fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/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/xattr.h>
  56 #include <linux/utsname.h>
  57 #include <linux/freezer.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4session.h"
  67 #include "fscache.h"
  68
  69 #define NFSDBG_FACILITY         NFSDBG_PROC
  70
  71 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  72 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  73
  74 struct nfs4_opendata;
  75 static int _nfs4_proc_open(struct nfs4_opendata *data);
  76 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  77 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  78 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  79 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  80 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *);
  81 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  82 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  83                             struct nfs_fattr *fattr, struct iattr *sattr,
  84                             struct nfs4_state *state);
  85 #ifdef CONFIG_NFS_V4_1
  86 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
  87 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
  88 #endif
  89 /* Prevent leaks of NFSv4 errors into userland */
  90 static int nfs4_map_errors(int err)
  91 {
  92         if (err >= -1000)
  93                 return err;
  94         switch (err) {
  95         case -NFS4ERR_RESOURCE:
  96         case -NFS4ERR_LAYOUTTRYLATER:
  97         case -NFS4ERR_RECALLCONFLICT:
  98                 return -EREMOTEIO;
  99         case -NFS4ERR_WRONGSEC:
 100                 return -EPERM;
 101         case -NFS4ERR_BADOWNER:
 102         case -NFS4ERR_BADNAME:
 103                 return -EINVAL;
 104         case -NFS4ERR_SHARE_DENIED:
 105                 return -EACCES;
 106         case -NFS4ERR_MINOR_VERS_MISMATCH:
 107                 return -EPROTONOSUPPORT;
 108         case -NFS4ERR_ACCESS:
 109                 return -EACCES;
 110         case -NFS4ERR_FILE_OPEN:
 111                 return -EBUSY;
 112         default:
 113                 dprintk("%s could not handle NFSv4 error %d\n",
 114                                 __func__, -err);
 115                 break;
 116         }
 117         return -EIO;
 118 }
 119
 120 /*
 121  * This is our standard bitmap for GETATTR requests.
 122  */
 123 const u32 nfs4_fattr_bitmap[3] = {
 124         FATTR4_WORD0_TYPE
 125         | FATTR4_WORD0_CHANGE
 126         | FATTR4_WORD0_SIZE
 127         | FATTR4_WORD0_FSID
 128         | FATTR4_WORD0_FILEID,
 129         FATTR4_WORD1_MODE
 130         | FATTR4_WORD1_NUMLINKS
 131         | FATTR4_WORD1_OWNER
 132         | FATTR4_WORD1_OWNER_GROUP
 133         | FATTR4_WORD1_RAWDEV
 134         | FATTR4_WORD1_SPACE_USED
 135         | FATTR4_WORD1_TIME_ACCESS
 136         | FATTR4_WORD1_TIME_METADATA
 137         | FATTR4_WORD1_TIME_MODIFY
 138 };
 139
 140 static const u32 nfs4_pnfs_open_bitmap[3] = {
 141         FATTR4_WORD0_TYPE
 142         | FATTR4_WORD0_CHANGE
 143         | FATTR4_WORD0_SIZE
 144         | FATTR4_WORD0_FSID
 145         | FATTR4_WORD0_FILEID,
 146         FATTR4_WORD1_MODE
 147         | FATTR4_WORD1_NUMLINKS
 148         | FATTR4_WORD1_OWNER
 149         | FATTR4_WORD1_OWNER_GROUP
 150         | FATTR4_WORD1_RAWDEV
 151         | FATTR4_WORD1_SPACE_USED
 152         | FATTR4_WORD1_TIME_ACCESS
 153         | FATTR4_WORD1_TIME_METADATA
 154         | FATTR4_WORD1_TIME_MODIFY,
 155         FATTR4_WORD2_MDSTHRESHOLD
 156 };
 157
 158 static const u32 nfs4_open_noattr_bitmap[3] = {
 159         FATTR4_WORD0_TYPE
 160         | FATTR4_WORD0_CHANGE
 161         | FATTR4_WORD0_FILEID,
 162 };
 163
 164 const u32 nfs4_statfs_bitmap[2] = {
 165         FATTR4_WORD0_FILES_AVAIL
 166         | FATTR4_WORD0_FILES_FREE
 167         | FATTR4_WORD0_FILES_TOTAL,
 168         FATTR4_WORD1_SPACE_AVAIL
 169         | FATTR4_WORD1_SPACE_FREE
 170         | FATTR4_WORD1_SPACE_TOTAL
 171 };
 172
 173 const u32 nfs4_pathconf_bitmap[2] = {
 174         FATTR4_WORD0_MAXLINK
 175         | FATTR4_WORD0_MAXNAME,
 176         0
 177 };
 178
 179 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 180                         | FATTR4_WORD0_MAXREAD
 181                         | FATTR4_WORD0_MAXWRITE
 182                         | FATTR4_WORD0_LEASE_TIME,
 183                         FATTR4_WORD1_TIME_DELTA
 184                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 185                         FATTR4_WORD2_LAYOUT_BLKSIZE
 186 };
 187
 188 const u32 nfs4_fs_locations_bitmap[2] = {
 189         FATTR4_WORD0_TYPE
 190         | FATTR4_WORD0_CHANGE
 191         | FATTR4_WORD0_SIZE
 192         | FATTR4_WORD0_FSID
 193         | FATTR4_WORD0_FILEID
 194         | FATTR4_WORD0_FS_LOCATIONS,
 195         FATTR4_WORD1_MODE
 196         | FATTR4_WORD1_NUMLINKS
 197         | FATTR4_WORD1_OWNER
 198         | FATTR4_WORD1_OWNER_GROUP
 199         | FATTR4_WORD1_RAWDEV
 200         | FATTR4_WORD1_SPACE_USED
 201         | FATTR4_WORD1_TIME_ACCESS
 202         | FATTR4_WORD1_TIME_METADATA
 203         | FATTR4_WORD1_TIME_MODIFY
 204         | FATTR4_WORD1_MOUNTED_ON_FILEID
 205 };
 206
 207 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 208                 struct nfs4_readdir_arg *readdir)
 209 {
 210         __be32 *start, *p;
 211
 212         if (cookie > 2) {
 213                 readdir->cookie = cookie;
 214                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 215                 return;
 216         }
 217
 218         readdir->cookie = 0;
 219         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 220         if (cookie == 2)
 221                 return;
 222
 223         /*
 224          * NFSv4 servers do not return entries for '.' and '..'
 225          * Therefore, we fake these entries here.  We let '.'
 226          * have cookie 0 and '..' have cookie 1.  Note that
 227          * when talking to the server, we always send cookie 0
 228          * instead of 1 or 2.
 229          */
 230         start = p = kmap_atomic(*readdir->pages);
 231
 232         if (cookie == 0) {
 233                 *p++ = xdr_one;                                  /* next */
 234                 *p++ = xdr_zero;                   /* cookie, first word */
 235                 *p++ = xdr_one;                   /* cookie, second word */
 236                 *p++ = xdr_one;                             /* entry len */
 237                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 238                 p++;
 239                 *p++ = xdr_one;                         /* bitmap length */
 240                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 241                 *p++ = htonl(8);              /* attribute buffer length */
 242                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 243         }
 244
 245         *p++ = xdr_one;                                  /* next */
 246         *p++ = xdr_zero;                   /* cookie, first word */
 247         *p++ = xdr_two;                   /* cookie, second word */
 248         *p++ = xdr_two;                             /* entry len */
 249         memcpy(p, "..\0\0", 4);                         /* entry */
 250         p++;
 251         *p++ = xdr_one;                         /* bitmap length */
 252         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 253         *p++ = htonl(8);              /* attribute buffer length */
 254         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 255
 256         readdir->pgbase = (char *)p - (char *)start;
 257         readdir->count -= readdir->pgbase;
 258         kunmap_atomic(start);
 259 }
 260
 261 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 262 {
 263         int res = 0;
 264
 265         might_sleep();
 266
 267         if (*timeout <= 0)
 268                 *timeout = NFS4_POLL_RETRY_MIN;
 269         if (*timeout > NFS4_POLL_RETRY_MAX)
 270                 *timeout = NFS4_POLL_RETRY_MAX;
 271         freezable_schedule_timeout_killable(*timeout);
 272         if (fatal_signal_pending(current))
 273                 res = -ERESTARTSYS;
 274         *timeout <<= 1;
 275         return res;
 276 }
 277
 278 /* This is the error handling routine for processes that are allowed
 279  * to sleep.
 280  */
 281 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 282 {
 283         struct nfs_client *clp = server->nfs_client;
 284         struct nfs4_state *state = exception->state;
 285         struct inode *inode = exception->inode;
 286         int ret = errorcode;
 287
 288         exception->retry = 0;
 289         switch(errorcode) {
 290                 case 0:
 291                         return 0;
 292                 case -NFS4ERR_OPENMODE:
 293                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 294                                 nfs4_inode_return_delegation(inode);
 295                                 exception->retry = 1;
 296                                 return 0;
 297                         }
 298                         if (state == NULL)
 299                                 break;
 300                         ret = nfs4_schedule_stateid_recovery(server, state);
 301                         if (ret < 0)
 302                                 break;
 303                         goto wait_on_recovery;
 304                 case -NFS4ERR_DELEG_REVOKED:
 305                 case -NFS4ERR_ADMIN_REVOKED:
 306                 case -NFS4ERR_BAD_STATEID:
 307                         if (inode != NULL && nfs4_have_delegation(inode, FMODE_READ)) {
 308                                 nfs_remove_bad_delegation(inode);
 309                                 exception->retry = 1;
 310                                 break;
 311                         }
 312                         if (state == NULL)
 313                                 break;
 314                         ret = nfs4_schedule_stateid_recovery(server, state);
 315                         if (ret < 0)
 316                                 break;
 317                         goto wait_on_recovery;
 318                 case -NFS4ERR_EXPIRED:
 319                         if (state != NULL) {
 320                                 ret = nfs4_schedule_stateid_recovery(server, state);
 321                                 if (ret < 0)
 322                                         break;
 323                         }
 324                 case -NFS4ERR_STALE_STATEID:
 325                 case -NFS4ERR_STALE_CLIENTID:
 326                         nfs4_schedule_lease_recovery(clp);
 327                         goto wait_on_recovery;
 328 #if defined(CONFIG_NFS_V4_1)
 329                 case -NFS4ERR_BADSESSION:
 330                 case -NFS4ERR_BADSLOT:
 331                 case -NFS4ERR_BAD_HIGH_SLOT:
 332                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 333                 case -NFS4ERR_DEADSESSION:
 334                 case -NFS4ERR_SEQ_FALSE_RETRY:
 335                 case -NFS4ERR_SEQ_MISORDERED:
 336                         dprintk("%s ERROR: %d Reset session\n", __func__,
 337                                 errorcode);
 338                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 339                         goto wait_on_recovery;
 340 #endif /* defined(CONFIG_NFS_V4_1) */
 341                 case -NFS4ERR_FILE_OPEN:
 342                         if (exception->timeout > HZ) {
 343                                 /* We have retried a decent amount, time to
 344                                  * fail
 345                                  */
 346                                 ret = -EBUSY;
 347                                 break;
 348                         }
 349                 case -NFS4ERR_GRACE:
 350                 case -NFS4ERR_DELAY:
 351                         ret = nfs4_delay(server->client, &exception->timeout);
 352                         if (ret != 0)
 353                                 break;
 354                 case -NFS4ERR_RETRY_UNCACHED_REP:
 355                 case -NFS4ERR_OLD_STATEID:
 356                         exception->retry = 1;
 357                         break;
 358                 case -NFS4ERR_BADOWNER:
 359                         /* The following works around a Linux server bug! */
 360                 case -NFS4ERR_BADNAME:
 361                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 362                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 363                                 exception->retry = 1;
 364                                 printk(KERN_WARNING "NFS: v4 server %s "
 365                                                 "does not accept raw "
 366                                                 "uid/gids. "
 367                                                 "Reenabling the idmapper.\n",
 368                                                 server->nfs_client->cl_hostname);
 369                         }
 370         }
 371         /* We failed to handle the error */
 372         return nfs4_map_errors(ret);
 373 wait_on_recovery:
 374         ret = nfs4_wait_clnt_recover(clp);
 375         if (ret == 0)
 376                 exception->retry = 1;
 377         return ret;
 378 }
 379
 380
 381 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 382 {
 383         spin_lock(&clp->cl_lock);
 384         if (time_before(clp->cl_last_renewal,timestamp))
 385                 clp->cl_last_renewal = timestamp;
 386         spin_unlock(&clp->cl_lock);
 387 }
 388
 389 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 390 {
 391         do_renew_lease(server->nfs_client, timestamp);
 392 }
 393
 394 #if defined(CONFIG_NFS_V4_1)
 395
 396 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 397 {
 398         struct nfs4_session *session;
 399         struct nfs4_slot_table *tbl;
 400         bool send_new_highest_used_slotid = false;
 401
 402         if (!res->sr_slot) {
 403                 /* just wake up the next guy waiting since
 404                  * we may have not consumed a slot after all */
 405                 dprintk("%s: No slot\n", __func__);
 406                 return;
 407         }
 408         tbl = res->sr_slot->table;
 409         session = tbl->session;
 410
 411         spin_lock(&tbl->slot_tbl_lock);
 412         /* Be nice to the server: try to ensure that the last transmitted
 413          * value for highest_user_slotid <= target_highest_slotid
 414          */
 415         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 416                 send_new_highest_used_slotid = true;
 417
 418         if (nfs41_wake_and_assign_slot(tbl, res->sr_slot)) {
 419                 send_new_highest_used_slotid = false;
 420                 goto out_unlock;
 421         }
 422         nfs4_free_slot(tbl, res->sr_slot);
 423
 424         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 425                 send_new_highest_used_slotid = false;
 426 out_unlock:
 427         spin_unlock(&tbl->slot_tbl_lock);
 428         res->sr_slot = NULL;
 429         if (send_new_highest_used_slotid)
 430                 nfs41_server_notify_highest_slotid_update(session->clp);
 431 }
 432
 433 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 434 {
 435         struct nfs4_session *session;
 436         struct nfs4_slot *slot;
 437         struct nfs_client *clp;
 438         bool interrupted = false;
 439         int ret = 1;
 440
 441         /* don't increment the sequence number if the task wasn't sent */
 442         if (!RPC_WAS_SENT(task))
 443                 goto out;
 444
 445         slot = res->sr_slot;
 446         session = slot->table->session;
 447
 448         if (slot->interrupted) {
 449                 slot->interrupted = 0;
 450                 interrupted = true;
 451         }
 452
 453         /* Check the SEQUENCE operation status */
 454         switch (res->sr_status) {
 455         case 0:
 456                 /* Update the slot's sequence and clientid lease timer */
 457                 ++slot->seq_nr;
 458                 clp = session->clp;
 459                 do_renew_lease(clp, res->sr_timestamp);
 460                 /* Check sequence flags */
 461                 if (res->sr_status_flags != 0)
 462                         nfs4_schedule_lease_recovery(clp);
 463                 nfs41_update_target_slotid(slot->table, slot, res);
 464                 break;
 465         case 1:
 466                 /*
 467                  * sr_status remains 1 if an RPC level error occurred.
 468                  * The server may or may not have processed the sequence
 469                  * operation..
 470                  * Mark the slot as having hosted an interrupted RPC call.
 471                  */
 472                 slot->interrupted = 1;
 473                 goto out;
 474         case -NFS4ERR_DELAY:
 475                 /* The server detected a resend of the RPC call and
 476                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 477                  * of RFC5661.
 478                  */
 479                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 480                         __func__,
 481                         slot->slot_nr,
 482                         slot->seq_nr);
 483                 goto out_retry;
 484         case -NFS4ERR_BADSLOT:
 485                 /*
 486                  * The slot id we used was probably retired. Try again
 487                  * using a different slot id.
 488                  */
 489                 goto retry_nowait;
 490         case -NFS4ERR_SEQ_MISORDERED:
 491                 /*
 492                  * Was the last operation on this sequence interrupted?
 493                  * If so, retry after bumping the sequence number.
 494                  */
 495                 if (interrupted) {
 496                         ++slot->seq_nr;
 497                         goto retry_nowait;
 498                 }
 499                 /*
 500                  * Could this slot have been previously retired?
 501                  * If so, then the server may be expecting seq_nr = 1!
 502                  */
 503                 if (slot->seq_nr != 1) {
 504                         slot->seq_nr = 1;
 505                         goto retry_nowait;
 506                 }
 507                 break;
 508         case -NFS4ERR_SEQ_FALSE_RETRY:
 509                 ++slot->seq_nr;
 510                 goto retry_nowait;
 511         default:
 512                 /* Just update the slot sequence no. */
 513                 ++slot->seq_nr;
 514         }
 515 out:
 516         /* The session may be reset by one of the error handlers. */
 517         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 518         nfs41_sequence_free_slot(res);
 519         return ret;
 520 retry_nowait:
 521         if (rpc_restart_call_prepare(task)) {
 522                 task->tk_status = 0;
 523                 ret = 0;
 524         }
 525         goto out;
 526 out_retry:
 527         if (!rpc_restart_call(task))
 528                 goto out;
 529         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 530         return 0;
 531 }
 532
 533 static int nfs4_sequence_done(struct rpc_task *task,
 534                                struct nfs4_sequence_res *res)
 535 {
 536         if (res->sr_slot == NULL)
 537                 return 1;
 538         return nfs41_sequence_done(task, res);
 539 }
 540
 541 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 542                 struct nfs4_sequence_res *res, int cache_reply)
 543 {
 544         args->sa_slot = NULL;
 545         args->sa_cache_this = 0;
 546         args->sa_privileged = 0;
 547         if (cache_reply)
 548                 args->sa_cache_this = 1;
 549         res->sr_slot = NULL;
 550 }
 551
 552 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 553 {
 554         args->sa_privileged = 1;
 555 }
 556
 557 int nfs41_setup_sequence(struct nfs4_session *session,
 558                                 struct nfs4_sequence_args *args,
 559                                 struct nfs4_sequence_res *res,
 560                                 struct rpc_task *task)
 561 {
 562         struct nfs4_slot *slot;
 563         struct nfs4_slot_table *tbl;
 564
 565         dprintk("--> %s\n", __func__);
 566         /* slot already allocated? */
 567         if (res->sr_slot != NULL)
 568                 goto out_success;
 569
 570         tbl = &session->fc_slot_table;
 571
 572         task->tk_timeout = 0;
 573
 574         spin_lock(&tbl->slot_tbl_lock);
 575         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 576             !args->sa_privileged) {
 577                 /* The state manager will wait until the slot table is empty */
 578                 dprintk("%s session is draining\n", __func__);
 579                 goto out_sleep;
 580         }
 581
 582         slot = nfs4_alloc_slot(tbl);
 583         if (IS_ERR(slot)) {
 584                 /* If out of memory, try again in 1/4 second */
 585                 if (slot == ERR_PTR(-ENOMEM))
 586                         task->tk_timeout = HZ >> 2;
 587                 dprintk("<-- %s: no free slots\n", __func__);
 588                 goto out_sleep;
 589         }
 590         spin_unlock(&tbl->slot_tbl_lock);
 591
 592         args->sa_slot = slot;
 593
 594         dprintk("<-- %s slotid=%d seqid=%d\n", __func__,
 595                         slot->slot_nr, slot->seq_nr);
 596
 597         res->sr_slot = slot;
 598         res->sr_timestamp = jiffies;
 599         res->sr_status_flags = 0;
 600         /*
 601          * sr_status is only set in decode_sequence, and so will remain
 602          * set to 1 if an rpc level failure occurs.
 603          */
 604         res->sr_status = 1;
 605 out_success:
 606         rpc_call_start(task);
 607         return 0;
 608 out_sleep:
 609         /* Privileged tasks are queued with top priority */
 610         if (args->sa_privileged)
 611                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 612                                 NULL, RPC_PRIORITY_PRIVILEGED);
 613         else
 614                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 615         spin_unlock(&tbl->slot_tbl_lock);
 616         return -EAGAIN;
 617 }
 618 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 619
 620 int nfs4_setup_sequence(const struct nfs_server *server,
 621                         struct nfs4_sequence_args *args,
 622                         struct nfs4_sequence_res *res,
 623                         struct rpc_task *task)
 624 {
 625         struct nfs4_session *session = nfs4_get_session(server);
 626         int ret = 0;
 627
 628         if (session == NULL) {
 629                 rpc_call_start(task);
 630                 goto out;
 631         }
 632
 633         dprintk("--> %s clp %p session %p sr_slot %d\n",
 634                 __func__, session->clp, session, res->sr_slot ?
 635                         res->sr_slot->slot_nr : -1);
 636
 637         ret = nfs41_setup_sequence(session, args, res, task);
 638 out:
 639         dprintk("<-- %s status=%d\n", __func__, ret);
 640         return ret;
 641 }
 642
 643 struct nfs41_call_sync_data {
 644         const struct nfs_server *seq_server;
 645         struct nfs4_sequence_args *seq_args;
 646         struct nfs4_sequence_res *seq_res;
 647 };
 648
 649 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 650 {
 651         struct nfs41_call_sync_data *data = calldata;
 652         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 653
 654         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 655
 656         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 657 }
 658
 659 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 660 {
 661         struct nfs41_call_sync_data *data = calldata;
 662
 663         nfs41_sequence_done(task, data->seq_res);
 664 }
 665
 666 static const struct rpc_call_ops nfs41_call_sync_ops = {
 667         .rpc_call_prepare = nfs41_call_sync_prepare,
 668         .rpc_call_done = nfs41_call_sync_done,
 669 };
 670
 671 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 672                                    struct nfs_server *server,
 673                                    struct rpc_message *msg,
 674                                    struct nfs4_sequence_args *args,
 675                                    struct nfs4_sequence_res *res)
 676 {
 677         int ret;
 678         struct rpc_task *task;
 679         struct nfs41_call_sync_data data = {
 680                 .seq_server = server,
 681                 .seq_args = args,
 682                 .seq_res = res,
 683         };
 684         struct rpc_task_setup task_setup = {
 685                 .rpc_client = clnt,
 686                 .rpc_message = msg,
 687                 .callback_ops = &nfs41_call_sync_ops,
 688                 .callback_data = &data
 689         };
 690
 691         task = rpc_run_task(&task_setup);
 692         if (IS_ERR(task))
 693                 ret = PTR_ERR(task);
 694         else {
 695                 ret = task->tk_status;
 696                 rpc_put_task(task);
 697         }
 698         return ret;
 699 }
 700
 701 #else
 702 static
 703 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 704                 struct nfs4_sequence_res *res, int cache_reply)
 705 {
 706 }
 707
 708 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 709 {
 710 }
 711
 712
 713 static int nfs4_sequence_done(struct rpc_task *task,
 714                                struct nfs4_sequence_res *res)
 715 {
 716         return 1;
 717 }
 718 #endif /* CONFIG_NFS_V4_1 */
 719
 720 static
 721 int _nfs4_call_sync(struct rpc_clnt *clnt,
 722                     struct nfs_server *server,
 723                     struct rpc_message *msg,
 724                     struct nfs4_sequence_args *args,
 725                     struct nfs4_sequence_res *res)
 726 {
 727         return rpc_call_sync(clnt, msg, 0);
 728 }
 729
 730 static
 731 int nfs4_call_sync(struct rpc_clnt *clnt,
 732                    struct nfs_server *server,
 733                    struct rpc_message *msg,
 734                    struct nfs4_sequence_args *args,
 735                    struct nfs4_sequence_res *res,
 736                    int cache_reply)
 737 {
 738         nfs41_init_sequence(args, res, cache_reply);
 739         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 740                                                 args, res);
 741 }
 742
 743 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 744 {
 745         struct nfs_inode *nfsi = NFS_I(dir);
 746
 747         spin_lock(&dir->i_lock);
 748         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 749         if (!cinfo->atomic || cinfo->before != dir->i_version)
 750                 nfs_force_lookup_revalidate(dir);
 751         dir->i_version = cinfo->after;
 752         nfs_fscache_invalidate(dir);
 753         spin_unlock(&dir->i_lock);
 754 }
 755
 756 struct nfs4_opendata {
 757         struct kref kref;
 758         struct nfs_openargs o_arg;
 759         struct nfs_openres o_res;
 760         struct nfs_open_confirmargs c_arg;
 761         struct nfs_open_confirmres c_res;
 762         struct nfs4_string owner_name;
 763         struct nfs4_string group_name;
 764         struct nfs_fattr f_attr;
 765         struct dentry *dir;
 766         struct dentry *dentry;
 767         struct nfs4_state_owner *owner;
 768         struct nfs4_state *state;
 769         struct iattr attrs;
 770         unsigned long timestamp;
 771         unsigned int rpc_done : 1;
 772         unsigned int is_recover : 1;
 773         int rpc_status;
 774         int cancelled;
 775 };
 776
 777 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 778                 int err, struct nfs4_exception *exception)
 779 {
 780         if (err != -EINVAL)
 781                 return false;
 782         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 783                 return false;
 784         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 785         exception->retry = 1;
 786         return true;
 787 }
 788
 789 static enum open_claim_type4
 790 nfs4_map_atomic_open_claim(struct nfs_server *server,
 791                 enum open_claim_type4 claim)
 792 {
 793         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 794                 return claim;
 795         switch (claim) {
 796         default:
 797                 return claim;
 798         case NFS4_OPEN_CLAIM_FH:
 799                 return NFS4_OPEN_CLAIM_NULL;
 800         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 801                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 802         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 803                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 804         }
 805 }
 806
 807 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 808 {
 809         p->o_res.f_attr = &p->f_attr;
 810         p->o_res.seqid = p->o_arg.seqid;
 811         p->c_res.seqid = p->c_arg.seqid;
 812         p->o_res.server = p->o_arg.server;
 813         p->o_res.access_request = p->o_arg.access;
 814         nfs_fattr_init(&p->f_attr);
 815         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 816 }
 817
 818 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 819                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 820                 const struct iattr *attrs,
 821                 enum open_claim_type4 claim,
 822                 gfp_t gfp_mask)
 823 {
 824         struct dentry *parent = dget_parent(dentry);
 825         struct inode *dir = parent->d_inode;
 826         struct nfs_server *server = NFS_SERVER(dir);
 827         struct nfs4_opendata *p;
 828
 829         p = kzalloc(sizeof(*p), gfp_mask);
 830         if (p == NULL)
 831                 goto err;
 832         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 833         if (p->o_arg.seqid == NULL)
 834                 goto err_free;
 835         nfs_sb_active(dentry->d_sb);
 836         p->dentry = dget(dentry);
 837         p->dir = parent;
 838         p->owner = sp;
 839         atomic_inc(&sp->so_count);
 840         p->o_arg.open_flags = flags;
 841         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 842         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
 843          * will return permission denied for all bits until close */
 844         if (!(flags & O_EXCL)) {
 845                 /* ask server to check for all possible rights as results
 846                  * are cached */
 847                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
 848                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
 849         }
 850         p->o_arg.clientid = server->nfs_client->cl_clientid;
 851         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
 852         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 853         p->o_arg.name = &dentry->d_name;
 854         p->o_arg.server = server;
 855         p->o_arg.bitmask = server->attr_bitmask;
 856         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
 857         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
 858         switch (p->o_arg.claim) {
 859         case NFS4_OPEN_CLAIM_NULL:
 860         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
 861         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
 862                 p->o_arg.fh = NFS_FH(dir);
 863                 break;
 864         case NFS4_OPEN_CLAIM_PREVIOUS:
 865         case NFS4_OPEN_CLAIM_FH:
 866         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 867         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 868                 p->o_arg.fh = NFS_FH(dentry->d_inode);
 869         }
 870         if (attrs != NULL && attrs->ia_valid != 0) {
 871                 __be32 verf[2];
 872
 873                 p->o_arg.u.attrs = &p->attrs;
 874                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 875
 876                 verf[0] = jiffies;
 877                 verf[1] = current->pid;
 878                 memcpy(p->o_arg.u.verifier.data, verf,
 879                                 sizeof(p->o_arg.u.verifier.data));
 880         }
 881         p->c_arg.fh = &p->o_res.fh;
 882         p->c_arg.stateid = &p->o_res.stateid;
 883         p->c_arg.seqid = p->o_arg.seqid;
 884         nfs4_init_opendata_res(p);
 885         kref_init(&p->kref);
 886         return p;
 887 err_free:
 888         kfree(p);
 889 err:
 890         dput(parent);
 891         return NULL;
 892 }
 893
 894 static void nfs4_opendata_free(struct kref *kref)
 895 {
 896         struct nfs4_opendata *p = container_of(kref,
 897                         struct nfs4_opendata, kref);
 898         struct super_block *sb = p->dentry->d_sb;
 899
 900         nfs_free_seqid(p->o_arg.seqid);
 901         if (p->state != NULL)
 902                 nfs4_put_open_state(p->state);
 903         nfs4_put_state_owner(p->owner);
 904         dput(p->dir);
 905         dput(p->dentry);
 906         nfs_sb_deactive(sb);
 907         nfs_fattr_free_names(&p->f_attr);
 908         kfree(p);
 909 }
 910
 911 static void nfs4_opendata_put(struct nfs4_opendata *p)
 912 {
 913         if (p != NULL)
 914                 kref_put(&p->kref, nfs4_opendata_free);
 915 }
 916
 917 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 918 {
 919         int ret;
 920
 921         ret = rpc_wait_for_completion_task(task);
 922         return ret;
 923 }
 924
 925 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 926 {
 927         int ret = 0;
 928
 929         if (open_mode & (O_EXCL|O_TRUNC))
 930                 goto out;
 931         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 932                 case FMODE_READ:
 933                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 934                                 && state->n_rdonly != 0;
 935                         break;
 936                 case FMODE_WRITE:
 937                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 938                                 && state->n_wronly != 0;
 939                         break;
 940                 case FMODE_READ|FMODE_WRITE:
 941                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 942                                 && state->n_rdwr != 0;
 943         }
 944 out:
 945         return ret;
 946 }
 947
 948 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 949 {
 950         if (delegation == NULL)
 951                 return 0;
 952         if ((delegation->type & fmode) != fmode)
 953                 return 0;
 954         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 955                 return 0;
 956         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
 957                 return 0;
 958         nfs_mark_delegation_referenced(delegation);
 959         return 1;
 960 }
 961
 962 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 963 {
 964         switch (fmode) {
 965                 case FMODE_WRITE:
 966                         state->n_wronly++;
 967                         break;
 968                 case FMODE_READ:
 969                         state->n_rdonly++;
 970                         break;
 971                 case FMODE_READ|FMODE_WRITE:
 972                         state->n_rdwr++;
 973         }
 974         nfs4_state_set_mode_locked(state, state->state | fmode);
 975 }
 976
 977 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 978 {
 979         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 980                 nfs4_stateid_copy(&state->stateid, stateid);
 981         nfs4_stateid_copy(&state->open_stateid, stateid);
 982         set_bit(NFS_OPEN_STATE, &state->flags);
 983         switch (fmode) {
 984                 case FMODE_READ:
 985                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 986                         break;
 987                 case FMODE_WRITE:
 988                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 989                         break;
 990                 case FMODE_READ|FMODE_WRITE:
 991                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 992         }
 993 }
 994
 995 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 996 {
 997         write_seqlock(&state->seqlock);
 998         nfs_set_open_stateid_locked(state, stateid, fmode);
 999         write_sequnlock(&state->seqlock);
1000 }
1001
1002 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1003 {
1004         /*
1005          * Protect the call to nfs4_state_set_mode_locked and
1006          * serialise the stateid update
1007          */
1008         write_seqlock(&state->seqlock);
1009         if (deleg_stateid != NULL) {
1010                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1011                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1012         }
1013         if (open_stateid != NULL)
1014                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1015         write_sequnlock(&state->seqlock);
1016         spin_lock(&state->owner->so_lock);
1017         update_open_stateflags(state, fmode);
1018         spin_unlock(&state->owner->so_lock);
1019 }
1020
1021 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1022 {
1023         struct nfs_inode *nfsi = NFS_I(state->inode);
1024         struct nfs_delegation *deleg_cur;
1025         int ret = 0;
1026
1027         fmode &= (FMODE_READ|FMODE_WRITE);
1028
1029         rcu_read_lock();
1030         deleg_cur = rcu_dereference(nfsi->delegation);
1031         if (deleg_cur == NULL)
1032                 goto no_delegation;
1033
1034         spin_lock(&deleg_cur->lock);
1035         if (nfsi->delegation != deleg_cur ||
1036            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1037             (deleg_cur->type & fmode) != fmode)
1038                 goto no_delegation_unlock;
1039
1040         if (delegation == NULL)
1041                 delegation = &deleg_cur->stateid;
1042         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1043                 goto no_delegation_unlock;
1044
1045         nfs_mark_delegation_referenced(deleg_cur);
1046         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1047         ret = 1;
1048 no_delegation_unlock:
1049         spin_unlock(&deleg_cur->lock);
1050 no_delegation:
1051         rcu_read_unlock();
1052
1053         if (!ret && open_stateid != NULL) {
1054                 __update_open_stateid(state, open_stateid, NULL, fmode);
1055                 ret = 1;
1056         }
1057
1058         return ret;
1059 }
1060
1061
1062 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1063 {
1064         struct nfs_delegation *delegation;
1065
1066         rcu_read_lock();
1067         delegation = rcu_dereference(NFS_I(inode)->delegation);
1068         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1069                 rcu_read_unlock();
1070                 return;
1071         }
1072         rcu_read_unlock();
1073         nfs4_inode_return_delegation(inode);
1074 }
1075
1076 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1077 {
1078         struct nfs4_state *state = opendata->state;
1079         struct nfs_inode *nfsi = NFS_I(state->inode);
1080         struct nfs_delegation *delegation;
1081         int open_mode = opendata->o_arg.open_flags;
1082         fmode_t fmode = opendata->o_arg.fmode;
1083         nfs4_stateid stateid;
1084         int ret = -EAGAIN;
1085
1086         for (;;) {
1087                 if (can_open_cached(state, fmode, open_mode)) {
1088                         spin_lock(&state->owner->so_lock);
1089                         if (can_open_cached(state, fmode, open_mode)) {
1090                                 update_open_stateflags(state, fmode);
1091                                 spin_unlock(&state->owner->so_lock);
1092                                 goto out_return_state;
1093                         }
1094                         spin_unlock(&state->owner->so_lock);
1095                 }
1096                 rcu_read_lock();
1097                 delegation = rcu_dereference(nfsi->delegation);
1098                 if (!can_open_delegated(delegation, fmode)) {
1099                         rcu_read_unlock();
1100                         break;
1101                 }
1102                 /* Save the delegation */
1103                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1104                 rcu_read_unlock();
1105                 nfs_release_seqid(opendata->o_arg.seqid);
1106                 if (!opendata->is_recover) {
1107                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1108                         if (ret != 0)
1109                                 goto out;
1110                 }
1111                 ret = -EAGAIN;
1112
1113                 /* Try to update the stateid using the delegation */
1114                 if (update_open_stateid(state, NULL, &stateid, fmode))
1115                         goto out_return_state;
1116         }
1117 out:
1118         return ERR_PTR(ret);
1119 out_return_state:
1120         atomic_inc(&state->count);
1121         return state;
1122 }
1123
1124 static void
1125 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1126 {
1127         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1128         struct nfs_delegation *delegation;
1129         int delegation_flags = 0;
1130
1131         rcu_read_lock();
1132         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1133         if (delegation)
1134                 delegation_flags = delegation->flags;
1135         rcu_read_unlock();
1136         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1137                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1138                                    "returning a delegation for "
1139                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1140                                    clp->cl_hostname);
1141         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1142                 nfs_inode_set_delegation(state->inode,
1143                                          data->owner->so_cred,
1144                                          &data->o_res);
1145         else
1146                 nfs_inode_reclaim_delegation(state->inode,
1147                                              data->owner->so_cred,
1148                                              &data->o_res);
1149 }
1150
1151 /*
1152  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1153  * and update the nfs4_state.
1154  */
1155 static struct nfs4_state *
1156 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1157 {
1158         struct inode *inode = data->state->inode;
1159         struct nfs4_state *state = data->state;
1160         int ret;
1161
1162         if (!data->rpc_done) {
1163                 if (data->rpc_status) {
1164                         ret = data->rpc_status;
1165                         goto err;
1166                 }
1167                 /* cached opens have already been processed */
1168                 goto update;
1169         }
1170
1171         ret = nfs_refresh_inode(inode, &data->f_attr);
1172         if (ret)
1173                 goto err;
1174
1175         if (data->o_res.delegation_type != 0)
1176                 nfs4_opendata_check_deleg(data, state);
1177 update:
1178         update_open_stateid(state, &data->o_res.stateid, NULL,
1179                             data->o_arg.fmode);
1180         atomic_inc(&state->count);
1181
1182         return state;
1183 err:
1184         return ERR_PTR(ret);
1185
1186 }
1187
1188 static struct nfs4_state *
1189 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1190 {
1191         struct inode *inode;
1192         struct nfs4_state *state = NULL;
1193         int ret;
1194
1195         if (!data->rpc_done) {
1196                 state = nfs4_try_open_cached(data);
1197                 goto out;
1198         }
1199
1200         ret = -EAGAIN;
1201         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1202                 goto err;
1203         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1204         ret = PTR_ERR(inode);
1205         if (IS_ERR(inode))
1206                 goto err;
1207         ret = -ENOMEM;
1208         state = nfs4_get_open_state(inode, data->owner);
1209         if (state == NULL)
1210                 goto err_put_inode;
1211         if (data->o_res.delegation_type != 0)
1212                 nfs4_opendata_check_deleg(data, state);
1213         update_open_stateid(state, &data->o_res.stateid, NULL,
1214                         data->o_arg.fmode);
1215         iput(inode);
1216 out:
1217         nfs_release_seqid(data->o_arg.seqid);
1218         return state;
1219 err_put_inode:
1220         iput(inode);
1221 err:
1222         return ERR_PTR(ret);
1223 }
1224
1225 static struct nfs4_state *
1226 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1227 {
1228         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1229                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1230         return _nfs4_opendata_to_nfs4_state(data);
1231 }
1232
1233 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1234 {
1235         struct nfs_inode *nfsi = NFS_I(state->inode);
1236         struct nfs_open_context *ctx;
1237
1238         spin_lock(&state->inode->i_lock);
1239         list_for_each_entry(ctx, &nfsi->open_files, list) {
1240                 if (ctx->state != state)
1241                         continue;
1242                 get_nfs_open_context(ctx);
1243                 spin_unlock(&state->inode->i_lock);
1244                 return ctx;
1245         }
1246         spin_unlock(&state->inode->i_lock);
1247         return ERR_PTR(-ENOENT);
1248 }
1249
1250 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1251                 struct nfs4_state *state, enum open_claim_type4 claim)
1252 {
1253         struct nfs4_opendata *opendata;
1254
1255         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1256                         NULL, claim, GFP_NOFS);
1257         if (opendata == NULL)
1258                 return ERR_PTR(-ENOMEM);
1259         opendata->state = state;
1260         atomic_inc(&state->count);
1261         return opendata;
1262 }
1263
1264 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1265 {
1266         struct nfs4_state *newstate;
1267         int ret;
1268
1269         opendata->o_arg.open_flags = 0;
1270         opendata->o_arg.fmode = fmode;
1271         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1272         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1273         nfs4_init_opendata_res(opendata);
1274         ret = _nfs4_recover_proc_open(opendata);
1275         if (ret != 0)
1276                 return ret;
1277         newstate = nfs4_opendata_to_nfs4_state(opendata);
1278         if (IS_ERR(newstate))
1279                 return PTR_ERR(newstate);
1280         nfs4_close_state(newstate, fmode);
1281         *res = newstate;
1282         return 0;
1283 }
1284
1285 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1286 {
1287         struct nfs4_state *newstate;
1288         int ret;
1289
1290         /* memory barrier prior to reading state->n_* */
1291         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1292         clear_bit(NFS_OPEN_STATE, &state->flags);
1293         smp_rmb();
1294         if (state->n_rdwr != 0) {
1295                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1296                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1297                 if (ret != 0)
1298                         return ret;
1299                 if (newstate != state)
1300                         return -ESTALE;
1301         }
1302         if (state->n_wronly != 0) {
1303                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1304                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1305                 if (ret != 0)
1306                         return ret;
1307                 if (newstate != state)
1308                         return -ESTALE;
1309         }
1310         if (state->n_rdonly != 0) {
1311                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1312                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1313                 if (ret != 0)
1314                         return ret;
1315                 if (newstate != state)
1316                         return -ESTALE;
1317         }
1318         /*
1319          * We may have performed cached opens for all three recoveries.
1320          * Check if we need to update the current stateid.
1321          */
1322         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1323             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1324                 write_seqlock(&state->seqlock);
1325                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1326                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1327                 write_sequnlock(&state->seqlock);
1328         }
1329         return 0;
1330 }
1331
1332 /*
1333  * OPEN_RECLAIM:
1334  *      reclaim state on the server after a reboot.
1335  */
1336 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1337 {
1338         struct nfs_delegation *delegation;
1339         struct nfs4_opendata *opendata;
1340         fmode_t delegation_type = 0;
1341         int status;
1342
1343         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1344                         NFS4_OPEN_CLAIM_PREVIOUS);
1345         if (IS_ERR(opendata))
1346                 return PTR_ERR(opendata);
1347         rcu_read_lock();
1348         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1349         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1350                 delegation_type = delegation->type;
1351         rcu_read_unlock();
1352         opendata->o_arg.u.delegation_type = delegation_type;
1353         status = nfs4_open_recover(opendata, state);
1354         nfs4_opendata_put(opendata);
1355         return status;
1356 }
1357
1358 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1359 {
1360         struct nfs_server *server = NFS_SERVER(state->inode);
1361         struct nfs4_exception exception = { };
1362         int err;
1363         do {
1364                 err = _nfs4_do_open_reclaim(ctx, state);
1365                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1366                         continue;
1367                 if (err != -NFS4ERR_DELAY)
1368                         break;
1369                 nfs4_handle_exception(server, err, &exception);
1370         } while (exception.retry);
1371         return err;
1372 }
1373
1374 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1375 {
1376         struct nfs_open_context *ctx;
1377         int ret;
1378
1379         ctx = nfs4_state_find_open_context(state);
1380         if (IS_ERR(ctx))
1381                 return -EAGAIN;
1382         ret = nfs4_do_open_reclaim(ctx, state);
1383         put_nfs_open_context(ctx);
1384         return ret;
1385 }
1386
1387 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1388 {
1389         switch (err) {
1390                 default:
1391                         printk(KERN_ERR "NFS: %s: unhandled error "
1392                                         "%d.\n", __func__, err);
1393                 case 0:
1394                 case -ENOENT:
1395                 case -ESTALE:
1396                         break;
1397                 case -NFS4ERR_BADSESSION:
1398                 case -NFS4ERR_BADSLOT:
1399                 case -NFS4ERR_BAD_HIGH_SLOT:
1400                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1401                 case -NFS4ERR_DEADSESSION:
1402                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1403                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1404                         return -EAGAIN;
1405                 case -NFS4ERR_STALE_CLIENTID:
1406                 case -NFS4ERR_STALE_STATEID:
1407                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1408                 case -NFS4ERR_EXPIRED:
1409                         /* Don't recall a delegation if it was lost */
1410                         nfs4_schedule_lease_recovery(server->nfs_client);
1411                         return -EAGAIN;
1412                 case -NFS4ERR_DELEG_REVOKED:
1413                 case -NFS4ERR_ADMIN_REVOKED:
1414                 case -NFS4ERR_BAD_STATEID:
1415                 case -NFS4ERR_OPENMODE:
1416                         nfs_inode_find_state_and_recover(state->inode,
1417                                         stateid);
1418                         nfs4_schedule_stateid_recovery(server, state);
1419                         return -EAGAIN;
1420                 case -NFS4ERR_DELAY:
1421                 case -NFS4ERR_GRACE:
1422                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1423                         ssleep(1);
1424                         return -EAGAIN;
1425                 case -ENOMEM:
1426                 case -NFS4ERR_DENIED:
1427                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1428                         return 0;
1429         }
1430         return err;
1431 }
1432
1433 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1434 {
1435         struct nfs_server *server = NFS_SERVER(state->inode);
1436         struct nfs4_opendata *opendata;
1437         int err;
1438
1439         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1440                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1441         if (IS_ERR(opendata))
1442                 return PTR_ERR(opendata);
1443         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1444         err = nfs4_open_recover(opendata, state);
1445         nfs4_opendata_put(opendata);
1446         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1447 }
1448
1449 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1450 {
1451         struct nfs4_opendata *data = calldata;
1452
1453         data->rpc_status = task->tk_status;
1454         if (data->rpc_status == 0) {
1455                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1456                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1457                 renew_lease(data->o_res.server, data->timestamp);
1458                 data->rpc_done = 1;
1459         }
1460 }
1461
1462 static void nfs4_open_confirm_release(void *calldata)
1463 {
1464         struct nfs4_opendata *data = calldata;
1465         struct nfs4_state *state = NULL;
1466
1467         /* If this request hasn't been cancelled, do nothing */
1468         if (data->cancelled == 0)
1469                 goto out_free;
1470         /* In case of error, no cleanup! */
1471         if (!data->rpc_done)
1472                 goto out_free;
1473         state = nfs4_opendata_to_nfs4_state(data);
1474         if (!IS_ERR(state))
1475                 nfs4_close_state(state, data->o_arg.fmode);
1476 out_free:
1477         nfs4_opendata_put(data);
1478 }
1479
1480 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1481         .rpc_call_done = nfs4_open_confirm_done,
1482         .rpc_release = nfs4_open_confirm_release,
1483 };
1484
1485 /*
1486  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1487  */
1488 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1489 {
1490         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1491         struct rpc_task *task;
1492         struct  rpc_message msg = {
1493                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1494                 .rpc_argp = &data->c_arg,
1495                 .rpc_resp = &data->c_res,
1496                 .rpc_cred = data->owner->so_cred,
1497         };
1498         struct rpc_task_setup task_setup_data = {
1499                 .rpc_client = server->client,
1500                 .rpc_message = &msg,
1501                 .callback_ops = &nfs4_open_confirm_ops,
1502                 .callback_data = data,
1503                 .workqueue = nfsiod_workqueue,
1504                 .flags = RPC_TASK_ASYNC,
1505         };
1506         int status;
1507
1508         kref_get(&data->kref);
1509         data->rpc_done = 0;
1510         data->rpc_status = 0;
1511         data->timestamp = jiffies;
1512         task = rpc_run_task(&task_setup_data);
1513         if (IS_ERR(task))
1514                 return PTR_ERR(task);
1515         status = nfs4_wait_for_completion_rpc_task(task);
1516         if (status != 0) {
1517                 data->cancelled = 1;
1518                 smp_wmb();
1519         } else
1520                 status = data->rpc_status;
1521         rpc_put_task(task);
1522         return status;
1523 }
1524
1525 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1526 {
1527         struct nfs4_opendata *data = calldata;
1528         struct nfs4_state_owner *sp = data->owner;
1529         struct nfs_client *clp = sp->so_server->nfs_client;
1530
1531         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1532                 goto out_wait;
1533         /*
1534          * Check if we still need to send an OPEN call, or if we can use
1535          * a delegation instead.
1536          */
1537         if (data->state != NULL) {
1538                 struct nfs_delegation *delegation;
1539
1540                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1541                         goto out_no_action;
1542                 rcu_read_lock();
1543                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1544                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1545                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1546                     can_open_delegated(delegation, data->o_arg.fmode))
1547                         goto unlock_no_action;
1548                 rcu_read_unlock();
1549         }
1550         /* Update client id. */
1551         data->o_arg.clientid = clp->cl_clientid;
1552         switch (data->o_arg.claim) {
1553         case NFS4_OPEN_CLAIM_PREVIOUS:
1554         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1555         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1556                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1557         case NFS4_OPEN_CLAIM_FH:
1558                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1559                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1560         }
1561         data->timestamp = jiffies;
1562         if (nfs4_setup_sequence(data->o_arg.server,
1563                                 &data->o_arg.seq_args,
1564                                 &data->o_res.seq_res,
1565                                 task) != 0)
1566                 nfs_release_seqid(data->o_arg.seqid);
1567
1568         /* Set the create mode (note dependency on the session type) */
1569         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1570         if (data->o_arg.open_flags & O_EXCL) {
1571                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1572                 if (nfs4_has_persistent_session(clp))
1573                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1574                 else if (clp->cl_mvops->minor_version > 0)
1575                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1576         }
1577         return;
1578 unlock_no_action:
1579         rcu_read_unlock();
1580 out_no_action:
1581         task->tk_action = NULL;
1582 out_wait:
1583         nfs4_sequence_done(task, &data->o_res.seq_res);
1584 }
1585
1586 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1587 {
1588         struct nfs4_opendata *data = calldata;
1589
1590         data->rpc_status = task->tk_status;
1591
1592         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1593                 return;
1594
1595         if (task->tk_status == 0) {
1596                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1597                         switch (data->o_res.f_attr->mode & S_IFMT) {
1598                         case S_IFREG:
1599                                 break;
1600                         case S_IFLNK:
1601                                 data->rpc_status = -ELOOP;
1602                                 break;
1603                         case S_IFDIR:
1604                                 data->rpc_status = -EISDIR;
1605                                 break;
1606                         default:
1607                                 data->rpc_status = -ENOTDIR;
1608                         }
1609                 }
1610                 renew_lease(data->o_res.server, data->timestamp);
1611                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1612                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1613         }
1614         data->rpc_done = 1;
1615 }
1616
1617 static void nfs4_open_release(void *calldata)
1618 {
1619         struct nfs4_opendata *data = calldata;
1620         struct nfs4_state *state = NULL;
1621
1622         /* If this request hasn't been cancelled, do nothing */
1623         if (data->cancelled == 0)
1624                 goto out_free;
1625         /* In case of error, no cleanup! */
1626         if (data->rpc_status != 0 || !data->rpc_done)
1627                 goto out_free;
1628         /* In case we need an open_confirm, no cleanup! */
1629         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1630                 goto out_free;
1631         state = nfs4_opendata_to_nfs4_state(data);
1632         if (!IS_ERR(state))
1633                 nfs4_close_state(state, data->o_arg.fmode);
1634 out_free:
1635         nfs4_opendata_put(data);
1636 }
1637
1638 static const struct rpc_call_ops nfs4_open_ops = {
1639         .rpc_call_prepare = nfs4_open_prepare,
1640         .rpc_call_done = nfs4_open_done,
1641         .rpc_release = nfs4_open_release,
1642 };
1643
1644 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1645 {
1646         struct inode *dir = data->dir->d_inode;
1647         struct nfs_server *server = NFS_SERVER(dir);
1648         struct nfs_openargs *o_arg = &data->o_arg;
1649         struct nfs_openres *o_res = &data->o_res;
1650         struct rpc_task *task;
1651         struct rpc_message msg = {
1652                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1653                 .rpc_argp = o_arg,
1654                 .rpc_resp = o_res,
1655                 .rpc_cred = data->owner->so_cred,
1656         };
1657         struct rpc_task_setup task_setup_data = {
1658                 .rpc_client = server->client,
1659                 .rpc_message = &msg,
1660                 .callback_ops = &nfs4_open_ops,
1661                 .callback_data = data,
1662                 .workqueue = nfsiod_workqueue,
1663                 .flags = RPC_TASK_ASYNC,
1664         };
1665         int status;
1666
1667         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1668         kref_get(&data->kref);
1669         data->rpc_done = 0;
1670         data->rpc_status = 0;
1671         data->cancelled = 0;
1672         data->is_recover = 0;
1673         if (isrecover) {
1674                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1675                 data->is_recover = 1;
1676         }
1677         task = rpc_run_task(&task_setup_data);
1678         if (IS_ERR(task))
1679                 return PTR_ERR(task);
1680         status = nfs4_wait_for_completion_rpc_task(task);
1681         if (status != 0) {
1682                 data->cancelled = 1;
1683                 smp_wmb();
1684         } else
1685                 status = data->rpc_status;
1686         rpc_put_task(task);
1687
1688         return status;
1689 }
1690
1691 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1692 {
1693         struct inode *dir = data->dir->d_inode;
1694         struct nfs_openres *o_res = &data->o_res;
1695         int status;
1696
1697         status = nfs4_run_open_task(data, 1);
1698         if (status != 0 || !data->rpc_done)
1699                 return status;
1700
1701         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1702
1703         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1704                 status = _nfs4_proc_open_confirm(data);
1705                 if (status != 0)
1706                         return status;
1707         }
1708
1709         return status;
1710 }
1711
1712 static int nfs4_opendata_access(struct rpc_cred *cred,
1713                                 struct nfs4_opendata *opendata,
1714                                 struct nfs4_state *state, fmode_t fmode,
1715                                 int openflags)
1716 {
1717         struct nfs_access_entry cache;
1718         u32 mask;
1719
1720         /* access call failed or for some reason the server doesn't
1721          * support any access modes -- defer access call until later */
1722         if (opendata->o_res.access_supported == 0)
1723                 return 0;
1724
1725         mask = 0;
1726         /* don't check MAY_WRITE - a newly created file may not have
1727          * write mode bits, but POSIX allows the creating process to write.
1728          * use openflags to check for exec, because fmode won't
1729          * always have FMODE_EXEC set when file open for exec. */
1730         if (openflags & __FMODE_EXEC) {
1731                 /* ONLY check for exec rights */
1732                 mask = MAY_EXEC;
1733         } else if (fmode & FMODE_READ)
1734                 mask = MAY_READ;
1735
1736         cache.cred = cred;
1737         cache.jiffies = jiffies;
1738         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1739         nfs_access_add_cache(state->inode, &cache);
1740
1741         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1742                 return 0;
1743
1744         /* even though OPEN succeeded, access is denied. Close the file */
1745         nfs4_close_state(state, fmode);
1746         return -EACCES;
1747 }
1748
1749 /*
1750  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1751  */
1752 static int _nfs4_proc_open(struct nfs4_opendata *data)
1753 {
1754         struct inode *dir = data->dir->d_inode;
1755         struct nfs_server *server = NFS_SERVER(dir);
1756         struct nfs_openargs *o_arg = &data->o_arg;
1757         struct nfs_openres *o_res = &data->o_res;
1758         int status;
1759
1760         status = nfs4_run_open_task(data, 0);
1761         if (!data->rpc_done)
1762                 return status;
1763         if (status != 0) {
1764                 if (status == -NFS4ERR_BADNAME &&
1765                                 !(o_arg->open_flags & O_CREAT))
1766                         return -ENOENT;
1767                 return status;
1768         }
1769
1770         nfs_fattr_map_and_free_names(server, &data->f_attr);
1771
1772         if (o_arg->open_flags & O_CREAT)
1773                 update_changeattr(dir, &o_res->cinfo);
1774         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1775                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1776         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1777                 status = _nfs4_proc_open_confirm(data);
1778                 if (status != 0)
1779                         return status;
1780         }
1781         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1782                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1783         return 0;
1784 }
1785
1786 static int nfs4_recover_expired_lease(struct nfs_server *server)
1787 {
1788         return nfs4_client_recover_expired_lease(server->nfs_client);
1789 }
1790
1791 /*
1792  * OPEN_EXPIRED:
1793  *      reclaim state on the server after a network partition.
1794  *      Assumes caller holds the appropriate lock
1795  */
1796 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1797 {
1798         struct nfs4_opendata *opendata;
1799         int ret;
1800
1801         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1802                         NFS4_OPEN_CLAIM_FH);
1803         if (IS_ERR(opendata))
1804                 return PTR_ERR(opendata);
1805         ret = nfs4_open_recover(opendata, state);
1806         if (ret == -ESTALE)
1807                 d_drop(ctx->dentry);
1808         nfs4_opendata_put(opendata);
1809         return ret;
1810 }
1811
1812 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1813 {
1814         struct nfs_server *server = NFS_SERVER(state->inode);
1815         struct nfs4_exception exception = { };
1816         int err;
1817
1818         do {
1819                 err = _nfs4_open_expired(ctx, state);
1820                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1821                         continue;
1822                 switch (err) {
1823                 default:
1824                         goto out;
1825                 case -NFS4ERR_GRACE:
1826                 case -NFS4ERR_DELAY:
1827                         nfs4_handle_exception(server, err, &exception);
1828                         err = 0;
1829                 }
1830         } while (exception.retry);
1831 out:
1832         return err;
1833 }
1834
1835 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1836 {
1837         struct nfs_open_context *ctx;
1838         int ret;
1839
1840         ctx = nfs4_state_find_open_context(state);
1841         if (IS_ERR(ctx))
1842                 return -EAGAIN;
1843         ret = nfs4_do_open_expired(ctx, state);
1844         put_nfs_open_context(ctx);
1845         return ret;
1846 }
1847
1848 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
1849 {
1850         nfs_remove_bad_delegation(state->inode);
1851         write_seqlock(&state->seqlock);
1852         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1853         write_sequnlock(&state->seqlock);
1854         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1855 }
1856
1857 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
1858 {
1859         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
1860                 nfs_finish_clear_delegation_stateid(state);
1861 }
1862
1863 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1864 {
1865         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
1866         nfs40_clear_delegation_stateid(state);
1867         return nfs4_open_expired(sp, state);
1868 }
1869
1870 #if defined(CONFIG_NFS_V4_1)
1871 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1872 {
1873         struct nfs_server *server = NFS_SERVER(state->inode);
1874         nfs4_stateid *stateid = &state->stateid;
1875         int status;
1876
1877         /* If a state reset has been done, test_stateid is unneeded */
1878         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1879                 return;
1880
1881         status = nfs41_test_stateid(server, stateid);
1882         if (status != NFS_OK) {
1883                 /* Free the stateid unless the server explicitly
1884                  * informs us the stateid is unrecognized. */
1885                 if (status != -NFS4ERR_BAD_STATEID)
1886                         nfs41_free_stateid(server, stateid);
1887                 nfs_remove_bad_delegation(state->inode);
1888
1889                 write_seqlock(&state->seqlock);
1890                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1891                 write_sequnlock(&state->seqlock);
1892                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1893         }
1894 }
1895
1896 /**
1897  * nfs41_check_open_stateid - possibly free an open stateid
1898  *
1899  * @state: NFSv4 state for an inode
1900  *
1901  * Returns NFS_OK if recovery for this stateid is now finished.
1902  * Otherwise a negative NFS4ERR value is returned.
1903  */
1904 static int nfs41_check_open_stateid(struct nfs4_state *state)
1905 {
1906         struct nfs_server *server = NFS_SERVER(state->inode);
1907         nfs4_stateid *stateid = &state->open_stateid;
1908         int status;
1909
1910         /* If a state reset has been done, test_stateid is unneeded */
1911         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1912             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1913             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1914                 return -NFS4ERR_BAD_STATEID;
1915
1916         status = nfs41_test_stateid(server, stateid);
1917         if (status != NFS_OK) {
1918                 /* Free the stateid unless the server explicitly
1919                  * informs us the stateid is unrecognized. */
1920                 if (status != -NFS4ERR_BAD_STATEID)
1921                         nfs41_free_stateid(server, stateid);
1922
1923                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1924                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1925                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1926                 clear_bit(NFS_OPEN_STATE, &state->flags);
1927         }
1928         return status;
1929 }
1930
1931 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1932 {
1933         int status;
1934
1935         nfs41_clear_delegation_stateid(state);
1936         status = nfs41_check_open_stateid(state);
1937         if (status != NFS_OK)
1938                 status = nfs4_open_expired(sp, state);
1939         return status;
1940 }
1941 #endif
1942
1943 /*
1944  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1945  * fields corresponding to attributes that were used to store the verifier.
1946  * Make sure we clobber those fields in the later setattr call
1947  */
1948 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1949 {
1950         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1951             !(sattr->ia_valid & ATTR_ATIME_SET))
1952                 sattr->ia_valid |= ATTR_ATIME;
1953
1954         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1955             !(sattr->ia_valid & ATTR_MTIME_SET))
1956                 sattr->ia_valid |= ATTR_MTIME;
1957 }
1958
1959 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
1960                 fmode_t fmode,
1961                 int flags,
1962                 struct nfs4_state **res)
1963 {
1964         struct nfs4_state_owner *sp = opendata->owner;
1965         struct nfs_server *server = sp->so_server;
1966         struct nfs4_state *state;
1967         unsigned int seq;
1968         int ret;
1969
1970         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
1971
1972         ret = _nfs4_proc_open(opendata);
1973         if (ret != 0)
1974                 goto out;
1975
1976         state = nfs4_opendata_to_nfs4_state(opendata);
1977         ret = PTR_ERR(state);
1978         if (IS_ERR(state))
1979                 goto out;
1980         if (server->caps & NFS_CAP_POSIX_LOCK)
1981                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1982
1983         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
1984         if (ret != 0)
1985                 goto out;
1986
1987         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
1988                 nfs4_schedule_stateid_recovery(server, state);
1989         *res = state;
1990 out:
1991         return ret;
1992 }
1993
1994 /*
1995  * Returns a referenced nfs4_state
1996  */
1997 static int _nfs4_do_open(struct inode *dir,
1998                         struct dentry *dentry,
1999                         fmode_t fmode,
2000                         int flags,
2001                         struct iattr *sattr,
2002                         struct rpc_cred *cred,
2003                         struct nfs4_state **res,
2004                         struct nfs4_threshold **ctx_th)
2005 {
2006         struct nfs4_state_owner  *sp;
2007         struct nfs4_state     *state = NULL;
2008         struct nfs_server       *server = NFS_SERVER(dir);
2009         struct nfs4_opendata *opendata;
2010         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2011         int status;
2012
2013         /* Protect against reboot recovery conflicts */
2014         status = -ENOMEM;
2015         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2016         if (sp == NULL) {
2017                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2018                 goto out_err;
2019         }
2020         status = nfs4_recover_expired_lease(server);
2021         if (status != 0)
2022                 goto err_put_state_owner;
2023         if (dentry->d_inode != NULL)
2024                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2025         status = -ENOMEM;
2026         if (dentry->d_inode)
2027                 claim = NFS4_OPEN_CLAIM_FH;
2028         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2029                         claim, GFP_KERNEL);
2030         if (opendata == NULL)
2031                 goto err_put_state_owner;
2032
2033         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2034                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2035                 if (!opendata->f_attr.mdsthreshold)
2036                         goto err_opendata_put;
2037                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2038         }
2039         if (dentry->d_inode != NULL)
2040                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2041
2042         status = _nfs4_open_and_get_state(opendata, fmode, flags, &state);
2043         if (status != 0)
2044                 goto err_opendata_put;
2045
2046         if ((opendata->o_arg.open_flags & O_EXCL) &&
2047             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2048                 nfs4_exclusive_attrset(opendata, sattr);
2049
2050                 nfs_fattr_init(opendata->o_res.f_attr);
2051                 status = nfs4_do_setattr(state->inode, cred,
2052                                 opendata->o_res.f_attr, sattr,
2053                                 state);
2054                 if (status == 0)
2055                         nfs_setattr_update_inode(state->inode, sattr);
2056                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2057         }
2058
2059         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2060                 *ctx_th = opendata->f_attr.mdsthreshold;
2061         else
2062                 kfree(opendata->f_attr.mdsthreshold);
2063         opendata->f_attr.mdsthreshold = NULL;
2064
2065         nfs4_opendata_put(opendata);
2066         nfs4_put_state_owner(sp);
2067         *res = state;
2068         return 0;
2069 err_opendata_put:
2070         kfree(opendata->f_attr.mdsthreshold);
2071         nfs4_opendata_put(opendata);
2072 err_put_state_owner:
2073         nfs4_put_state_owner(sp);
2074 out_err:
2075         *res = NULL;
2076         return status;
2077 }
2078
2079
2080 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2081                                         struct dentry *dentry,
2082                                         fmode_t fmode,
2083                                         int flags,
2084                                         struct iattr *sattr,
2085                                         struct rpc_cred *cred,
2086                                         struct nfs4_threshold **ctx_th)
2087 {
2088         struct nfs_server *server = NFS_SERVER(dir);
2089         struct nfs4_exception exception = { };
2090         struct nfs4_state *res;
2091         int status;
2092
2093         fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
2094         do {
2095                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
2096                                        &res, ctx_th);
2097                 if (status == 0)
2098                         break;
2099                 /* NOTE: BAD_SEQID means the server and client disagree about the
2100                  * book-keeping w.r.t. state-changing operations
2101                  * (OPEN/CLOSE/LOCK/LOCKU...)
2102                  * It is actually a sign of a bug on the client or on the server.
2103                  *
2104                  * If we receive a BAD_SEQID error in the particular case of
2105                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2106                  * have unhashed the old state_owner for us, and that we can
2107                  * therefore safely retry using a new one. We should still warn
2108                  * the user though...
2109                  */
2110                 if (status == -NFS4ERR_BAD_SEQID) {
2111                         pr_warn_ratelimited("NFS: v4 server %s "
2112                                         " returned a bad sequence-id error!\n",
2113                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2114                         exception.retry = 1;
2115                         continue;
2116                 }
2117                 /*
2118                  * BAD_STATEID on OPEN means that the server cancelled our
2119                  * state before it received the OPEN_CONFIRM.
2120                  * Recover by retrying the request as per the discussion
2121                  * on Page 181 of RFC3530.
2122                  */
2123                 if (status == -NFS4ERR_BAD_STATEID) {
2124                         exception.retry = 1;
2125                         continue;
2126                 }
2127                 if (status == -EAGAIN) {
2128                         /* We must have found a delegation */
2129                         exception.retry = 1;
2130                         continue;
2131                 }
2132                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2133                         continue;
2134                 res = ERR_PTR(nfs4_handle_exception(server,
2135                                         status, &exception));
2136         } while (exception.retry);
2137         return res;
2138 }
2139
2140 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2141                             struct nfs_fattr *fattr, struct iattr *sattr,
2142                             struct nfs4_state *state)
2143 {
2144         struct nfs_server *server = NFS_SERVER(inode);
2145         struct nfs_setattrargs  arg = {
2146                 .fh             = NFS_FH(inode),
2147                 .iap            = sattr,
2148                 .server         = server,
2149                 .bitmask = server->attr_bitmask,
2150         };
2151         struct nfs_setattrres  res = {
2152                 .fattr          = fattr,
2153                 .server         = server,
2154         };
2155         struct rpc_message msg = {
2156                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2157                 .rpc_argp       = &arg,
2158                 .rpc_resp       = &res,
2159                 .rpc_cred       = cred,
2160         };
2161         unsigned long timestamp = jiffies;
2162         fmode_t fmode;
2163         bool truncate;
2164         int status;
2165
2166         nfs_fattr_init(fattr);
2167
2168         /* Servers should only apply open mode checks for file size changes */
2169         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2170         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2171
2172         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2173                 /* Use that stateid */
2174         } else if (truncate && state != NULL && nfs4_valid_open_stateid(state)) {
2175                 struct nfs_lockowner lockowner = {
2176                         .l_owner = current->files,
2177                         .l_pid = current->tgid,
2178                 };
2179                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2180                                 &lockowner);
2181         } else
2182                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2183
2184         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2185         if (status == 0 && state != NULL)
2186                 renew_lease(server, timestamp);
2187         return status;
2188 }
2189
2190 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2191                            struct nfs_fattr *fattr, struct iattr *sattr,
2192                            struct nfs4_state *state)
2193 {
2194         struct nfs_server *server = NFS_SERVER(inode);
2195         struct nfs4_exception exception = {
2196                 .state = state,
2197                 .inode = inode,
2198         };
2199         int err;
2200         do {
2201                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2202                 switch (err) {
2203                 case -NFS4ERR_OPENMODE:
2204                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2205                                 pr_warn_once("NFSv4: server %s is incorrectly "
2206                                                 "applying open mode checks to "
2207                                                 "a SETATTR that is not "
2208                                                 "changing file size.\n",
2209                                                 server->nfs_client->cl_hostname);
2210                         }
2211                         if (state && !(state->state & FMODE_WRITE)) {
2212                                 err = -EBADF;
2213                                 if (sattr->ia_valid & ATTR_OPEN)
2214                                         err = -EACCES;
2215                                 goto out;
2216                         }
2217                 }
2218                 err = nfs4_handle_exception(server, err, &exception);
2219         } while (exception.retry);
2220 out:
2221         return err;
2222 }
2223
2224 struct nfs4_closedata {
2225         struct inode *inode;
2226         struct nfs4_state *state;
2227         struct nfs_closeargs arg;
2228         struct nfs_closeres res;
2229         struct nfs_fattr fattr;
2230         unsigned long timestamp;
2231         bool roc;
2232         u32 roc_barrier;
2233 };
2234
2235 static void nfs4_free_closedata(void *data)
2236 {
2237         struct nfs4_closedata *calldata = data;
2238         struct nfs4_state_owner *sp = calldata->state->owner;
2239         struct super_block *sb = calldata->state->inode->i_sb;
2240
2241         if (calldata->roc)
2242                 pnfs_roc_release(calldata->state->inode);
2243         nfs4_put_open_state(calldata->state);
2244         nfs_free_seqid(calldata->arg.seqid);
2245         nfs4_put_state_owner(sp);
2246         nfs_sb_deactive(sb);
2247         kfree(calldata);
2248 }
2249
2250 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2251                 fmode_t fmode)
2252 {
2253         spin_lock(&state->owner->so_lock);
2254         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2255         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
2256         case FMODE_WRITE:
2257                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2258                 break;
2259         case FMODE_READ:
2260                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2261                 break;
2262         case 0:
2263                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2264                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2265                 clear_bit(NFS_OPEN_STATE, &state->flags);
2266         }
2267         spin_unlock(&state->owner->so_lock);
2268 }
2269
2270 static void nfs4_close_done(struct rpc_task *task, void *data)
2271 {
2272         struct nfs4_closedata *calldata = data;
2273         struct nfs4_state *state = calldata->state;
2274         struct nfs_server *server = NFS_SERVER(calldata->inode);
2275
2276         dprintk("%s: begin!\n", __func__);
2277         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2278                 return;
2279         /* hmm. we are done with the inode, and in the process of freeing
2280          * the state_owner. we keep this around to process errors
2281          */
2282         switch (task->tk_status) {
2283                 case 0:
2284                         if (calldata->roc)
2285                                 pnfs_roc_set_barrier(state->inode,
2286                                                      calldata->roc_barrier);
2287                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2288                         renew_lease(server, calldata->timestamp);
2289                         nfs4_close_clear_stateid_flags(state,
2290                                         calldata->arg.fmode);
2291                         break;
2292                 case -NFS4ERR_STALE_STATEID:
2293                 case -NFS4ERR_OLD_STATEID:
2294                 case -NFS4ERR_BAD_STATEID:
2295                 case -NFS4ERR_EXPIRED:
2296                         if (calldata->arg.fmode == 0)
2297                                 break;
2298                 default:
2299                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2300                                 rpc_restart_call_prepare(task);
2301         }
2302         nfs_release_seqid(calldata->arg.seqid);
2303         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2304         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2305 }
2306
2307 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2308 {
2309         struct nfs4_closedata *calldata = data;
2310         struct nfs4_state *state = calldata->state;
2311         struct inode *inode = calldata->inode;
2312         bool is_rdonly, is_wronly, is_rdwr;
2313         int call_close = 0;
2314
2315         dprintk("%s: begin!\n", __func__);
2316         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2317                 goto out_wait;
2318
2319         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2320         spin_lock(&state->owner->so_lock);
2321         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2322         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2323         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2324         /* Calculate the change in open mode */
2325         calldata->arg.fmode = 0;
2326         if (state->n_rdwr == 0) {
2327                 if (state->n_rdonly == 0)
2328                         call_close |= is_rdonly;
2329                 else if (is_rdonly)
2330                         calldata->arg.fmode |= FMODE_READ;
2331                 if (state->n_wronly == 0)
2332                         call_close |= is_wronly;
2333                 else if (is_wronly)
2334                         calldata->arg.fmode |= FMODE_WRITE;
2335         } else if (is_rdwr)
2336                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2337
2338         if (calldata->arg.fmode == 0)
2339                 call_close |= is_rdwr;
2340
2341         if (!nfs4_valid_open_stateid(state))
2342                 call_close = 0;
2343         spin_unlock(&state->owner->so_lock);
2344
2345         if (!call_close) {
2346                 /* Note: exit _without_ calling nfs4_close_done */
2347                 goto out_no_action;
2348         }
2349
2350         if (calldata->arg.fmode == 0) {
2351                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2352                 if (calldata->roc &&
2353                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2354                         nfs_release_seqid(calldata->arg.seqid);
2355                         goto out_wait;
2356                     }
2357         }
2358
2359         nfs_fattr_init(calldata->res.fattr);
2360         calldata->timestamp = jiffies;
2361         if (nfs4_setup_sequence(NFS_SERVER(inode),
2362                                 &calldata->arg.seq_args,
2363                                 &calldata->res.seq_res,
2364                                 task) != 0)
2365                 nfs_release_seqid(calldata->arg.seqid);
2366         dprintk("%s: done!\n", __func__);
2367         return;
2368 out_no_action:
2369         task->tk_action = NULL;
2370 out_wait:
2371         nfs4_sequence_done(task, &calldata->res.seq_res);
2372 }
2373
2374 static const struct rpc_call_ops nfs4_close_ops = {
2375         .rpc_call_prepare = nfs4_close_prepare,
2376         .rpc_call_done = nfs4_close_done,
2377         .rpc_release = nfs4_free_closedata,
2378 };
2379
2380 /*
2381  * It is possible for data to be read/written from a mem-mapped file
2382  * after the sys_close call (which hits the vfs layer as a flush).
2383  * This means that we can't safely call nfsv4 close on a file until
2384  * the inode is cleared. This in turn means that we are not good
2385  * NFSv4 citizens - we do not indicate to the server to update the file's
2386  * share state even when we are done with one of the three share
2387  * stateid's in the inode.
2388  *
2389  * NOTE: Caller must be holding the sp->so_owner semaphore!
2390  */
2391 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2392 {
2393         struct nfs_server *server = NFS_SERVER(state->inode);
2394         struct nfs4_closedata *calldata;
2395         struct nfs4_state_owner *sp = state->owner;
2396         struct rpc_task *task;
2397         struct rpc_message msg = {
2398                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2399                 .rpc_cred = state->owner->so_cred,
2400         };
2401         struct rpc_task_setup task_setup_data = {
2402                 .rpc_client = server->client,
2403                 .rpc_message = &msg,
2404                 .callback_ops = &nfs4_close_ops,
2405                 .workqueue = nfsiod_workqueue,
2406                 .flags = RPC_TASK_ASYNC,
2407         };
2408         int status = -ENOMEM;
2409
2410         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2411         if (calldata == NULL)
2412                 goto out;
2413         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2414         calldata->inode = state->inode;
2415         calldata->state = state;
2416         calldata->arg.fh = NFS_FH(state->inode);
2417         calldata->arg.stateid = &state->open_stateid;
2418         /* Serialization for the sequence id */
2419         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2420         if (calldata->arg.seqid == NULL)
2421                 goto out_free_calldata;
2422         calldata->arg.fmode = 0;
2423         calldata->arg.bitmask = server->cache_consistency_bitmask;
2424         calldata->res.fattr = &calldata->fattr;
2425         calldata->res.seqid = calldata->arg.seqid;
2426         calldata->res.server = server;
2427         calldata->roc = pnfs_roc(state->inode);
2428         nfs_sb_active(calldata->inode->i_sb);
2429
2430         msg.rpc_argp = &calldata->arg;
2431         msg.rpc_resp = &calldata->res;
2432         task_setup_data.callback_data = calldata;
2433         task = rpc_run_task(&task_setup_data);
2434         if (IS_ERR(task))
2435                 return PTR_ERR(task);
2436         status = 0;
2437         if (wait)
2438                 status = rpc_wait_for_completion_task(task);
2439         rpc_put_task(task);
2440         return status;
2441 out_free_calldata:
2442         kfree(calldata);
2443 out:
2444         nfs4_put_open_state(state);
2445         nfs4_put_state_owner(sp);
2446         return status;
2447 }
2448
2449 static struct inode *
2450 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2451 {
2452         struct nfs4_state *state;
2453
2454         /* Protect against concurrent sillydeletes */
2455         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2456                              ctx->cred, &ctx->mdsthreshold);
2457         if (IS_ERR(state))
2458                 return ERR_CAST(state);
2459         ctx->state = state;
2460         return igrab(state->inode);
2461 }
2462
2463 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2464 {
2465         if (ctx->state == NULL)
2466                 return;
2467         if (is_sync)
2468                 nfs4_close_sync(ctx->state, ctx->mode);
2469         else
2470                 nfs4_close_state(ctx->state, ctx->mode);
2471 }
2472
2473 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2474 {
2475         struct nfs4_server_caps_arg args = {
2476                 .fhandle = fhandle,
2477         };
2478         struct nfs4_server_caps_res res = {};
2479         struct rpc_message msg = {
2480                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2481                 .rpc_argp = &args,
2482                 .rpc_resp = &res,
2483         };
2484         int status;
2485
2486         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2487         if (status == 0) {
2488                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2489                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2490                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2491                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2492                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2493                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2494                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2495                         server->caps |= NFS_CAP_ACLS;
2496                 if (res.has_links != 0)
2497                         server->caps |= NFS_CAP_HARDLINKS;
2498                 if (res.has_symlinks != 0)
2499                         server->caps |= NFS_CAP_SYMLINKS;
2500                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2501                         server->caps |= NFS_CAP_FILEID;
2502                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2503                         server->caps |= NFS_CAP_MODE;
2504                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2505                         server->caps |= NFS_CAP_NLINK;
2506                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2507                         server->caps |= NFS_CAP_OWNER;
2508                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2509                         server->caps |= NFS_CAP_OWNER_GROUP;
2510                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2511                         server->caps |= NFS_CAP_ATIME;
2512                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2513                         server->caps |= NFS_CAP_CTIME;
2514                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2515                         server->caps |= NFS_CAP_MTIME;
2516
2517                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2518                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2519                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2520                 server->acl_bitmask = res.acl_bitmask;
2521                 server->fh_expire_type = res.fh_expire_type;
2522         }
2523
2524         return status;
2525 }
2526
2527 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2528 {
2529         struct nfs4_exception exception = { };
2530         int err;
2531         do {
2532                 err = nfs4_handle_exception(server,
2533                                 _nfs4_server_capabilities(server, fhandle),
2534                                 &exception);
2535         } while (exception.retry);
2536         return err;
2537 }
2538
2539 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2540                 struct nfs_fsinfo *info)
2541 {
2542         struct nfs4_lookup_root_arg args = {
2543                 .bitmask = nfs4_fattr_bitmap,
2544         };
2545         struct nfs4_lookup_res res = {
2546                 .server = server,
2547                 .fattr = info->fattr,
2548                 .fh = fhandle,
2549         };
2550         struct rpc_message msg = {
2551                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2552                 .rpc_argp = &args,
2553                 .rpc_resp = &res,
2554         };
2555
2556         nfs_fattr_init(info->fattr);
2557         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2558 }
2559
2560 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2561                 struct nfs_fsinfo *info)
2562 {
2563         struct nfs4_exception exception = { };
2564         int err;
2565         do {
2566                 err = _nfs4_lookup_root(server, fhandle, info);
2567                 switch (err) {
2568                 case 0:
2569                 case -NFS4ERR_WRONGSEC:
2570                         goto out;
2571                 default:
2572                         err = nfs4_handle_exception(server, err, &exception);
2573                 }
2574         } while (exception.retry);
2575 out:
2576         return err;
2577 }
2578
2579 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2580                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2581 {
2582         struct rpc_auth *auth;
2583         int ret;
2584
2585         auth = rpcauth_create(flavor, server->client);
2586         if (IS_ERR(auth)) {
2587                 ret = -EACCES;
2588                 goto out;
2589         }
2590         ret = nfs4_lookup_root(server, fhandle, info);
2591 out:
2592         return ret;
2593 }
2594
2595 /*
2596  * Retry pseudoroot lookup with various security flavors.  We do this when:
2597  *
2598  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2599  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2600  *
2601  * Returns zero on success, or a negative NFS4ERR value, or a
2602  * negative errno value.
2603  */
2604 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2605                               struct nfs_fsinfo *info)
2606 {
2607         /* Per 3530bis 15.33.5 */
2608         static const rpc_authflavor_t flav_array[] = {
2609                 RPC_AUTH_GSS_KRB5P,
2610                 RPC_AUTH_GSS_KRB5I,
2611                 RPC_AUTH_GSS_KRB5,
2612                 RPC_AUTH_UNIX,                  /* courtesy */
2613                 RPC_AUTH_NULL,
2614         };
2615         int status = -EPERM;
2616         size_t i;
2617
2618         for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
2619                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2620                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2621                         continue;
2622                 break;
2623         }
2624
2625         /*
2626          * -EACCESS could mean that the user doesn't have correct permissions
2627          * to access the mount.  It could also mean that we tried to mount
2628          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2629          * existing mount programs don't handle -EACCES very well so it should
2630          * be mapped to -EPERM instead.
2631          */
2632         if (status == -EACCES)
2633                 status = -EPERM;
2634         return status;
2635 }
2636
2637 static int nfs4_do_find_root_sec(struct nfs_server *server,
2638                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
2639 {
2640         int mv = server->nfs_client->cl_minorversion;
2641         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
2642 }
2643
2644 /**
2645  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
2646  * @server: initialized nfs_server handle
2647  * @fhandle: we fill in the pseudo-fs root file handle
2648  * @info: we fill in an FSINFO struct
2649  *
2650  * Returns zero on success, or a negative errno.
2651  */
2652 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2653                          struct nfs_fsinfo *info)
2654 {
2655         int status;
2656
2657         status = nfs4_lookup_root(server, fhandle, info);
2658         if ((status == -NFS4ERR_WRONGSEC) &&
2659             !(server->flags & NFS_MOUNT_SECFLAVOUR))
2660                 status = nfs4_do_find_root_sec(server, fhandle, info);
2661
2662         if (status == 0)
2663                 status = nfs4_server_capabilities(server, fhandle);
2664         if (status == 0)
2665                 status = nfs4_do_fsinfo(server, fhandle, info);
2666
2667         return nfs4_map_errors(status);
2668 }
2669
2670 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2671                               struct nfs_fsinfo *info)
2672 {
2673         int error;
2674         struct nfs_fattr *fattr = info->fattr;
2675
2676         error = nfs4_server_capabilities(server, mntfh);
2677         if (error < 0) {
2678                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2679                 return error;
2680         }
2681
2682         error = nfs4_proc_getattr(server, mntfh, fattr);
2683         if (error < 0) {
2684                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2685                 return error;
2686         }
2687
2688         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2689             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2690                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2691
2692         return error;
2693 }
2694
2695 /*
2696  * Get locations and (maybe) other attributes of a referral.
2697  * Note that we'll actually follow the referral later when
2698  * we detect fsid mismatch in inode revalidation
2699  */
2700 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2701                              const struct qstr *name, struct nfs_fattr *fattr,
2702                              struct nfs_fh *fhandle)
2703 {
2704         int status = -ENOMEM;
2705         struct page *page = NULL;
2706         struct nfs4_fs_locations *locations = NULL;
2707
2708         page = alloc_page(GFP_KERNEL);
2709         if (page == NULL)
2710                 goto out;
2711         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2712         if (locations == NULL)
2713                 goto out;
2714
2715         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2716         if (status != 0)
2717                 goto out;
2718         /* Make sure server returned a different fsid for the referral */
2719         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2720                 dprintk("%s: server did not return a different fsid for"
2721                         " a referral at %s\n", __func__, name->name);
2722                 status = -EIO;
2723                 goto out;
2724         }
2725         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2726         nfs_fixup_referral_attributes(&locations->fattr);
2727
2728         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2729         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2730         memset(fhandle, 0, sizeof(struct nfs_fh));
2731 out:
2732         if (page)
2733                 __free_page(page);
2734         kfree(locations);
2735         return status;
2736 }
2737
2738 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2739 {
2740         struct nfs4_getattr_arg args = {
2741                 .fh = fhandle,
2742                 .bitmask = server->attr_bitmask,
2743         };
2744         struct nfs4_getattr_res res = {
2745                 .fattr = fattr,
2746                 .server = server,
2747         };
2748         struct rpc_message msg = {
2749                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2750                 .rpc_argp = &args,
2751                 .rpc_resp = &res,
2752         };
2753
2754         nfs_fattr_init(fattr);
2755         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2756 }
2757
2758 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2759 {
2760         struct nfs4_exception exception = { };
2761         int err;
2762         do {
2763                 err = nfs4_handle_exception(server,
2764                                 _nfs4_proc_getattr(server, fhandle, fattr),
2765                                 &exception);
2766         } while (exception.retry);
2767         return err;
2768 }
2769
2770 /*
2771  * The file is not closed if it is opened due to the a request to change
2772  * the size of the file. The open call will not be needed once the
2773  * VFS layer lookup-intents are implemented.
2774  *
2775  * Close is called when the inode is destroyed.
2776  * If we haven't opened the file for O_WRONLY, we
2777  * need to in the size_change case to obtain a stateid.
2778  *
2779  * Got race?
2780  * Because OPEN is always done by name in nfsv4, it is
2781  * possible that we opened a different file by the same
2782  * name.  We can recognize this race condition, but we
2783  * can't do anything about it besides returning an error.
2784  *
2785  * This will be fixed with VFS changes (lookup-intent).
2786  */
2787 static int
2788 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2789                   struct iattr *sattr)
2790 {
2791         struct inode *inode = dentry->d_inode;
2792         struct rpc_cred *cred = NULL;
2793         struct nfs4_state *state = NULL;
2794         int status;
2795
2796         if (pnfs_ld_layoutret_on_setattr(inode))
2797                 pnfs_commit_and_return_layout(inode);
2798
2799         nfs_fattr_init(fattr);
2800
2801         /* Deal with open(O_TRUNC) */
2802         if (sattr->ia_valid & ATTR_OPEN)
2803                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2804
2805         /* Optimization: if the end result is no change, don't RPC */
2806         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2807                 return 0;
2808
2809         /* Search for an existing open(O_WRITE) file */
2810         if (sattr->ia_valid & ATTR_FILE) {
2811                 struct nfs_open_context *ctx;
2812
2813                 ctx = nfs_file_open_context(sattr->ia_file);
2814                 if (ctx) {
2815                         cred = ctx->cred;
2816                         state = ctx->state;
2817                 }
2818         }
2819
2820         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2821         if (status == 0)
2822                 nfs_setattr_update_inode(inode, sattr);
2823         return status;
2824 }
2825
2826 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2827                 const struct qstr *name, struct nfs_fh *fhandle,
2828                 struct nfs_fattr *fattr)
2829 {
2830         struct nfs_server *server = NFS_SERVER(dir);
2831         int                    status;
2832         struct nfs4_lookup_arg args = {
2833                 .bitmask = server->attr_bitmask,
2834                 .dir_fh = NFS_FH(dir),
2835                 .name = name,
2836         };
2837         struct nfs4_lookup_res res = {
2838                 .server = server,
2839                 .fattr = fattr,
2840                 .fh = fhandle,
2841         };
2842         struct rpc_message msg = {
2843                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2844                 .rpc_argp = &args,
2845                 .rpc_resp = &res,
2846         };
2847
2848         nfs_fattr_init(fattr);
2849
2850         dprintk("NFS call  lookup %s\n", name->name);
2851         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2852         dprintk("NFS reply lookup: %d\n", status);
2853         return status;
2854 }
2855
2856 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2857 {
2858         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2859                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2860         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2861         fattr->nlink = 2;
2862 }
2863
2864 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2865                                    struct qstr *name, struct nfs_fh *fhandle,
2866                                    struct nfs_fattr *fattr)
2867 {
2868         struct nfs4_exception exception = { };
2869         struct rpc_clnt *client = *clnt;
2870         int err;
2871         do {
2872                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2873                 switch (err) {
2874                 case -NFS4ERR_BADNAME:
2875                         err = -ENOENT;
2876                         goto out;
2877                 case -NFS4ERR_MOVED:
2878                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2879                         goto out;
2880                 case -NFS4ERR_WRONGSEC:
2881                         err = -EPERM;
2882                         if (client != *clnt)
2883                                 goto out;
2884
2885                         client = nfs4_create_sec_client(client, dir, name);
2886                         if (IS_ERR(client))
2887                                 return PTR_ERR(client);
2888
2889                         exception.retry = 1;
2890                         break;
2891                 default:
2892                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2893                 }
2894         } while (exception.retry);
2895
2896 out:
2897         if (err == 0)
2898                 *clnt = client;
2899         else if (client != *clnt)
2900                 rpc_shutdown_client(client);
2901
2902         return err;
2903 }
2904
2905 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2906                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2907 {
2908         int status;
2909         struct rpc_clnt *client = NFS_CLIENT(dir);
2910
2911         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2912         if (client != NFS_CLIENT(dir)) {
2913                 rpc_shutdown_client(client);
2914                 nfs_fixup_secinfo_attributes(fattr);
2915         }
2916         return status;
2917 }
2918
2919 struct rpc_clnt *
2920 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2921                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2922 {
2923         int status;
2924         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2925
2926         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2927         if (status < 0) {
2928                 rpc_shutdown_client(client);
2929                 return ERR_PTR(status);
2930         }
2931         return client;
2932 }
2933
2934 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2935 {
2936         struct nfs_server *server = NFS_SERVER(inode);
2937         struct nfs4_accessargs args = {
2938                 .fh = NFS_FH(inode),
2939                 .bitmask = server->cache_consistency_bitmask,
2940         };
2941         struct nfs4_accessres res = {
2942                 .server = server,
2943         };
2944         struct rpc_message msg = {
2945                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2946                 .rpc_argp = &args,
2947                 .rpc_resp = &res,
2948                 .rpc_cred = entry->cred,
2949         };
2950         int mode = entry->mask;
2951         int status;
2952
2953         /*
2954          * Determine which access bits we want to ask for...
2955          */
2956         if (mode & MAY_READ)
2957                 args.access |= NFS4_ACCESS_READ;
2958         if (S_ISDIR(inode->i_mode)) {
2959                 if (mode & MAY_WRITE)
2960                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2961                 if (mode & MAY_EXEC)
2962                         args.access |= NFS4_ACCESS_LOOKUP;
2963         } else {
2964                 if (mode & MAY_WRITE)
2965                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2966                 if (mode & MAY_EXEC)
2967                         args.access |= NFS4_ACCESS_EXECUTE;
2968         }
2969
2970         res.fattr = nfs_alloc_fattr();
2971         if (res.fattr == NULL)
2972                 return -ENOMEM;
2973
2974         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2975         if (!status) {
2976                 nfs_access_set_mask(entry, res.access);
2977                 nfs_refresh_inode(inode, res.fattr);
2978         }
2979         nfs_free_fattr(res.fattr);
2980         return status;
2981 }
2982
2983 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2984 {
2985         struct nfs4_exception exception = { };
2986         int err;
2987         do {
2988                 err = nfs4_handle_exception(NFS_SERVER(inode),
2989                                 _nfs4_proc_access(inode, entry),
2990                                 &exception);
2991         } while (exception.retry);
2992         return err;
2993 }
2994
2995 /*
2996  * TODO: For the time being, we don't try to get any attributes
2997  * along with any of the zero-copy operations READ, READDIR,
2998  * READLINK, WRITE.
2999  *
3000  * In the case of the first three, we want to put the GETATTR
3001  * after the read-type operation -- this is because it is hard
3002  * to predict the length of a GETATTR response in v4, and thus
3003  * align the READ data correctly.  This means that the GETATTR
3004  * may end up partially falling into the page cache, and we should
3005  * shift it into the 'tail' of the xdr_buf before processing.
3006  * To do this efficiently, we need to know the total length
3007  * of data received, which doesn't seem to be available outside
3008  * of the RPC layer.
3009  *
3010  * In the case of WRITE, we also want to put the GETATTR after
3011  * the operation -- in this case because we want to make sure
3012  * we get the post-operation mtime and size.
3013  *
3014  * Both of these changes to the XDR layer would in fact be quite
3015  * minor, but I decided to leave them for a subsequent patch.
3016  */
3017 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3018                 unsigned int pgbase, unsigned int pglen)
3019 {
3020         struct nfs4_readlink args = {
3021                 .fh       = NFS_FH(inode),
3022                 .pgbase   = pgbase,
3023                 .pglen    = pglen,
3024                 .pages    = &page,
3025         };
3026         struct nfs4_readlink_res res;
3027         struct rpc_message msg = {
3028                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3029                 .rpc_argp = &args,
3030                 .rpc_resp = &res,
3031         };
3032
3033         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3034 }
3035
3036 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3037                 unsigned int pgbase, unsigned int pglen)
3038 {
3039         struct nfs4_exception exception = { };
3040         int err;
3041         do {
3042                 err = nfs4_handle_exception(NFS_SERVER(inode),
3043                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
3044                                 &exception);
3045         } while (exception.retry);
3046         return err;
3047 }
3048
3049 /*
3050  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3051  */
3052 static int
3053 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3054                  int flags)
3055 {
3056         struct nfs_open_context *ctx;
3057         struct nfs4_state *state;
3058         int status = 0;
3059
3060         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3061         if (IS_ERR(ctx))
3062                 return PTR_ERR(ctx);
3063
3064         sattr->ia_mode &= ~current_umask();
3065         state = nfs4_do_open(dir, dentry, ctx->mode,
3066                         flags, sattr, ctx->cred,
3067                         &ctx->mdsthreshold);
3068         d_drop(dentry);
3069         if (IS_ERR(state)) {
3070                 status = PTR_ERR(state);
3071                 goto out;
3072         }
3073         d_add(dentry, igrab(state->inode));
3074         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
3075         ctx->state = state;
3076 out:
3077         put_nfs_open_context(ctx);
3078         return status;
3079 }
3080
3081 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3082 {
3083         struct nfs_server *server = NFS_SERVER(dir);
3084         struct nfs_removeargs args = {
3085                 .fh = NFS_FH(dir),
3086                 .name = *name,
3087         };
3088         struct nfs_removeres res = {
3089                 .server = server,
3090         };
3091         struct rpc_message msg = {
3092                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3093                 .rpc_argp = &args,
3094                 .rpc_resp = &res,
3095         };
3096         int status;
3097
3098         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3099         if (status == 0)
3100                 update_changeattr(dir, &res.cinfo);
3101         return status;
3102 }
3103
3104 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3105 {
3106         struct nfs4_exception exception = { };
3107         int err;
3108         do {
3109                 err = nfs4_handle_exception(NFS_SERVER(dir),
3110                                 _nfs4_proc_remove(dir, name),
3111                                 &exception);
3112         } while (exception.retry);
3113         return err;
3114 }
3115
3116 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3117 {
3118         struct nfs_server *server = NFS_SERVER(dir);
3119         struct nfs_removeargs *args = msg->rpc_argp;
3120         struct nfs_removeres *res = msg->rpc_resp;
3121
3122         res->server = server;
3123         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3124         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3125 }
3126
3127 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3128 {
3129         nfs4_setup_sequence(NFS_SERVER(data->dir),
3130                         &data->args.seq_args,
3131                         &data->res.seq_res,
3132                         task);
3133 }
3134
3135 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3136 {
3137         struct nfs_removeres *res = task->tk_msg.rpc_resp;
3138
3139         if (!nfs4_sequence_done(task, &res->seq_res))
3140                 return 0;
3141         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3142                 return 0;
3143         update_changeattr(dir, &res->cinfo);
3144         return 1;
3145 }
3146
3147 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3148 {
3149         struct nfs_server *server = NFS_SERVER(dir);
3150         struct nfs_renameargs *arg = msg->rpc_argp;
3151         struct nfs_renameres *res = msg->rpc_resp;
3152
3153         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3154         res->server = server;
3155         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3156 }
3157
3158 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3159 {
3160         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3161                         &data->args.seq_args,
3162                         &data->res.seq_res,
3163                         task);
3164 }
3165
3166 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3167                                  struct inode *new_dir)
3168 {
3169         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3170
3171         if (!nfs4_sequence_done(task, &res->seq_res))
3172                 return 0;
3173         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3174                 return 0;
3175
3176         update_changeattr(old_dir, &res->old_cinfo);
3177         update_changeattr(new_dir, &res->new_cinfo);
3178         return 1;
3179 }
3180
3181 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3182                 struct inode *new_dir, struct qstr *new_name)
3183 {
3184         struct nfs_server *server = NFS_SERVER(old_dir);
3185         struct nfs_renameargs arg = {
3186                 .old_dir = NFS_FH(old_dir),
3187                 .new_dir = NFS_FH(new_dir),
3188                 .old_name = old_name,
3189                 .new_name = new_name,
3190         };
3191         struct nfs_renameres res = {
3192                 .server = server,
3193         };
3194         struct rpc_message msg = {
3195                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3196                 .rpc_argp = &arg,
3197                 .rpc_resp = &res,
3198         };
3199         int status = -ENOMEM;
3200
3201         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3202         if (!status) {
3203                 update_changeattr(old_dir, &res.old_cinfo);
3204                 update_changeattr(new_dir, &res.new_cinfo);
3205         }
3206         return status;
3207 }
3208
3209 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3210                 struct inode *new_dir, struct qstr *new_name)
3211 {
3212         struct nfs4_exception exception = { };
3213         int err;
3214         do {
3215                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3216                                 _nfs4_proc_rename(old_dir, old_name,
3217                                         new_dir, new_name),
3218                                 &exception);
3219         } while (exception.retry);
3220         return err;
3221 }
3222
3223 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3224 {
3225         struct nfs_server *server = NFS_SERVER(inode);
3226         struct nfs4_link_arg arg = {
3227                 .fh     = NFS_FH(inode),
3228                 .dir_fh = NFS_FH(dir),
3229                 .name   = name,
3230                 .bitmask = server->attr_bitmask,
3231         };
3232         struct nfs4_link_res res = {
3233                 .server = server,
3234         };
3235         struct rpc_message msg = {
3236                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3237                 .rpc_argp = &arg,
3238                 .rpc_resp = &res,
3239         };
3240         int status = -ENOMEM;
3241
3242         res.fattr = nfs_alloc_fattr();
3243         if (res.fattr == NULL)
3244                 goto out;
3245
3246         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3247         if (!status) {
3248                 update_changeattr(dir, &res.cinfo);
3249                 nfs_post_op_update_inode(inode, res.fattr);
3250         }
3251 out:
3252         nfs_free_fattr(res.fattr);
3253         return status;
3254 }
3255
3256 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3257 {
3258         struct nfs4_exception exception = { };
3259         int err;
3260         do {
3261                 err = nfs4_handle_exception(NFS_SERVER(inode),
3262                                 _nfs4_proc_link(inode, dir, name),
3263                                 &exception);
3264         } while (exception.retry);
3265         return err;
3266 }
3267
3268 struct nfs4_createdata {
3269         struct rpc_message msg;
3270         struct nfs4_create_arg arg;
3271         struct nfs4_create_res res;
3272         struct nfs_fh fh;
3273         struct nfs_fattr fattr;
3274 };
3275
3276 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3277                 struct qstr *name, struct iattr *sattr, u32 ftype)
3278 {
3279         struct nfs4_createdata *data;
3280
3281         data = kzalloc(sizeof(*data), GFP_KERNEL);
3282         if (data != NULL) {
3283                 struct nfs_server *server = NFS_SERVER(dir);
3284
3285                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3286                 data->msg.rpc_argp = &data->arg;
3287                 data->msg.rpc_resp = &data->res;
3288                 data->arg.dir_fh = NFS_FH(dir);
3289                 data->arg.server = server;
3290                 data->arg.name = name;
3291                 data->arg.attrs = sattr;
3292                 data->arg.ftype = ftype;
3293                 data->arg.bitmask = server->attr_bitmask;
3294                 data->res.server = server;
3295                 data->res.fh = &data->fh;
3296                 data->res.fattr = &data->fattr;
3297                 nfs_fattr_init(data->res.fattr);
3298         }
3299         return data;
3300 }
3301
3302 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3303 {
3304         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3305                                     &data->arg.seq_args, &data->res.seq_res, 1);
3306         if (status == 0) {
3307                 update_changeattr(dir, &data->res.dir_cinfo);
3308                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3309         }
3310         return status;
3311 }
3312
3313 static void nfs4_free_createdata(struct nfs4_createdata *data)
3314 {
3315         kfree(data);
3316 }
3317
3318 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3319                 struct page *page, unsigned int len, struct iattr *sattr)
3320 {
3321         struct nfs4_createdata *data;
3322         int status = -ENAMETOOLONG;
3323
3324         if (len > NFS4_MAXPATHLEN)
3325                 goto out;
3326
3327         status = -ENOMEM;
3328         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3329         if (data == NULL)
3330                 goto out;
3331
3332         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3333         data->arg.u.symlink.pages = &page;
3334         data->arg.u.symlink.len = len;
3335
3336         status = nfs4_do_create(dir, dentry, data);
3337
3338         nfs4_free_createdata(data);
3339 out:
3340         return status;
3341 }
3342
3343 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3344                 struct page *page, unsigned int len, struct iattr *sattr)
3345 {
3346         struct nfs4_exception exception = { };
3347         int err;
3348         do {
3349                 err = nfs4_handle_exception(NFS_SERVER(dir),
3350                                 _nfs4_proc_symlink(dir, dentry, page,
3351                                                         len, sattr),
3352                                 &exception);
3353         } while (exception.retry);
3354         return err;
3355 }
3356
3357 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3358                 struct iattr *sattr)
3359 {
3360         struct nfs4_createdata *data;
3361         int status = -ENOMEM;
3362
3363         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3364         if (data == NULL)
3365                 goto out;
3366
3367         status = nfs4_do_create(dir, dentry, data);
3368
3369         nfs4_free_createdata(data);
3370 out:
3371         return status;
3372 }
3373
3374 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3375                 struct iattr *sattr)
3376 {
3377         struct nfs4_exception exception = { };
3378         int err;
3379
3380         sattr->ia_mode &= ~current_umask();
3381         do {
3382                 err = nfs4_handle_exception(NFS_SERVER(dir),
3383                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3384                                 &exception);
3385         } while (exception.retry);
3386         return err;
3387 }
3388
3389 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3390                 u64 cookie, struct page **pages, unsigned int count, int plus)
3391 {
3392         struct inode            *dir = dentry->d_inode;
3393         struct nfs4_readdir_arg args = {
3394                 .fh = NFS_FH(dir),
3395                 .pages = pages,
3396                 .pgbase = 0,
3397                 .count = count,
3398                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3399                 .plus = plus,
3400         };
3401         struct nfs4_readdir_res res;
3402         struct rpc_message msg = {
3403                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3404                 .rpc_argp = &args,
3405                 .rpc_resp = &res,
3406                 .rpc_cred = cred,
3407         };
3408         int                     status;
3409
3410         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3411                         dentry->d_parent->d_name.name,
3412                         dentry->d_name.name,
3413                         (unsigned long long)cookie);
3414         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3415         res.pgbase = args.pgbase;
3416         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3417         if (status >= 0) {
3418                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3419                 status += args.pgbase;
3420         }
3421
3422         nfs_invalidate_atime(dir);
3423
3424         dprintk("%s: returns %d\n", __func__, status);
3425         return status;
3426 }
3427
3428 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3429                 u64 cookie, struct page **pages, unsigned int count, int plus)
3430 {
3431         struct nfs4_exception exception = { };
3432         int err;
3433         do {
3434                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3435                                 _nfs4_proc_readdir(dentry, cred, cookie,
3436                                         pages, count, plus),
3437                                 &exception);
3438         } while (exception.retry);
3439         return err;
3440 }
3441
3442 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3443                 struct iattr *sattr, dev_t rdev)
3444 {
3445         struct nfs4_createdata *data;
3446         int mode = sattr->ia_mode;
3447         int status = -ENOMEM;
3448
3449         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3450         if (data == NULL)
3451                 goto out;
3452
3453         if (S_ISFIFO(mode))
3454                 data->arg.ftype = NF4FIFO;
3455         else if (S_ISBLK(mode)) {
3456                 data->arg.ftype = NF4BLK;
3457                 data->arg.u.device.specdata1 = MAJOR(rdev);
3458                 data->arg.u.device.specdata2 = MINOR(rdev);
3459         }
3460         else if (S_ISCHR(mode)) {
3461                 data->arg.ftype = NF4CHR;
3462                 data->arg.u.device.specdata1 = MAJOR(rdev);
3463                 data->arg.u.device.specdata2 = MINOR(rdev);
3464         } else if (!S_ISSOCK(mode)) {
3465                 status = -EINVAL;
3466                 goto out_free;
3467         }
3468
3469         status = nfs4_do_create(dir, dentry, data);
3470 out_free:
3471         nfs4_free_createdata(data);
3472 out:
3473         return status;
3474 }
3475
3476 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3477                 struct iattr *sattr, dev_t rdev)
3478 {
3479         struct nfs4_exception exception = { };
3480         int err;
3481
3482         sattr->ia_mode &= ~current_umask();
3483         do {
3484                 err = nfs4_handle_exception(NFS_SERVER(dir),
3485                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3486                                 &exception);
3487         } while (exception.retry);
3488         return err;
3489 }
3490
3491 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3492                  struct nfs_fsstat *fsstat)
3493 {
3494         struct nfs4_statfs_arg args = {
3495                 .fh = fhandle,
3496                 .bitmask = server->attr_bitmask,
3497         };
3498         struct nfs4_statfs_res res = {
3499                 .fsstat = fsstat,
3500         };
3501         struct rpc_message msg = {
3502                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3503                 .rpc_argp = &args,
3504                 .rpc_resp = &res,
3505         };
3506
3507         nfs_fattr_init(fsstat->fattr);
3508         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3509 }
3510
3511 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3512 {
3513         struct nfs4_exception exception = { };
3514         int err;
3515         do {
3516                 err = nfs4_handle_exception(server,
3517                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3518                                 &exception);
3519         } while (exception.retry);
3520         return err;
3521 }
3522
3523 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3524                 struct nfs_fsinfo *fsinfo)
3525 {
3526         struct nfs4_fsinfo_arg args = {
3527                 .fh = fhandle,
3528                 .bitmask = server->attr_bitmask,
3529         };
3530         struct nfs4_fsinfo_res res = {
3531                 .fsinfo = fsinfo,
3532         };
3533         struct rpc_message msg = {
3534                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3535                 .rpc_argp = &args,
3536                 .rpc_resp = &res,
3537         };
3538
3539         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3540 }
3541
3542 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3543 {
3544         struct nfs4_exception exception = { };
3545         unsigned long now = jiffies;
3546         int err;
3547
3548         do {
3549                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
3550                 if (err == 0) {
3551                         struct nfs_client *clp = server->nfs_client;
3552
3553                         spin_lock(&clp->cl_lock);
3554                         clp->cl_lease_time = fsinfo->lease_time * HZ;
3555                         clp->cl_last_renewal = now;
3556                         spin_unlock(&clp->cl_lock);
3557                         break;
3558                 }
3559                 err = nfs4_handle_exception(server, err, &exception);
3560         } while (exception.retry);
3561         return err;
3562 }
3563
3564 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3565 {
3566         int error;
3567
3568         nfs_fattr_init(fsinfo->fattr);
3569         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3570         if (error == 0) {
3571                 /* block layout checks this! */
3572                 server->pnfs_blksize = fsinfo->blksize;
3573                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3574         }
3575
3576         return error;
3577 }
3578
3579 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3580                 struct nfs_pathconf *pathconf)
3581 {
3582         struct nfs4_pathconf_arg args = {
3583                 .fh = fhandle,
3584                 .bitmask = server->attr_bitmask,
3585         };
3586         struct nfs4_pathconf_res res = {
3587                 .pathconf = pathconf,
3588         };
3589         struct rpc_message msg = {
3590                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3591                 .rpc_argp = &args,
3592                 .rpc_resp = &res,
3593         };
3594
3595         /* None of the pathconf attributes are mandatory to implement */
3596         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3597                 memset(pathconf, 0, sizeof(*pathconf));
3598                 return 0;
3599         }
3600
3601         nfs_fattr_init(pathconf->fattr);
3602         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3603 }
3604
3605 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3606                 struct nfs_pathconf *pathconf)
3607 {
3608         struct nfs4_exception exception = { };
3609         int err;
3610
3611         do {
3612                 err = nfs4_handle_exception(server,
3613                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3614                                 &exception);
3615         } while (exception.retry);
3616         return err;
3617 }
3618
3619 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
3620                 const struct nfs_open_context *ctx,
3621                 const struct nfs_lock_context *l_ctx,
3622                 fmode_t fmode)
3623 {
3624         const struct nfs_lockowner *lockowner = NULL;
3625
3626         if (l_ctx != NULL)
3627                 lockowner = &l_ctx->lockowner;
3628         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
3629 }
3630 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
3631
3632 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
3633                 const struct nfs_open_context *ctx,
3634                 const struct nfs_lock_context *l_ctx,
3635                 fmode_t fmode)
3636 {
3637         nfs4_stateid current_stateid;
3638
3639         /* If the current stateid represents a lost lock, then exit */
3640         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
3641                 return true;
3642         return nfs4_stateid_match(stateid, &current_stateid);
3643 }
3644
3645 static bool nfs4_error_stateid_expired(int err)
3646 {
3647         switch (err) {
3648         case -NFS4ERR_DELEG_REVOKED:
3649         case -NFS4ERR_ADMIN_REVOKED:
3650         case -NFS4ERR_BAD_STATEID:
3651         case -NFS4ERR_STALE_STATEID:
3652         case -NFS4ERR_OLD_STATEID:
3653         case -NFS4ERR_OPENMODE:
3654         case -NFS4ERR_EXPIRED:
3655                 return true;
3656         }
3657         return false;
3658 }
3659
3660 void __nfs4_read_done_cb(struct nfs_read_data *data)
3661 {
3662         nfs_invalidate_atime(data->header->inode);
3663 }
3664
3665 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3666 {
3667         struct nfs_server *server = NFS_SERVER(data->header->inode);
3668
3669         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3670                 rpc_restart_call_prepare(task);
3671                 return -EAGAIN;
3672         }
3673
3674         __nfs4_read_done_cb(data);
3675         if (task->tk_status > 0)
3676                 renew_lease(server, data->timestamp);
3677         return 0;
3678 }
3679
3680 static bool nfs4_read_stateid_changed(struct rpc_task *task,
3681                 struct nfs_readargs *args)
3682 {
3683
3684         if (!nfs4_error_stateid_expired(task->tk_status) ||
3685                 nfs4_stateid_is_current(&args->stateid,
3686                                 args->context,
3687                                 args->lock_context,
3688                                 FMODE_READ))
3689                 return false;
3690         rpc_restart_call_prepare(task);
3691         return true;
3692 }
3693
3694 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3695 {
3696
3697         dprintk("--> %s\n", __func__);
3698
3699         if (!nfs4_sequence_done(task, &data->res.seq_res))
3700                 return -EAGAIN;
3701         if (nfs4_read_stateid_changed(task, &data->args))
3702                 return -EAGAIN;
3703         return data->read_done_cb ? data->read_done_cb(task, data) :
3704                                     nfs4_read_done_cb(task, data);
3705 }
3706
3707 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3708 {
3709         data->timestamp   = jiffies;
3710         data->read_done_cb = nfs4_read_done_cb;
3711         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3712         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3713 }
3714
3715 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3716 {
3717         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3718                         &data->args.seq_args,
3719                         &data->res.seq_res,
3720                         task))
3721                 return;
3722         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3723                         data->args.lock_context, FMODE_READ);
3724 }
3725
3726 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3727 {
3728         struct inode *inode = data->header->inode;
3729
3730         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3731                 rpc_restart_call_prepare(task);
3732                 return -EAGAIN;
3733         }
3734         if (task->tk_status >= 0) {
3735                 renew_lease(NFS_SERVER(inode), data->timestamp);
3736                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3737         }
3738         return 0;
3739 }
3740
3741 static bool nfs4_write_stateid_changed(struct rpc_task *task,
3742                 struct nfs_writeargs *args)
3743 {
3744
3745         if (!nfs4_error_stateid_expired(task->tk_status) ||
3746                 nfs4_stateid_is_current(&args->stateid,
3747                                 args->context,
3748                                 args->lock_context,
3749                                 FMODE_WRITE))
3750                 return false;
3751         rpc_restart_call_prepare(task);
3752         return true;
3753 }
3754
3755 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3756 {
3757         if (!nfs4_sequence_done(task, &data->res.seq_res))
3758                 return -EAGAIN;
3759         if (nfs4_write_stateid_changed(task, &data->args))
3760                 return -EAGAIN;
3761         return data->write_done_cb ? data->write_done_cb(task, data) :
3762                 nfs4_write_done_cb(task, data);
3763 }
3764
3765 static
3766 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3767 {
3768         const struct nfs_pgio_header *hdr = data->header;
3769
3770         /* Don't request attributes for pNFS or O_DIRECT writes */
3771         if (data->ds_clp != NULL || hdr->dreq != NULL)
3772                 return false;
3773         /* Otherwise, request attributes if and only if we don't hold
3774          * a delegation
3775          */
3776         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3777 }
3778
3779 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3780 {
3781         struct nfs_server *server = NFS_SERVER(data->header->inode);
3782
3783         if (!nfs4_write_need_cache_consistency_data(data)) {
3784                 data->args.bitmask = NULL;
3785                 data->res.fattr = NULL;
3786         } else
3787                 data->args.bitmask = server->cache_consistency_bitmask;
3788
3789         if (!data->write_done_cb)
3790                 data->write_done_cb = nfs4_write_done_cb;
3791         data->res.server = server;
3792         data->timestamp   = jiffies;
3793
3794         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3795         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3796 }
3797
3798 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3799 {
3800         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3801                         &data->args.seq_args,
3802                         &data->res.seq_res,
3803                         task))
3804                 return;
3805         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3806                         data->args.lock_context, FMODE_WRITE);
3807 }
3808
3809 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3810 {
3811         nfs4_setup_sequence(NFS_SERVER(data->inode),
3812                         &data->args.seq_args,
3813                         &data->res.seq_res,
3814                         task);
3815 }
3816
3817 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3818 {
3819         struct inode *inode = data->inode;
3820
3821         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3822                 rpc_restart_call_prepare(task);
3823                 return -EAGAIN;
3824         }
3825         return 0;
3826 }
3827
3828 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3829 {
3830         if (!nfs4_sequence_done(task, &data->res.seq_res))
3831                 return -EAGAIN;
3832         return data->commit_done_cb(task, data);
3833 }
3834
3835 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3836 {
3837         struct nfs_server *server = NFS_SERVER(data->inode);
3838
3839         if (data->commit_done_cb == NULL)
3840                 data->commit_done_cb = nfs4_commit_done_cb;
3841         data->res.server = server;
3842         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3843         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3844 }
3845
3846 struct nfs4_renewdata {
3847         struct nfs_client       *client;
3848         unsigned long           timestamp;
3849 };
3850
3851 /*
3852  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3853  * standalone procedure for queueing an asynchronous RENEW.
3854  */
3855 static void nfs4_renew_release(void *calldata)
3856 {
3857         struct nfs4_renewdata *data = calldata;
3858         struct nfs_client *clp = data->client;
3859
3860         if (atomic_read(&clp->cl_count) > 1)
3861                 nfs4_schedule_state_renewal(clp);
3862         nfs_put_client(clp);
3863         kfree(data);
3864 }
3865
3866 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3867 {
3868         struct nfs4_renewdata *data = calldata;
3869         struct nfs_client *clp = data->client;
3870         unsigned long timestamp = data->timestamp;
3871
3872         if (task->tk_status < 0) {
3873                 /* Unless we're shutting down, schedule state recovery! */
3874                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3875                         return;
3876                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3877                         nfs4_schedule_lease_recovery(clp);
3878                         return;
3879                 }
3880                 nfs4_schedule_path_down_recovery(clp);
3881         }
3882         do_renew_lease(clp, timestamp);
3883 }
3884
3885 static const struct rpc_call_ops nfs4_renew_ops = {
3886         .rpc_call_done = nfs4_renew_done,
3887         .rpc_release = nfs4_renew_release,
3888 };
3889
3890 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3891 {
3892         struct rpc_message msg = {
3893                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3894                 .rpc_argp       = clp,
3895                 .rpc_cred       = cred,
3896         };
3897         struct nfs4_renewdata *data;
3898
3899         if (renew_flags == 0)
3900                 return 0;
3901         if (!atomic_inc_not_zero(&clp->cl_count))
3902                 return -EIO;
3903         data = kmalloc(sizeof(*data), GFP_NOFS);
3904         if (data == NULL)
3905                 return -ENOMEM;
3906         data->client = clp;
3907         data->timestamp = jiffies;
3908         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
3909                         &nfs4_renew_ops, data);
3910 }
3911
3912 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3913 {
3914         struct rpc_message msg = {
3915                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3916                 .rpc_argp       = clp,
3917                 .rpc_cred       = cred,
3918         };
3919         unsigned long now = jiffies;
3920         int status;
3921
3922         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3923         if (status < 0)
3924                 return status;
3925         do_renew_lease(clp, now);
3926         return 0;
3927 }
3928
3929 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3930 {
3931         return (server->caps & NFS_CAP_ACLS)
3932                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3933                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3934 }
3935
3936 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3937  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3938  * the stack.
3939  */
3940 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3941
3942 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3943                 struct page **pages, unsigned int *pgbase)
3944 {
3945         struct page *newpage, **spages;
3946         int rc = 0;
3947         size_t len;
3948         spages = pages;
3949
3950         do {
3951                 len = min_t(size_t, PAGE_SIZE, buflen);
3952                 newpage = alloc_page(GFP_KERNEL);
3953
3954                 if (newpage == NULL)
3955                         goto unwind;
3956                 memcpy(page_address(newpage), buf, len);
3957                 buf += len;
3958                 buflen -= len;
3959                 *pages++ = newpage;
3960                 rc++;
3961         } while (buflen != 0);
3962
3963         return rc;
3964
3965 unwind:
3966         for(; rc > 0; rc--)
3967                 __free_page(spages[rc-1]);
3968         return -ENOMEM;
3969 }
3970
3971 struct nfs4_cached_acl {
3972         int cached;
3973         size_t len;
3974         char data[0];
3975 };
3976
3977 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3978 {
3979         struct nfs_inode *nfsi = NFS_I(inode);
3980
3981         spin_lock(&inode->i_lock);
3982         kfree(nfsi->nfs4_acl);
3983         nfsi->nfs4_acl = acl;
3984         spin_unlock(&inode->i_lock);
3985 }
3986
3987 static void nfs4_zap_acl_attr(struct inode *inode)
3988 {
3989         nfs4_set_cached_acl(inode, NULL);
3990 }
3991
3992 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3993 {
3994         struct nfs_inode *nfsi = NFS_I(inode);
3995         struct nfs4_cached_acl *acl;
3996         int ret = -ENOENT;
3997
3998         spin_lock(&inode->i_lock);
3999         acl = nfsi->nfs4_acl;
4000         if (acl == NULL)
4001                 goto out;
4002         if (buf == NULL) /* user is just asking for length */
4003                 goto out_len;
4004         if (acl->cached == 0)
4005                 goto out;
4006         ret = -ERANGE; /* see getxattr(2) man page */
4007         if (acl->len > buflen)
4008                 goto out;
4009         memcpy(buf, acl->data, acl->len);
4010 out_len:
4011         ret = acl->len;
4012 out:
4013         spin_unlock(&inode->i_lock);
4014         return ret;
4015 }
4016
4017 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4018 {
4019         struct nfs4_cached_acl *acl;
4020         size_t buflen = sizeof(*acl) + acl_len;
4021
4022         if (buflen <= PAGE_SIZE) {
4023                 acl = kmalloc(buflen, GFP_KERNEL);
4024                 if (acl == NULL)
4025                         goto out;
4026                 acl->cached = 1;
4027                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4028         } else {
4029                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4030                 if (acl == NULL)
4031                         goto out;
4032                 acl->cached = 0;
4033         }
4034         acl->len = acl_len;
4035 out:
4036         nfs4_set_cached_acl(inode, acl);
4037 }
4038
4039 /*
4040  * The getxattr API returns the required buffer length when called with a
4041  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4042  * the required buf.  On a NULL buf, we send a page of data to the server
4043  * guessing that the ACL request can be serviced by a page. If so, we cache
4044  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4045  * the cache. If not so, we throw away the page, and cache the required
4046  * length. The next getxattr call will then produce another round trip to
4047  * the server, this time with the input buf of the required size.
4048  */
4049 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4050 {
4051         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4052         struct nfs_getaclargs args = {
4053                 .fh = NFS_FH(inode),
4054                 .acl_pages = pages,
4055                 .acl_len = buflen,
4056         };
4057         struct nfs_getaclres res = {
4058                 .acl_len = buflen,
4059         };
4060         struct rpc_message msg = {
4061                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4062                 .rpc_argp = &args,
4063                 .rpc_resp = &res,
4064         };
4065         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4066         int ret = -ENOMEM, i;
4067
4068         /* As long as we're doing a round trip to the server anyway,
4069          * let's be prepared for a page of acl data. */
4070         if (npages == 0)
4071                 npages = 1;
4072         if (npages > ARRAY_SIZE(pages))
4073                 return -ERANGE;
4074
4075         for (i = 0; i < npages; i++) {
4076                 pages[i] = alloc_page(GFP_KERNEL);
4077                 if (!pages[i])
4078                         goto out_free;
4079         }
4080
4081         /* for decoding across pages */
4082         res.acl_scratch = alloc_page(GFP_KERNEL);
4083         if (!res.acl_scratch)
4084                 goto out_free;
4085
4086         args.acl_len = npages * PAGE_SIZE;
4087         args.acl_pgbase = 0;
4088
4089         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4090                 __func__, buf, buflen, npages, args.acl_len);
4091         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4092                              &msg, &args.seq_args, &res.seq_res, 0);
4093         if (ret)
4094                 goto out_free;
4095
4096         /* Handle the case where the passed-in buffer is too short */
4097         if (res.acl_flags & NFS4_ACL_TRUNC) {
4098                 /* Did the user only issue a request for the acl length? */
4099                 if (buf == NULL)
4100                         goto out_ok;
4101                 ret = -ERANGE;
4102                 goto out_free;
4103         }
4104         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4105         if (buf) {
4106                 if (res.acl_len > buflen) {
4107                         ret = -ERANGE;
4108                         goto out_free;
4109                 }
4110                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4111         }
4112 out_ok:
4113         ret = res.acl_len;
4114 out_free:
4115         for (i = 0; i < npages; i++)
4116                 if (pages[i])
4117                         __free_page(pages[i]);
4118         if (res.acl_scratch)
4119                 __free_page(res.acl_scratch);
4120         return ret;
4121 }
4122
4123 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4124 {
4125         struct nfs4_exception exception = { };
4126         ssize_t ret;
4127         do {
4128                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4129                 if (ret >= 0)
4130                         break;
4131                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4132         } while (exception.retry);
4133         return ret;
4134 }
4135
4136 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4137 {
4138         struct nfs_server *server = NFS_SERVER(inode);
4139         int ret;
4140
4141         if (!nfs4_server_supports_acls(server))
4142                 return -EOPNOTSUPP;
4143         ret = nfs_revalidate_inode(server, inode);
4144         if (ret < 0)
4145                 return ret;
4146         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4147                 nfs_zap_acl_cache(inode);
4148         ret = nfs4_read_cached_acl(inode, buf, buflen);
4149         if (ret != -ENOENT)
4150                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4151                  * but no cached acl data, just the acl length */
4152                 return ret;
4153         return nfs4_get_acl_uncached(inode, buf, buflen);
4154 }
4155
4156 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4157 {
4158         struct nfs_server *server = NFS_SERVER(inode);
4159         struct page *pages[NFS4ACL_MAXPAGES];
4160         struct nfs_setaclargs arg = {
4161                 .fh             = NFS_FH(inode),
4162                 .acl_pages      = pages,
4163                 .acl_len        = buflen,
4164         };
4165         struct nfs_setaclres res;
4166         struct rpc_message msg = {
4167                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4168                 .rpc_argp       = &arg,
4169                 .rpc_resp       = &res,
4170         };
4171         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4172         int ret, i;
4173
4174         if (!nfs4_server_supports_acls(server))
4175                 return -EOPNOTSUPP;
4176         if (npages > ARRAY_SIZE(pages))
4177                 return -ERANGE;
4178         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4179         if (i < 0)
4180                 return i;
4181         nfs4_inode_return_delegation(inode);
4182         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4183
4184         /*
4185          * Free each page after tx, so the only ref left is
4186          * held by the network stack
4187          */
4188         for (; i > 0; i--)
4189                 put_page(pages[i-1]);
4190
4191         /*
4192          * Acl update can result in inode attribute update.
4193          * so mark the attribute cache invalid.
4194          */
4195         spin_lock(&inode->i_lock);
4196         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4197         spin_unlock(&inode->i_lock);
4198         nfs_access_zap_cache(inode);
4199         nfs_zap_acl_cache(inode);
4200         return ret;
4201 }
4202
4203 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4204 {
4205         struct nfs4_exception exception = { };
4206         int err;
4207         do {
4208                 err = nfs4_handle_exception(NFS_SERVER(inode),
4209                                 __nfs4_proc_set_acl(inode, buf, buflen),
4210                                 &exception);
4211         } while (exception.retry);
4212         return err;
4213 }
4214
4215 static int
4216 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4217 {
4218         struct nfs_client *clp = server->nfs_client;
4219
4220         if (task->tk_status >= 0)
4221                 return 0;
4222         switch(task->tk_status) {
4223                 case -NFS4ERR_DELEG_REVOKED:
4224                 case -NFS4ERR_ADMIN_REVOKED:
4225                 case -NFS4ERR_BAD_STATEID:
4226                         if (state == NULL)
4227                                 break;
4228                         nfs_remove_bad_delegation(state->inode);
4229                 case -NFS4ERR_OPENMODE:
4230                         if (state == NULL)
4231                                 break;
4232                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4233                                 goto stateid_invalid;
4234                         goto wait_on_recovery;
4235                 case -NFS4ERR_EXPIRED:
4236                         if (state != NULL) {
4237                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4238                                         goto stateid_invalid;
4239                         }
4240                 case -NFS4ERR_STALE_STATEID:
4241                 case -NFS4ERR_STALE_CLIENTID:
4242                         nfs4_schedule_lease_recovery(clp);
4243                         goto wait_on_recovery;
4244 #if defined(CONFIG_NFS_V4_1)
4245                 case -NFS4ERR_BADSESSION:
4246                 case -NFS4ERR_BADSLOT:
4247                 case -NFS4ERR_BAD_HIGH_SLOT:
4248                 case -NFS4ERR_DEADSESSION:
4249                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4250                 case -NFS4ERR_SEQ_FALSE_RETRY:
4251                 case -NFS4ERR_SEQ_MISORDERED:
4252                         dprintk("%s ERROR %d, Reset session\n", __func__,
4253                                 task->tk_status);
4254                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4255                         goto wait_on_recovery;
4256 #endif /* CONFIG_NFS_V4_1 */
4257                 case -NFS4ERR_DELAY:
4258                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4259                 case -NFS4ERR_GRACE:
4260                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4261                         task->tk_status = 0;
4262                         return -EAGAIN;
4263                 case -NFS4ERR_RETRY_UNCACHED_REP:
4264                 case -NFS4ERR_OLD_STATEID:
4265                         task->tk_status = 0;
4266                         return -EAGAIN;
4267         }
4268         task->tk_status = nfs4_map_errors(task->tk_status);
4269         return 0;
4270 stateid_invalid:
4271         task->tk_status = -EIO;
4272         return 0;
4273 wait_on_recovery:
4274         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4275         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4276                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4277         task->tk_status = 0;
4278         return -EAGAIN;
4279 }
4280
4281 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4282                                     nfs4_verifier *bootverf)
4283 {
4284         __be32 verf[2];
4285
4286         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4287                 /* An impossible timestamp guarantees this value
4288                  * will never match a generated boot time. */
4289                 verf[0] = 0;
4290                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4291         } else {
4292                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4293                 verf[0] = (__be32)nn->boot_time.tv_sec;
4294                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4295         }
4296         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4297 }
4298
4299 static unsigned int
4300 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4301                                    char *buf, size_t len)
4302 {
4303         unsigned int result;
4304
4305         rcu_read_lock();
4306         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4307                                 clp->cl_ipaddr,
4308                                 rpc_peeraddr2str(clp->cl_rpcclient,
4309                                                         RPC_DISPLAY_ADDR),
4310                                 rpc_peeraddr2str(clp->cl_rpcclient,
4311                                                         RPC_DISPLAY_PROTO));
4312         rcu_read_unlock();
4313         return result;
4314 }
4315
4316 static unsigned int
4317 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4318                                 char *buf, size_t len)
4319 {
4320         char *nodename = clp->cl_rpcclient->cl_nodename;
4321
4322         if (nfs4_client_id_uniquifier[0] != '\0')
4323                 nodename = nfs4_client_id_uniquifier;
4324         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4325                                 clp->rpc_ops->version, clp->cl_minorversion,
4326                                 nodename);
4327 }
4328
4329 /**
4330  * nfs4_proc_setclientid - Negotiate client ID
4331  * @clp: state data structure
4332  * @program: RPC program for NFSv4 callback service
4333  * @port: IP port number for NFS4 callback service
4334  * @cred: RPC credential to use for this call
4335  * @res: where to place the result
4336  *
4337  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4338  */
4339 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4340                 unsigned short port, struct rpc_cred *cred,
4341                 struct nfs4_setclientid_res *res)
4342 {
4343         nfs4_verifier sc_verifier;
4344         struct nfs4_setclientid setclientid = {
4345                 .sc_verifier = &sc_verifier,
4346                 .sc_prog = program,
4347                 .sc_cb_ident = clp->cl_cb_ident,
4348         };
4349         struct rpc_message msg = {
4350                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4351                 .rpc_argp = &setclientid,
4352                 .rpc_resp = res,
4353                 .rpc_cred = cred,
4354         };
4355         int status;
4356
4357         /* nfs_client_id4 */
4358         nfs4_init_boot_verifier(clp, &sc_verifier);
4359         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4360                 setclientid.sc_name_len =
4361                                 nfs4_init_uniform_client_string(clp,
4362                                                 setclientid.sc_name,
4363                                                 sizeof(setclientid.sc_name));
4364         else
4365                 setclientid.sc_name_len =
4366                                 nfs4_init_nonuniform_client_string(clp,
4367                                                 setclientid.sc_name,
4368                                                 sizeof(setclientid.sc_name));
4369         /* cb_client4 */
4370         rcu_read_lock();
4371         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4372                                 sizeof(setclientid.sc_netid),
4373                                 rpc_peeraddr2str(clp->cl_rpcclient,
4374                                                         RPC_DISPLAY_NETID));
4375         rcu_read_unlock();
4376         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4377                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4378                                 clp->cl_ipaddr, port >> 8, port & 255);
4379
4380         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4381                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4382                 setclientid.sc_name_len, setclientid.sc_name);
4383         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4384         dprintk("NFS reply setclientid: %d\n", status);
4385         return status;
4386 }
4387
4388 /**
4389  * nfs4_proc_setclientid_confirm - Confirm client ID
4390  * @clp: state data structure
4391  * @res: result of a previous SETCLIENTID
4392  * @cred: RPC credential to use for this call
4393  *
4394  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4395  */
4396 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4397                 struct nfs4_setclientid_res *arg,
4398                 struct rpc_cred *cred)
4399 {
4400         struct rpc_message msg = {
4401                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4402                 .rpc_argp = arg,
4403                 .rpc_cred = cred,
4404         };
4405         int status;
4406
4407         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4408                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4409                 clp->cl_clientid);
4410         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4411         dprintk("NFS reply setclientid_confirm: %d\n", status);
4412         return status;
4413 }
4414
4415 struct nfs4_delegreturndata {
4416         struct nfs4_delegreturnargs args;
4417         struct nfs4_delegreturnres res;
4418         struct nfs_fh fh;
4419         nfs4_stateid stateid;
4420         unsigned long timestamp;
4421         struct nfs_fattr fattr;
4422         int rpc_status;
4423 };
4424
4425 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4426 {
4427         struct nfs4_delegreturndata *data = calldata;
4428
4429         if (!nfs4_sequence_done(task, &data->res.seq_res))
4430                 return;
4431
4432         switch (task->tk_status) {
4433         case 0:
4434                 renew_lease(data->res.server, data->timestamp);
4435                 break;
4436         case -NFS4ERR_ADMIN_REVOKED:
4437         case -NFS4ERR_DELEG_REVOKED:
4438         case -NFS4ERR_BAD_STATEID:
4439         case -NFS4ERR_OLD_STATEID:
4440         case -NFS4ERR_STALE_STATEID:
4441         case -NFS4ERR_EXPIRED:
4442                 task->tk_status = 0;
4443                 break;
4444         default:
4445                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4446                                 -EAGAIN) {
4447                         rpc_restart_call_prepare(task);
4448                         return;
4449                 }
4450         }
4451         data->rpc_status = task->tk_status;
4452 }
4453
4454 static void nfs4_delegreturn_release(void *calldata)
4455 {
4456         kfree(calldata);
4457 }
4458
4459 #if defined(CONFIG_NFS_V4_1)
4460 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4461 {
4462         struct nfs4_delegreturndata *d_data;
4463
4464         d_data = (struct nfs4_delegreturndata *)data;
4465
4466         nfs4_setup_sequence(d_data->res.server,
4467                         &d_data->args.seq_args,
4468                         &d_data->res.seq_res,
4469                         task);
4470 }
4471 #endif /* CONFIG_NFS_V4_1 */
4472
4473 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4474 #if defined(CONFIG_NFS_V4_1)
4475         .rpc_call_prepare = nfs4_delegreturn_prepare,
4476 #endif /* CONFIG_NFS_V4_1 */
4477         .rpc_call_done = nfs4_delegreturn_done,
4478         .rpc_release = nfs4_delegreturn_release,
4479 };
4480
4481 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4482 {
4483         struct nfs4_delegreturndata *data;
4484         struct nfs_server *server = NFS_SERVER(inode);
4485         struct rpc_task *task;
4486         struct rpc_message msg = {
4487                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4488                 .rpc_cred = cred,
4489         };
4490         struct rpc_task_setup task_setup_data = {
4491                 .rpc_client = server->client,
4492                 .rpc_message = &msg,
4493                 .callback_ops = &nfs4_delegreturn_ops,
4494                 .flags = RPC_TASK_ASYNC,
4495         };
4496         int status = 0;
4497
4498         data = kzalloc(sizeof(*data), GFP_NOFS);
4499         if (data == NULL)
4500                 return -ENOMEM;
4501         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4502         data->args.fhandle = &data->fh;
4503         data->args.stateid = &data->stateid;
4504         data->args.bitmask = server->cache_consistency_bitmask;
4505         nfs_copy_fh(&data->fh, NFS_FH(inode));
4506         nfs4_stateid_copy(&data->stateid, stateid);
4507         data->res.fattr = &data->fattr;
4508         data->res.server = server;
4509         nfs_fattr_init(data->res.fattr);
4510         data->timestamp = jiffies;
4511         data->rpc_status = 0;
4512
4513         task_setup_data.callback_data = data;
4514         msg.rpc_argp = &data->args;
4515         msg.rpc_resp = &data->res;
4516         task = rpc_run_task(&task_setup_data);
4517         if (IS_ERR(task))
4518                 return PTR_ERR(task);
4519         if (!issync)
4520                 goto out;
4521         status = nfs4_wait_for_completion_rpc_task(task);
4522         if (status != 0)
4523                 goto out;
4524         status = data->rpc_status;
4525         if (status == 0)
4526                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4527         else
4528                 nfs_refresh_inode(inode, &data->fattr);
4529 out:
4530         rpc_put_task(task);
4531         return status;
4532 }
4533
4534 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4535 {
4536         struct nfs_server *server = NFS_SERVER(inode);
4537         struct nfs4_exception exception = { };
4538         int err;
4539         do {
4540                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4541                 switch (err) {
4542                         case -NFS4ERR_STALE_STATEID:
4543                         case -NFS4ERR_EXPIRED:
4544                         case 0:
4545                                 return 0;
4546                 }
4547                 err = nfs4_handle_exception(server, err, &exception);
4548         } while (exception.retry);
4549         return err;
4550 }
4551
4552 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4553 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4554
4555 /*
4556  * sleep, with exponential backoff, and retry the LOCK operation.
4557  */
4558 static unsigned long
4559 nfs4_set_lock_task_retry(unsigned long timeout)
4560 {
4561         freezable_schedule_timeout_killable(timeout);
4562         timeout <<= 1;
4563         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4564                 return NFS4_LOCK_MAXTIMEOUT;
4565         return timeout;
4566 }
4567
4568 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4569 {
4570         struct inode *inode = state->inode;
4571         struct nfs_server *server = NFS_SERVER(inode);
4572         struct nfs_client *clp = server->nfs_client;
4573         struct nfs_lockt_args arg = {
4574                 .fh = NFS_FH(inode),
4575                 .fl = request,
4576         };
4577         struct nfs_lockt_res res = {
4578                 .denied = request,
4579         };
4580         struct rpc_message msg = {
4581                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4582                 .rpc_argp       = &arg,
4583                 .rpc_resp       = &res,
4584                 .rpc_cred       = state->owner->so_cred,
4585         };
4586         struct nfs4_lock_state *lsp;
4587         int status;
4588
4589         arg.lock_owner.clientid = clp->cl_clientid;
4590         status = nfs4_set_lock_state(state, request);
4591         if (status != 0)
4592                 goto out;
4593         lsp = request->fl_u.nfs4_fl.owner;
4594         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4595         arg.lock_owner.s_dev = server->s_dev;
4596         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4597         switch (status) {
4598                 case 0:
4599                         request->fl_type = F_UNLCK;
4600                         break;
4601                 case -NFS4ERR_DENIED:
4602                         status = 0;
4603         }
4604         request->fl_ops->fl_release_private(request);
4605         request->fl_ops = NULL;
4606 out:
4607         return status;
4608 }
4609
4610 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4611 {
4612         struct nfs4_exception exception = { };
4613         int err;
4614
4615         do {
4616                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4617                                 _nfs4_proc_getlk(state, cmd, request),
4618                                 &exception);
4619         } while (exception.retry);
4620         return err;
4621 }
4622
4623 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4624 {
4625         int res = 0;
4626         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4627                 case FL_POSIX:
4628                         res = posix_lock_file_wait(file, fl);
4629                         break;
4630                 case FL_FLOCK:
4631                         res = flock_lock_file_wait(file, fl);
4632                         break;
4633                 default:
4634                         BUG();
4635         }
4636         return res;
4637 }
4638
4639 struct nfs4_unlockdata {
4640         struct nfs_locku_args arg;
4641         struct nfs_locku_res res;
4642         struct nfs4_lock_state *lsp;
4643         struct nfs_open_context *ctx;
4644         struct file_lock fl;
4645         const struct nfs_server *server;
4646         unsigned long timestamp;
4647 };
4648
4649 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4650                 struct nfs_open_context *ctx,
4651                 struct nfs4_lock_state *lsp,
4652                 struct nfs_seqid *seqid)
4653 {
4654         struct nfs4_unlockdata *p;
4655         struct inode *inode = lsp->ls_state->inode;
4656
4657         p = kzalloc(sizeof(*p), GFP_NOFS);
4658         if (p == NULL)
4659                 return NULL;
4660         p->arg.fh = NFS_FH(inode);
4661         p->arg.fl = &p->fl;
4662         p->arg.seqid = seqid;
4663         p->res.seqid = seqid;
4664         p->arg.stateid = &lsp->ls_stateid;
4665         p->lsp = lsp;
4666         atomic_inc(&lsp->ls_count);
4667         /* Ensure we don't close file until we're done freeing locks! */
4668         p->ctx = get_nfs_open_context(ctx);
4669         memcpy(&p->fl, fl, sizeof(p->fl));
4670         p->server = NFS_SERVER(inode);
4671         return p;
4672 }
4673
4674 static void nfs4_locku_release_calldata(void *data)
4675 {
4676         struct nfs4_unlockdata *calldata = data;
4677         nfs_free_seqid(calldata->arg.seqid);
4678         nfs4_put_lock_state(calldata->lsp);
4679         put_nfs_open_context(calldata->ctx);
4680         kfree(calldata);
4681 }
4682
4683 static void nfs4_locku_done(struct rpc_task *task, void *data)
4684 {
4685         struct nfs4_unlockdata *calldata = data;
4686
4687         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4688                 return;
4689         switch (task->tk_status) {
4690                 case 0:
4691                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4692                                         &calldata->res.stateid);
4693                         renew_lease(calldata->server, calldata->timestamp);
4694                         break;
4695                 case -NFS4ERR_BAD_STATEID:
4696                 case -NFS4ERR_OLD_STATEID:
4697                 case -NFS4ERR_STALE_STATEID:
4698                 case -NFS4ERR_EXPIRED:
4699                         break;
4700                 default:
4701                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4702                                 rpc_restart_call_prepare(task);
4703         }
4704         nfs_release_seqid(calldata->arg.seqid);
4705 }
4706
4707 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4708 {
4709         struct nfs4_unlockdata *calldata = data;
4710
4711         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4712                 goto out_wait;
4713         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4714                 /* Note: exit _without_ running nfs4_locku_done */
4715                 goto out_no_action;
4716         }
4717         calldata->timestamp = jiffies;
4718         if (nfs4_setup_sequence(calldata->server,
4719                                 &calldata->arg.seq_args,
4720                                 &calldata->res.seq_res,
4721                                 task) != 0)
4722                 nfs_release_seqid(calldata->arg.seqid);
4723         return;
4724 out_no_action:
4725         task->tk_action = NULL;
4726 out_wait:
4727         nfs4_sequence_done(task, &calldata->res.seq_res);
4728 }
4729
4730 static const struct rpc_call_ops nfs4_locku_ops = {
4731         .rpc_call_prepare = nfs4_locku_prepare,
4732         .rpc_call_done = nfs4_locku_done,
4733         .rpc_release = nfs4_locku_release_calldata,
4734 };
4735
4736 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4737                 struct nfs_open_context *ctx,
4738                 struct nfs4_lock_state *lsp,
4739                 struct nfs_seqid *seqid)
4740 {
4741         struct nfs4_unlockdata *data;
4742         struct rpc_message msg = {
4743                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4744                 .rpc_cred = ctx->cred,
4745         };
4746         struct rpc_task_setup task_setup_data = {
4747                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4748                 .rpc_message = &msg,
4749                 .callback_ops = &nfs4_locku_ops,
4750                 .workqueue = nfsiod_workqueue,
4751                 .flags = RPC_TASK_ASYNC,
4752         };
4753
4754         /* Ensure this is an unlock - when canceling a lock, the
4755          * canceled lock is passed in, and it won't be an unlock.
4756          */
4757         fl->fl_type = F_UNLCK;
4758
4759         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4760         if (data == NULL) {
4761                 nfs_free_seqid(seqid);
4762                 return ERR_PTR(-ENOMEM);
4763         }
4764
4765         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4766         msg.rpc_argp = &data->arg;
4767         msg.rpc_resp = &data->res;
4768         task_setup_data.callback_data = data;
4769         return rpc_run_task(&task_setup_data);
4770 }
4771
4772 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4773 {
4774         struct inode *inode = state->inode;
4775         struct nfs4_state_owner *sp = state->owner;
4776         struct nfs_inode *nfsi = NFS_I(inode);
4777         struct nfs_seqid *seqid;
4778         struct nfs4_lock_state *lsp;
4779         struct rpc_task *task;
4780         int status = 0;
4781         unsigned char fl_flags = request->fl_flags;
4782
4783         status = nfs4_set_lock_state(state, request);
4784         /* Unlock _before_ we do the RPC call */
4785         request->fl_flags |= FL_EXISTS;
4786         /* Exclude nfs_delegation_claim_locks() */
4787         mutex_lock(&sp->so_delegreturn_mutex);
4788         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
4789         down_read(&nfsi->rwsem);
4790         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4791                 up_read(&nfsi->rwsem);
4792                 mutex_unlock(&sp->so_delegreturn_mutex);
4793                 goto out;
4794         }
4795         up_read(&nfsi->rwsem);
4796         mutex_unlock(&sp->so_delegreturn_mutex);
4797         if (status != 0)
4798                 goto out;
4799         /* Is this a delegated lock? */
4800         lsp = request->fl_u.nfs4_fl.owner;
4801         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
4802                 goto out;
4803         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4804         status = -ENOMEM;
4805         if (seqid == NULL)
4806                 goto out;
4807         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4808         status = PTR_ERR(task);
4809         if (IS_ERR(task))
4810                 goto out;
4811         status = nfs4_wait_for_completion_rpc_task(task);
4812         rpc_put_task(task);
4813 out:
4814         request->fl_flags = fl_flags;
4815         return status;
4816 }
4817
4818 struct nfs4_lockdata {
4819         struct nfs_lock_args arg;
4820         struct nfs_lock_res res;
4821         struct nfs4_lock_state *lsp;
4822         struct nfs_open_context *ctx;
4823         struct file_lock fl;
4824         unsigned long timestamp;
4825         int rpc_status;
4826         int cancelled;
4827         struct nfs_server *server;
4828 };
4829
4830 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4831                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4832                 gfp_t gfp_mask)
4833 {
4834         struct nfs4_lockdata *p;
4835         struct inode *inode = lsp->ls_state->inode;
4836         struct nfs_server *server = NFS_SERVER(inode);
4837
4838         p = kzalloc(sizeof(*p), gfp_mask);
4839         if (p == NULL)
4840                 return NULL;
4841
4842         p->arg.fh = NFS_FH(inode);
4843         p->arg.fl = &p->fl;
4844         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4845         if (p->arg.open_seqid == NULL)
4846                 goto out_free;
4847         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4848         if (p->arg.lock_seqid == NULL)
4849                 goto out_free_seqid;
4850         p->arg.lock_stateid = &lsp->ls_stateid;
4851         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4852         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4853         p->arg.lock_owner.s_dev = server->s_dev;
4854         p->res.lock_seqid = p->arg.lock_seqid;
4855         p->lsp = lsp;
4856         p->server = server;
4857         atomic_inc(&lsp->ls_count);
4858         p->ctx = get_nfs_open_context(ctx);
4859         memcpy(&p->fl, fl, sizeof(p->fl));
4860         return p;
4861 out_free_seqid:
4862         nfs_free_seqid(p->arg.open_seqid);
4863 out_free:
4864         kfree(p);
4865         return NULL;
4866 }
4867
4868 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4869 {
4870         struct nfs4_lockdata *data = calldata;
4871         struct nfs4_state *state = data->lsp->ls_state;
4872
4873         dprintk("%s: begin!\n", __func__);
4874         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4875                 goto out_wait;
4876         /* Do we need to do an open_to_lock_owner? */
4877         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4878                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
4879                         goto out_release_lock_seqid;
4880                 }
4881                 data->arg.open_stateid = &state->open_stateid;
4882                 data->arg.new_lock_owner = 1;
4883                 data->res.open_seqid = data->arg.open_seqid;
4884         } else
4885                 data->arg.new_lock_owner = 0;
4886         if (!nfs4_valid_open_stateid(state)) {
4887                 data->rpc_status = -EBADF;
4888                 task->tk_action = NULL;
4889                 goto out_release_open_seqid;
4890         }
4891         data->timestamp = jiffies;
4892         if (nfs4_setup_sequence(data->server,
4893                                 &data->arg.seq_args,
4894                                 &data->res.seq_res,
4895                                 task) == 0)
4896                 return;
4897 out_release_open_seqid:
4898         nfs_release_seqid(data->arg.open_seqid);
4899 out_release_lock_seqid:
4900         nfs_release_seqid(data->arg.lock_seqid);
4901 out_wait:
4902         nfs4_sequence_done(task, &data->res.seq_res);
4903         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4904 }
4905
4906 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4907 {
4908         struct nfs4_lockdata *data = calldata;
4909
4910         dprintk("%s: begin!\n", __func__);
4911
4912         if (!nfs4_sequence_done(task, &data->res.seq_res))
4913                 return;
4914
4915         data->rpc_status = task->tk_status;
4916         if (data->arg.new_lock_owner != 0) {
4917                 if (data->rpc_status == 0)
4918                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4919                 else
4920                         goto out;
4921         }
4922         if (data->rpc_status == 0) {
4923                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4924                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4925                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4926         }
4927 out:
4928         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4929 }
4930
4931 static void nfs4_lock_release(void *calldata)
4932 {
4933         struct nfs4_lockdata *data = calldata;
4934
4935         dprintk("%s: begin!\n", __func__);
4936         nfs_free_seqid(data->arg.open_seqid);
4937         if (data->cancelled != 0) {
4938                 struct rpc_task *task;
4939                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4940                                 data->arg.lock_seqid);
4941                 if (!IS_ERR(task))
4942                         rpc_put_task_async(task);
4943                 dprintk("%s: cancelling lock!\n", __func__);
4944         } else
4945                 nfs_free_seqid(data->arg.lock_seqid);
4946         nfs4_put_lock_state(data->lsp);
4947         put_nfs_open_context(data->ctx);
4948         kfree(data);
4949         dprintk("%s: done!\n", __func__);
4950 }
4951
4952 static const struct rpc_call_ops nfs4_lock_ops = {
4953         .rpc_call_prepare = nfs4_lock_prepare,
4954         .rpc_call_done = nfs4_lock_done,
4955         .rpc_release = nfs4_lock_release,
4956 };
4957
4958 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4959 {
4960         switch (error) {
4961         case -NFS4ERR_ADMIN_REVOKED:
4962         case -NFS4ERR_BAD_STATEID:
4963                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4964                 if (new_lock_owner != 0 ||
4965                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4966                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4967                 break;
4968         case -NFS4ERR_STALE_STATEID:
4969                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4970         case -NFS4ERR_EXPIRED:
4971                 nfs4_schedule_lease_recovery(server->nfs_client);
4972         };
4973 }
4974
4975 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4976 {
4977         struct nfs4_lockdata *data;
4978         struct rpc_task *task;
4979         struct rpc_message msg = {
4980                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4981                 .rpc_cred = state->owner->so_cred,
4982         };
4983         struct rpc_task_setup task_setup_data = {
4984                 .rpc_client = NFS_CLIENT(state->inode),
4985                 .rpc_message = &msg,
4986                 .callback_ops = &nfs4_lock_ops,
4987                 .workqueue = nfsiod_workqueue,
4988                 .flags = RPC_TASK_ASYNC,
4989         };
4990         int ret;
4991
4992         dprintk("%s: begin!\n", __func__);
4993         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4994                         fl->fl_u.nfs4_fl.owner,
4995                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4996         if (data == NULL)
4997                 return -ENOMEM;
4998         if (IS_SETLKW(cmd))
4999                 data->arg.block = 1;
5000         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5001         msg.rpc_argp = &data->arg;
5002         msg.rpc_resp = &data->res;
5003         task_setup_data.callback_data = data;
5004         if (recovery_type > NFS_LOCK_NEW) {
5005                 if (recovery_type == NFS_LOCK_RECLAIM)
5006                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5007                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5008         }
5009         task = rpc_run_task(&task_setup_data);
5010         if (IS_ERR(task))
5011                 return PTR_ERR(task);
5012         ret = nfs4_wait_for_completion_rpc_task(task);
5013         if (ret == 0) {
5014                 ret = data->rpc_status;
5015                 if (ret)
5016                         nfs4_handle_setlk_error(data->server, data->lsp,
5017                                         data->arg.new_lock_owner, ret);
5018         } else
5019                 data->cancelled = 1;
5020         rpc_put_task(task);
5021         dprintk("%s: done, ret = %d!\n", __func__, ret);
5022         return ret;
5023 }
5024
5025 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5026 {
5027         struct nfs_server *server = NFS_SERVER(state->inode);
5028         struct nfs4_exception exception = {
5029                 .inode = state->inode,
5030         };
5031         int err;
5032
5033         do {
5034                 /* Cache the lock if possible... */
5035                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5036                         return 0;
5037                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5038                 if (err != -NFS4ERR_DELAY)
5039                         break;
5040                 nfs4_handle_exception(server, err, &exception);
5041         } while (exception.retry);
5042         return err;
5043 }
5044
5045 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5046 {
5047         struct nfs_server *server = NFS_SERVER(state->inode);
5048         struct nfs4_exception exception = {
5049                 .inode = state->inode,
5050         };
5051         int err;
5052
5053         err = nfs4_set_lock_state(state, request);
5054         if (err != 0)
5055                 return err;
5056         do {
5057                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5058                         return 0;
5059                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5060                 switch (err) {
5061                 default:
5062                         goto out;
5063                 case -NFS4ERR_GRACE:
5064                 case -NFS4ERR_DELAY:
5065                         nfs4_handle_exception(server, err, &exception);
5066                         err = 0;
5067                 }
5068         } while (exception.retry);
5069 out:
5070         return err;
5071 }
5072
5073 #if defined(CONFIG_NFS_V4_1)
5074 /**
5075  * nfs41_check_expired_locks - possibly free a lock stateid
5076  *
5077  * @state: NFSv4 state for an inode
5078  *
5079  * Returns NFS_OK if recovery for this stateid is now finished.
5080  * Otherwise a negative NFS4ERR value is returned.
5081  */
5082 static int nfs41_check_expired_locks(struct nfs4_state *state)
5083 {
5084         int status, ret = -NFS4ERR_BAD_STATEID;
5085         struct nfs4_lock_state *lsp;
5086         struct nfs_server *server = NFS_SERVER(state->inode);
5087
5088         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5089                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5090                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
5091                         if (status != NFS_OK) {
5092                                 /* Free the stateid unless the server
5093                                  * informs us the stateid is unrecognized. */
5094                                 if (status != -NFS4ERR_BAD_STATEID)
5095                                         nfs41_free_stateid(server,
5096                                                         &lsp->ls_stateid);
5097                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5098                                 ret = status;
5099                         }
5100                 }
5101         };
5102
5103         return ret;
5104 }
5105
5106 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5107 {
5108         int status = NFS_OK;
5109
5110         if (test_bit(LK_STATE_IN_USE, &state->flags))
5111                 status = nfs41_check_expired_locks(state);
5112         if (status != NFS_OK)
5113                 status = nfs4_lock_expired(state, request);
5114         return status;
5115 }
5116 #endif
5117
5118 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5119 {
5120         struct nfs4_state_owner *sp = state->owner;
5121         struct nfs_inode *nfsi = NFS_I(state->inode);
5122         unsigned char fl_flags = request->fl_flags;
5123         unsigned int seq;
5124         int status = -ENOLCK;
5125
5126         if ((fl_flags & FL_POSIX) &&
5127                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5128                 goto out;
5129         /* Is this a delegated open? */
5130         status = nfs4_set_lock_state(state, request);
5131         if (status != 0)
5132                 goto out;
5133         request->fl_flags |= FL_ACCESS;
5134         status = do_vfs_lock(request->fl_file, request);
5135         if (status < 0)
5136                 goto out;
5137         down_read(&nfsi->rwsem);
5138         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5139                 /* Yes: cache locks! */
5140                 /* ...but avoid races with delegation recall... */
5141                 request->fl_flags = fl_flags & ~FL_SLEEP;
5142                 status = do_vfs_lock(request->fl_file, request);
5143                 goto out_unlock;
5144         }
5145         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5146         up_read(&nfsi->rwsem);
5147         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5148         if (status != 0)
5149                 goto out;
5150         down_read(&nfsi->rwsem);
5151         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5152                 status = -NFS4ERR_DELAY;
5153                 goto out_unlock;
5154         }
5155         /* Note: we always want to sleep here! */
5156         request->fl_flags = fl_flags | FL_SLEEP;
5157         if (do_vfs_lock(request->fl_file, request) < 0)
5158                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5159                         "manager!\n", __func__);
5160 out_unlock:
5161         up_read(&nfsi->rwsem);
5162 out:
5163         request->fl_flags = fl_flags;
5164         return status;
5165 }
5166
5167 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5168 {
5169         struct nfs4_exception exception = {
5170                 .state = state,
5171                 .inode = state->inode,
5172         };
5173         int err;
5174
5175         do {
5176                 err = _nfs4_proc_setlk(state, cmd, request);
5177                 if (err == -NFS4ERR_DENIED)
5178                         err = -EAGAIN;
5179                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5180                                 err, &exception);
5181         } while (exception.retry);
5182         return err;
5183 }
5184
5185 static int
5186 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5187 {
5188         struct nfs_open_context *ctx;
5189         struct nfs4_state *state;
5190         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5191         int status;
5192
5193         /* verify open state */
5194         ctx = nfs_file_open_context(filp);
5195         state = ctx->state;
5196
5197         if (request->fl_start < 0 || request->fl_end < 0)
5198                 return -EINVAL;
5199
5200         if (IS_GETLK(cmd)) {
5201                 if (state != NULL)
5202                         return nfs4_proc_getlk(state, F_GETLK, request);
5203                 return 0;
5204         }
5205
5206         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5207                 return -EINVAL;
5208
5209         if (request->fl_type == F_UNLCK) {
5210                 if (state != NULL)
5211                         return nfs4_proc_unlck(state, cmd, request);
5212                 return 0;
5213         }
5214
5215         if (state == NULL)
5216                 return -ENOLCK;
5217         /*
5218          * Don't rely on the VFS having checked the file open mode,
5219          * since it won't do this for flock() locks.
5220          */
5221         switch (request->fl_type) {
5222         case F_RDLCK:
5223                 if (!(filp->f_mode & FMODE_READ))
5224                         return -EBADF;
5225                 break;
5226         case F_WRLCK:
5227                 if (!(filp->f_mode & FMODE_WRITE))
5228                         return -EBADF;
5229         }
5230
5231         do {
5232                 status = nfs4_proc_setlk(state, cmd, request);
5233                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5234                         break;
5235                 timeout = nfs4_set_lock_task_retry(timeout);
5236                 status = -ERESTARTSYS;
5237                 if (signalled())
5238                         break;
5239         } while(status < 0);
5240         return status;
5241 }
5242
5243 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5244 {
5245         struct nfs_server *server = NFS_SERVER(state->inode);
5246         int err;
5247
5248         err = nfs4_set_lock_state(state, fl);
5249         if (err != 0)
5250                 return err;
5251         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5252         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5253 }
5254
5255 struct nfs_release_lockowner_data {
5256         struct nfs4_lock_state *lsp;
5257         struct nfs_server *server;
5258         struct nfs_release_lockowner_args args;
5259 };
5260
5261 static void nfs4_release_lockowner_release(void *calldata)
5262 {
5263         struct nfs_release_lockowner_data *data = calldata;
5264         nfs4_free_lock_state(data->server, data->lsp);
5265         kfree(calldata);
5266 }
5267
5268 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5269         .rpc_release = nfs4_release_lockowner_release,
5270 };
5271
5272 static int nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
5273 {
5274         struct nfs_release_lockowner_data *data;
5275         struct rpc_message msg = {
5276                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5277         };
5278
5279         if (server->nfs_client->cl_mvops->minor_version != 0)
5280                 return -EINVAL;
5281         data = kmalloc(sizeof(*data), GFP_NOFS);
5282         if (!data)
5283                 return -ENOMEM;
5284         data->lsp = lsp;
5285         data->server = server;
5286         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5287         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5288         data->args.lock_owner.s_dev = server->s_dev;
5289         msg.rpc_argp = &data->args;
5290         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5291         return 0;
5292 }
5293
5294 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5295
5296 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5297                                    const void *buf, size_t buflen,
5298                                    int flags, int type)
5299 {
5300         if (strcmp(key, "") != 0)
5301                 return -EINVAL;
5302
5303         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5304 }
5305
5306 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5307                                    void *buf, size_t buflen, int type)
5308 {
5309         if (strcmp(key, "") != 0)
5310                 return -EINVAL;
5311
5312         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5313 }
5314
5315 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5316                                        size_t list_len, const char *name,
5317                                        size_t name_len, int type)
5318 {
5319         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5320
5321         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5322                 return 0;
5323
5324         if (list && len <= list_len)
5325                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5326         return len;
5327 }
5328
5329 /*
5330  * nfs_fhget will use either the mounted_on_fileid or the fileid
5331  */
5332 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5333 {
5334         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5335                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5336               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5337               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5338                 return;
5339
5340         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5341                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5342         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5343         fattr->nlink = 2;
5344 }
5345
5346 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5347                                    const struct qstr *name,
5348                                    struct nfs4_fs_locations *fs_locations,
5349                                    struct page *page)
5350 {
5351         struct nfs_server *server = NFS_SERVER(dir);
5352         u32 bitmask[2] = {
5353                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5354         };
5355         struct nfs4_fs_locations_arg args = {
5356                 .dir_fh = NFS_FH(dir),
5357                 .name = name,
5358                 .page = page,
5359                 .bitmask = bitmask,
5360         };
5361         struct nfs4_fs_locations_res res = {
5362                 .fs_locations = fs_locations,
5363         };
5364         struct rpc_message msg = {
5365                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5366                 .rpc_argp = &args,
5367                 .rpc_resp = &res,
5368         };
5369         int status;
5370
5371         dprintk("%s: start\n", __func__);
5372
5373         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5374          * is not supported */
5375         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5376                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5377         else
5378                 bitmask[0] |= FATTR4_WORD0_FILEID;
5379
5380         nfs_fattr_init(&fs_locations->fattr);
5381         fs_locations->server = server;
5382         fs_locations->nlocations = 0;
5383         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5384         dprintk("%s: returned status = %d\n", __func__, status);
5385         return status;
5386 }
5387
5388 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5389                            const struct qstr *name,
5390                            struct nfs4_fs_locations *fs_locations,
5391                            struct page *page)
5392 {
5393         struct nfs4_exception exception = { };
5394         int err;
5395         do {
5396                 err = nfs4_handle_exception(NFS_SERVER(dir),
5397                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5398                                 &exception);
5399         } while (exception.retry);
5400         return err;
5401 }
5402
5403 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5404 {
5405         int status;
5406         struct nfs4_secinfo_arg args = {
5407                 .dir_fh = NFS_FH(dir),
5408                 .name   = name,
5409         };
5410         struct nfs4_secinfo_res res = {
5411                 .flavors     = flavors,
5412         };
5413         struct rpc_message msg = {
5414                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5415                 .rpc_argp = &args,
5416                 .rpc_resp = &res,
5417         };
5418
5419         dprintk("NFS call  secinfo %s\n", name->name);
5420         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5421         dprintk("NFS reply  secinfo: %d\n", status);
5422         return status;
5423 }
5424
5425 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5426                       struct nfs4_secinfo_flavors *flavors)
5427 {
5428         struct nfs4_exception exception = { };
5429         int err;
5430         do {
5431                 err = nfs4_handle_exception(NFS_SERVER(dir),
5432                                 _nfs4_proc_secinfo(dir, name, flavors),
5433                                 &exception);
5434         } while (exception.retry);
5435         return err;
5436 }
5437
5438 #ifdef CONFIG_NFS_V4_1
5439 /*
5440  * Check the exchange flags returned by the server for invalid flags, having
5441  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5442  * DS flags set.
5443  */
5444 static int nfs4_check_cl_exchange_flags(u32 flags)
5445 {
5446         if (flags & ~EXCHGID4_FLAG_MASK_R)
5447                 goto out_inval;
5448         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5449             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5450                 goto out_inval;
5451         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5452                 goto out_inval;
5453         return NFS_OK;
5454 out_inval:
5455         return -NFS4ERR_INVAL;
5456 }
5457
5458 static bool
5459 nfs41_same_server_scope(struct nfs41_server_scope *a,
5460                         struct nfs41_server_scope *b)
5461 {
5462         if (a->server_scope_sz == b->server_scope_sz &&
5463             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5464                 return true;
5465
5466         return false;
5467 }
5468
5469 /*
5470  * nfs4_proc_bind_conn_to_session()
5471  *
5472  * The 4.1 client currently uses the same TCP connection for the
5473  * fore and backchannel.
5474  */
5475 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5476 {
5477         int status;
5478         struct nfs41_bind_conn_to_session_res res;
5479         struct rpc_message msg = {
5480                 .rpc_proc =
5481                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5482                 .rpc_argp = clp,
5483                 .rpc_resp = &res,
5484                 .rpc_cred = cred,
5485         };
5486
5487         dprintk("--> %s\n", __func__);
5488
5489         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5490         if (unlikely(res.session == NULL)) {
5491                 status = -ENOMEM;
5492                 goto out;
5493         }
5494
5495         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5496         if (status == 0) {
5497                 if (memcmp(res.session->sess_id.data,
5498                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5499                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5500                         status = -EIO;
5501                         goto out_session;
5502                 }
5503                 if (res.dir != NFS4_CDFS4_BOTH) {
5504                         dprintk("NFS: %s: Unexpected direction from server\n",
5505                                 __func__);
5506                         status = -EIO;
5507                         goto out_session;
5508                 }
5509                 if (res.use_conn_in_rdma_mode) {
5510                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5511                                 __func__);
5512                         status = -EIO;
5513                         goto out_session;
5514                 }
5515         }
5516 out_session:
5517         kfree(res.session);
5518 out:
5519         dprintk("<-- %s status= %d\n", __func__, status);
5520         return status;
5521 }
5522
5523 /*
5524  * nfs4_proc_exchange_id()
5525  *
5526  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5527  *
5528  * Since the clientid has expired, all compounds using sessions
5529  * associated with the stale clientid will be returning
5530  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5531  * be in some phase of session reset.
5532  */
5533 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5534 {
5535         nfs4_verifier verifier;
5536         struct nfs41_exchange_id_args args = {
5537                 .verifier = &verifier,
5538                 .client = clp,
5539                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5540         };
5541         struct nfs41_exchange_id_res res = {
5542                 0
5543         };
5544         int status;
5545         struct rpc_message msg = {
5546                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5547                 .rpc_argp = &args,
5548                 .rpc_resp = &res,
5549                 .rpc_cred = cred,
5550         };
5551
5552         nfs4_init_boot_verifier(clp, &verifier);
5553         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5554                                                         sizeof(args.id));
5555         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5556                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5557                 args.id_len, args.id);
5558
5559         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5560                                         GFP_NOFS);
5561         if (unlikely(res.server_owner == NULL)) {
5562                 status = -ENOMEM;
5563                 goto out;
5564         }
5565
5566         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5567                                         GFP_NOFS);
5568         if (unlikely(res.server_scope == NULL)) {
5569                 status = -ENOMEM;
5570                 goto out_server_owner;
5571         }
5572
5573         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5574         if (unlikely(res.impl_id == NULL)) {
5575                 status = -ENOMEM;
5576                 goto out_server_scope;
5577         }
5578
5579         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5580         if (status == 0)
5581                 status = nfs4_check_cl_exchange_flags(res.flags);
5582
5583         if (status == 0) {
5584                 clp->cl_clientid = res.clientid;
5585                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5586                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5587                         clp->cl_seqid = res.seqid;
5588
5589                 kfree(clp->cl_serverowner);
5590                 clp->cl_serverowner = res.server_owner;
5591                 res.server_owner = NULL;
5592
5593                 /* use the most recent implementation id */
5594                 kfree(clp->cl_implid);
5595                 clp->cl_implid = res.impl_id;
5596
5597                 if (clp->cl_serverscope != NULL &&
5598                     !nfs41_same_server_scope(clp->cl_serverscope,
5599                                              res.server_scope)) {
5600                         dprintk("%s: server_scope mismatch detected\n",
5601                                 __func__);
5602                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5603                         kfree(clp->cl_serverscope);
5604                         clp->cl_serverscope = NULL;
5605                 }
5606
5607                 if (clp->cl_serverscope == NULL) {
5608                         clp->cl_serverscope = res.server_scope;
5609                         goto out;
5610                 }
5611         } else
5612                 kfree(res.impl_id);
5613
5614 out_server_owner:
5615         kfree(res.server_owner);
5616 out_server_scope:
5617         kfree(res.server_scope);
5618 out:
5619         if (clp->cl_implid != NULL)
5620                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5621                         "domain: %s, name: %s, date: %llu,%u\n",
5622                         clp->cl_implid->domain, clp->cl_implid->name,
5623                         clp->cl_implid->date.seconds,
5624                         clp->cl_implid->date.nseconds);
5625         dprintk("NFS reply exchange_id: %d\n", status);
5626         return status;
5627 }
5628
5629 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5630                 struct rpc_cred *cred)
5631 {
5632         struct rpc_message msg = {
5633                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5634                 .rpc_argp = clp,
5635                 .rpc_cred = cred,
5636         };
5637         int status;
5638
5639         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5640         if (status)
5641                 dprintk("NFS: Got error %d from the server %s on "
5642                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5643         return status;
5644 }
5645
5646 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5647                 struct rpc_cred *cred)
5648 {
5649         unsigned int loop;
5650         int ret;
5651
5652         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5653                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5654                 switch (ret) {
5655                 case -NFS4ERR_DELAY:
5656                 case -NFS4ERR_CLIENTID_BUSY:
5657                         ssleep(1);
5658                         break;
5659                 default:
5660                         return ret;
5661                 }
5662         }
5663         return 0;
5664 }
5665
5666 int nfs4_destroy_clientid(struct nfs_client *clp)
5667 {
5668         struct rpc_cred *cred;
5669         int ret = 0;
5670
5671         if (clp->cl_mvops->minor_version < 1)
5672                 goto out;
5673         if (clp->cl_exchange_flags == 0)
5674                 goto out;
5675         if (clp->cl_preserve_clid)
5676                 goto out;
5677         cred = nfs4_get_exchange_id_cred(clp);
5678         ret = nfs4_proc_destroy_clientid(clp, cred);
5679         if (cred)
5680                 put_rpccred(cred);
5681         switch (ret) {
5682         case 0:
5683         case -NFS4ERR_STALE_CLIENTID:
5684                 clp->cl_exchange_flags = 0;
5685         }
5686 out:
5687         return ret;
5688 }
5689
5690 struct nfs4_get_lease_time_data {
5691         struct nfs4_get_lease_time_args *args;
5692         struct nfs4_get_lease_time_res *res;
5693         struct nfs_client *clp;
5694 };
5695
5696 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5697                                         void *calldata)
5698 {
5699         struct nfs4_get_lease_time_data *data =
5700                         (struct nfs4_get_lease_time_data *)calldata;
5701
5702         dprintk("--> %s\n", __func__);
5703         /* just setup sequence, do not trigger session recovery
5704            since we're invoked within one */
5705         nfs41_setup_sequence(data->clp->cl_session,
5706                         &data->args->la_seq_args,
5707                         &data->res->lr_seq_res,
5708                         task);
5709         dprintk("<-- %s\n", __func__);
5710 }
5711
5712 /*
5713  * Called from nfs4_state_manager thread for session setup, so don't recover
5714  * from sequence operation or clientid errors.
5715  */
5716 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5717 {
5718         struct nfs4_get_lease_time_data *data =
5719                         (struct nfs4_get_lease_time_data *)calldata;
5720
5721         dprintk("--> %s\n", __func__);
5722         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5723                 return;
5724         switch (task->tk_status) {
5725         case -NFS4ERR_DELAY:
5726         case -NFS4ERR_GRACE:
5727                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5728                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5729                 task->tk_status = 0;
5730                 /* fall through */
5731         case -NFS4ERR_RETRY_UNCACHED_REP:
5732                 rpc_restart_call_prepare(task);
5733                 return;
5734         }
5735         dprintk("<-- %s\n", __func__);
5736 }
5737
5738 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5739         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5740         .rpc_call_done = nfs4_get_lease_time_done,
5741 };
5742
5743 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5744 {
5745         struct rpc_task *task;
5746         struct nfs4_get_lease_time_args args;
5747         struct nfs4_get_lease_time_res res = {
5748                 .lr_fsinfo = fsinfo,
5749         };
5750         struct nfs4_get_lease_time_data data = {
5751                 .args = &args,
5752                 .res = &res,
5753                 .clp = clp,
5754         };
5755         struct rpc_message msg = {
5756                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5757                 .rpc_argp = &args,
5758                 .rpc_resp = &res,
5759         };
5760         struct rpc_task_setup task_setup = {
5761                 .rpc_client = clp->cl_rpcclient,
5762                 .rpc_message = &msg,
5763                 .callback_ops = &nfs4_get_lease_time_ops,
5764                 .callback_data = &data,
5765                 .flags = RPC_TASK_TIMEOUT,
5766         };
5767         int status;
5768
5769         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5770         nfs4_set_sequence_privileged(&args.la_seq_args);
5771         dprintk("--> %s\n", __func__);
5772         task = rpc_run_task(&task_setup);
5773
5774         if (IS_ERR(task))
5775                 status = PTR_ERR(task);
5776         else {
5777                 status = task->tk_status;
5778                 rpc_put_task(task);
5779         }
5780         dprintk("<-- %s return %d\n", __func__, status);
5781
5782         return status;
5783 }
5784
5785 /*
5786  * Initialize the values to be used by the client in CREATE_SESSION
5787  * If nfs4_init_session set the fore channel request and response sizes,
5788  * use them.
5789  *
5790  * Set the back channel max_resp_sz_cached to zero to force the client to
5791  * always set csa_cachethis to FALSE because the current implementation
5792  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5793  */
5794 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5795 {
5796         struct nfs4_session *session = args->client->cl_session;
5797         unsigned int mxrqst_sz = session->fc_target_max_rqst_sz,
5798                      mxresp_sz = session->fc_target_max_resp_sz;
5799
5800         if (mxrqst_sz == 0)
5801                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5802         if (mxresp_sz == 0)
5803                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5804         /* Fore channel attributes */
5805         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5806         args->fc_attrs.max_resp_sz = mxresp_sz;
5807         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5808         args->fc_attrs.max_reqs = max_session_slots;
5809
5810         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5811                 "max_ops=%u max_reqs=%u\n",
5812                 __func__,
5813                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5814                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5815
5816         /* Back channel attributes */
5817         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5818         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5819         args->bc_attrs.max_resp_sz_cached = 0;
5820         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5821         args->bc_attrs.max_reqs = 1;
5822
5823         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5824                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5825                 __func__,
5826                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5827                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5828                 args->bc_attrs.max_reqs);
5829 }
5830
5831 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5832 {
5833         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5834         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5835
5836         if (rcvd->max_resp_sz > sent->max_resp_sz)
5837                 return -EINVAL;
5838         /*
5839          * Our requested max_ops is the minimum we need; we're not
5840          * prepared to break up compounds into smaller pieces than that.
5841          * So, no point even trying to continue if the server won't
5842          * cooperate:
5843          */
5844         if (rcvd->max_ops < sent->max_ops)
5845                 return -EINVAL;
5846         if (rcvd->max_reqs == 0)
5847                 return -EINVAL;
5848         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5849                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5850         return 0;
5851 }
5852
5853 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5854 {
5855         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5856         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5857
5858         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5859                 return -EINVAL;
5860         if (rcvd->max_resp_sz < sent->max_resp_sz)
5861                 return -EINVAL;
5862         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5863                 return -EINVAL;
5864         /* These would render the backchannel useless: */
5865         if (rcvd->max_ops != sent->max_ops)
5866                 return -EINVAL;
5867         if (rcvd->max_reqs != sent->max_reqs)
5868                 return -EINVAL;
5869         return 0;
5870 }
5871
5872 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5873                                      struct nfs4_session *session)
5874 {
5875         int ret;
5876
5877         ret = nfs4_verify_fore_channel_attrs(args, session);
5878         if (ret)
5879                 return ret;
5880         return nfs4_verify_back_channel_attrs(args, session);
5881 }
5882
5883 static int _nfs4_proc_create_session(struct nfs_client *clp,
5884                 struct rpc_cred *cred)
5885 {
5886         struct nfs4_session *session = clp->cl_session;
5887         struct nfs41_create_session_args args = {
5888                 .client = clp,
5889                 .cb_program = NFS4_CALLBACK,
5890         };
5891         struct nfs41_create_session_res res = {
5892                 .client = clp,
5893         };
5894         struct rpc_message msg = {
5895                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5896                 .rpc_argp = &args,
5897                 .rpc_resp = &res,
5898                 .rpc_cred = cred,
5899         };
5900         int status;
5901
5902         nfs4_init_channel_attrs(&args);
5903         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5904
5905         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5906
5907         if (!status) {
5908                 /* Verify the session's negotiated channel_attrs values */
5909                 status = nfs4_verify_channel_attrs(&args, session);
5910                 /* Increment the clientid slot sequence id */
5911                 clp->cl_seqid++;
5912         }
5913
5914         return status;
5915 }
5916
5917 /*
5918  * Issues a CREATE_SESSION operation to the server.
5919  * It is the responsibility of the caller to verify the session is
5920  * expired before calling this routine.
5921  */
5922 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5923 {
5924         int status;
5925         unsigned *ptr;
5926         struct nfs4_session *session = clp->cl_session;
5927
5928         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5929
5930         status = _nfs4_proc_create_session(clp, cred);
5931         if (status)
5932                 goto out;
5933
5934         /* Init or reset the session slot tables */
5935         status = nfs4_setup_session_slot_tables(session);
5936         dprintk("slot table setup returned %d\n", status);
5937         if (status)
5938                 goto out;
5939
5940         ptr = (unsigned *)&session->sess_id.data[0];
5941         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5942                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5943 out:
5944         dprintk("<-- %s\n", __func__);
5945         return status;
5946 }
5947
5948 /*
5949  * Issue the over-the-wire RPC DESTROY_SESSION.
5950  * The caller must serialize access to this routine.
5951  */
5952 int nfs4_proc_destroy_session(struct nfs4_session *session,
5953                 struct rpc_cred *cred)
5954 {
5955         struct rpc_message msg = {
5956                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5957                 .rpc_argp = session,
5958                 .rpc_cred = cred,
5959         };
5960         int status = 0;
5961
5962         dprintk("--> nfs4_proc_destroy_session\n");
5963
5964         /* session is still being setup */
5965         if (session->clp->cl_cons_state != NFS_CS_READY)
5966                 return status;
5967
5968         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5969
5970         if (status)
5971                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5972                         "Session has been destroyed regardless...\n", status);
5973
5974         dprintk("<-- nfs4_proc_destroy_session\n");
5975         return status;
5976 }
5977
5978 /*
5979  * Renew the cl_session lease.
5980  */
5981 struct nfs4_sequence_data {
5982         struct nfs_client *clp;
5983         struct nfs4_sequence_args args;
5984         struct nfs4_sequence_res res;
5985 };
5986
5987 static void nfs41_sequence_release(void *data)
5988 {
5989         struct nfs4_sequence_data *calldata = data;
5990         struct nfs_client *clp = calldata->clp;
5991
5992         if (atomic_read(&clp->cl_count) > 1)
5993                 nfs4_schedule_state_renewal(clp);
5994         nfs_put_client(clp);
5995         kfree(calldata);
5996 }
5997
5998 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5999 {
6000         switch(task->tk_status) {
6001         case -NFS4ERR_DELAY:
6002                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6003                 return -EAGAIN;
6004         default:
6005                 nfs4_schedule_lease_recovery(clp);
6006         }
6007         return 0;
6008 }
6009
6010 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
6011 {
6012         struct nfs4_sequence_data *calldata = data;
6013         struct nfs_client *clp = calldata->clp;
6014
6015         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
6016                 return;
6017
6018         if (task->tk_status < 0) {
6019                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
6020                 if (atomic_read(&clp->cl_count) == 1)
6021                         goto out;
6022
6023                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
6024                         rpc_restart_call_prepare(task);
6025                         return;
6026                 }
6027         }
6028         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
6029 out:
6030         dprintk("<-- %s\n", __func__);
6031 }
6032
6033 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
6034 {
6035         struct nfs4_sequence_data *calldata = data;
6036         struct nfs_client *clp = calldata->clp;
6037         struct nfs4_sequence_args *args;
6038         struct nfs4_sequence_res *res;
6039
6040         args = task->tk_msg.rpc_argp;
6041         res = task->tk_msg.rpc_resp;
6042
6043         nfs41_setup_sequence(clp->cl_session, args, res, task);
6044 }
6045
6046 static const struct rpc_call_ops nfs41_sequence_ops = {
6047         .rpc_call_done = nfs41_sequence_call_done,
6048         .rpc_call_prepare = nfs41_sequence_prepare,
6049         .rpc_release = nfs41_sequence_release,
6050 };
6051
6052 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
6053                 struct rpc_cred *cred,
6054                 bool is_privileged)
6055 {
6056         struct nfs4_sequence_data *calldata;
6057         struct rpc_message msg = {
6058                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
6059                 .rpc_cred = cred,
6060         };
6061         struct rpc_task_setup task_setup_data = {
6062                 .rpc_client = clp->cl_rpcclient,
6063                 .rpc_message = &msg,
6064                 .callback_ops = &nfs41_sequence_ops,
6065                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6066         };
6067
6068         if (!atomic_inc_not_zero(&clp->cl_count))
6069                 return ERR_PTR(-EIO);
6070         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6071         if (calldata == NULL) {
6072                 nfs_put_client(clp);
6073                 return ERR_PTR(-ENOMEM);
6074         }
6075         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6076         if (is_privileged)
6077                 nfs4_set_sequence_privileged(&calldata->args);
6078         msg.rpc_argp = &calldata->args;
6079         msg.rpc_resp = &calldata->res;
6080         calldata->clp = clp;
6081         task_setup_data.callback_data = calldata;
6082
6083         return rpc_run_task(&task_setup_data);
6084 }
6085
6086 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6087 {
6088         struct rpc_task *task;
6089         int ret = 0;
6090
6091         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6092                 return -EAGAIN;
6093         task = _nfs41_proc_sequence(clp, cred, false);
6094         if (IS_ERR(task))
6095                 ret = PTR_ERR(task);
6096         else
6097                 rpc_put_task_async(task);
6098         dprintk("<-- %s status=%d\n", __func__, ret);
6099         return ret;
6100 }
6101
6102 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6103 {
6104         struct rpc_task *task;
6105         int ret;
6106
6107         task = _nfs41_proc_sequence(clp, cred, true);
6108         if (IS_ERR(task)) {
6109                 ret = PTR_ERR(task);
6110                 goto out;
6111         }
6112         ret = rpc_wait_for_completion_task(task);
6113         if (!ret) {
6114                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6115
6116                 if (task->tk_status == 0)
6117                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6118                 ret = task->tk_status;
6119         }
6120         rpc_put_task(task);
6121 out:
6122         dprintk("<-- %s status=%d\n", __func__, ret);
6123         return ret;
6124 }
6125
6126 struct nfs4_reclaim_complete_data {
6127         struct nfs_client *clp;
6128         struct nfs41_reclaim_complete_args arg;
6129         struct nfs41_reclaim_complete_res res;
6130 };
6131
6132 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6133 {
6134         struct nfs4_reclaim_complete_data *calldata = data;
6135
6136         nfs41_setup_sequence(calldata->clp->cl_session,
6137                         &calldata->arg.seq_args,
6138                         &calldata->res.seq_res,
6139                         task);
6140 }
6141
6142 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6143 {
6144         switch(task->tk_status) {
6145         case 0:
6146         case -NFS4ERR_COMPLETE_ALREADY:
6147         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6148                 break;
6149         case -NFS4ERR_DELAY:
6150                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6151                 /* fall through */
6152         case -NFS4ERR_RETRY_UNCACHED_REP:
6153                 return -EAGAIN;
6154         default:
6155                 nfs4_schedule_lease_recovery(clp);
6156         }
6157         return 0;
6158 }
6159
6160 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6161 {
6162         struct nfs4_reclaim_complete_data *calldata = data;
6163         struct nfs_client *clp = calldata->clp;
6164         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6165
6166         dprintk("--> %s\n", __func__);
6167         if (!nfs41_sequence_done(task, res))
6168                 return;
6169
6170         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6171                 rpc_restart_call_prepare(task);
6172                 return;
6173         }
6174         dprintk("<-- %s\n", __func__);
6175 }
6176
6177 static void nfs4_free_reclaim_complete_data(void *data)
6178 {
6179         struct nfs4_reclaim_complete_data *calldata = data;
6180
6181         kfree(calldata);
6182 }
6183
6184 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6185         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6186         .rpc_call_done = nfs4_reclaim_complete_done,
6187         .rpc_release = nfs4_free_reclaim_complete_data,
6188 };
6189
6190 /*
6191  * Issue a global reclaim complete.
6192  */
6193 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6194 {
6195         struct nfs4_reclaim_complete_data *calldata;
6196         struct rpc_task *task;
6197         struct rpc_message msg = {
6198                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6199         };
6200         struct rpc_task_setup task_setup_data = {
6201                 .rpc_client = clp->cl_rpcclient,
6202                 .rpc_message = &msg,
6203                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6204                 .flags = RPC_TASK_ASYNC,
6205         };
6206         int status = -ENOMEM;
6207
6208         dprintk("--> %s\n", __func__);
6209         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6210         if (calldata == NULL)
6211                 goto out;
6212         calldata->clp = clp;
6213         calldata->arg.one_fs = 0;
6214
6215         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6216         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
6217         msg.rpc_argp = &calldata->arg;
6218         msg.rpc_resp = &calldata->res;
6219         task_setup_data.callback_data = calldata;
6220         task = rpc_run_task(&task_setup_data);
6221         if (IS_ERR(task)) {
6222                 status = PTR_ERR(task);
6223                 goto out;
6224         }
6225         status = nfs4_wait_for_completion_rpc_task(task);
6226         if (status == 0)
6227                 status = task->tk_status;
6228         rpc_put_task(task);
6229         return 0;
6230 out:
6231         dprintk("<-- %s status=%d\n", __func__, status);
6232         return status;
6233 }
6234
6235 static void
6236 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6237 {
6238         struct nfs4_layoutget *lgp = calldata;
6239         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6240         struct nfs4_session *session = nfs4_get_session(server);
6241
6242         dprintk("--> %s\n", __func__);
6243         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6244          * right now covering the LAYOUTGET we are about to send.
6245          * However, that is not so catastrophic, and there seems
6246          * to be no way to prevent it completely.
6247          */
6248         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
6249                                 &lgp->res.seq_res, task))
6250                 return;
6251         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6252                                           NFS_I(lgp->args.inode)->layout,
6253                                           lgp->args.ctx->state)) {
6254                 rpc_exit(task, NFS4_OK);
6255         }
6256 }
6257
6258 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6259 {
6260         struct nfs4_layoutget *lgp = calldata;
6261         struct inode *inode = lgp->args.inode;
6262         struct nfs_server *server = NFS_SERVER(inode);
6263         struct pnfs_layout_hdr *lo;
6264         struct nfs4_state *state = NULL;
6265         unsigned long timeo, now, giveup;
6266
6267         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
6268
6269         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
6270                 goto out;
6271
6272         switch (task->tk_status) {
6273         case 0:
6274                 goto out;
6275         /*
6276          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
6277          * (or clients) writing to the same RAID stripe
6278          */
6279         case -NFS4ERR_LAYOUTTRYLATER:
6280         /*
6281          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
6282          * existing layout before getting a new one).
6283          */
6284         case -NFS4ERR_RECALLCONFLICT:
6285                 timeo = rpc_get_timeout(task->tk_client);
6286                 giveup = lgp->args.timestamp + timeo;
6287                 now = jiffies;
6288                 if (time_after(giveup, now)) {
6289                         unsigned long delay;
6290
6291                         /* Delay for:
6292                          * - Not less then NFS4_POLL_RETRY_MIN.
6293                          * - One last time a jiffie before we give up
6294                          * - exponential backoff (time_now minus start_attempt)
6295                          */
6296                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
6297                                     min((giveup - now - 1),
6298                                         now - lgp->args.timestamp));
6299
6300                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
6301                                 __func__, delay);
6302                         rpc_delay(task, delay);
6303                         task->tk_status = 0;
6304                         rpc_restart_call_prepare(task);
6305                         goto out; /* Do not call nfs4_async_handle_error() */
6306                 }
6307                 break;
6308         case -NFS4ERR_EXPIRED:
6309         case -NFS4ERR_BAD_STATEID:
6310                 spin_lock(&inode->i_lock);
6311                 lo = NFS_I(inode)->layout;
6312                 if (!lo || list_empty(&lo->plh_segs)) {
6313                         spin_unlock(&inode->i_lock);
6314                         /* If the open stateid was bad, then recover it. */
6315                         state = lgp->args.ctx->state;
6316                 } else {
6317                         LIST_HEAD(head);
6318
6319                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6320                         spin_unlock(&inode->i_lock);
6321                         /* Mark the bad layout state as invalid, then
6322                          * retry using the open stateid. */
6323                         pnfs_free_lseg_list(&head);
6324                 }
6325         }
6326         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6327                 rpc_restart_call_prepare(task);
6328 out:
6329         dprintk("<-- %s\n", __func__);
6330 }
6331
6332 static size_t max_response_pages(struct nfs_server *server)
6333 {
6334         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6335         return nfs_page_array_len(0, max_resp_sz);
6336 }
6337
6338 static void nfs4_free_pages(struct page **pages, size_t size)
6339 {
6340         int i;
6341
6342         if (!pages)
6343                 return;
6344
6345         for (i = 0; i < size; i++) {
6346                 if (!pages[i])
6347                         break;
6348                 __free_page(pages[i]);
6349         }
6350         kfree(pages);
6351 }
6352
6353 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6354 {
6355         struct page **pages;
6356         int i;
6357
6358         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6359         if (!pages) {
6360                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6361                 return NULL;
6362         }
6363
6364         for (i = 0; i < size; i++) {
6365                 pages[i] = alloc_page(gfp_flags);
6366                 if (!pages[i]) {
6367                         dprintk("%s: failed to allocate page\n", __func__);
6368                         nfs4_free_pages(pages, size);
6369                         return NULL;
6370                 }
6371         }
6372
6373         return pages;
6374 }
6375
6376 static void nfs4_layoutget_release(void *calldata)
6377 {
6378         struct nfs4_layoutget *lgp = calldata;
6379         struct inode *inode = lgp->args.inode;
6380         struct nfs_server *server = NFS_SERVER(inode);
6381         size_t max_pages = max_response_pages(server);
6382
6383         dprintk("--> %s\n", __func__);
6384         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6385         pnfs_put_layout_hdr(NFS_I(inode)->layout);
6386         put_nfs_open_context(lgp->args.ctx);
6387         kfree(calldata);
6388         dprintk("<-- %s\n", __func__);
6389 }
6390
6391 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6392         .rpc_call_prepare = nfs4_layoutget_prepare,
6393         .rpc_call_done = nfs4_layoutget_done,
6394         .rpc_release = nfs4_layoutget_release,
6395 };
6396
6397 struct pnfs_layout_segment *
6398 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6399 {
6400         struct inode *inode = lgp->args.inode;
6401         struct nfs_server *server = NFS_SERVER(inode);
6402         size_t max_pages = max_response_pages(server);
6403         struct rpc_task *task;
6404         struct rpc_message msg = {
6405                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6406                 .rpc_argp = &lgp->args,
6407                 .rpc_resp = &lgp->res,
6408         };
6409         struct rpc_task_setup task_setup_data = {
6410                 .rpc_client = server->client,
6411                 .rpc_message = &msg,
6412                 .callback_ops = &nfs4_layoutget_call_ops,
6413                 .callback_data = lgp,
6414                 .flags = RPC_TASK_ASYNC,
6415         };
6416         struct pnfs_layout_segment *lseg = NULL;
6417         int status = 0;
6418
6419         dprintk("--> %s\n", __func__);
6420
6421         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
6422         pnfs_get_layout_hdr(NFS_I(inode)->layout);
6423
6424         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6425         if (!lgp->args.layout.pages) {
6426                 nfs4_layoutget_release(lgp);
6427                 return ERR_PTR(-ENOMEM);
6428         }
6429         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6430         lgp->args.timestamp = jiffies;
6431
6432         lgp->res.layoutp = &lgp->args.layout;
6433         lgp->res.seq_res.sr_slot = NULL;
6434         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6435
6436         task = rpc_run_task(&task_setup_data);
6437         if (IS_ERR(task))
6438                 return ERR_CAST(task);
6439         status = nfs4_wait_for_completion_rpc_task(task);
6440         if (status == 0)
6441                 status = task->tk_status;
6442         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
6443         if (status == 0 && lgp->res.layoutp->len)
6444                 lseg = pnfs_layout_process(lgp);
6445         rpc_put_task(task);
6446         dprintk("<-- %s status=%d\n", __func__, status);
6447         if (status)
6448                 return ERR_PTR(status);
6449         return lseg;
6450 }
6451
6452 static void
6453 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6454 {
6455         struct nfs4_layoutreturn *lrp = calldata;
6456
6457         dprintk("--> %s\n", __func__);
6458         nfs41_setup_sequence(lrp->clp->cl_session,
6459                         &lrp->args.seq_args,
6460                         &lrp->res.seq_res,
6461                         task);
6462 }
6463
6464 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6465 {
6466         struct nfs4_layoutreturn *lrp = calldata;
6467         struct nfs_server *server;
6468
6469         dprintk("--> %s\n", __func__);
6470
6471         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
6472                 return;
6473
6474         server = NFS_SERVER(lrp->args.inode);
6475         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6476                 rpc_restart_call_prepare(task);
6477                 return;
6478         }
6479         dprintk("<-- %s\n", __func__);
6480 }
6481
6482 static void nfs4_layoutreturn_release(void *calldata)
6483 {
6484         struct nfs4_layoutreturn *lrp = calldata;
6485         struct pnfs_layout_hdr *lo = lrp->args.layout;
6486
6487         dprintk("--> %s\n", __func__);
6488         spin_lock(&lo->plh_inode->i_lock);
6489         if (lrp->res.lrs_present)
6490                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6491         lo->plh_block_lgets--;
6492         spin_unlock(&lo->plh_inode->i_lock);
6493         pnfs_put_layout_hdr(lrp->args.layout);
6494         kfree(calldata);
6495         dprintk("<-- %s\n", __func__);
6496 }
6497
6498 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6499         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6500         .rpc_call_done = nfs4_layoutreturn_done,
6501         .rpc_release = nfs4_layoutreturn_release,
6502 };
6503
6504 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6505 {
6506         struct rpc_task *task;
6507         struct rpc_message msg = {
6508                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6509                 .rpc_argp = &lrp->args,
6510                 .rpc_resp = &lrp->res,
6511         };
6512         struct rpc_task_setup task_setup_data = {
6513                 .rpc_client = lrp->clp->cl_rpcclient,
6514                 .rpc_message = &msg,
6515                 .callback_ops = &nfs4_layoutreturn_call_ops,
6516                 .callback_data = lrp,
6517         };
6518         int status;
6519
6520         dprintk("--> %s\n", __func__);
6521         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6522         task = rpc_run_task(&task_setup_data);
6523         if (IS_ERR(task))
6524                 return PTR_ERR(task);
6525         status = task->tk_status;
6526         dprintk("<-- %s status=%d\n", __func__, status);
6527         rpc_put_task(task);
6528         return status;
6529 }
6530
6531 /*
6532  * Retrieve the list of Data Server devices from the MDS.
6533  */
6534 static int _nfs4_getdevicelist(struct nfs_server *server,
6535                                     const struct nfs_fh *fh,
6536                                     struct pnfs_devicelist *devlist)
6537 {
6538         struct nfs4_getdevicelist_args args = {
6539                 .fh = fh,
6540                 .layoutclass = server->pnfs_curr_ld->id,
6541         };
6542         struct nfs4_getdevicelist_res res = {
6543                 .devlist = devlist,
6544         };
6545         struct rpc_message msg = {
6546                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6547                 .rpc_argp = &args,
6548                 .rpc_resp = &res,
6549         };
6550         int status;
6551
6552         dprintk("--> %s\n", __func__);
6553         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6554                                 &res.seq_res, 0);
6555         dprintk("<-- %s status=%d\n", __func__, status);
6556         return status;
6557 }
6558
6559 int nfs4_proc_getdevicelist(struct nfs_server *server,
6560                             const struct nfs_fh *fh,
6561                             struct pnfs_devicelist *devlist)
6562 {
6563         struct nfs4_exception exception = { };
6564         int err;
6565
6566         do {
6567                 err = nfs4_handle_exception(server,
6568                                 _nfs4_getdevicelist(server, fh, devlist),
6569                                 &exception);
6570         } while (exception.retry);
6571
6572         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6573                 err, devlist->num_devs);
6574
6575         return err;
6576 }
6577 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6578
6579 static int
6580 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6581 {
6582         struct nfs4_getdeviceinfo_args args = {
6583                 .pdev = pdev,
6584         };
6585         struct nfs4_getdeviceinfo_res res = {
6586                 .pdev = pdev,
6587         };
6588         struct rpc_message msg = {
6589                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6590                 .rpc_argp = &args,
6591                 .rpc_resp = &res,
6592         };
6593         int status;
6594
6595         dprintk("--> %s\n", __func__);
6596         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6597         dprintk("<-- %s status=%d\n", __func__, status);
6598
6599         return status;
6600 }
6601
6602 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6603 {
6604         struct nfs4_exception exception = { };
6605         int err;
6606
6607         do {
6608                 err = nfs4_handle_exception(server,
6609                                         _nfs4_proc_getdeviceinfo(server, pdev),
6610                                         &exception);
6611         } while (exception.retry);
6612         return err;
6613 }
6614 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6615
6616 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6617 {
6618         struct nfs4_layoutcommit_data *data = calldata;
6619         struct nfs_server *server = NFS_SERVER(data->args.inode);
6620         struct nfs4_session *session = nfs4_get_session(server);
6621
6622         nfs41_setup_sequence(session,
6623                         &data->args.seq_args,
6624                         &data->res.seq_res,
6625                         task);
6626 }
6627
6628 static void
6629 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6630 {
6631         struct nfs4_layoutcommit_data *data = calldata;
6632         struct nfs_server *server = NFS_SERVER(data->args.inode);
6633
6634         if (!nfs41_sequence_done(task, &data->res.seq_res))
6635                 return;
6636
6637         switch (task->tk_status) { /* Just ignore these failures */
6638         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6639         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6640         case -NFS4ERR_BADLAYOUT:     /* no layout */
6641         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6642                 task->tk_status = 0;
6643                 break;
6644         case 0:
6645                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6646                                                    data->res.fattr);
6647                 break;
6648         default:
6649                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6650                         rpc_restart_call_prepare(task);
6651                         return;
6652                 }
6653         }
6654 }
6655
6656 static void nfs4_layoutcommit_release(void *calldata)
6657 {
6658         struct nfs4_layoutcommit_data *data = calldata;
6659
6660         pnfs_cleanup_layoutcommit(data);
6661         put_rpccred(data->cred);
6662         kfree(data);
6663 }
6664
6665 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6666         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6667         .rpc_call_done = nfs4_layoutcommit_done,
6668         .rpc_release = nfs4_layoutcommit_release,
6669 };
6670
6671 int
6672 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6673 {
6674         struct rpc_message msg = {
6675                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6676                 .rpc_argp = &data->args,
6677                 .rpc_resp = &data->res,
6678                 .rpc_cred = data->cred,
6679         };
6680         struct rpc_task_setup task_setup_data = {
6681                 .task = &data->task,
6682                 .rpc_client = NFS_CLIENT(data->args.inode),
6683                 .rpc_message = &msg,
6684                 .callback_ops = &nfs4_layoutcommit_ops,
6685                 .callback_data = data,
6686                 .flags = RPC_TASK_ASYNC,
6687         };
6688         struct rpc_task *task;
6689         int status = 0;
6690
6691         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6692                 "lbw: %llu inode %lu\n",
6693                 data->task.tk_pid, sync,
6694                 data->args.lastbytewritten,
6695                 data->args.inode->i_ino);
6696
6697         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6698         task = rpc_run_task(&task_setup_data);
6699         if (IS_ERR(task))
6700                 return PTR_ERR(task);
6701         if (sync == false)
6702                 goto out;
6703         status = nfs4_wait_for_completion_rpc_task(task);
6704         if (status != 0)
6705                 goto out;
6706         status = task->tk_status;
6707 out:
6708         dprintk("%s: status %d\n", __func__, status);
6709         rpc_put_task(task);
6710         return status;
6711 }
6712
6713 static int
6714 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6715                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6716 {
6717         struct nfs41_secinfo_no_name_args args = {
6718                 .style = SECINFO_STYLE_CURRENT_FH,
6719         };
6720         struct nfs4_secinfo_res res = {
6721                 .flavors = flavors,
6722         };
6723         struct rpc_message msg = {
6724                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6725                 .rpc_argp = &args,
6726                 .rpc_resp = &res,
6727         };
6728         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6729 }
6730
6731 static int
6732 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6733                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6734 {
6735         struct nfs4_exception exception = { };
6736         int err;
6737         do {
6738                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6739                 switch (err) {
6740                 case 0:
6741                 case -NFS4ERR_WRONGSEC:
6742                 case -ENOTSUPP:
6743                         goto out;
6744                 default:
6745                         err = nfs4_handle_exception(server, err, &exception);
6746                 }
6747         } while (exception.retry);
6748 out:
6749         return err;
6750 }
6751
6752 static int
6753 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6754                     struct nfs_fsinfo *info)
6755 {
6756         int err;
6757         struct page *page;
6758         rpc_authflavor_t flavor;
6759         struct nfs4_secinfo_flavors *flavors;
6760
6761         page = alloc_page(GFP_KERNEL);
6762         if (!page) {
6763                 err = -ENOMEM;
6764                 goto out;
6765         }
6766
6767         flavors = page_address(page);
6768         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6769
6770         /*
6771          * Fall back on "guess and check" method if
6772          * the server doesn't support SECINFO_NO_NAME
6773          */
6774         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
6775                 err = nfs4_find_root_sec(server, fhandle, info);
6776                 goto out_freepage;
6777         }
6778         if (err)
6779                 goto out_freepage;
6780
6781         flavor = nfs_find_best_sec(flavors);
6782         if (err == 0)
6783                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6784
6785 out_freepage:
6786         put_page(page);
6787         if (err == -EACCES)
6788                 return -EPERM;
6789 out:
6790         return err;
6791 }
6792
6793 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6794 {
6795         int status;
6796         struct nfs41_test_stateid_args args = {
6797                 .stateid = stateid,
6798         };
6799         struct nfs41_test_stateid_res res;
6800         struct rpc_message msg = {
6801                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6802                 .rpc_argp = &args,
6803                 .rpc_resp = &res,
6804         };
6805
6806         dprintk("NFS call  test_stateid %p\n", stateid);
6807         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6808         nfs4_set_sequence_privileged(&args.seq_args);
6809         status = nfs4_call_sync_sequence(server->client, server, &msg,
6810                         &args.seq_args, &res.seq_res);
6811         if (status != NFS_OK) {
6812                 dprintk("NFS reply test_stateid: failed, %d\n", status);
6813                 return status;
6814         }
6815         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
6816         return -res.status;
6817 }
6818
6819 /**
6820  * nfs41_test_stateid - perform a TEST_STATEID operation
6821  *
6822  * @server: server / transport on which to perform the operation
6823  * @stateid: state ID to test
6824  *
6825  * Returns NFS_OK if the server recognizes that "stateid" is valid.
6826  * Otherwise a negative NFS4ERR value is returned if the operation
6827  * failed or the state ID is not currently valid.
6828  */
6829 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6830 {
6831         struct nfs4_exception exception = { };
6832         int err;
6833         do {
6834                 err = _nfs41_test_stateid(server, stateid);
6835                 if (err != -NFS4ERR_DELAY)
6836                         break;
6837                 nfs4_handle_exception(server, err, &exception);
6838         } while (exception.retry);
6839         return err;
6840 }
6841
6842 struct nfs_free_stateid_data {
6843         struct nfs_server *server;
6844         struct nfs41_free_stateid_args args;
6845         struct nfs41_free_stateid_res res;
6846 };
6847
6848 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
6849 {
6850         struct nfs_free_stateid_data *data = calldata;
6851         nfs41_setup_sequence(nfs4_get_session(data->server),
6852                         &data->args.seq_args,
6853                         &data->res.seq_res,
6854                         task);
6855 }
6856
6857 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
6858 {
6859         struct nfs_free_stateid_data *data = calldata;
6860
6861         nfs41_sequence_done(task, &data->res.seq_res);
6862
6863         switch (task->tk_status) {
6864         case -NFS4ERR_DELAY:
6865                 if (nfs4_async_handle_error(task, data->server, NULL) == -EAGAIN)
6866                         rpc_restart_call_prepare(task);
6867         }
6868 }
6869
6870 static void nfs41_free_stateid_release(void *calldata)
6871 {
6872         kfree(calldata);
6873 }
6874
6875 const struct rpc_call_ops nfs41_free_stateid_ops = {
6876         .rpc_call_prepare = nfs41_free_stateid_prepare,
6877         .rpc_call_done = nfs41_free_stateid_done,
6878         .rpc_release = nfs41_free_stateid_release,
6879 };
6880
6881 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
6882                 nfs4_stateid *stateid,
6883                 bool privileged)
6884 {
6885         struct rpc_message msg = {
6886                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6887         };
6888         struct rpc_task_setup task_setup = {
6889                 .rpc_client = server->client,
6890                 .rpc_message = &msg,
6891                 .callback_ops = &nfs41_free_stateid_ops,
6892                 .flags = RPC_TASK_ASYNC,
6893         };
6894         struct nfs_free_stateid_data *data;
6895
6896         dprintk("NFS call  free_stateid %p\n", stateid);
6897         data = kmalloc(sizeof(*data), GFP_NOFS);
6898         if (!data)
6899                 return ERR_PTR(-ENOMEM);
6900         data->server = server;
6901         nfs4_stateid_copy(&data->args.stateid, stateid);
6902
6903         task_setup.callback_data = data;
6904
6905         msg.rpc_argp = &data->args;
6906         msg.rpc_resp = &data->res;
6907         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6908         if (privileged)
6909                 nfs4_set_sequence_privileged(&data->args.seq_args);
6910
6911         return rpc_run_task(&task_setup);
6912 }
6913
6914 /**
6915  * nfs41_free_stateid - perform a FREE_STATEID operation
6916  *
6917  * @server: server / transport on which to perform the operation
6918  * @stateid: state ID to release
6919  *
6920  * Returns NFS_OK if the server freed "stateid".  Otherwise a
6921  * negative NFS4ERR value is returned.
6922  */
6923 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6924 {
6925         struct rpc_task *task;
6926         int ret;
6927
6928         task = _nfs41_free_stateid(server, stateid, true);
6929         if (IS_ERR(task))
6930                 return PTR_ERR(task);
6931         ret = rpc_wait_for_completion_task(task);
6932         if (!ret)
6933                 ret = task->tk_status;
6934         rpc_put_task(task);
6935         return ret;
6936 }
6937
6938 static int nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
6939 {
6940         struct rpc_task *task;
6941
6942         task = _nfs41_free_stateid(server, &lsp->ls_stateid, false);
6943         nfs4_free_lock_state(server, lsp);
6944         if (IS_ERR(task))
6945                 return PTR_ERR(task);
6946         rpc_put_task(task);
6947         return 0;
6948 }
6949
6950 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6951                 const nfs4_stateid *s2)
6952 {
6953         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6954                 return false;
6955
6956         if (s1->seqid == s2->seqid)
6957                 return true;
6958         if (s1->seqid == 0 || s2->seqid == 0)
6959                 return true;
6960
6961         return false;
6962 }
6963
6964 #endif /* CONFIG_NFS_V4_1 */
6965
6966 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6967                 const nfs4_stateid *s2)
6968 {
6969         return nfs4_stateid_match(s1, s2);
6970 }
6971
6972
6973 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6974         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6975         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6976         .recover_open   = nfs4_open_reclaim,
6977         .recover_lock   = nfs4_lock_reclaim,
6978         .establish_clid = nfs4_init_clientid,
6979         .get_clid_cred  = nfs4_get_setclientid_cred,
6980         .detect_trunking = nfs40_discover_server_trunking,
6981 };
6982
6983 #if defined(CONFIG_NFS_V4_1)
6984 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6985         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6986         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6987         .recover_open   = nfs4_open_reclaim,
6988         .recover_lock   = nfs4_lock_reclaim,
6989         .establish_clid = nfs41_init_clientid,
6990         .get_clid_cred  = nfs4_get_exchange_id_cred,
6991         .reclaim_complete = nfs41_proc_reclaim_complete,
6992         .detect_trunking = nfs41_discover_server_trunking,
6993 };
6994 #endif /* CONFIG_NFS_V4_1 */
6995
6996 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6997         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6998         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6999         .recover_open   = nfs40_open_expired,
7000         .recover_lock   = nfs4_lock_expired,
7001         .establish_clid = nfs4_init_clientid,
7002         .get_clid_cred  = nfs4_get_setclientid_cred,
7003 };
7004
7005 #if defined(CONFIG_NFS_V4_1)
7006 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
7007         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
7008         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
7009         .recover_open   = nfs41_open_expired,
7010         .recover_lock   = nfs41_lock_expired,
7011         .establish_clid = nfs41_init_clientid,
7012         .get_clid_cred  = nfs4_get_exchange_id_cred,
7013 };
7014 #endif /* CONFIG_NFS_V4_1 */
7015
7016 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
7017         .sched_state_renewal = nfs4_proc_async_renew,
7018         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
7019         .renew_lease = nfs4_proc_renew,
7020 };
7021
7022 #if defined(CONFIG_NFS_V4_1)
7023 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
7024         .sched_state_renewal = nfs41_proc_async_sequence,
7025         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
7026         .renew_lease = nfs4_proc_sequence,
7027 };
7028 #endif
7029
7030 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
7031         .minor_version = 0,
7032         .init_caps = NFS_CAP_READDIRPLUS
7033                 | NFS_CAP_ATOMIC_OPEN
7034                 | NFS_CAP_CHANGE_ATTR
7035                 | NFS_CAP_POSIX_LOCK,
7036         .call_sync = _nfs4_call_sync,
7037         .match_stateid = nfs4_match_stateid,
7038         .find_root_sec = nfs4_find_root_sec,
7039         .free_lock_state = nfs4_release_lockowner,
7040         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
7041         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
7042         .state_renewal_ops = &nfs40_state_renewal_ops,
7043 };
7044
7045 #if defined(CONFIG_NFS_V4_1)
7046 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
7047         .minor_version = 1,
7048         .init_caps = NFS_CAP_READDIRPLUS
7049                 | NFS_CAP_ATOMIC_OPEN
7050                 | NFS_CAP_CHANGE_ATTR
7051                 | NFS_CAP_POSIX_LOCK
7052                 | NFS_CAP_STATEID_NFSV41
7053                 | NFS_CAP_ATOMIC_OPEN_V1,
7054         .call_sync = nfs4_call_sync_sequence,
7055         .match_stateid = nfs41_match_stateid,
7056         .find_root_sec = nfs41_find_root_sec,
7057         .free_lock_state = nfs41_free_lock_state,
7058         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
7059         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
7060         .state_renewal_ops = &nfs41_state_renewal_ops,
7061 };
7062 #endif
7063
7064 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
7065         [0] = &nfs_v4_0_minor_ops,
7066 #if defined(CONFIG_NFS_V4_1)
7067         [1] = &nfs_v4_1_minor_ops,
7068 #endif
7069 };
7070
7071 const struct inode_operations nfs4_dir_inode_operations = {
7072         .create         = nfs_create,
7073         .lookup         = nfs_lookup,
7074         .atomic_open    = nfs_atomic_open,
7075         .link           = nfs_link,
7076         .unlink         = nfs_unlink,
7077         .symlink        = nfs_symlink,
7078         .mkdir          = nfs_mkdir,
7079         .rmdir          = nfs_rmdir,
7080         .mknod          = nfs_mknod,
7081         .rename         = nfs_rename,
7082         .permission     = nfs_permission,
7083         .getattr        = nfs_getattr,
7084         .setattr        = nfs_setattr,
7085         .getxattr       = generic_getxattr,
7086         .setxattr       = generic_setxattr,
7087         .listxattr      = generic_listxattr,
7088         .removexattr    = generic_removexattr,
7089 };
7090
7091 static const struct inode_operations nfs4_file_inode_operations = {
7092         .permission     = nfs_permission,
7093         .getattr        = nfs_getattr,
7094         .setattr        = nfs_setattr,
7095         .getxattr       = generic_getxattr,
7096         .setxattr       = generic_setxattr,
7097         .listxattr      = generic_listxattr,
7098         .removexattr    = generic_removexattr,
7099 };
7100
7101 const struct nfs_rpc_ops nfs_v4_clientops = {
7102         .version        = 4,                    /* protocol version */
7103         .dentry_ops     = &nfs4_dentry_operations,
7104         .dir_inode_ops  = &nfs4_dir_inode_operations,
7105         .file_inode_ops = &nfs4_file_inode_operations,
7106         .file_ops       = &nfs4_file_operations,
7107         .getroot        = nfs4_proc_get_root,
7108         .submount       = nfs4_submount,
7109         .try_mount      = nfs4_try_mount,
7110         .getattr        = nfs4_proc_getattr,
7111         .setattr        = nfs4_proc_setattr,
7112         .lookup         = nfs4_proc_lookup,
7113         .access         = nfs4_proc_access,
7114         .readlink       = nfs4_proc_readlink,
7115         .create         = nfs4_proc_create,
7116         .remove         = nfs4_proc_remove,
7117         .unlink_setup   = nfs4_proc_unlink_setup,
7118         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
7119         .unlink_done    = nfs4_proc_unlink_done,
7120         .rename         = nfs4_proc_rename,
7121         .rename_setup   = nfs4_proc_rename_setup,
7122         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
7123         .rename_done    = nfs4_proc_rename_done,
7124         .link           = nfs4_proc_link,
7125         .symlink        = nfs4_proc_symlink,
7126         .mkdir          = nfs4_proc_mkdir,
7127         .rmdir          = nfs4_proc_remove,
7128         .readdir        = nfs4_proc_readdir,
7129         .mknod          = nfs4_proc_mknod,
7130         .statfs         = nfs4_proc_statfs,
7131         .fsinfo         = nfs4_proc_fsinfo,
7132         .pathconf       = nfs4_proc_pathconf,
7133         .set_capabilities = nfs4_server_capabilities,
7134         .decode_dirent  = nfs4_decode_dirent,
7135         .read_setup     = nfs4_proc_read_setup,
7136         .read_pageio_init = pnfs_pageio_init_read,
7137         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
7138         .read_done      = nfs4_read_done,
7139         .write_setup    = nfs4_proc_write_setup,
7140         .write_pageio_init = pnfs_pageio_init_write,
7141         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
7142         .write_done     = nfs4_write_done,
7143         .commit_setup   = nfs4_proc_commit_setup,
7144         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
7145         .commit_done    = nfs4_commit_done,
7146         .lock           = nfs4_proc_lock,
7147         .clear_acl_cache = nfs4_zap_acl_attr,
7148         .close_context  = nfs4_close_context,
7149         .open_context   = nfs4_atomic_open,
7150         .have_delegation = nfs4_have_delegation,
7151         .return_delegation = nfs4_inode_return_delegation,
7152         .alloc_client   = nfs4_alloc_client,
7153         .init_client    = nfs4_init_client,
7154         .free_client    = nfs4_free_client,
7155         .create_server  = nfs4_create_server,
7156         .clone_server   = nfs_clone_server,
7157 };
7158
7159 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7160         .prefix = XATTR_NAME_NFSV4_ACL,
7161         .list   = nfs4_xattr_list_nfs4_acl,
7162         .get    = nfs4_xattr_get_nfs4_acl,
7163         .set    = nfs4_xattr_set_nfs4_acl,
7164 };
7165
7166 const struct xattr_handler *nfs4_xattr_handlers[] = {
7167         &nfs4_xattr_nfs4_acl_handler,
7168         NULL
7169 };
7170
7171 /*
7172  * Local variables:
7173  *  c-basic-offset: 8
7174  * End:
7175  */