);
4220        struct nfs4_exception exception = { };
4221        int err;
4222        do {
4223                err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4224                switch (err) {
4225                        case -NFS4ERR_STALE_STATEID:
4226                        case -NFS4ERR_EXPIRED:
4227                        case 0:
4228                                return 0;
4229                }
4230                err = nfs4_handle_exception(server, err, &exception);
4231        } while (exception.retry);
4232        return err;
4233}
4234
4235#define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4236#define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4237
4238/* 
4239 * sleep, with exponential backoff, and retry the LOCK operation. 
4240 */
4241static unsigned long
4242nfs4_set_lock_task_retry(unsigned long timeout)
4243{
4244        freezable_schedule_timeout_killable(timeout);
4245        timeout <<= 1;
4246        if (timeout > NFS4_LOCK_MAXTIMEOUT)
4247                return NFS4_LOCK_MAXTIMEOUT;
4248        return timeout;
4249}
4250
4251static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4252{
4253        struct inode *inode = state->inode;
4254        struct nfs_server *server = NFS_SERVER(inode);
4255        struct nfs_client *clp = server->nfs_client;
4256        struct nfs_lockt_args arg = {
4257                .fh = NFS_FH(inode),
4258                .fl = request,
4259        };
4260        struct nfs_lockt_res res = {
4261                .denied = request,
4262        };
4263        struct rpc_message msg = {
4264                .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4265                .rpc_argp       = &arg,
4266                .rpc_resp       = &res,
4267                .rpc_cred       = state->owner->so_cred,
4268        };
4269        struct nfs4_lock_state *lsp;
4270        int status;
4271
4272        arg.lock_owner.clientid = clp->cl_clientid;
4273        status = nfs4_set_lock_state(state, request);
4274        if (status != 0)
4275                goto out;
4276        lsp = request->fl_u.nfs4_fl.owner;
4277        arg.lock_owner.id = lsp->ls_seqid.owner_id;
4278        arg.lock_owner.s_dev = server->s_dev;
4279        status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4280        switch (status) {
4281                case 0:
4282                        request->fl_type = F_UNLCK;
4283                        break;
4284                case -NFS4ERR_DENIED:
4285                        status = 0;
4286        }
4287        request->fl_ops->fl_release_private(request);
4288out:
4289        return status;
4290}
4291
4292static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4293{
4294        struct nfs4_exception exception = { };
4295        int err;
4296
4297        do {
4298                err = nfs4_handle_exception(NFS_SERVER(state->inode),
4299                                _nfs4_proc_getlk(state, cmd, request),
4300                                &exception);
4301        } while (exception.retry);
4302        return err;
4303}
4304
4305static int do_vfs_lock(struct file *file, struct file_lock *fl)
4306{
4307        int res = 0;
4308        switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4309                case FL_POSIX:
4310                        res = posix_lock_file_wait(file, fl);
4311                        break;
4312                case FL_FLOCK:
4313                        res = flock_lock_file_wait(file, fl);
4314                        break;
4315                default:
4316                        BUG();
4317        }
4318        return res;
4319}
4320
4321struct nfs4_unlockdata {
4322        struct nfs_locku_args arg;
4323        struct nfs_locku_res res;
4324        struct nfs4_lock_state *lsp;
4325        struct nfs_open_context *ctx;
4326        struct file_lock fl;
4327        const struct nfs_server *server;
4328        unsigned long timestamp;
4329};
4330
4331static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4332                struct nfs_open_context *ctx,
4333                struct nfs4_lock_state *lsp,
4334                struct nfs_seqid *seqid)
4335{
4336        struct nfs4_unlockdata *p;
4337        struct inode *inode = lsp->ls_state->inode;
4338
4339        p = kzalloc(sizeof(*p), GFP_NOFS);
4340        if (p == NULL)
4341                return NULL;
4342        p->arg.fh = NFS_FH(inode);
4343        p->arg.fl = &p->fl;
4344        p->arg.seqid = seqid;
4345        p->res.seqid = seqid;
4346        p->arg.stateid = &lsp->ls_stateid;
4347        p->lsp = lsp;
4348        atomic_inc(&lsp->ls_count);
4349        /* Ensure we don't close file until we're done freeing locks! */
4350        p->ctx = get_nfs_open_context(ctx);
4351        memcpy(&p->fl, fl, sizeof(p->fl));
4352        p->server = NFS_SERVER(inode);
4353        return p;
4354}
4355
4356static void nfs4_locku_release_calldata(void *data)
4357{
4358        struct nfs4_unlockdata *calldata = data;
4359        nfs_free_seqid(calldata->arg.seqid);
4360        nfs4_put_lock_state(calldata->lsp);
4361        put_nfs_open_context(calldata->ctx);
4362        kfree(calldata);
4363}
4364
4365static void nfs4_locku_done(struct rpc_task *task, void *data)
4366{
4367        struct nfs4_unlockdata *calldata = data;
4368
4369        if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4370                return;
4371        switch (task->tk_status) {
4372                case 0:
4373                        nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4374                                        &calldata->res.stateid);
4375                        renew_lease(calldata->server, calldata->timestamp);
4376                        break;
4377                case -NFS4ERR_BAD_STATEID:
4378                case -NFS4ERR_OLD_STATEID:
4379                case -NFS4ERR_STALE_STATEID:
4380                case -NFS4ERR_EXPIRED:
4381                        break;
4382                default:
4383                        if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4384                                rpc_restart_call_prepare(task);
4385        }
4386}
4387
4388static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4389{
4390        struct nfs4_unlockdata *calldata = data;
4391
4392        if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4393                return;
4394        if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4395                /* Note: exit _without_ running nfs4_locku_done */
4396                task->tk_action = NULL;
4397                return;
4398        }
4399        calldata->timestamp = jiffies;
4400        if (nfs4_setup_sequence(calldata->server,
4401                                &calldata->arg.seq_args,
4402                                &calldata->res.seq_res, task))
4403                return;
4404        rpc_call_start(task);
4405}
4406
4407static const struct rpc_call_ops nfs4_locku_ops = {
4408        .rpc_call_prepare = nfs4_locku_prepare,
4409        .rpc_call_done = nfs4_locku_done,
4410        .rpc_release = nfs4_locku_release_calldata,
4411};
4412
4413static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4414                struct nfs_open_context *ctx,
4415                struct nfs4_lock_state *lsp,
4416                struct nfs_seqid *seqid)
4417{
4418        struct nfs4_unlockdata *data;
4419        struct rpc_message msg = {
4420                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4421                .rpc_cred = ctx->cred,
4422        };
4423        struct rpc_task_setup task_setup_data = {
4424                .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4425                .rpc_message = &msg,
4426                .callback_ops = &nfs4_locku_ops,
4427                .workqueue = nfsiod_workqueue,
4428                .flags = RPC_TASK_ASYNC,
4429        };
4430
4431        /* Ensure this is an unlock - when canceling a lock, the
4432         * canceled lock is passed in, and it won't be an unlock.
4433         */
4434        fl->fl_type = F_UNLCK;
4435
4436        data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4437        if (data == NULL) {
4438                nfs_free_seqid(seqid);
4439                return ERR_PTR(-ENOMEM);
4440        }
4441
4442        nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4443        msg.rpc_argp = &data->arg;
4444        msg.rpc_resp = &data->res;
4445        task_setup_data.callback_data = data;
4446        return rpc_run_task(&task_setup_data);
4447}
4448
4449static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4450{
4451        struct nfs_inode *nfsi = NFS_I(state->inode);
4452        struct nfs_seqid *seqid;
4453        struct nfs4_lock_state *lsp;
4454        struct rpc_task *task;
4455        int status = 0;
4456        unsigned char fl_flags = request->fl_flags;
4457
4458        status = nfs4_set_lock_state(state, request);
4459        /* Unlock _before_ we do the RPC call */
4460        request->fl_flags |= FL_EXISTS;
4461        down_read(&nfsi->rwsem);
4462        if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4463                up_read(&nfsi->rwsem);
4464                goto out;
4465        }
4466        up_read(&nfsi->rwsem);
4467        if (status != 0)
4468                goto out;
4469        /* Is this a delegated lock? */
4470        if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4471                goto out;
4472        lsp = request->fl_u.nfs4_fl.owner;
4473        seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4474        status = -ENOMEM;
4475        if (seqid == NULL)
4476                goto out;
4477        task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4478        status = PTR_ERR(task);
4479        if (IS_ERR(task))
4480                goto out;
4481        status = nfs4_wait_for_completion_rpc_task(task);
4482        rpc_put_task(task);
4483out:
4484        request->fl_flags = fl_flags;
4485        return status;
4486}
4487
4488struct nfs4_lockdata {
4489        struct nfs_lock_args arg;
4490        struct nfs_lock_res res;
4491        struct nfs4_lock_state *lsp;
4492        struct nfs_open_context *ctx;
4493        struct file_lock fl;
4494        unsigned long timestamp;
4495        int rpc_status;
4496        int cancelled;
4497        struct nfs_server *server;
4498};
4499
4500static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4501                struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4502                gfp_t gfp_mask)
4503{
4504        struct nfs4_lockdata *p;
4505        struct inode *inode = lsp->ls_state->inode;
4506        struct nfs_server *server = NFS_SERVER(inode);
4507
4508        p = kzalloc(sizeof(*p), gfp_mask);
4509        if (p == NULL)
4510                return NULL;
4511
4512        p->arg.fh = NFS_FH(inode);
4513        p->arg.fl = &p->fl;
4514        p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4515        if (p->arg.open_seqid == NULL)
4516                goto out_free;
4517        p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4518        if (p->arg.lock_seqid == NULL)
4519                goto out_free_seqid;
4520        p->arg.lock_stateid = &lsp->ls_stateid;
4521        p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4522        p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4523        p->arg.lock_owner.s_dev = server->s_dev;
4524        p->res.lock_seqid = p->arg.lock_seqid;
4525        p->lsp = lsp;
4526        p->server = server;
4527        atomic_inc(&lsp->ls_count);
4528        p->ctx = get_nfs_open_context(ctx);
4529        memcpy(&p->fl, fl, sizeof(p->fl));
4530        return p;
4531out_free_seqid:
4532        nfs_free_seqid(p->arg.open_seqid);
4533out_free:
4534        kfree(p);
4535        return NULL;
4536}
4537
4538static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4539{
4540        struct nfs4_lockdata *data = calldata;
4541        struct nfs4_state *state = data->lsp->ls_state;
4542
4543        dprintk("%s: begin!\n", __func__);
4544        if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4545                return;
4546        /* Do we need to do an open_to_lock_owner? */
4547        if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4548                if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4549                        return;
4550                data->arg.open_stateid = &state->stateid;
4551                data->arg.new_lock_owner = 1;
4552                data->res.open_seqid = data->arg.open_seqid;
4553        } else
4554                data->arg.new_lock_owner = 0;
4555        data->timestamp = jiffies;
4556        if (nfs4_setup_sequence(data->server,
4557                                &data->arg.seq_args,
4558                                &data->res.seq_res, task))
4559                return;
4560        rpc_call_start(task);
4561        dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4562}
4563
4564static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4565{
4566        rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4567        nfs4_lock_prepare(task, calldata);
4568}
4569
4570static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4571{
4572        struct nfs4_lockdata *data = calldata;
4573
4574        dprintk("%s: begin!\n", __func__);
4575
4576        if (!nfs4_sequence_done(task, &data->res.seq_res))
4577                return;
4578
4579        data->rpc_status = task->tk_status;
4580        if (data->arg.new_lock_owner != 0) {
4581                if (data->rpc_status == 0)
4582                        nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4583                else
4584                        goto out;
4585        }
4586        if (data->rpc_status == 0) {
4587                nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4588                data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4589                renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4590        }
4591out:
4592        dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4593}
4594
4595static void nfs4_lock_release(void *calldata)
4596{
4597        struct nfs4_lockdata *data = calldata;
4598
4599        dprintk("%s: begin!\n", __func__);
4600        nfs_free_seqid(data->arg.open_seqid);
4601        if (data->cancelled != 0) {
4602                struct rpc_task *task;
4603                task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4604                                data->arg.lock_seqid);
4605                if (!IS_ERR(task))
4606                        rpc_put_task_async(task);
4607                dprintk("%s: cancelling lock!\n", __func__);
4608        } else
4609                nfs_free_seqid(data->arg.lock_seqid);
4610        nfs4_put_lock_state(data->lsp);
4611        put_nfs_open_context(data->ctx);
4612        kfree(data);
4613        dprintk("%s: done!\n", __func__);
4614}
4615
4616static const struct rpc_call_ops nfs4_lock_ops = {
4617        .rpc_call_prepare = nfs4_lock_prepare,
4618        .rpc_call_done = nfs4_lock_done,
4619        .rpc_release = nfs4_lock_release,
4620};
4621
4622static const struct rpc_call_ops nfs4_recover_lock_ops = {
4623        .rpc_call_prepare = nfs4_recover_lock_prepare,
4624        .rpc_call_done = nfs4_lock_done,
4625        .rpc_release = nfs4_lock_release,
4626};
4627
4628static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4629{
4630        switch (error) {
4631        case -NFS4ERR_ADMIN_REVOKED:
4632        case -NFS4ERR_BAD_STATEID:
4633                lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4634                if (new_lock_owner != 0 ||
4635                   (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4636                        nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4637                break;
4638        case -NFS4ERR_STALE_STATEID:
4639                lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4640        case -NFS4ERR_EXPIRED:
4641                nfs4_schedule_lease_recovery(server->nfs_client);
4642        };
4643}
4644
4645static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4646{
4647        struct nfs4_lockdata *data;
4648        struct rpc_task *task;
4649        struct rpc_message msg = {
4650                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4651                .rpc_cred = state->owner->so_cred,
4652        };
4653        struct rpc_task_setup task_setup_data = {
4654                .rpc_client = NFS_CLIENT(state->inode),
4655                .rpc_message = &msg,
4656                .callback_ops = &nfs4_lock_ops,
4657                .workqueue = nfsiod_workqueue,
4658                .flags = RPC_TASK_ASYNC,
4659        };
4660        int ret;
4661
4662        dprintk("%s: begin!\n", __func__);
4663        data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4664                        fl->fl_u.nfs4_fl.owner,
4665                        recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4666        if (data == NULL)
4667                return -ENOMEM;
4668        if (IS_SETLKW(cmd))
4669                data->arg.block = 1;
4670        if (recovery_type > NFS_LOCK_NEW) {
4671                if (recovery_type == NFS_LOCK_RECLAIM)
4672                        data->arg.reclaim = NFS_LOCK_RECLAIM;
4673                task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4674        }
4675        nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4676        msg.rpc_argp = &data->arg;
4677        msg.rpc_resp = &data->res;
4678        task_setup_data.callback_data = data;
4679        task = rpc_run_task(&task_setup_data);
4680        if (IS_ERR(task))
4681                return PTR_ERR(task);
4682        ret = nfs4_wait_for_completion_rpc_task(task);
4683        if (ret == 0) {
4684                ret = data->rpc_status;
4685                if (ret)
4686                        nfs4_handle_setlk_error(data->server, data->lsp,
4687                                        data->arg.new_lock_owner, ret);
4688        } else
4689                data->cancelled = 1;
4690        rpc_put_task(task);
4691        dprintk("%s: done, ret = %d!\n", __func__, ret);
4692        return ret;
4693}
4694
4695static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4696{
4697        struct nfs_server *server = NFS_SERVER(state->inode);
4698        struct nfs4_exception exception = {
4699                .inode = state->inode,
4700        };
4701        int err;
4702
4703        do {
4704                /* Cache the lock if possible... */
4705                if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4706                        return 0;
4707                err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4708                if (err != -NFS4ERR_DELAY)
4709                        break;
4710                nfs4_handle_exception(server, err, &exception);
4711        } while (exception.retry);
4712        return err;
4713}
4714
4715static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4716{
4717        struct nfs_server *server = NFS_SERVER(state->inode);
4718        struct nfs4_exception exception = {
4719                .inode = state->inode,
4720        };
4721        int err;
4722
4723        err = nfs4_set_lock_state(state, request);
4724        if (err != 0)
4725                return err;
4726        do {
4727                if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4728                        return 0;
4729                err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4730                switch (err) {
4731                default:
4732                        goto out;
4733                case -NFS4ERR_GRACE:
4734                case -NFS4ERR_DELAY:
4735                        nfs4_handle_exception(server, err, &exception);
4736                        err = 0;
4737                }
4738        } while (exception.retry);
4739out:
4740        return err;
4741}
4742
4743#if defined(CONFIG_NFS_V4_1)
4744/**
4745 * nfs41_check_expired_locks - possibly free a lock stateid
4746 *
4747 * @state: NFSv4 state for an inode
4748 *
4749 * Returns NFS_OK if recovery for this stateid is now finished.
4750 * Otherwise a negative NFS4ERR value is returned.
4751 */
4752static int nfs41_check_expired_locks(struct nfs4_state *state)
4753{
4754        int status, ret = -NFS4ERR_BAD_STATEID;
4755        struct nfs4_lock_state *lsp;
4756        struct nfs_server *server = NFS_SERVER(state->inode);
4757
4758        list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4759                if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4760                        status = nfs41_test_stateid(server, &lsp->ls_stateid);
4761                        if (status != NFS_OK) {
4762                                /* Free the stateid unless the server
4763                                 * informs us the stateid is unrecognized. */
4764                                if (status != -NFS4ERR_BAD_STATEID)
4765                                        nfs41_free_stateid(server,
4766                                                        &lsp->ls_stateid);
4767                                lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4768                                ret = status;
4769                        }
4770                }
4771        };
4772
4773        return ret;
4774}
4775
4776static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4777{
4778        int status = NFS_OK;
4779
4780        if (test_bit(LK_STATE_IN_USE, &state->flags))
4781                status = nfs41_check_expired_locks(state);
4782        if (status != NFS_OK)
4783                status = nfs4_lock_expired(state, request);
4784        return status;
4785}
4786#endif
4787
4788static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4789{
4790        struct nfs_inode *nfsi = NFS_I(state->inode);
4791        unsigned char fl_flags = request->fl_flags;
4792        int status = -ENOLCK;
4793
4794        if ((fl_flags & FL_POSIX) &&
4795                        !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4796                goto out;
4797        /* Is this a delegated open? */
4798        status = nfs4_set_lock_state(state, request);
4799        if (status != 0)
4800                goto out;
4801        request->fl_flags |= FL_ACCESS;
4802        status = do_vfs_lock(request->fl_file, request);
4803        if (status < 0)
4804                goto out;
4805        down_read(&nfsi->rwsem);
4806        if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4807                /* Yes: cache locks! */
4808                /* ...but avoid races with delegation recall... */
4809                request->fl_flags = fl_flags & ~FL_SLEEP;
4810                status = do_vfs_lock(request->fl_file, request);
4811                goto out_unlock;
4812        }
4813        status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4814        if (status != 0)
4815                goto out_unlock;
4816        /* Note: we always want to sleep here! */
4817        request->fl_flags = fl_flags | FL_SLEEP;
4818        if (do_vfs_lock(request->fl_file, request) < 0)
4819                printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4820                        "manager!\n", __func__);
4821out_unlock:
4822        up_read(&nfsi->rwsem);
4823out:
4824        request->fl_flags = fl_flags;
4825        return status;
4826}
4827
4828static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4829{
4830        struct nfs4_exception exception = {
4831                .state = state,
4832                .inode = state->inode,
4833        };
4834        int err;
4835
4836        do {
4837                err = _nfs4_proc_setlk(state, cmd, request);
4838                if (err == -NFS4ERR_DENIED)
4839                        err = -EAGAIN;
4840                err = nfs4_handle_exception(NFS_SERVER(state->inode),
4841                                err, &exception);
4842        } while (exception.retry);
4843        return err;
4844}
4845
4846static int
4847nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4848{
4849        struct nfs_open_context *ctx;
4850        struct nfs4_state *state;
4851        unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4852        int status;
4853
4854        /* verify open state */
4855        ctx = nfs_file_open_context(filp);
4856        state = ctx->state;
4857
4858        if (request->fl_start < 0 || request->fl_end < 0)
4859                return -EINVAL;
4860
4861        if (IS_GETLK(cmd)) {
4862                if (state != NULL)
4863                        return nfs4_proc_getlk(state, F_GETLK, request);
4864                return 0;
4865        }
4866
4867        if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4868                return -EINVAL;
4869
4870        if (request->fl_type == F_UNLCK) {
4871                if (state != NULL)
4872                        return nfs4_proc_unlck(state, cmd, request);
4873                return 0;
4874        }
4875
4876        if (state == NULL)
4877                return -ENOLCK;
4878        /*
4879         * Don't rely on the VFS having checked the file open mode,
4880         * since it won't do this for flock() locks.
4881         */
4882        switch (request->fl_type) {
4883        case F_RDLCK:
4884                if (!(filp->f_mode & FMODE_READ))
4885                        return -EBADF;
4886                break;
4887        case F_WRLCK:
4888                if (!(filp->f_mode & FMODE_WRITE))
4889                        return -EBADF;
4890        }
4891
4892        do {
4893                status = nfs4_proc_setlk(state, cmd, request);
4894                if ((status != -EAGAIN) || IS_SETLK(cmd))
4895                        break;
4896                timeout = nfs4_set_lock_task_retry(timeout);
4897                status = -ERESTARTSYS;
4898                if (signalled())
4899                        break;
4900        } while(status < 0);
4901        return status;
4902}
4903
4904int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4905{
4906        struct nfs_server *server = NFS_SERVER(state->inode);
4907        struct nfs4_exception exception = { };
4908        int err;
4909
4910        err = nfs4_set_lock_state(state, fl);
4911        if (err != 0)
4912                goto out;
4913        do {
4914                err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4915                switch (err) {
4916                        default:
4917                                printk(KERN_ERR "NFS: %s: unhandled error "
4918                                        "%d.\n", __func__, err);
4919                        case 0:
4920                        case -ESTALE:
4921                                goto out;
4922                        case -NFS4ERR_EXPIRED:
4923                                nfs4_schedule_stateid_recovery(server, state);
4924                        case -NFS4ERR_STALE_CLIENTID:
4925                        case -NFS4ERR_STALE_STATEID:
4926                                nfs4_schedule_lease_recovery(server->nfs_client);
4927                                goto out;
4928                        case -NFS4ERR_BADSESSION:
4929                        case -NFS4ERR_BADSLOT:
4930                        case -NFS4ERR_BAD_HIGH_SLOT:
4931                        case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4932                        case -NFS4ERR_DEADSESSION:
4933                                nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
4934                                goto out;
4935                        case -ERESTARTSYS:
4936                                /*
4937                                 * The show must go on: exit, but mark the
4938                                 * stateid as needing recovery.
4939                                 */
4940                        case -NFS4ERR_DELEG_REVOKED:
4941                        case -NFS4ERR_ADMIN_REVOKED:
4942                        case -NFS4ERR_BAD_STATEID:
4943                        case -NFS4ERR_OPENMODE:
4944                                nfs4_schedule_stateid_recovery(server, state);
4945                                err = 0;
4946                                goto out;
4947                        case -EKEYEXPIRED:
4948                                /*
4949                                 * User RPCSEC_GSS context has expired.
4950                                 * We cannot recover this stateid now, so
4951                                 * skip it and allow recovery thread to
4952                                 * proceed.
4953                                 */
4954                                err = 0;
4955                                goto out;
4956                        case -ENOMEM:
4957                        case -NFS4ERR_DENIED:
4958                                /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4959                                err = 0;
4960                                goto out;
4961                        case -NFS4ERR_DELAY:
4962                                break;
4963                }
4964                err = nfs4_handle_exception(server, err, &exception);
4965        } while (exception.retry);
4966out:
4967        return err;
4968}
4969
4970struct nfs_release_lockowner_data {
4971        struct nfs4_lock_state *lsp;
4972        struct nfs_server *server;
4973        struct nfs_release_lockowner_args args;
4974};
4975
4976static void nfs4_release_lockowner_release(void *calldata)
4977{
4978        struct nfs_release_lockowner_data *data = calldata;
4979        nfs4_free_lock_state(data->server, data->lsp);
4980        kfree(calldata);
4981}
4982
4983static const struct rpc_call_ops nfs4_release_lockowner_ops = {
4984        .rpc_release = nfs4_release_lockowner_release,
4985};
4986
4987int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
4988{
4989        struct nfs_server *server = lsp->ls_state->owner->so_server;
4990        struct nfs_release_lockowner_data *data;
4991        struct rpc_message msg = {
4992                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4993        };
4994
4995        if (server->nfs_client->cl_mvops->minor_version != 0)
4996                return -EINVAL;
4997        data = kmalloc(sizeof(*data), GFP_NOFS);
4998        if (!data)
4999                return -ENOMEM;
5000        data->lsp = lsp;
5001        data->server = server;
5002        data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5003        data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5004        data->args.lock_owner.s_dev = server->s_dev;
5005        msg.rpc_argp = &data->args;
5006        rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5007        return 0;
5008}
5009
5010#define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5011
5012static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5013                                   const void *buf, size_t buflen,
5014                                   int flags, int type)
5015{
5016        if (strcmp(key, "") != 0)
5017                return -EINVAL;
5018
5019        return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5020}
5021
5022static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5023                                   void *buf, size_t buflen, int type)
5024{
5025        if (strcmp(key, "") != 0)
5026                return -EINVAL;
5027
5028        return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5029}
5030
5031static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5032                                       size_t list_len, const char *name,
5033                                       size_t name_len, int type)
5034{
5035        size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5036
5037        if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5038                return 0;
5039
5040        if (list && len <= list_len)
5041                memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5042        return len;
5043}
5044
5045/*
5046 * nfs_fhget will use either the mounted_on_fileid or the fileid
5047 */
5048static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5049{
5050        if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5051               (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5052              (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5053              (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5054                return;
5055
5056        fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5057                NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5058        fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5059        fattr->nlink = 2;
5060}
5061
5062static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5063                                   const struct qstr *name,
5064                                   struct nfs4_fs_locations *fs_locations,
5065                                   struct page *page)
5066{
5067        struct nfs_server *server = NFS_SERVER(dir);
5068        u32 bitmask[2] = {
5069                [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5070        };
5071        struct nfs4_fs_locations_arg args = {
5072                .dir_fh = NFS_FH(dir),
5073                .name = name,
5074                .page = page,
5075                .bitmask = bitmask,
5076        };
5077        struct nfs4_fs_locations_res res = {
5078                .fs_locations = fs_locations,
5079        };
5080        struct rpc_message msg = {
5081                .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5082                .rpc_argp = &args,
5083                .rpc_resp = &res,
5084        };
5085        int status;