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 0;
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 #if defined(CONFIG_NFS_V4_1)
1849 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1850 {
1851         struct nfs_server *server = NFS_SERVER(state->inode);
1852         nfs4_stateid *stateid = &state->stateid;
1853         int status;
1854
1855         /* If a state reset has been done, test_stateid is unneeded */
1856         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1857                 return;
1858
1859         status = nfs41_test_stateid(server, stateid);
1860         if (status != NFS_OK) {
1861                 /* Free the stateid unless the server explicitly
1862                  * informs us the stateid is unrecognized. */
1863                 if (status != -NFS4ERR_BAD_STATEID)
1864                         nfs41_free_stateid(server, stateid);
1865                 nfs_remove_bad_delegation(state->inode);
1866
1867                 write_seqlock(&state->seqlock);
1868                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1869                 write_sequnlock(&state->seqlock);
1870                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1871         }
1872 }
1873
1874 /**
1875  * nfs41_check_open_stateid - possibly free an open stateid
1876  *
1877  * @state: NFSv4 state for an inode
1878  *
1879  * Returns NFS_OK if recovery for this stateid is now finished.
1880  * Otherwise a negative NFS4ERR value is returned.
1881  */
1882 static int nfs41_check_open_stateid(struct nfs4_state *state)
1883 {
1884         struct nfs_server *server = NFS_SERVER(state->inode);
1885         nfs4_stateid *stateid = &state->open_stateid;
1886         int status;
1887
1888         /* If a state reset has been done, test_stateid is unneeded */
1889         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1890             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1891             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1892                 return -NFS4ERR_BAD_STATEID;
1893
1894         status = nfs41_test_stateid(server, stateid);
1895         if (status != NFS_OK) {
1896                 /* Free the stateid unless the server explicitly
1897                  * informs us the stateid is unrecognized. */
1898                 if (status != -NFS4ERR_BAD_STATEID)
1899                         nfs41_free_stateid(server, stateid);
1900
1901                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1902                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1903                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1904                 clear_bit(NFS_OPEN_STATE, &state->flags);
1905         }
1906         return status;
1907 }
1908
1909 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1910 {
1911         int status;
1912
1913         nfs41_clear_delegation_stateid(state);
1914         status = nfs41_check_open_stateid(state);
1915         if (status != NFS_OK)
1916                 status = nfs4_open_expired(sp, state);
1917         return status;
1918 }
1919 #endif
1920
1921 /*
1922  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1923  * fields corresponding to attributes that were used to store the verifier.
1924  * Make sure we clobber those fields in the later setattr call
1925  */
1926 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1927 {
1928         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1929             !(sattr->ia_valid & ATTR_ATIME_SET))
1930                 sattr->ia_valid |= ATTR_ATIME;
1931
1932         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1933             !(sattr->ia_valid & ATTR_MTIME_SET))
1934                 sattr->ia_valid |= ATTR_MTIME;
1935 }
1936
1937 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
1938                 fmode_t fmode,
1939                 int flags,
1940                 struct nfs4_state **res)
1941 {
1942         struct nfs4_state_owner *sp = opendata->owner;
1943         struct nfs_server *server = sp->so_server;
1944         struct nfs4_state *state;
1945         unsigned int seq;
1946         int ret;
1947
1948         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
1949
1950         ret = _nfs4_proc_open(opendata);
1951         if (ret != 0)
1952                 goto out;
1953
1954         state = nfs4_opendata_to_nfs4_state(opendata);
1955         ret = PTR_ERR(state);
1956         if (IS_ERR(state))
1957                 goto out;
1958         if (server->caps & NFS_CAP_POSIX_LOCK)
1959                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1960
1961         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
1962         if (ret != 0)
1963                 goto out;
1964
1965         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
1966                 nfs4_schedule_stateid_recovery(server, state);
1967         *res = state;
1968 out:
1969         return ret;
1970 }
1971
1972 /*
1973  * Returns a referenced nfs4_state
1974  */
1975 static int _nfs4_do_open(struct inode *dir,
1976                         struct dentry *dentry,
1977                         fmode_t fmode,
1978                         int flags,
1979                         struct iattr *sattr,
1980                         struct rpc_cred *cred,
1981                         struct nfs4_state **res,
1982                         struct nfs4_threshold **ctx_th)
1983 {
1984         struct nfs4_state_owner  *sp;
1985         struct nfs4_state     *state = NULL;
1986         struct nfs_server       *server = NFS_SERVER(dir);
1987         struct nfs4_opendata *opendata;
1988         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
1989         int status;
1990
1991         /* Protect against reboot recovery conflicts */
1992         status = -ENOMEM;
1993         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1994         if (sp == NULL) {
1995                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1996                 goto out_err;
1997         }
1998         status = nfs4_recover_expired_lease(server);
1999         if (status != 0)
2000                 goto err_put_state_owner;
2001         if (dentry->d_inode != NULL)
2002                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2003         status = -ENOMEM;
2004         if (dentry->d_inode)
2005                 claim = NFS4_OPEN_CLAIM_FH;
2006         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2007                         claim, GFP_KERNEL);
2008         if (opendata == NULL)
2009                 goto err_put_state_owner;
2010
2011         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2012                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2013                 if (!opendata->f_attr.mdsthreshold)
2014                         goto err_opendata_put;
2015                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2016         }
2017         if (dentry->d_inode != NULL)
2018                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2019
2020         status = _nfs4_open_and_get_state(opendata, fmode, flags, &state);
2021         if (status != 0)
2022                 goto err_opendata_put;
2023
2024         if ((opendata->o_arg.open_flags & O_EXCL) &&
2025             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2026                 nfs4_exclusive_attrset(opendata, sattr);
2027
2028                 nfs_fattr_init(opendata->o_res.f_attr);
2029                 status = nfs4_do_setattr(state->inode, cred,
2030                                 opendata->o_res.f_attr, sattr,
2031                                 state);
2032                 if (status == 0)
2033                         nfs_setattr_update_inode(state->inode, sattr);
2034                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2035         }
2036
2037         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2038                 *ctx_th = opendata->f_attr.mdsthreshold;
2039         else
2040                 kfree(opendata->f_attr.mdsthreshold);
2041         opendata->f_attr.mdsthreshold = NULL;
2042
2043         nfs4_opendata_put(opendata);
2044         nfs4_put_state_owner(sp);
2045         *res = state;
2046         return 0;
2047 err_opendata_put:
2048         kfree(opendata->f_attr.mdsthreshold);
2049         nfs4_opendata_put(opendata);
2050 err_put_state_owner:
2051         nfs4_put_state_owner(sp);
2052 out_err:
2053         *res = NULL;
2054         return status;
2055 }
2056
2057
2058 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2059                                         struct dentry *dentry,
2060                                         fmode_t fmode,
2061                                         int flags,
2062                                         struct iattr *sattr,
2063                                         struct rpc_cred *cred,
2064                                         struct nfs4_threshold **ctx_th)
2065 {
2066         struct nfs_server *server = NFS_SERVER(dir);
2067         struct nfs4_exception exception = { };
2068         struct nfs4_state *res;
2069         int status;
2070
2071         fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
2072         do {
2073                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
2074                                        &res, ctx_th);
2075                 if (status == 0)
2076                         break;
2077                 /* NOTE: BAD_SEQID means the server and client disagree about the
2078                  * book-keeping w.r.t. state-changing operations
2079                  * (OPEN/CLOSE/LOCK/LOCKU...)
2080                  * It is actually a sign of a bug on the client or on the server.
2081                  *
2082                  * If we receive a BAD_SEQID error in the particular case of
2083                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2084                  * have unhashed the old state_owner for us, and that we can
2085                  * therefore safely retry using a new one. We should still warn
2086                  * the user though...
2087                  */
2088                 if (status == -NFS4ERR_BAD_SEQID) {
2089                         pr_warn_ratelimited("NFS: v4 server %s "
2090                                         " returned a bad sequence-id error!\n",
2091                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2092                         exception.retry = 1;
2093                         continue;
2094                 }
2095                 /*
2096                  * BAD_STATEID on OPEN means that the server cancelled our
2097                  * state before it received the OPEN_CONFIRM.
2098                  * Recover by retrying the request as per the discussion
2099                  * on Page 181 of RFC3530.
2100                  */
2101                 if (status == -NFS4ERR_BAD_STATEID) {
2102                         exception.retry = 1;
2103                         continue;
2104                 }
2105                 if (status == -EAGAIN) {
2106                         /* We must have found a delegation */
2107                         exception.retry = 1;
2108                         continue;
2109                 }
2110                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2111                         continue;
2112                 res = ERR_PTR(nfs4_handle_exception(server,
2113                                         status, &exception));
2114         } while (exception.retry);
2115         return res;
2116 }
2117
2118 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2119                             struct nfs_fattr *fattr, struct iattr *sattr,
2120                             struct nfs4_state *state)
2121 {
2122         struct nfs_server *server = NFS_SERVER(inode);
2123         struct nfs_setattrargs  arg = {
2124                 .fh             = NFS_FH(inode),
2125                 .iap            = sattr,
2126                 .server         = server,
2127                 .bitmask = server->attr_bitmask,
2128         };
2129         struct nfs_setattrres  res = {
2130                 .fattr          = fattr,
2131                 .server         = server,
2132         };
2133         struct rpc_message msg = {
2134                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2135                 .rpc_argp       = &arg,
2136                 .rpc_resp       = &res,
2137                 .rpc_cred       = cred,
2138         };
2139         unsigned long timestamp = jiffies;
2140         fmode_t fmode;
2141         bool truncate;
2142         int status;
2143
2144         nfs_fattr_init(fattr);
2145
2146         /* Servers should only apply open mode checks for file size changes */
2147         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2148         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2149
2150         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2151                 /* Use that stateid */
2152         } else if (truncate && state != NULL && nfs4_valid_open_stateid(state)) {
2153                 struct nfs_lockowner lockowner = {
2154                         .l_owner = current->files,
2155                         .l_pid = current->tgid,
2156                 };
2157                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2158                                 &lockowner);
2159         } else
2160                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2161
2162         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2163         if (status == 0 && state != NULL)
2164                 renew_lease(server, timestamp);
2165         return status;
2166 }
2167
2168 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2169                            struct nfs_fattr *fattr, struct iattr *sattr,
2170                            struct nfs4_state *state)
2171 {
2172         struct nfs_server *server = NFS_SERVER(inode);
2173         struct nfs4_exception exception = {
2174                 .state = state,
2175                 .inode = inode,
2176         };
2177         int err;
2178         do {
2179                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2180                 switch (err) {
2181                 case -NFS4ERR_OPENMODE:
2182                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2183                                 pr_warn_once("NFSv4: server %s is incorrectly "
2184                                                 "applying open mode checks to "
2185                                                 "a SETATTR that is not "
2186                                                 "changing file size.\n",
2187                                                 server->nfs_client->cl_hostname);
2188                         }
2189                         if (state && !(state->state & FMODE_WRITE)) {
2190                                 err = -EBADF;
2191                                 if (sattr->ia_valid & ATTR_OPEN)
2192                                         err = -EACCES;
2193                                 goto out;
2194                         }
2195                 }
2196                 err = nfs4_handle_exception(server, err, &exception);
2197         } while (exception.retry);
2198 out:
2199         return err;
2200 }
2201
2202 struct nfs4_closedata {
2203         struct inode *inode;
2204         struct nfs4_state *state;
2205         struct nfs_closeargs arg;
2206         struct nfs_closeres res;
2207         struct nfs_fattr fattr;
2208         unsigned long timestamp;
2209         bool roc;
2210         u32 roc_barrier;
2211 };
2212
2213 static void nfs4_free_closedata(void *data)
2214 {
2215         struct nfs4_closedata *calldata = data;
2216         struct nfs4_state_owner *sp = calldata->state->owner;
2217         struct super_block *sb = calldata->state->inode->i_sb;
2218
2219         if (calldata->roc)
2220                 pnfs_roc_release(calldata->state->inode);
2221         nfs4_put_open_state(calldata->state);
2222         nfs_free_seqid(calldata->arg.seqid);
2223         nfs4_put_state_owner(sp);
2224         nfs_sb_deactive(sb);
2225         kfree(calldata);
2226 }
2227
2228 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2229                 fmode_t fmode)
2230 {
2231         spin_lock(&state->owner->so_lock);
2232         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2233         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
2234         case FMODE_WRITE:
2235                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2236                 break;
2237         case FMODE_READ:
2238                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2239                 break;
2240         case 0:
2241                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2242                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2243                 clear_bit(NFS_OPEN_STATE, &state->flags);
2244         }
2245         spin_unlock(&state->owner->so_lock);
2246 }
2247
2248 static void nfs4_close_done(struct rpc_task *task, void *data)
2249 {
2250         struct nfs4_closedata *calldata = data;
2251         struct nfs4_state *state = calldata->state;
2252         struct nfs_server *server = NFS_SERVER(calldata->inode);
2253
2254         dprintk("%s: begin!\n", __func__);
2255         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2256                 return;
2257         /* hmm. we are done with the inode, and in the process of freeing
2258          * the state_owner. we keep this around to process errors
2259          */
2260         switch (task->tk_status) {
2261                 case 0:
2262                         if (calldata->roc)
2263                                 pnfs_roc_set_barrier(state->inode,
2264                                                      calldata->roc_barrier);
2265                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2266                         renew_lease(server, calldata->timestamp);
2267                         nfs4_close_clear_stateid_flags(state,
2268                                         calldata->arg.fmode);
2269                         break;
2270                 case -NFS4ERR_STALE_STATEID:
2271                 case -NFS4ERR_OLD_STATEID:
2272                 case -NFS4ERR_BAD_STATEID:
2273                 case -NFS4ERR_EXPIRED:
2274                         if (calldata->arg.fmode == 0)
2275                                 break;
2276                 default:
2277                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2278                                 rpc_restart_call_prepare(task);
2279         }
2280         nfs_release_seqid(calldata->arg.seqid);
2281         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2282         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2283 }
2284
2285 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2286 {
2287         struct nfs4_closedata *calldata = data;
2288         struct nfs4_state *state = calldata->state;
2289         struct inode *inode = calldata->inode;
2290         bool is_rdonly, is_wronly, is_rdwr;
2291         int call_close = 0;
2292
2293         dprintk("%s: begin!\n", __func__);
2294         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2295                 goto out_wait;
2296
2297         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2298         spin_lock(&state->owner->so_lock);
2299         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2300         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2301         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2302         /* Calculate the change in open mode */
2303         calldata->arg.fmode = 0;
2304         if (state->n_rdwr == 0) {
2305                 if (state->n_rdonly == 0)
2306                         call_close |= is_rdonly;
2307                 else if (is_rdonly)
2308                         calldata->arg.fmode |= FMODE_READ;
2309                 if (state->n_wronly == 0)
2310                         call_close |= is_wronly;
2311                 else if (is_wronly)
2312                         calldata->arg.fmode |= FMODE_WRITE;
2313         } else if (is_rdwr)
2314                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2315
2316         if (calldata->arg.fmode == 0)
2317                 call_close |= is_rdwr;
2318
2319         if (!nfs4_valid_open_stateid(state))
2320                 call_close = 0;
2321         spin_unlock(&state->owner->so_lock);
2322
2323         if (!call_close) {
2324                 /* Note: exit _without_ calling nfs4_close_done */
2325                 goto out_no_action;
2326         }
2327
2328         if (calldata->arg.fmode == 0) {
2329                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2330                 if (calldata->roc &&
2331                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2332                         nfs_release_seqid(calldata->arg.seqid);
2333                         goto out_wait;
2334                     }
2335         }
2336
2337         nfs_fattr_init(calldata->res.fattr);
2338         calldata->timestamp = jiffies;
2339         if (nfs4_setup_sequence(NFS_SERVER(inode),
2340                                 &calldata->arg.seq_args,
2341                                 &calldata->res.seq_res,
2342                                 task) != 0)
2343                 nfs_release_seqid(calldata->arg.seqid);
2344         dprintk("%s: done!\n", __func__);
2345         return;
2346 out_no_action:
2347         task->tk_action = NULL;
2348 out_wait:
2349         nfs4_sequence_done(task, &calldata->res.seq_res);
2350 }
2351
2352 static const struct rpc_call_ops nfs4_close_ops = {
2353         .rpc_call_prepare = nfs4_close_prepare,
2354         .rpc_call_done = nfs4_close_done,
2355         .rpc_release = nfs4_free_closedata,
2356 };
2357
2358 /*
2359  * It is possible for data to be read/written from a mem-mapped file
2360  * after the sys_close call (which hits the vfs layer as a flush).
2361  * This means that we can't safely call nfsv4 close on a file until
2362  * the inode is cleared. This in turn means that we are not good
2363  * NFSv4 citizens - we do not indicate to the server to update the file's
2364  * share state even when we are done with one of the three share
2365  * stateid's in the inode.
2366  *
2367  * NOTE: Caller must be holding the sp->so_owner semaphore!
2368  */
2369 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2370 {
2371         struct nfs_server *server = NFS_SERVER(state->inode);
2372         struct nfs4_closedata *calldata;
2373         struct nfs4_state_owner *sp = state->owner;
2374         struct rpc_task *task;
2375         struct rpc_message msg = {
2376                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2377                 .rpc_cred = state->owner->so_cred,
2378         };
2379         struct rpc_task_setup task_setup_data = {
2380                 .rpc_client = server->client,
2381                 .rpc_message = &msg,
2382                 .callback_ops = &nfs4_close_ops,
2383                 .workqueue = nfsiod_workqueue,
2384                 .flags = RPC_TASK_ASYNC,
2385         };
2386         int status = -ENOMEM;
2387
2388         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2389         if (calldata == NULL)
2390                 goto out;
2391         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2392         calldata->inode = state->inode;
2393         calldata->state = state;
2394         calldata->arg.fh = NFS_FH(state->inode);
2395         calldata->arg.stateid = &state->open_stateid;
2396         /* Serialization for the sequence id */
2397         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2398         if (calldata->arg.seqid == NULL)
2399                 goto out_free_calldata;
2400         calldata->arg.fmode = 0;
2401         calldata->arg.bitmask = server->cache_consistency_bitmask;
2402         calldata->res.fattr = &calldata->fattr;
2403         calldata->res.seqid = calldata->arg.seqid;
2404         calldata->res.server = server;
2405         calldata->roc = pnfs_roc(state->inode);
2406         nfs_sb_active(calldata->inode->i_sb);
2407
2408         msg.rpc_argp = &calldata->arg;
2409         msg.rpc_resp = &calldata->res;
2410         task_setup_data.callback_data = calldata;
2411         task = rpc_run_task(&task_setup_data);
2412         if (IS_ERR(task))
2413                 return PTR_ERR(task);
2414         status = 0;
2415         if (wait)
2416                 status = rpc_wait_for_completion_task(task);
2417         rpc_put_task(task);
2418         return status;
2419 out_free_calldata:
2420         kfree(calldata);
2421 out:
2422         nfs4_put_open_state(state);
2423         nfs4_put_state_owner(sp);
2424         return status;
2425 }
2426
2427 static struct inode *
2428 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2429 {
2430         struct nfs4_state *state;
2431
2432         /* Protect against concurrent sillydeletes */
2433         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2434                              ctx->cred, &ctx->mdsthreshold);
2435         if (IS_ERR(state))
2436                 return ERR_CAST(state);
2437         ctx->state = state;
2438         return igrab(state->inode);
2439 }
2440
2441 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2442 {
2443         if (ctx->state == NULL)
2444                 return;
2445         if (is_sync)
2446                 nfs4_close_sync(ctx->state, ctx->mode);
2447         else
2448                 nfs4_close_state(ctx->state, ctx->mode);
2449 }
2450
2451 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2452 {
2453         struct nfs4_server_caps_arg args = {
2454                 .fhandle = fhandle,
2455         };
2456         struct nfs4_server_caps_res res = {};
2457         struct rpc_message msg = {
2458                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2459                 .rpc_argp = &args,
2460                 .rpc_resp = &res,
2461         };
2462         int status;
2463
2464         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2465         if (status == 0) {
2466                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2467                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2468                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2469                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2470                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2471                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2472                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2473                         server->caps |= NFS_CAP_ACLS;
2474                 if (res.has_links != 0)
2475                         server->caps |= NFS_CAP_HARDLINKS;
2476                 if (res.has_symlinks != 0)
2477                         server->caps |= NFS_CAP_SYMLINKS;
2478                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2479                         server->caps |= NFS_CAP_FILEID;
2480                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2481                         server->caps |= NFS_CAP_MODE;
2482                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2483                         server->caps |= NFS_CAP_NLINK;
2484                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2485                         server->caps |= NFS_CAP_OWNER;
2486                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2487                         server->caps |= NFS_CAP_OWNER_GROUP;
2488                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2489                         server->caps |= NFS_CAP_ATIME;
2490                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2491                         server->caps |= NFS_CAP_CTIME;
2492                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2493                         server->caps |= NFS_CAP_MTIME;
2494
2495                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2496                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2497                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2498                 server->acl_bitmask = res.acl_bitmask;
2499                 server->fh_expire_type = res.fh_expire_type;
2500         }
2501
2502         return status;
2503 }
2504
2505 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2506 {
2507         struct nfs4_exception exception = { };
2508         int err;
2509         do {
2510                 err = nfs4_handle_exception(server,
2511                                 _nfs4_server_capabilities(server, fhandle),
2512                                 &exception);
2513         } while (exception.retry);
2514         return err;
2515 }
2516
2517 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2518                 struct nfs_fsinfo *info)
2519 {
2520         struct nfs4_lookup_root_arg args = {
2521                 .bitmask = nfs4_fattr_bitmap,
2522         };
2523         struct nfs4_lookup_res res = {
2524                 .server = server,
2525                 .fattr = info->fattr,
2526                 .fh = fhandle,
2527         };
2528         struct rpc_message msg = {
2529                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2530                 .rpc_argp = &args,
2531                 .rpc_resp = &res,
2532         };
2533
2534         nfs_fattr_init(info->fattr);
2535         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2536 }
2537
2538 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2539                 struct nfs_fsinfo *info)
2540 {
2541         struct nfs4_exception exception = { };
2542         int err;
2543         do {
2544                 err = _nfs4_lookup_root(server, fhandle, info);
2545                 switch (err) {
2546                 case 0:
2547                 case -NFS4ERR_WRONGSEC:
2548                         goto out;
2549                 default:
2550                         err = nfs4_handle_exception(server, err, &exception);
2551                 }
2552         } while (exception.retry);
2553 out:
2554         return err;
2555 }
2556
2557 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2558                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2559 {
2560         struct rpc_auth *auth;
2561         int ret;
2562
2563         auth = rpcauth_create(flavor, server->client);
2564         if (IS_ERR(auth)) {
2565                 ret = -EACCES;
2566                 goto out;
2567         }
2568         ret = nfs4_lookup_root(server, fhandle, info);
2569 out:
2570         return ret;
2571 }
2572
2573 /*
2574  * Retry pseudoroot lookup with various security flavors.  We do this when:
2575  *
2576  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2577  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2578  *
2579  * Returns zero on success, or a negative NFS4ERR value, or a
2580  * negative errno value.
2581  */
2582 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2583                               struct nfs_fsinfo *info)
2584 {
2585         /* Per 3530bis 15.33.5 */
2586         static const rpc_authflavor_t flav_array[] = {
2587                 RPC_AUTH_GSS_KRB5P,
2588                 RPC_AUTH_GSS_KRB5I,
2589                 RPC_AUTH_GSS_KRB5,
2590                 RPC_AUTH_UNIX,                  /* courtesy */
2591                 RPC_AUTH_NULL,
2592         };
2593         int status = -EPERM;
2594         size_t i;
2595
2596         for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
2597                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2598                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2599                         continue;
2600                 break;
2601         }
2602
2603         /*
2604          * -EACCESS could mean that the user doesn't have correct permissions
2605          * to access the mount.  It could also mean that we tried to mount
2606          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2607          * existing mount programs don't handle -EACCES very well so it should
2608          * be mapped to -EPERM instead.
2609          */
2610         if (status == -EACCES)
2611                 status = -EPERM;
2612         return status;
2613 }
2614
2615 static int nfs4_do_find_root_sec(struct nfs_server *server,
2616                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
2617 {
2618         int mv = server->nfs_client->cl_minorversion;
2619         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
2620 }
2621
2622 /**
2623  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
2624  * @server: initialized nfs_server handle
2625  * @fhandle: we fill in the pseudo-fs root file handle
2626  * @info: we fill in an FSINFO struct
2627  *
2628  * Returns zero on success, or a negative errno.
2629  */
2630 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2631                          struct nfs_fsinfo *info)
2632 {
2633         int status;
2634
2635         status = nfs4_lookup_root(server, fhandle, info);
2636         if ((status == -NFS4ERR_WRONGSEC) &&
2637             !(server->flags & NFS_MOUNT_SECFLAVOUR))
2638                 status = nfs4_do_find_root_sec(server, fhandle, info);
2639
2640         if (status == 0)
2641                 status = nfs4_server_capabilities(server, fhandle);
2642         if (status == 0)
2643                 status = nfs4_do_fsinfo(server, fhandle, info);
2644
2645         return nfs4_map_errors(status);
2646 }
2647
2648 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2649                               struct nfs_fsinfo *info)
2650 {
2651         int error;
2652         struct nfs_fattr *fattr = info->fattr;
2653
2654         error = nfs4_server_capabilities(server, mntfh);
2655         if (error < 0) {
2656                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2657                 return error;
2658         }
2659
2660         error = nfs4_proc_getattr(server, mntfh, fattr);
2661         if (error < 0) {
2662                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2663                 return error;
2664         }
2665
2666         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2667             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2668                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2669
2670         return error;
2671 }
2672
2673 /*
2674  * Get locations and (maybe) other attributes of a referral.
2675  * Note that we'll actually follow the referral later when
2676  * we detect fsid mismatch in inode revalidation
2677  */
2678 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2679                              const struct qstr *name, struct nfs_fattr *fattr,
2680                              struct nfs_fh *fhandle)
2681 {
2682         int status = -ENOMEM;
2683         struct page *page = NULL;
2684         struct nfs4_fs_locations *locations = NULL;
2685
2686         page = alloc_page(GFP_KERNEL);
2687         if (page == NULL)
2688                 goto out;
2689         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2690         if (locations == NULL)
2691                 goto out;
2692
2693         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2694         if (status != 0)
2695                 goto out;
2696         /* Make sure server returned a different fsid for the referral */
2697         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2698                 dprintk("%s: server did not return a different fsid for"
2699                         " a referral at %s\n", __func__, name->name);
2700                 status = -EIO;
2701                 goto out;
2702         }
2703         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2704         nfs_fixup_referral_attributes(&locations->fattr);
2705
2706         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2707         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2708         memset(fhandle, 0, sizeof(struct nfs_fh));
2709 out:
2710         if (page)
2711                 __free_page(page);
2712         kfree(locations);
2713         return status;
2714 }
2715
2716 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2717 {
2718         struct nfs4_getattr_arg args = {
2719                 .fh = fhandle,
2720                 .bitmask = server->attr_bitmask,
2721         };
2722         struct nfs4_getattr_res res = {
2723                 .fattr = fattr,
2724                 .server = server,
2725         };
2726         struct rpc_message msg = {
2727                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2728                 .rpc_argp = &args,
2729                 .rpc_resp = &res,
2730         };
2731
2732         nfs_fattr_init(fattr);
2733         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2734 }
2735
2736 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2737 {
2738         struct nfs4_exception exception = { };
2739         int err;
2740         do {
2741                 err = nfs4_handle_exception(server,
2742                                 _nfs4_proc_getattr(server, fhandle, fattr),
2743                                 &exception);
2744         } while (exception.retry);
2745         return err;
2746 }
2747
2748 /*
2749  * The file is not closed if it is opened due to the a request to change
2750  * the size of the file. The open call will not be needed once the
2751  * VFS layer lookup-intents are implemented.
2752  *
2753  * Close is called when the inode is destroyed.
2754  * If we haven't opened the file for O_WRONLY, we
2755  * need to in the size_change case to obtain a stateid.
2756  *
2757  * Got race?
2758  * Because OPEN is always done by name in nfsv4, it is
2759  * possible that we opened a different file by the same
2760  * name.  We can recognize this race condition, but we
2761  * can't do anything about it besides returning an error.
2762  *
2763  * This will be fixed with VFS changes (lookup-intent).
2764  */
2765 static int
2766 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2767                   struct iattr *sattr)
2768 {
2769         struct inode *inode = dentry->d_inode;
2770         struct rpc_cred *cred = NULL;
2771         struct nfs4_state *state = NULL;
2772         int status;
2773
2774         if (pnfs_ld_layoutret_on_setattr(inode))
2775                 pnfs_commit_and_return_layout(inode);
2776
2777         nfs_fattr_init(fattr);
2778
2779         /* Deal with open(O_TRUNC) */
2780         if (sattr->ia_valid & ATTR_OPEN)
2781                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2782
2783         /* Optimization: if the end result is no change, don't RPC */
2784         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2785                 return 0;
2786
2787         /* Search for an existing open(O_WRITE) file */
2788         if (sattr->ia_valid & ATTR_FILE) {
2789                 struct nfs_open_context *ctx;
2790
2791                 ctx = nfs_file_open_context(sattr->ia_file);
2792                 if (ctx) {
2793                         cred = ctx->cred;
2794                         state = ctx->state;
2795                 }
2796         }
2797
2798         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2799         if (status == 0)
2800                 nfs_setattr_update_inode(inode, sattr);
2801         return status;
2802 }
2803
2804 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2805                 const struct qstr *name, struct nfs_fh *fhandle,
2806                 struct nfs_fattr *fattr)
2807 {
2808         struct nfs_server *server = NFS_SERVER(dir);
2809         int                    status;
2810         struct nfs4_lookup_arg args = {
2811                 .bitmask = server->attr_bitmask,
2812                 .dir_fh = NFS_FH(dir),
2813                 .name = name,
2814         };
2815         struct nfs4_lookup_res res = {
2816                 .server = server,
2817                 .fattr = fattr,
2818                 .fh = fhandle,
2819         };
2820         struct rpc_message msg = {
2821                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2822                 .rpc_argp = &args,
2823                 .rpc_resp = &res,
2824         };
2825
2826         nfs_fattr_init(fattr);
2827
2828         dprintk("NFS call  lookup %s\n", name->name);
2829         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2830         dprintk("NFS reply lookup: %d\n", status);
2831         return status;
2832 }
2833
2834 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2835 {
2836         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2837                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2838         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2839         fattr->nlink = 2;
2840 }
2841
2842 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2843                                    struct qstr *name, struct nfs_fh *fhandle,
2844                                    struct nfs_fattr *fattr)
2845 {
2846         struct nfs4_exception exception = { };
2847         struct rpc_clnt *client = *clnt;
2848         int err;
2849         do {
2850                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2851                 switch (err) {
2852                 case -NFS4ERR_BADNAME:
2853                         err = -ENOENT;
2854                         goto out;
2855                 case -NFS4ERR_MOVED:
2856                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2857                         goto out;
2858                 case -NFS4ERR_WRONGSEC:
2859                         err = -EPERM;
2860                         if (client != *clnt)
2861                                 goto out;
2862
2863                         client = nfs4_create_sec_client(client, dir, name);
2864                         if (IS_ERR(client))
2865                                 return PTR_ERR(client);
2866
2867                         exception.retry = 1;
2868                         break;
2869                 default:
2870                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2871                 }
2872         } while (exception.retry);
2873
2874 out:
2875         if (err == 0)
2876                 *clnt = client;
2877         else if (client != *clnt)
2878                 rpc_shutdown_client(client);
2879
2880         return err;
2881 }
2882
2883 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2884                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2885 {
2886         int status;
2887         struct rpc_clnt *client = NFS_CLIENT(dir);
2888
2889         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2890         if (client != NFS_CLIENT(dir)) {
2891                 rpc_shutdown_client(client);
2892                 nfs_fixup_secinfo_attributes(fattr);
2893         }
2894         return status;
2895 }
2896
2897 struct rpc_clnt *
2898 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2899                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2900 {
2901         int status;
2902         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2903
2904         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2905         if (status < 0) {
2906                 rpc_shutdown_client(client);
2907                 return ERR_PTR(status);
2908         }
2909         return client;
2910 }
2911
2912 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2913 {
2914         struct nfs_server *server = NFS_SERVER(inode);
2915         struct nfs4_accessargs args = {
2916                 .fh = NFS_FH(inode),
2917                 .bitmask = server->cache_consistency_bitmask,
2918         };
2919         struct nfs4_accessres res = {
2920                 .server = server,
2921         };
2922         struct rpc_message msg = {
2923                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2924                 .rpc_argp = &args,
2925                 .rpc_resp = &res,
2926                 .rpc_cred = entry->cred,
2927         };
2928         int mode = entry->mask;
2929         int status;
2930
2931         /*
2932          * Determine which access bits we want to ask for...
2933          */
2934         if (mode & MAY_READ)
2935                 args.access |= NFS4_ACCESS_READ;
2936         if (S_ISDIR(inode->i_mode)) {
2937                 if (mode & MAY_WRITE)
2938                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2939                 if (mode & MAY_EXEC)
2940                         args.access |= NFS4_ACCESS_LOOKUP;
2941         } else {
2942                 if (mode & MAY_WRITE)
2943                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2944                 if (mode & MAY_EXEC)
2945                         args.access |= NFS4_ACCESS_EXECUTE;
2946         }
2947
2948         res.fattr = nfs_alloc_fattr();
2949         if (res.fattr == NULL)
2950                 return -ENOMEM;
2951
2952         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2953         if (!status) {
2954                 nfs_access_set_mask(entry, res.access);
2955                 nfs_refresh_inode(inode, res.fattr);
2956         }
2957         nfs_free_fattr(res.fattr);
2958         return status;
2959 }
2960
2961 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2962 {
2963         struct nfs4_exception exception = { };
2964         int err;
2965         do {
2966                 err = nfs4_handle_exception(NFS_SERVER(inode),
2967                                 _nfs4_proc_access(inode, entry),
2968                                 &exception);
2969         } while (exception.retry);
2970         return err;
2971 }
2972
2973 /*
2974  * TODO: For the time being, we don't try to get any attributes
2975  * along with any of the zero-copy operations READ, READDIR,
2976  * READLINK, WRITE.
2977  *
2978  * In the case of the first three, we want to put the GETATTR
2979  * after the read-type operation -- this is because it is hard
2980  * to predict the length of a GETATTR response in v4, and thus
2981  * align the READ data correctly.  This means that the GETATTR
2982  * may end up partially falling into the page cache, and we should
2983  * shift it into the 'tail' of the xdr_buf before processing.
2984  * To do this efficiently, we need to know the total length
2985  * of data received, which doesn't seem to be available outside
2986  * of the RPC layer.
2987  *
2988  * In the case of WRITE, we also want to put the GETATTR after
2989  * the operation -- in this case because we want to make sure
2990  * we get the post-operation mtime and size.
2991  *
2992  * Both of these changes to the XDR layer would in fact be quite
2993  * minor, but I decided to leave them for a subsequent patch.
2994  */
2995 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2996                 unsigned int pgbase, unsigned int pglen)
2997 {
2998         struct nfs4_readlink args = {
2999                 .fh       = NFS_FH(inode),
3000                 .pgbase   = pgbase,
3001                 .pglen    = pglen,
3002                 .pages    = &page,
3003         };
3004         struct nfs4_readlink_res res;
3005         struct rpc_message msg = {
3006                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3007                 .rpc_argp = &args,
3008                 .rpc_resp = &res,
3009         };
3010
3011         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3012 }
3013
3014 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3015                 unsigned int pgbase, unsigned int pglen)
3016 {
3017         struct nfs4_exception exception = { };
3018         int err;
3019         do {
3020                 err = nfs4_handle_exception(NFS_SERVER(inode),
3021                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
3022                                 &exception);
3023         } while (exception.retry);
3024         return err;
3025 }
3026
3027 /*
3028  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3029  */
3030 static int
3031 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3032                  int flags)
3033 {
3034         struct nfs_open_context *ctx;
3035         struct nfs4_state *state;
3036         int status = 0;
3037
3038         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3039         if (IS_ERR(ctx))
3040                 return PTR_ERR(ctx);
3041
3042         sattr->ia_mode &= ~current_umask();
3043         state = nfs4_do_open(dir, dentry, ctx->mode,
3044                         flags, sattr, ctx->cred,
3045                         &ctx->mdsthreshold);
3046         d_drop(dentry);
3047         if (IS_ERR(state)) {
3048                 status = PTR_ERR(state);
3049                 goto out;
3050         }
3051         d_add(dentry, igrab(state->inode));
3052         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
3053         ctx->state = state;
3054 out:
3055         put_nfs_open_context(ctx);
3056         return status;
3057 }
3058
3059 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3060 {
3061         struct nfs_server *server = NFS_SERVER(dir);
3062         struct nfs_removeargs args = {
3063                 .fh = NFS_FH(dir),
3064                 .name = *name,
3065         };
3066         struct nfs_removeres res = {
3067                 .server = server,
3068         };
3069         struct rpc_message msg = {
3070                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3071                 .rpc_argp = &args,
3072                 .rpc_resp = &res,
3073         };
3074         int status;
3075
3076         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3077         if (status == 0)
3078                 update_changeattr(dir, &res.cinfo);
3079         return status;
3080 }
3081
3082 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3083 {
3084         struct nfs4_exception exception = { };
3085         int err;
3086         do {
3087                 err = nfs4_handle_exception(NFS_SERVER(dir),
3088                                 _nfs4_proc_remove(dir, name),
3089                                 &exception);
3090         } while (exception.retry);
3091         return err;
3092 }
3093
3094 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3095 {
3096         struct nfs_server *server = NFS_SERVER(dir);
3097         struct nfs_removeargs *args = msg->rpc_argp;
3098         struct nfs_removeres *res = msg->rpc_resp;
3099
3100         res->server = server;
3101         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3102         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3103 }
3104
3105 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3106 {
3107         nfs4_setup_sequence(NFS_SERVER(data->dir),
3108                         &data->args.seq_args,
3109                         &data->res.seq_res,
3110                         task);
3111 }
3112
3113 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3114 {
3115         struct nfs_removeres *res = task->tk_msg.rpc_resp;
3116
3117         if (!nfs4_sequence_done(task, &res->seq_res))
3118                 return 0;
3119         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3120                 return 0;
3121         update_changeattr(dir, &res->cinfo);
3122         return 1;
3123 }
3124
3125 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3126 {
3127         struct nfs_server *server = NFS_SERVER(dir);
3128         struct nfs_renameargs *arg = msg->rpc_argp;
3129         struct nfs_renameres *res = msg->rpc_resp;
3130
3131         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3132         res->server = server;
3133         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3134 }
3135
3136 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3137 {
3138         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3139                         &data->args.seq_args,
3140                         &data->res.seq_res,
3141                         task);
3142 }
3143
3144 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3145                                  struct inode *new_dir)
3146 {
3147         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3148
3149         if (!nfs4_sequence_done(task, &res->seq_res))
3150                 return 0;
3151         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3152                 return 0;
3153
3154         update_changeattr(old_dir, &res->old_cinfo);
3155         update_changeattr(new_dir, &res->new_cinfo);
3156         return 1;
3157 }
3158
3159 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3160                 struct inode *new_dir, struct qstr *new_name)
3161 {
3162         struct nfs_server *server = NFS_SERVER(old_dir);
3163         struct nfs_renameargs arg = {
3164                 .old_dir = NFS_FH(old_dir),
3165                 .new_dir = NFS_FH(new_dir),
3166                 .old_name = old_name,
3167                 .new_name = new_name,
3168         };
3169         struct nfs_renameres res = {
3170                 .server = server,
3171         };
3172         struct rpc_message msg = {
3173                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3174                 .rpc_argp = &arg,
3175                 .rpc_resp = &res,
3176         };
3177         int status = -ENOMEM;
3178
3179         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3180         if (!status) {
3181                 update_changeattr(old_dir, &res.old_cinfo);
3182                 update_changeattr(new_dir, &res.new_cinfo);
3183         }
3184         return status;
3185 }
3186
3187 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3188                 struct inode *new_dir, struct qstr *new_name)
3189 {
3190         struct nfs4_exception exception = { };
3191         int err;
3192         do {
3193                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3194                                 _nfs4_proc_rename(old_dir, old_name,
3195                                         new_dir, new_name),
3196                                 &exception);
3197         } while (exception.retry);
3198         return err;
3199 }
3200
3201 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3202 {
3203         struct nfs_server *server = NFS_SERVER(inode);
3204         struct nfs4_link_arg arg = {
3205                 .fh     = NFS_FH(inode),
3206                 .dir_fh = NFS_FH(dir),
3207                 .name   = name,
3208                 .bitmask = server->attr_bitmask,
3209         };
3210         struct nfs4_link_res res = {
3211                 .server = server,
3212         };
3213         struct rpc_message msg = {
3214                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3215                 .rpc_argp = &arg,
3216                 .rpc_resp = &res,
3217         };
3218         int status = -ENOMEM;
3219
3220         res.fattr = nfs_alloc_fattr();
3221         if (res.fattr == NULL)
3222                 goto out;
3223
3224         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3225         if (!status) {
3226                 update_changeattr(dir, &res.cinfo);
3227                 nfs_post_op_update_inode(inode, res.fattr);
3228         }
3229 out:
3230         nfs_free_fattr(res.fattr);
3231         return status;
3232 }
3233
3234 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3235 {
3236         struct nfs4_exception exception = { };
3237         int err;
3238         do {
3239                 err = nfs4_handle_exception(NFS_SERVER(inode),
3240                                 _nfs4_proc_link(inode, dir, name),
3241                                 &exception);
3242         } while (exception.retry);
3243         return err;
3244 }
3245
3246 struct nfs4_createdata {
3247         struct rpc_message msg;
3248         struct nfs4_create_arg arg;
3249         struct nfs4_create_res res;
3250         struct nfs_fh fh;
3251         struct nfs_fattr fattr;
3252 };
3253
3254 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3255                 struct qstr *name, struct iattr *sattr, u32 ftype)
3256 {
3257         struct nfs4_createdata *data;
3258
3259         data = kzalloc(sizeof(*data), GFP_KERNEL);
3260         if (data != NULL) {
3261                 struct nfs_server *server = NFS_SERVER(dir);
3262
3263                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3264                 data->msg.rpc_argp = &data->arg;
3265                 data->msg.rpc_resp = &data->res;
3266                 data->arg.dir_fh = NFS_FH(dir);
3267                 data->arg.server = server;
3268                 data->arg.name = name;
3269                 data->arg.attrs = sattr;
3270                 data->arg.ftype = ftype;
3271                 data->arg.bitmask = server->attr_bitmask;
3272                 data->res.server = server;
3273                 data->res.fh = &data->fh;
3274                 data->res.fattr = &data->fattr;
3275                 nfs_fattr_init(data->res.fattr);
3276         }
3277         return data;
3278 }
3279
3280 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3281 {
3282         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3283                                     &data->arg.seq_args, &data->res.seq_res, 1);
3284         if (status == 0) {
3285                 update_changeattr(dir, &data->res.dir_cinfo);
3286                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3287         }
3288         return status;
3289 }
3290
3291 static void nfs4_free_createdata(struct nfs4_createdata *data)
3292 {
3293         kfree(data);
3294 }
3295
3296 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3297                 struct page *page, unsigned int len, struct iattr *sattr)
3298 {
3299         struct nfs4_createdata *data;
3300         int status = -ENAMETOOLONG;
3301
3302         if (len > NFS4_MAXPATHLEN)
3303                 goto out;
3304
3305         status = -ENOMEM;
3306         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3307         if (data == NULL)
3308                 goto out;
3309
3310         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3311         data->arg.u.symlink.pages = &page;
3312         data->arg.u.symlink.len = len;
3313
3314         status = nfs4_do_create(dir, dentry, data);
3315
3316         nfs4_free_createdata(data);
3317 out:
3318         return status;
3319 }
3320
3321 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3322                 struct page *page, unsigned int len, struct iattr *sattr)
3323 {
3324         struct nfs4_exception exception = { };
3325         int err;
3326         do {
3327                 err = nfs4_handle_exception(NFS_SERVER(dir),
3328                                 _nfs4_proc_symlink(dir, dentry, page,
3329                                                         len, sattr),
3330                                 &exception);
3331         } while (exception.retry);
3332         return err;
3333 }
3334
3335 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3336                 struct iattr *sattr)
3337 {
3338         struct nfs4_createdata *data;
3339         int status = -ENOMEM;
3340
3341         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3342         if (data == NULL)
3343                 goto out;
3344
3345         status = nfs4_do_create(dir, dentry, data);
3346
3347         nfs4_free_createdata(data);
3348 out:
3349         return status;
3350 }
3351
3352 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3353                 struct iattr *sattr)
3354 {
3355         struct nfs4_exception exception = { };
3356         int err;
3357
3358         sattr->ia_mode &= ~current_umask();
3359         do {
3360                 err = nfs4_handle_exception(NFS_SERVER(dir),
3361                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3362                                 &exception);
3363         } while (exception.retry);
3364         return err;
3365 }
3366
3367 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3368                 u64 cookie, struct page **pages, unsigned int count, int plus)
3369 {
3370         struct inode            *dir = dentry->d_inode;
3371         struct nfs4_readdir_arg args = {
3372                 .fh = NFS_FH(dir),
3373                 .pages = pages,
3374                 .pgbase = 0,
3375                 .count = count,
3376                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3377                 .plus = plus,
3378         };
3379         struct nfs4_readdir_res res;
3380         struct rpc_message msg = {
3381                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3382                 .rpc_argp = &args,
3383                 .rpc_resp = &res,
3384                 .rpc_cred = cred,
3385         };
3386         int                     status;
3387
3388         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3389                         dentry->d_parent->d_name.name,
3390                         dentry->d_name.name,
3391                         (unsigned long long)cookie);
3392         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3393         res.pgbase = args.pgbase;
3394         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3395         if (status >= 0) {
3396                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3397                 status += args.pgbase;
3398         }
3399
3400         nfs_invalidate_atime(dir);
3401
3402         dprintk("%s: returns %d\n", __func__, status);
3403         return status;
3404 }
3405
3406 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3407                 u64 cookie, struct page **pages, unsigned int count, int plus)
3408 {
3409         struct nfs4_exception exception = { };
3410         int err;
3411         do {
3412                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3413                                 _nfs4_proc_readdir(dentry, cred, cookie,
3414                                         pages, count, plus),
3415                                 &exception);
3416         } while (exception.retry);
3417         return err;
3418 }
3419
3420 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3421                 struct iattr *sattr, dev_t rdev)
3422 {
3423         struct nfs4_createdata *data;
3424         int mode = sattr->ia_mode;
3425         int status = -ENOMEM;
3426
3427         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3428         if (data == NULL)
3429                 goto out;
3430
3431         if (S_ISFIFO(mode))
3432                 data->arg.ftype = NF4FIFO;
3433         else if (S_ISBLK(mode)) {
3434                 data->arg.ftype = NF4BLK;
3435                 data->arg.u.device.specdata1 = MAJOR(rdev);
3436                 data->arg.u.device.specdata2 = MINOR(rdev);
3437         }
3438         else if (S_ISCHR(mode)) {
3439                 data->arg.ftype = NF4CHR;
3440                 data->arg.u.device.specdata1 = MAJOR(rdev);
3441                 data->arg.u.device.specdata2 = MINOR(rdev);
3442         } else if (!S_ISSOCK(mode)) {
3443                 status = -EINVAL;
3444                 goto out_free;
3445         }
3446
3447         status = nfs4_do_create(dir, dentry, data);
3448 out_free:
3449         nfs4_free_createdata(data);
3450 out:
3451         return status;
3452 }
3453
3454 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3455                 struct iattr *sattr, dev_t rdev)
3456 {
3457         struct nfs4_exception exception = { };
3458         int err;
3459
3460         sattr->ia_mode &= ~current_umask();
3461         do {
3462                 err = nfs4_handle_exception(NFS_SERVER(dir),
3463                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3464                                 &exception);
3465         } while (exception.retry);
3466         return err;
3467 }
3468
3469 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3470                  struct nfs_fsstat *fsstat)
3471 {
3472         struct nfs4_statfs_arg args = {
3473                 .fh = fhandle,
3474                 .bitmask = server->attr_bitmask,
3475         };
3476         struct nfs4_statfs_res res = {
3477                 .fsstat = fsstat,
3478         };
3479         struct rpc_message msg = {
3480                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3481                 .rpc_argp = &args,
3482                 .rpc_resp = &res,
3483         };
3484
3485         nfs_fattr_init(fsstat->fattr);
3486         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3487 }
3488
3489 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3490 {
3491         struct nfs4_exception exception = { };
3492         int err;
3493         do {
3494                 err = nfs4_handle_exception(server,
3495                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3496                                 &exception);
3497         } while (exception.retry);
3498         return err;
3499 }
3500
3501 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3502                 struct nfs_fsinfo *fsinfo)
3503 {
3504         struct nfs4_fsinfo_arg args = {
3505                 .fh = fhandle,
3506                 .bitmask = server->attr_bitmask,
3507         };
3508         struct nfs4_fsinfo_res res = {
3509                 .fsinfo = fsinfo,
3510         };
3511         struct rpc_message msg = {
3512                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3513                 .rpc_argp = &args,
3514                 .rpc_resp = &res,
3515         };
3516
3517         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3518 }
3519
3520 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3521 {
3522         struct nfs4_exception exception = { };
3523         unsigned long now = jiffies;
3524         int err;
3525
3526         do {
3527                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
3528                 if (err == 0) {
3529                         struct nfs_client *clp = server->nfs_client;
3530
3531                         spin_lock(&clp->cl_lock);
3532                         clp->cl_lease_time = fsinfo->lease_time * HZ;
3533                         clp->cl_last_renewal = now;
3534                         spin_unlock(&clp->cl_lock);
3535                         break;
3536                 }
3537                 err = nfs4_handle_exception(server, err, &exception);
3538         } while (exception.retry);
3539         return err;
3540 }
3541
3542 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3543 {
3544         int error;
3545
3546         nfs_fattr_init(fsinfo->fattr);
3547         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3548         if (error == 0) {
3549                 /* block layout checks this! */
3550                 server->pnfs_blksize = fsinfo->blksize;
3551                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3552         }
3553
3554         return error;
3555 }
3556
3557 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3558                 struct nfs_pathconf *pathconf)
3559 {
3560         struct nfs4_pathconf_arg args = {
3561                 .fh = fhandle,
3562                 .bitmask = server->attr_bitmask,
3563         };
3564         struct nfs4_pathconf_res res = {
3565                 .pathconf = pathconf,
3566         };
3567         struct rpc_message msg = {
3568                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3569                 .rpc_argp = &args,
3570                 .rpc_resp = &res,
3571         };
3572
3573         /* None of the pathconf attributes are mandatory to implement */
3574         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3575                 memset(pathconf, 0, sizeof(*pathconf));
3576                 return 0;
3577         }
3578
3579         nfs_fattr_init(pathconf->fattr);
3580         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3581 }
3582
3583 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3584                 struct nfs_pathconf *pathconf)
3585 {
3586         struct nfs4_exception exception = { };
3587         int err;
3588
3589         do {
3590                 err = nfs4_handle_exception(server,
3591                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3592                                 &exception);
3593         } while (exception.retry);
3594         return err;
3595 }
3596
3597 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
3598                 const struct nfs_open_context *ctx,
3599                 const struct nfs_lock_context *l_ctx,
3600                 fmode_t fmode)
3601 {
3602         const struct nfs_lockowner *lockowner = NULL;
3603
3604         if (l_ctx != NULL)
3605                 lockowner = &l_ctx->lockowner;
3606         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
3607 }
3608 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
3609
3610 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
3611                 const struct nfs_open_context *ctx,
3612                 const struct nfs_lock_context *l_ctx,
3613                 fmode_t fmode)
3614 {
3615         nfs4_stateid current_stateid;
3616
3617         /* If the current stateid represents a lost lock, then exit */
3618         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
3619                 return true;
3620         return nfs4_stateid_match(stateid, &current_stateid);
3621 }
3622
3623 static bool nfs4_error_stateid_expired(int err)
3624 {
3625         switch (err) {
3626         case -NFS4ERR_DELEG_REVOKED:
3627         case -NFS4ERR_ADMIN_REVOKED:
3628         case -NFS4ERR_BAD_STATEID:
3629         case -NFS4ERR_STALE_STATEID:
3630         case -NFS4ERR_OLD_STATEID:
3631         case -NFS4ERR_OPENMODE:
3632         case -NFS4ERR_EXPIRED:
3633                 return true;
3634         }
3635         return false;
3636 }
3637
3638 void __nfs4_read_done_cb(struct nfs_read_data *data)
3639 {
3640         nfs_invalidate_atime(data->header->inode);
3641 }
3642
3643 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3644 {
3645         struct nfs_server *server = NFS_SERVER(data->header->inode);
3646
3647         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3648                 rpc_restart_call_prepare(task);
3649                 return -EAGAIN;
3650         }
3651
3652         __nfs4_read_done_cb(data);
3653         if (task->tk_status > 0)
3654                 renew_lease(server, data->timestamp);
3655         return 0;
3656 }
3657
3658 static bool nfs4_read_stateid_changed(struct rpc_task *task,
3659                 struct nfs_readargs *args)
3660 {
3661
3662         if (!nfs4_error_stateid_expired(task->tk_status) ||
3663                 nfs4_stateid_is_current(&args->stateid,
3664                                 args->context,
3665                                 args->lock_context,
3666                                 FMODE_READ))
3667                 return false;
3668         rpc_restart_call_prepare(task);
3669         return true;
3670 }
3671
3672 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3673 {
3674
3675         dprintk("--> %s\n", __func__);
3676
3677         if (!nfs4_sequence_done(task, &data->res.seq_res))
3678                 return -EAGAIN;
3679         if (nfs4_read_stateid_changed(task, &data->args))
3680                 return -EAGAIN;
3681         return data->read_done_cb ? data->read_done_cb(task, data) :
3682                                     nfs4_read_done_cb(task, data);
3683 }
3684
3685 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3686 {
3687         data->timestamp   = jiffies;
3688         data->read_done_cb = nfs4_read_done_cb;
3689         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3690         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3691 }
3692
3693 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3694 {
3695         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3696                         &data->args.seq_args,
3697                         &data->res.seq_res,
3698                         task))
3699                 return;
3700         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3701                         data->args.lock_context, FMODE_READ);
3702 }
3703
3704 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3705 {
3706         struct inode *inode = data->header->inode;
3707
3708         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3709                 rpc_restart_call_prepare(task);
3710                 return -EAGAIN;
3711         }
3712         if (task->tk_status >= 0) {
3713                 renew_lease(NFS_SERVER(inode), data->timestamp);
3714                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3715         }
3716         return 0;
3717 }
3718
3719 static bool nfs4_write_stateid_changed(struct rpc_task *task,
3720                 struct nfs_writeargs *args)
3721 {
3722
3723         if (!nfs4_error_stateid_expired(task->tk_status) ||
3724                 nfs4_stateid_is_current(&args->stateid,
3725                                 args->context,
3726                                 args->lock_context,
3727                                 FMODE_WRITE))
3728                 return false;
3729         rpc_restart_call_prepare(task);
3730         return true;
3731 }
3732
3733 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3734 {
3735         if (!nfs4_sequence_done(task, &data->res.seq_res))
3736                 return -EAGAIN;
3737         if (nfs4_write_stateid_changed(task, &data->args))
3738                 return -EAGAIN;
3739         return data->write_done_cb ? data->write_done_cb(task, data) :
3740                 nfs4_write_done_cb(task, data);
3741 }
3742
3743 static
3744 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3745 {
3746         const struct nfs_pgio_header *hdr = data->header;
3747
3748         /* Don't request attributes for pNFS or O_DIRECT writes */
3749         if (data->ds_clp != NULL || hdr->dreq != NULL)
3750                 return false;
3751         /* Otherwise, request attributes if and only if we don't hold
3752          * a delegation
3753          */
3754         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3755 }
3756
3757 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3758 {
3759         struct nfs_server *server = NFS_SERVER(data->header->inode);
3760
3761         if (!nfs4_write_need_cache_consistency_data(data)) {
3762                 data->args.bitmask = NULL;
3763                 data->res.fattr = NULL;
3764         } else
3765                 data->args.bitmask = server->cache_consistency_bitmask;
3766
3767         if (!data->write_done_cb)
3768                 data->write_done_cb = nfs4_write_done_cb;
3769         data->res.server = server;
3770         data->timestamp   = jiffies;
3771
3772         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3773         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3774 }
3775
3776 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3777 {
3778         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3779                         &data->args.seq_args,
3780                         &data->res.seq_res,
3781                         task))
3782                 return;
3783         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3784                         data->args.lock_context, FMODE_WRITE);
3785 }
3786
3787 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3788 {
3789         nfs4_setup_sequence(NFS_SERVER(data->inode),
3790                         &data->args.seq_args,
3791                         &data->res.seq_res,
3792                         task);
3793 }
3794
3795 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3796 {
3797         struct inode *inode = data->inode;
3798
3799         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3800                 rpc_restart_call_prepare(task);
3801                 return -EAGAIN;
3802         }
3803         return 0;
3804 }
3805
3806 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3807 {
3808         if (!nfs4_sequence_done(task, &data->res.seq_res))
3809                 return -EAGAIN;
3810         return data->commit_done_cb(task, data);
3811 }
3812
3813 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3814 {
3815         struct nfs_server *server = NFS_SERVER(data->inode);
3816
3817         if (data->commit_done_cb == NULL)
3818                 data->commit_done_cb = nfs4_commit_done_cb;
3819         data->res.server = server;
3820         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3821         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3822 }
3823
3824 struct nfs4_renewdata {
3825         struct nfs_client       *client;
3826         unsigned long           timestamp;
3827 };
3828
3829 /*
3830  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3831  * standalone procedure for queueing an asynchronous RENEW.
3832  */
3833 static void nfs4_renew_release(void *calldata)
3834 {
3835         struct nfs4_renewdata *data = calldata;
3836         struct nfs_client *clp = data->client;
3837
3838         if (atomic_read(&clp->cl_count) > 1)
3839                 nfs4_schedule_state_renewal(clp);
3840         nfs_put_client(clp);
3841         kfree(data);
3842 }
3843
3844 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3845 {
3846         struct nfs4_renewdata *data = calldata;
3847         struct nfs_client *clp = data->client;
3848         unsigned long timestamp = data->timestamp;
3849
3850         if (task->tk_status < 0) {
3851                 /* Unless we're shutting down, schedule state recovery! */
3852                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3853                         return;
3854                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3855                         nfs4_schedule_lease_recovery(clp);
3856                         return;
3857                 }
3858                 nfs4_schedule_path_down_recovery(clp);
3859         }
3860         do_renew_lease(clp, timestamp);
3861 }
3862
3863 static const struct rpc_call_ops nfs4_renew_ops = {
3864         .rpc_call_done = nfs4_renew_done,
3865         .rpc_release = nfs4_renew_release,
3866 };
3867
3868 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3869 {
3870         struct rpc_message msg = {
3871                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3872                 .rpc_argp       = clp,
3873                 .rpc_cred       = cred,
3874         };
3875         struct nfs4_renewdata *data;
3876
3877         if (renew_flags == 0)
3878                 return 0;
3879         if (!atomic_inc_not_zero(&clp->cl_count))
3880                 return -EIO;
3881         data = kmalloc(sizeof(*data), GFP_NOFS);
3882         if (data == NULL)
3883                 return -ENOMEM;
3884         data->client = clp;
3885         data->timestamp = jiffies;
3886         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
3887                         &nfs4_renew_ops, data);
3888 }
3889
3890 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3891 {
3892         struct rpc_message msg = {
3893                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3894                 .rpc_argp       = clp,
3895                 .rpc_cred       = cred,
3896         };
3897         unsigned long now = jiffies;
3898         int status;
3899
3900         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3901         if (status < 0)
3902                 return status;
3903         do_renew_lease(clp, now);
3904         return 0;
3905 }
3906
3907 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3908 {
3909         return (server->caps & NFS_CAP_ACLS)
3910                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3911                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3912 }
3913
3914 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3915  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3916  * the stack.
3917  */
3918 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3919
3920 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3921                 struct page **pages, unsigned int *pgbase)
3922 {
3923         struct page *newpage, **spages;
3924         int rc = 0;
3925         size_t len;
3926         spages = pages;
3927
3928         do {
3929                 len = min_t(size_t, PAGE_SIZE, buflen);
3930                 newpage = alloc_page(GFP_KERNEL);
3931
3932                 if (newpage == NULL)
3933                         goto unwind;
3934                 memcpy(page_address(newpage), buf, len);
3935                 buf += len;
3936                 buflen -= len;
3937                 *pages++ = newpage;
3938                 rc++;
3939         } while (buflen != 0);
3940
3941         return rc;
3942
3943 unwind:
3944         for(; rc > 0; rc--)
3945                 __free_page(spages[rc-1]);
3946         return -ENOMEM;
3947 }
3948
3949 struct nfs4_cached_acl {
3950         int cached;
3951         size_t len;
3952         char data[0];
3953 };
3954
3955 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3956 {
3957         struct nfs_inode *nfsi = NFS_I(inode);
3958
3959         spin_lock(&inode->i_lock);
3960         kfree(nfsi->nfs4_acl);
3961         nfsi->nfs4_acl = acl;
3962         spin_unlock(&inode->i_lock);
3963 }
3964
3965 static void nfs4_zap_acl_attr(struct inode *inode)
3966 {
3967         nfs4_set_cached_acl(inode, NULL);
3968 }
3969
3970 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3971 {
3972         struct nfs_inode *nfsi = NFS_I(inode);
3973         struct nfs4_cached_acl *acl;
3974         int ret = -ENOENT;
3975
3976         spin_lock(&inode->i_lock);
3977         acl = nfsi->nfs4_acl;
3978         if (acl == NULL)
3979                 goto out;
3980         if (buf == NULL) /* user is just asking for length */
3981                 goto out_len;
3982         if (acl->cached == 0)
3983                 goto out;
3984         ret = -ERANGE; /* see getxattr(2) man page */
3985         if (acl->len > buflen)
3986                 goto out;
3987         memcpy(buf, acl->data, acl->len);
3988 out_len:
3989         ret = acl->len;
3990 out:
3991         spin_unlock(&inode->i_lock);
3992         return ret;
3993 }
3994
3995 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3996 {
3997         struct nfs4_cached_acl *acl;
3998         size_t buflen = sizeof(*acl) + acl_len;
3999
4000         if (buflen <= PAGE_SIZE) {
4001                 acl = kmalloc(buflen, GFP_KERNEL);
4002                 if (acl == NULL)
4003                         goto out;
4004                 acl->cached = 1;
4005                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4006         } else {
4007                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4008                 if (acl == NULL)
4009                         goto out;
4010                 acl->cached = 0;
4011         }
4012         acl->len = acl_len;
4013 out:
4014         nfs4_set_cached_acl(inode, acl);
4015 }
4016
4017 /*
4018  * The getxattr API returns the required buffer length when called with a
4019  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4020  * the required buf.  On a NULL buf, we send a page of data to the server
4021  * guessing that the ACL request can be serviced by a page. If so, we cache
4022  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4023  * the cache. If not so, we throw away the page, and cache the required
4024  * length. The next getxattr call will then produce another round trip to
4025  * the server, this time with the input buf of the required size.
4026  */
4027 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4028 {
4029         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4030         struct nfs_getaclargs args = {
4031                 .fh = NFS_FH(inode),
4032                 .acl_pages = pages,
4033                 .acl_len = buflen,
4034         };
4035         struct nfs_getaclres res = {
4036                 .acl_len = buflen,
4037         };
4038         struct rpc_message msg = {
4039                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4040                 .rpc_argp = &args,
4041                 .rpc_resp = &res,
4042         };
4043         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4044         int ret = -ENOMEM, i;
4045
4046         /* As long as we're doing a round trip to the server anyway,
4047          * let's be prepared for a page of acl data. */
4048         if (npages == 0)
4049                 npages = 1;
4050         if (npages > ARRAY_SIZE(pages))
4051                 return -ERANGE;
4052
4053         for (i = 0; i < npages; i++) {
4054                 pages[i] = alloc_page(GFP_KERNEL);
4055                 if (!pages[i])
4056                         goto out_free;
4057         }
4058
4059         /* for decoding across pages */
4060         res.acl_scratch = alloc_page(GFP_KERNEL);
4061         if (!res.acl_scratch)
4062                 goto out_free;
4063
4064         args.acl_len = npages * PAGE_SIZE;
4065         args.acl_pgbase = 0;
4066
4067         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4068                 __func__, buf, buflen, npages, args.acl_len);
4069         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4070                              &msg, &args.seq_args, &res.seq_res, 0);
4071         if (ret)
4072                 goto out_free;
4073
4074         /* Handle the case where the passed-in buffer is too short */
4075         if (res.acl_flags & NFS4_ACL_TRUNC) {
4076                 /* Did the user only issue a request for the acl length? */
4077                 if (buf == NULL)
4078                         goto out_ok;
4079                 ret = -ERANGE;
4080                 goto out_free;
4081         }
4082         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4083         if (buf) {
4084                 if (res.acl_len > buflen) {
4085                         ret = -ERANGE;
4086                         goto out_free;
4087                 }
4088                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4089         }
4090 out_ok:
4091         ret = res.acl_len;
4092 out_free:
4093         for (i = 0; i < npages; i++)
4094                 if (pages[i])
4095                         __free_page(pages[i]);
4096         if (res.acl_scratch)
4097                 __free_page(res.acl_scratch);
4098         return ret;
4099 }
4100
4101 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4102 {
4103         struct nfs4_exception exception = { };
4104         ssize_t ret;
4105         do {
4106                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4107                 if (ret >= 0)
4108                         break;
4109                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4110         } while (exception.retry);
4111         return ret;
4112 }
4113
4114 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4115 {
4116         struct nfs_server *server = NFS_SERVER(inode);
4117         int ret;
4118
4119         if (!nfs4_server_supports_acls(server))
4120                 return -EOPNOTSUPP;
4121         ret = nfs_revalidate_inode(server, inode);
4122         if (ret < 0)
4123                 return ret;
4124         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4125                 nfs_zap_acl_cache(inode);
4126         ret = nfs4_read_cached_acl(inode, buf, buflen);
4127         if (ret != -ENOENT)
4128                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4129                  * but no cached acl data, just the acl length */
4130                 return ret;
4131         return nfs4_get_acl_uncached(inode, buf, buflen);
4132 }
4133
4134 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4135 {
4136         struct nfs_server *server = NFS_SERVER(inode);
4137         struct page *pages[NFS4ACL_MAXPAGES];
4138         struct nfs_setaclargs arg = {
4139                 .fh             = NFS_FH(inode),
4140                 .acl_pages      = pages,
4141                 .acl_len        = buflen,
4142         };
4143         struct nfs_setaclres res;
4144         struct rpc_message msg = {
4145                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4146                 .rpc_argp       = &arg,
4147                 .rpc_resp       = &res,
4148         };
4149         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4150         int ret, i;
4151
4152         if (!nfs4_server_supports_acls(server))
4153                 return -EOPNOTSUPP;
4154         if (npages > ARRAY_SIZE(pages))
4155                 return -ERANGE;
4156         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4157         if (i < 0)
4158                 return i;
4159         nfs4_inode_return_delegation(inode);
4160         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4161
4162         /*
4163          * Free each page after tx, so the only ref left is
4164          * held by the network stack
4165          */
4166         for (; i > 0; i--)
4167                 put_page(pages[i-1]);
4168
4169         /*
4170          * Acl update can result in inode attribute update.
4171          * so mark the attribute cache invalid.
4172          */
4173         spin_lock(&inode->i_lock);
4174         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4175         spin_unlock(&inode->i_lock);
4176         nfs_access_zap_cache(inode);
4177         nfs_zap_acl_cache(inode);
4178         return ret;
4179 }
4180
4181 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4182 {
4183         struct nfs4_exception exception = { };
4184         int err;
4185         do {
4186                 err = nfs4_handle_exception(NFS_SERVER(inode),
4187                                 __nfs4_proc_set_acl(inode, buf, buflen),
4188                                 &exception);
4189         } while (exception.retry);
4190         return err;
4191 }
4192
4193 static int
4194 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4195 {
4196         struct nfs_client *clp = server->nfs_client;
4197
4198         if (task->tk_status >= 0)
4199                 return 0;
4200         switch(task->tk_status) {
4201                 case -NFS4ERR_DELEG_REVOKED:
4202                 case -NFS4ERR_ADMIN_REVOKED:
4203                 case -NFS4ERR_BAD_STATEID:
4204                         if (state == NULL)
4205                                 break;
4206                         nfs_remove_bad_delegation(state->inode);
4207                 case -NFS4ERR_OPENMODE:
4208                         if (state == NULL)
4209                                 break;
4210                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4211                                 goto stateid_invalid;
4212                         goto wait_on_recovery;
4213                 case -NFS4ERR_EXPIRED:
4214                         if (state != NULL) {
4215                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4216                                         goto stateid_invalid;
4217                         }
4218                 case -NFS4ERR_STALE_STATEID:
4219                 case -NFS4ERR_STALE_CLIENTID:
4220                         nfs4_schedule_lease_recovery(clp);
4221                         goto wait_on_recovery;
4222 #if defined(CONFIG_NFS_V4_1)
4223                 case -NFS4ERR_BADSESSION:
4224                 case -NFS4ERR_BADSLOT:
4225                 case -NFS4ERR_BAD_HIGH_SLOT:
4226                 case -NFS4ERR_DEADSESSION:
4227                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4228                 case -NFS4ERR_SEQ_FALSE_RETRY:
4229                 case -NFS4ERR_SEQ_MISORDERED:
4230                         dprintk("%s ERROR %d, Reset session\n", __func__,
4231                                 task->tk_status);
4232                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4233                         goto wait_on_recovery;
4234 #endif /* CONFIG_NFS_V4_1 */
4235                 case -NFS4ERR_DELAY:
4236                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4237                 case -NFS4ERR_GRACE:
4238                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4239                         task->tk_status = 0;
4240                         return -EAGAIN;
4241                 case -NFS4ERR_RETRY_UNCACHED_REP:
4242                 case -NFS4ERR_OLD_STATEID:
4243                         task->tk_status = 0;
4244                         return -EAGAIN;
4245         }
4246         task->tk_status = nfs4_map_errors(task->tk_status);
4247         return 0;
4248 stateid_invalid:
4249         task->tk_status = -EIO;
4250         return 0;
4251 wait_on_recovery:
4252         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4253         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4254                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4255         task->tk_status = 0;
4256         return -EAGAIN;
4257 }
4258
4259 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4260                                     nfs4_verifier *bootverf)
4261 {
4262         __be32 verf[2];
4263
4264         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4265                 /* An impossible timestamp guarantees this value
4266                  * will never match a generated boot time. */
4267                 verf[0] = 0;
4268                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4269         } else {
4270                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4271                 verf[0] = (__be32)nn->boot_time.tv_sec;
4272                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4273         }
4274         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4275 }
4276
4277 static unsigned int
4278 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4279                                    char *buf, size_t len)
4280 {
4281         unsigned int result;
4282
4283         rcu_read_lock();
4284         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4285                                 clp->cl_ipaddr,
4286                                 rpc_peeraddr2str(clp->cl_rpcclient,
4287                                                         RPC_DISPLAY_ADDR),
4288                                 rpc_peeraddr2str(clp->cl_rpcclient,
4289                                                         RPC_DISPLAY_PROTO));
4290         rcu_read_unlock();
4291         return result;
4292 }
4293
4294 static unsigned int
4295 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4296                                 char *buf, size_t len)
4297 {
4298         char *nodename = clp->cl_rpcclient->cl_nodename;
4299
4300         if (nfs4_client_id_uniquifier[0] != '\0')
4301                 nodename = nfs4_client_id_uniquifier;
4302         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4303                                 clp->rpc_ops->version, clp->cl_minorversion,
4304                                 nodename);
4305 }
4306
4307 /**
4308  * nfs4_proc_setclientid - Negotiate client ID
4309  * @clp: state data structure
4310  * @program: RPC program for NFSv4 callback service
4311  * @port: IP port number for NFS4 callback service
4312  * @cred: RPC credential to use for this call
4313  * @res: where to place the result
4314  *
4315  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4316  */
4317 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4318                 unsigned short port, struct rpc_cred *cred,
4319                 struct nfs4_setclientid_res *res)
4320 {
4321         nfs4_verifier sc_verifier;
4322         struct nfs4_setclientid setclientid = {
4323                 .sc_verifier = &sc_verifier,
4324                 .sc_prog = program,
4325                 .sc_cb_ident = clp->cl_cb_ident,
4326         };
4327         struct rpc_message msg = {
4328                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4329                 .rpc_argp = &setclientid,
4330                 .rpc_resp = res,
4331                 .rpc_cred = cred,
4332         };
4333         int status;
4334
4335         /* nfs_client_id4 */
4336         nfs4_init_boot_verifier(clp, &sc_verifier);
4337         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4338                 setclientid.sc_name_len =
4339                                 nfs4_init_uniform_client_string(clp,
4340                                                 setclientid.sc_name,
4341                                                 sizeof(setclientid.sc_name));
4342         else
4343                 setclientid.sc_name_len =
4344                                 nfs4_init_nonuniform_client_string(clp,
4345                                                 setclientid.sc_name,
4346                                                 sizeof(setclientid.sc_name));
4347         /* cb_client4 */
4348         rcu_read_lock();
4349         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4350                                 sizeof(setclientid.sc_netid),
4351                                 rpc_peeraddr2str(clp->cl_rpcclient,
4352                                                         RPC_DISPLAY_NETID));
4353         rcu_read_unlock();
4354         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4355                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4356                                 clp->cl_ipaddr, port >> 8, port & 255);
4357
4358         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4359                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4360                 setclientid.sc_name_len, setclientid.sc_name);
4361         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4362         dprintk("NFS reply setclientid: %d\n", status);
4363         return status;
4364 }
4365
4366 /**
4367  * nfs4_proc_setclientid_confirm - Confirm client ID
4368  * @clp: state data structure
4369  * @res: result of a previous SETCLIENTID
4370  * @cred: RPC credential to use for this call
4371  *
4372  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4373  */
4374 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4375                 struct nfs4_setclientid_res *arg,
4376                 struct rpc_cred *cred)
4377 {
4378         struct rpc_message msg = {
4379                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4380                 .rpc_argp = arg,
4381                 .rpc_cred = cred,
4382         };
4383         int status;
4384
4385         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4386                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4387                 clp->cl_clientid);
4388         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4389         dprintk("NFS reply setclientid_confirm: %d\n", status);
4390         return status;
4391 }
4392
4393 struct nfs4_delegreturndata {
4394         struct nfs4_delegreturnargs args;
4395         struct nfs4_delegreturnres res;
4396         struct nfs_fh fh;
4397         nfs4_stateid stateid;
4398         unsigned long timestamp;
4399         struct nfs_fattr fattr;
4400         int rpc_status;
4401 };
4402
4403 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4404 {
4405         struct nfs4_delegreturndata *data = calldata;
4406
4407         if (!nfs4_sequence_done(task, &data->res.seq_res))
4408                 return;
4409
4410         switch (task->tk_status) {
4411         case 0:
4412                 renew_lease(data->res.server, data->timestamp);
4413                 break;
4414         case -NFS4ERR_ADMIN_REVOKED:
4415         case -NFS4ERR_DELEG_REVOKED:
4416         case -NFS4ERR_BAD_STATEID:
4417         case -NFS4ERR_OLD_STATEID:
4418         case -NFS4ERR_STALE_STATEID:
4419         case -NFS4ERR_EXPIRED:
4420                 task->tk_status = 0;
4421                 break;
4422         default:
4423                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4424                                 -EAGAIN) {
4425                         rpc_restart_call_prepare(task);
4426                         return;
4427                 }
4428         }
4429         data->rpc_status = task->tk_status;
4430 }
4431
4432 static void nfs4_delegreturn_release(void *calldata)
4433 {
4434         kfree(calldata);
4435 }
4436
4437 #if defined(CONFIG_NFS_V4_1)
4438 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4439 {
4440         struct nfs4_delegreturndata *d_data;
4441
4442         d_data = (struct nfs4_delegreturndata *)data;
4443
4444         nfs4_setup_sequence(d_data->res.server,
4445                         &d_data->args.seq_args,
4446                         &d_data->res.seq_res,
4447                         task);
4448 }
4449 #endif /* CONFIG_NFS_V4_1 */
4450
4451 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4452 #if defined(CONFIG_NFS_V4_1)
4453         .rpc_call_prepare = nfs4_delegreturn_prepare,
4454 #endif /* CONFIG_NFS_V4_1 */
4455         .rpc_call_done = nfs4_delegreturn_done,
4456         .rpc_release = nfs4_delegreturn_release,
4457 };
4458
4459 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4460 {
4461         struct nfs4_delegreturndata *data;
4462         struct nfs_server *server = NFS_SERVER(inode);
4463         struct rpc_task *task;
4464         struct rpc_message msg = {
4465                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4466                 .rpc_cred = cred,
4467         };
4468         struct rpc_task_setup task_setup_data = {
4469                 .rpc_client = server->client,
4470                 .rpc_message = &msg,
4471                 .callback_ops = &nfs4_delegreturn_ops,
4472                 .flags = RPC_TASK_ASYNC,
4473         };
4474         int status = 0;
4475
4476         data = kzalloc(sizeof(*data), GFP_NOFS);
4477         if (data == NULL)
4478                 return -ENOMEM;
4479         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4480         data->args.fhandle = &data->fh;
4481         data->args.stateid = &data->stateid;
4482         data->args.bitmask = server->cache_consistency_bitmask;
4483         nfs_copy_fh(&data->fh, NFS_FH(inode));
4484         nfs4_stateid_copy(&data->stateid, stateid);
4485         data->res.fattr = &data->fattr;
4486         data->res.server = server;
4487         nfs_fattr_init(data->res.fattr);
4488         data->timestamp = jiffies;
4489         data->rpc_status = 0;
4490
4491         task_setup_data.callback_data = data;
4492         msg.rpc_argp = &data->args;
4493         msg.rpc_resp = &data->res;
4494         task = rpc_run_task(&task_setup_data);
4495         if (IS_ERR(task))
4496                 return PTR_ERR(task);
4497         if (!issync)
4498                 goto out;
4499         status = nfs4_wait_for_completion_rpc_task(task);
4500         if (status != 0)
4501                 goto out;
4502         status = data->rpc_status;
4503         if (status == 0)
4504                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4505         else
4506                 nfs_refresh_inode(inode, &data->fattr);
4507 out:
4508         rpc_put_task(task);
4509         return status;
4510 }
4511
4512 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4513 {
4514         struct nfs_server *server = NFS_SERVER(inode);
4515         struct nfs4_exception exception = { };
4516         int err;
4517         do {
4518                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4519                 switch (err) {
4520                         case -NFS4ERR_STALE_STATEID:
4521                         case -NFS4ERR_EXPIRED:
4522                         case 0:
4523                                 return 0;
4524                 }
4525                 err = nfs4_handle_exception(server, err, &exception);
4526         } while (exception.retry);
4527         return err;
4528 }
4529
4530 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4531 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4532
4533 /*
4534  * sleep, with exponential backoff, and retry the LOCK operation.
4535  */
4536 static unsigned long
4537 nfs4_set_lock_task_retry(unsigned long timeout)
4538 {
4539         freezable_schedule_timeout_killable(timeout);
4540         timeout <<= 1;
4541         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4542                 return NFS4_LOCK_MAXTIMEOUT;
4543         return timeout;
4544 }
4545
4546 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4547 {
4548         struct inode *inode = state->inode;
4549         struct nfs_server *server = NFS_SERVER(inode);
4550         struct nfs_client *clp = server->nfs_client;
4551         struct nfs_lockt_args arg = {
4552                 .fh = NFS_FH(inode),
4553                 .fl = request,
4554         };
4555         struct nfs_lockt_res res = {
4556                 .denied = request,
4557         };
4558         struct rpc_message msg = {
4559                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4560                 .rpc_argp       = &arg,
4561                 .rpc_resp       = &res,
4562                 .rpc_cred       = state->owner->so_cred,
4563         };
4564         struct nfs4_lock_state *lsp;
4565         int status;
4566
4567         arg.lock_owner.clientid = clp->cl_clientid;
4568         status = nfs4_set_lock_state(state, request);
4569         if (status != 0)
4570                 goto out;
4571         lsp = request->fl_u.nfs4_fl.owner;
4572         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4573         arg.lock_owner.s_dev = server->s_dev;
4574         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4575         switch (status) {
4576                 case 0:
4577                         request->fl_type = F_UNLCK;
4578                         break;
4579                 case -NFS4ERR_DENIED:
4580                         status = 0;
4581         }
4582         request->fl_ops->fl_release_private(request);
4583         request->fl_ops = NULL;
4584 out:
4585         return status;
4586 }
4587
4588 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4589 {
4590         struct nfs4_exception exception = { };
4591         int err;
4592
4593         do {
4594                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4595                                 _nfs4_proc_getlk(state, cmd, request),
4596                                 &exception);
4597         } while (exception.retry);
4598         return err;
4599 }
4600
4601 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4602 {
4603         int res = 0;
4604         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4605                 case FL_POSIX:
4606                         res = posix_lock_file_wait(file, fl);
4607                         break;
4608                 case FL_FLOCK:
4609                         res = flock_lock_file_wait(file, fl);
4610                         break;
4611                 default:
4612                         BUG();
4613         }
4614         return res;
4615 }
4616
4617 struct nfs4_unlockdata {
4618         struct nfs_locku_args arg;
4619         struct nfs_locku_res res;
4620         struct nfs4_lock_state *lsp;
4621         struct nfs_open_context *ctx;
4622         struct file_lock fl;
4623         const struct nfs_server *server;
4624         unsigned long timestamp;
4625 };
4626
4627 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4628                 struct nfs_open_context *ctx,
4629                 struct nfs4_lock_state *lsp,
4630                 struct nfs_seqid *seqid)
4631 {
4632         struct nfs4_unlockdata *p;
4633         struct inode *inode = lsp->ls_state->inode;
4634
4635         p = kzalloc(sizeof(*p), GFP_NOFS);
4636         if (p == NULL)
4637                 return NULL;
4638         p->arg.fh = NFS_FH(inode);
4639         p->arg.fl = &p->fl;
4640         p->arg.seqid = seqid;
4641         p->res.seqid = seqid;
4642         p->arg.stateid = &lsp->ls_stateid;
4643         p->lsp = lsp;
4644         atomic_inc(&lsp->ls_count);
4645         /* Ensure we don't close file until we're done freeing locks! */
4646         p->ctx = get_nfs_open_context(ctx);
4647         memcpy(&p->fl, fl, sizeof(p->fl));
4648         p->server = NFS_SERVER(inode);
4649         return p;
4650 }
4651
4652 static void nfs4_locku_release_calldata(void *data)
4653 {
4654         struct nfs4_unlockdata *calldata = data;
4655         nfs_free_seqid(calldata->arg.seqid);
4656         nfs4_put_lock_state(calldata->lsp);
4657         put_nfs_open_context(calldata->ctx);
4658         kfree(calldata);
4659 }
4660
4661 static void nfs4_locku_done(struct rpc_task *task, void *data)
4662 {
4663         struct nfs4_unlockdata *calldata = data;
4664
4665         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4666                 return;
4667         switch (task->tk_status) {
4668                 case 0:
4669                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4670                                         &calldata->res.stateid);
4671                         renew_lease(calldata->server, calldata->timestamp);
4672                         break;
4673                 case -NFS4ERR_BAD_STATEID:
4674                 case -NFS4ERR_OLD_STATEID:
4675                 case -NFS4ERR_STALE_STATEID:
4676                 case -NFS4ERR_EXPIRED:
4677                         break;
4678                 default:
4679                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4680                                 rpc_restart_call_prepare(task);
4681         }
4682         nfs_release_seqid(calldata->arg.seqid);
4683 }
4684
4685 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4686 {
4687         struct nfs4_unlockdata *calldata = data;
4688
4689         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4690                 goto out_wait;
4691         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4692                 /* Note: exit _without_ running nfs4_locku_done */
4693                 goto out_no_action;
4694         }
4695         calldata->timestamp = jiffies;
4696         if (nfs4_setup_sequence(calldata->server,
4697                                 &calldata->arg.seq_args,
4698                                 &calldata->res.seq_res,
4699                                 task) != 0)
4700                 nfs_release_seqid(calldata->arg.seqid);
4701         return;
4702 out_no_action:
4703         task->tk_action = NULL;
4704 out_wait:
4705         nfs4_sequence_done(task, &calldata->res.seq_res);
4706 }
4707
4708 static const struct rpc_call_ops nfs4_locku_ops = {
4709         .rpc_call_prepare = nfs4_locku_prepare,
4710         .rpc_call_done = nfs4_locku_done,
4711         .rpc_release = nfs4_locku_release_calldata,
4712 };
4713
4714 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4715                 struct nfs_open_context *ctx,
4716                 struct nfs4_lock_state *lsp,
4717                 struct nfs_seqid *seqid)
4718 {
4719         struct nfs4_unlockdata *data;
4720         struct rpc_message msg = {
4721                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4722                 .rpc_cred = ctx->cred,
4723         };
4724         struct rpc_task_setup task_setup_data = {
4725                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4726                 .rpc_message = &msg,
4727                 .callback_ops = &nfs4_locku_ops,
4728                 .workqueue = nfsiod_workqueue,
4729                 .flags = RPC_TASK_ASYNC,
4730         };
4731
4732         /* Ensure this is an unlock - when canceling a lock, the
4733          * canceled lock is passed in, and it won't be an unlock.
4734          */
4735         fl->fl_type = F_UNLCK;
4736
4737         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4738         if (data == NULL) {
4739                 nfs_free_seqid(seqid);
4740                 return ERR_PTR(-ENOMEM);
4741         }
4742
4743         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4744         msg.rpc_argp = &data->arg;
4745         msg.rpc_resp = &data->res;
4746         task_setup_data.callback_data = data;
4747         return rpc_run_task(&task_setup_data);
4748 }
4749
4750 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4751 {
4752         struct inode *inode = state->inode;
4753         struct nfs4_state_owner *sp = state->owner;
4754         struct nfs_inode *nfsi = NFS_I(inode);
4755         struct nfs_seqid *seqid;
4756         struct nfs4_lock_state *lsp;
4757         struct rpc_task *task;
4758         int status = 0;
4759         unsigned char fl_flags = request->fl_flags;
4760
4761         status = nfs4_set_lock_state(state, request);
4762         /* Unlock _before_ we do the RPC call */
4763         request->fl_flags |= FL_EXISTS;
4764         /* Exclude nfs_delegation_claim_locks() */
4765         mutex_lock(&sp->so_delegreturn_mutex);
4766         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
4767         down_read(&nfsi->rwsem);
4768         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4769                 up_read(&nfsi->rwsem);
4770                 mutex_unlock(&sp->so_delegreturn_mutex);
4771                 goto out;
4772         }
4773         up_read(&nfsi->rwsem);
4774         mutex_unlock(&sp->so_delegreturn_mutex);
4775         if (status != 0)
4776                 goto out;
4777         /* Is this a delegated lock? */
4778         lsp = request->fl_u.nfs4_fl.owner;
4779         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
4780                 goto out;
4781         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4782         status = -ENOMEM;
4783         if (seqid == NULL)
4784                 goto out;
4785         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4786         status = PTR_ERR(task);
4787         if (IS_ERR(task))
4788                 goto out;
4789         status = nfs4_wait_for_completion_rpc_task(task);
4790         rpc_put_task(task);
4791 out:
4792         request->fl_flags = fl_flags;
4793         return status;
4794 }
4795
4796 struct nfs4_lockdata {
4797         struct nfs_lock_args arg;
4798         struct nfs_lock_res res;
4799         struct nfs4_lock_state *lsp;
4800         struct nfs_open_context *ctx;
4801         struct file_lock fl;
4802         unsigned long timestamp;
4803         int rpc_status;
4804         int cancelled;
4805         struct nfs_server *server;
4806 };
4807
4808 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4809                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4810                 gfp_t gfp_mask)
4811 {
4812         struct nfs4_lockdata *p;
4813         struct inode *inode = lsp->ls_state->inode;
4814         struct nfs_server *server = NFS_SERVER(inode);
4815
4816         p = kzalloc(sizeof(*p), gfp_mask);
4817         if (p == NULL)
4818                 return NULL;
4819
4820         p->arg.fh = NFS_FH(inode);
4821         p->arg.fl = &p->fl;
4822         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4823         if (p->arg.open_seqid == NULL)
4824                 goto out_free;
4825         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4826         if (p->arg.lock_seqid == NULL)
4827                 goto out_free_seqid;
4828         p->arg.lock_stateid = &lsp->ls_stateid;
4829         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4830         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4831         p->arg.lock_owner.s_dev = server->s_dev;
4832         p->res.lock_seqid = p->arg.lock_seqid;
4833         p->lsp = lsp;
4834         p->server = server;
4835         atomic_inc(&lsp->ls_count);
4836         p->ctx = get_nfs_open_context(ctx);
4837         memcpy(&p->fl, fl, sizeof(p->fl));
4838         return p;
4839 out_free_seqid:
4840         nfs_free_seqid(p->arg.open_seqid);
4841 out_free:
4842         kfree(p);
4843         return NULL;
4844 }
4845
4846 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4847 {
4848         struct nfs4_lockdata *data = calldata;
4849         struct nfs4_state *state = data->lsp->ls_state;
4850
4851         dprintk("%s: begin!\n", __func__);
4852         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4853                 goto out_wait;
4854         /* Do we need to do an open_to_lock_owner? */
4855         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4856                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
4857                         goto out_release_lock_seqid;
4858                 }
4859                 data->arg.open_stateid = &state->open_stateid;
4860                 data->arg.new_lock_owner = 1;
4861                 data->res.open_seqid = data->arg.open_seqid;
4862         } else
4863                 data->arg.new_lock_owner = 0;
4864         if (!nfs4_valid_open_stateid(state)) {
4865                 data->rpc_status = -EBADF;
4866                 task->tk_action = NULL;
4867                 goto out_release_open_seqid;
4868         }
4869         data->timestamp = jiffies;
4870         if (nfs4_setup_sequence(data->server,
4871                                 &data->arg.seq_args,
4872                                 &data->res.seq_res,
4873                                 task) == 0)
4874                 return;
4875 out_release_open_seqid:
4876         nfs_release_seqid(data->arg.open_seqid);
4877 out_release_lock_seqid:
4878         nfs_release_seqid(data->arg.lock_seqid);
4879 out_wait:
4880         nfs4_sequence_done(task, &data->res.seq_res);
4881         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4882 }
4883
4884 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4885 {
4886         struct nfs4_lockdata *data = calldata;
4887
4888         dprintk("%s: begin!\n", __func__);
4889
4890         if (!nfs4_sequence_done(task, &data->res.seq_res))
4891                 return;
4892
4893         data->rpc_status = task->tk_status;
4894         if (data->arg.new_lock_owner != 0) {
4895                 if (data->rpc_status == 0)
4896                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4897                 else
4898                         goto out;
4899         }
4900         if (data->rpc_status == 0) {
4901                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4902                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4903                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4904         }
4905 out:
4906         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4907 }
4908
4909 static void nfs4_lock_release(void *calldata)
4910 {
4911         struct nfs4_lockdata *data = calldata;
4912
4913         dprintk("%s: begin!\n", __func__);
4914         nfs_free_seqid(data->arg.open_seqid);
4915         if (data->cancelled != 0) {
4916                 struct rpc_task *task;
4917                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4918                                 data->arg.lock_seqid);
4919                 if (!IS_ERR(task))
4920                         rpc_put_task_async(task);
4921                 dprintk("%s: cancelling lock!\n", __func__);
4922         } else
4923                 nfs_free_seqid(data->arg.lock_seqid);
4924         nfs4_put_lock_state(data->lsp);
4925         put_nfs_open_context(data->ctx);
4926         kfree(data);
4927         dprintk("%s: done!\n", __func__);
4928 }
4929
4930 static const struct rpc_call_ops nfs4_lock_ops = {
4931         .rpc_call_prepare = nfs4_lock_prepare,
4932         .rpc_call_done = nfs4_lock_done,
4933         .rpc_release = nfs4_lock_release,
4934 };
4935
4936 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4937 {
4938         switch (error) {
4939         case -NFS4ERR_ADMIN_REVOKED:
4940         case -NFS4ERR_BAD_STATEID:
4941                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4942                 if (new_lock_owner != 0 ||
4943                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4944                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4945                 break;
4946         case -NFS4ERR_STALE_STATEID:
4947                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4948         case -NFS4ERR_EXPIRED:
4949                 nfs4_schedule_lease_recovery(server->nfs_client);
4950         };
4951 }
4952
4953 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4954 {
4955         struct nfs4_lockdata *data;
4956         struct rpc_task *task;
4957         struct rpc_message msg = {
4958                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4959                 .rpc_cred = state->owner->so_cred,
4960         };
4961         struct rpc_task_setup task_setup_data = {
4962                 .rpc_client = NFS_CLIENT(state->inode),
4963                 .rpc_message = &msg,
4964                 .callback_ops = &nfs4_lock_ops,
4965                 .workqueue = nfsiod_workqueue,
4966                 .flags = RPC_TASK_ASYNC,
4967         };
4968         int ret;
4969
4970         dprintk("%s: begin!\n", __func__);
4971         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4972                         fl->fl_u.nfs4_fl.owner,
4973                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4974         if (data == NULL)
4975                 return -ENOMEM;
4976         if (IS_SETLKW(cmd))
4977                 data->arg.block = 1;
4978         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4979         msg.rpc_argp = &data->arg;
4980         msg.rpc_resp = &data->res;
4981         task_setup_data.callback_data = data;
4982         if (recovery_type > NFS_LOCK_NEW) {
4983                 if (recovery_type == NFS_LOCK_RECLAIM)
4984                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4985                 nfs4_set_sequence_privileged(&data->arg.seq_args);
4986         }
4987         task = rpc_run_task(&task_setup_data);
4988         if (IS_ERR(task))
4989                 return PTR_ERR(task);
4990         ret = nfs4_wait_for_completion_rpc_task(task);
4991         if (ret == 0) {
4992                 ret = data->rpc_status;
4993                 if (ret)
4994                         nfs4_handle_setlk_error(data->server, data->lsp,
4995                                         data->arg.new_lock_owner, ret);
4996         } else
4997                 data->cancelled = 1;
4998         rpc_put_task(task);
4999         dprintk("%s: done, ret = %d!\n", __func__, ret);
5000         return ret;
5001 }
5002
5003 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5004 {
5005         struct nfs_server *server = NFS_SERVER(state->inode);
5006         struct nfs4_exception exception = {
5007                 .inode = state->inode,
5008         };
5009         int err;
5010
5011         do {
5012                 /* Cache the lock if possible... */
5013                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5014                         return 0;
5015                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5016                 if (err != -NFS4ERR_DELAY)
5017                         break;
5018                 nfs4_handle_exception(server, err, &exception);
5019         } while (exception.retry);
5020         return err;
5021 }
5022
5023 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5024 {
5025         struct nfs_server *server = NFS_SERVER(state->inode);
5026         struct nfs4_exception exception = {
5027                 .inode = state->inode,
5028         };
5029         int err;
5030
5031         err = nfs4_set_lock_state(state, request);
5032         if (err != 0)
5033                 return err;
5034         do {
5035                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5036                         return 0;
5037                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5038                 switch (err) {
5039                 default:
5040                         goto out;
5041                 case -NFS4ERR_GRACE:
5042                 case -NFS4ERR_DELAY:
5043                         nfs4_handle_exception(server, err, &exception);
5044                         err = 0;
5045                 }
5046         } while (exception.retry);
5047 out:
5048         return err;
5049 }
5050
5051 #if defined(CONFIG_NFS_V4_1)
5052 /**
5053  * nfs41_check_expired_locks - possibly free a lock stateid
5054  *
5055  * @state: NFSv4 state for an inode
5056  *
5057  * Returns NFS_OK if recovery for this stateid is now finished.
5058  * Otherwise a negative NFS4ERR value is returned.
5059  */
5060 static int nfs41_check_expired_locks(struct nfs4_state *state)
5061 {
5062         int status, ret = -NFS4ERR_BAD_STATEID;
5063         struct nfs4_lock_state *lsp;
5064         struct nfs_server *server = NFS_SERVER(state->inode);
5065
5066         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5067                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5068                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
5069                         if (status != NFS_OK) {
5070                                 /* Free the stateid unless the server
5071                                  * informs us the stateid is unrecognized. */
5072                                 if (status != -NFS4ERR_BAD_STATEID)
5073                                         nfs41_free_stateid(server,
5074                                                         &lsp->ls_stateid);
5075                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5076                                 ret = status;
5077                         }
5078                 }
5079         };
5080
5081         return ret;
5082 }
5083
5084 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5085 {
5086         int status = NFS_OK;
5087
5088         if (test_bit(LK_STATE_IN_USE, &state->flags))
5089                 status = nfs41_check_expired_locks(state);
5090         if (status != NFS_OK)
5091                 status = nfs4_lock_expired(state, request);
5092         return status;
5093 }
5094 #endif
5095
5096 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5097 {
5098         struct nfs4_state_owner *sp = state->owner;
5099         struct nfs_inode *nfsi = NFS_I(state->inode);
5100         unsigned char fl_flags = request->fl_flags;
5101         unsigned int seq;
5102         int status = -ENOLCK;
5103
5104         if ((fl_flags & FL_POSIX) &&
5105                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5106                 goto out;
5107         /* Is this a delegated open? */
5108         status = nfs4_set_lock_state(state, request);
5109         if (status != 0)
5110                 goto out;
5111         request->fl_flags |= FL_ACCESS;
5112         status = do_vfs_lock(request->fl_file, request);
5113         if (status < 0)
5114                 goto out;
5115         down_read(&nfsi->rwsem);
5116         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5117                 /* Yes: cache locks! */
5118                 /* ...but avoid races with delegation recall... */
5119                 request->fl_flags = fl_flags & ~FL_SLEEP;
5120                 status = do_vfs_lock(request->fl_file, request);
5121                 goto out_unlock;
5122         }
5123         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5124         up_read(&nfsi->rwsem);
5125         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5126         if (status != 0)
5127                 goto out;
5128         down_read(&nfsi->rwsem);
5129         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5130                 status = -NFS4ERR_DELAY;
5131                 goto out_unlock;
5132         }
5133         /* Note: we always want to sleep here! */
5134         request->fl_flags = fl_flags | FL_SLEEP;
5135         if (do_vfs_lock(request->fl_file, request) < 0)
5136                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5137                         "manager!\n", __func__);
5138 out_unlock:
5139         up_read(&nfsi->rwsem);
5140 out:
5141         request->fl_flags = fl_flags;
5142         return status;
5143 }
5144
5145 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5146 {
5147         struct nfs4_exception exception = {
5148                 .state = state,
5149                 .inode = state->inode,
5150         };
5151         int err;
5152
5153         do {
5154                 err = _nfs4_proc_setlk(state, cmd, request);
5155                 if (err == -NFS4ERR_DENIED)
5156                         err = -EAGAIN;
5157                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5158                                 err, &exception);
5159         } while (exception.retry);
5160         return err;
5161 }
5162
5163 static int
5164 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5165 {
5166         struct nfs_open_context *ctx;
5167         struct nfs4_state *state;
5168         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5169         int status;
5170
5171         /* verify open state */
5172         ctx = nfs_file_open_context(filp);
5173         state = ctx->state;
5174
5175         if (request->fl_start < 0 || request->fl_end < 0)
5176                 return -EINVAL;
5177
5178         if (IS_GETLK(cmd)) {
5179                 if (state != NULL)
5180                         return nfs4_proc_getlk(state, F_GETLK, request);
5181                 return 0;
5182         }
5183
5184         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5185                 return -EINVAL;
5186
5187         if (request->fl_type == F_UNLCK) {
5188                 if (state != NULL)
5189                         return nfs4_proc_unlck(state, cmd, request);
5190                 return 0;
5191         }
5192
5193         if (state == NULL)
5194                 return -ENOLCK;
5195         /*
5196          * Don't rely on the VFS having checked the file open mode,
5197          * since it won't do this for flock() locks.
5198          */
5199         switch (request->fl_type) {
5200         case F_RDLCK:
5201                 if (!(filp->f_mode & FMODE_READ))
5202                         return -EBADF;
5203                 break;
5204         case F_WRLCK:
5205                 if (!(filp->f_mode & FMODE_WRITE))
5206                         return -EBADF;
5207         }
5208
5209         do {
5210                 status = nfs4_proc_setlk(state, cmd, request);
5211                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5212                         break;
5213                 timeout = nfs4_set_lock_task_retry(timeout);
5214                 status = -ERESTARTSYS;
5215                 if (signalled())
5216                         break;
5217         } while(status < 0);
5218         return status;
5219 }
5220
5221 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5222 {
5223         struct nfs_server *server = NFS_SERVER(state->inode);
5224         int err;
5225
5226         err = nfs4_set_lock_state(state, fl);
5227         if (err != 0)
5228                 return err;
5229         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5230         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5231 }
5232
5233 struct nfs_release_lockowner_data {
5234         struct nfs4_lock_state *lsp;
5235         struct nfs_server *server;
5236         struct nfs_release_lockowner_args args;
5237 };
5238
5239 static void nfs4_release_lockowner_release(void *calldata)
5240 {
5241         struct nfs_release_lockowner_data *data = calldata;
5242         nfs4_free_lock_state(data->server, data->lsp);
5243         kfree(calldata);
5244 }
5245
5246 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5247         .rpc_release = nfs4_release_lockowner_release,
5248 };
5249
5250 static int nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
5251 {
5252         struct nfs_release_lockowner_data *data;
5253         struct rpc_message msg = {
5254                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5255         };
5256
5257         if (server->nfs_client->cl_mvops->minor_version != 0)
5258                 return -EINVAL;
5259         data = kmalloc(sizeof(*data), GFP_NOFS);
5260         if (!data)
5261                 return -ENOMEM;
5262         data->lsp = lsp;
5263         data->server = server;
5264         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5265         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5266         data->args.lock_owner.s_dev = server->s_dev;
5267         msg.rpc_argp = &data->args;
5268         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5269         return 0;
5270 }
5271
5272 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5273
5274 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5275                                    const void *buf, size_t buflen,
5276                                    int flags, int type)
5277 {
5278         if (strcmp(key, "") != 0)
5279                 return -EINVAL;
5280
5281         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5282 }
5283
5284 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5285                                    void *buf, size_t buflen, int type)
5286 {
5287         if (strcmp(key, "") != 0)
5288                 return -EINVAL;
5289
5290         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5291 }
5292
5293 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5294                                        size_t list_len, const char *name,
5295                                        size_t name_len, int type)
5296 {
5297         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5298
5299         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5300                 return 0;
5301
5302         if (list && len <= list_len)
5303                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5304         return len;
5305 }
5306
5307 /*
5308  * nfs_fhget will use either the mounted_on_fileid or the fileid
5309  */
5310 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5311 {
5312         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5313                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5314               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5315               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5316                 return;
5317
5318         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5319                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5320         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5321         fattr->nlink = 2;
5322 }
5323
5324 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5325                                    const struct qstr *name,
5326                                    struct nfs4_fs_locations *fs_locations,
5327                                    struct page *page)
5328 {
5329         struct nfs_server *server = NFS_SERVER(dir);
5330         u32 bitmask[2] = {
5331                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5332         };
5333         struct nfs4_fs_locations_arg args = {
5334                 .dir_fh = NFS_FH(dir),
5335                 .name = name,
5336                 .page = page,
5337                 .bitmask = bitmask,
5338         };
5339         struct nfs4_fs_locations_res res = {
5340                 .fs_locations = fs_locations,
5341         };
5342         struct rpc_message msg = {
5343                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5344                 .rpc_argp = &args,
5345                 .rpc_resp = &res,
5346         };
5347         int status;
5348
5349         dprintk("%s: start\n", __func__);
5350
5351         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5352          * is not supported */
5353         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5354                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5355         else
5356                 bitmask[0] |= FATTR4_WORD0_FILEID;
5357
5358         nfs_fattr_init(&fs_locations->fattr);
5359         fs_locations->server = server;
5360         fs_locations->nlocations = 0;
5361         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5362         dprintk("%s: returned status = %d\n", __func__, status);
5363         return status;
5364 }
5365
5366 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5367                            const struct qstr *name,
5368                            struct nfs4_fs_locations *fs_locations,
5369                            struct page *page)
5370 {
5371         struct nfs4_exception exception = { };
5372         int err;
5373         do {
5374                 err = nfs4_handle_exception(NFS_SERVER(dir),
5375                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5376                                 &exception);
5377         } while (exception.retry);
5378         return err;
5379 }
5380
5381 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5382 {
5383         int status;
5384         struct nfs4_secinfo_arg args = {
5385                 .dir_fh = NFS_FH(dir),
5386                 .name   = name,
5387         };
5388         struct nfs4_secinfo_res res = {
5389                 .flavors     = flavors,
5390         };
5391         struct rpc_message msg = {
5392                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5393                 .rpc_argp = &args,
5394                 .rpc_resp = &res,
5395         };
5396
5397         dprintk("NFS call  secinfo %s\n", name->name);
5398         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5399         dprintk("NFS reply  secinfo: %d\n", status);
5400         return status;
5401 }
5402
5403 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5404                       struct nfs4_secinfo_flavors *flavors)
5405 {
5406         struct nfs4_exception exception = { };
5407         int err;
5408         do {
5409                 err = nfs4_handle_exception(NFS_SERVER(dir),
5410                                 _nfs4_proc_secinfo(dir, name, flavors),
5411                                 &exception);
5412         } while (exception.retry);
5413         return err;
5414 }
5415
5416 #ifdef CONFIG_NFS_V4_1
5417 /*
5418  * Check the exchange flags returned by the server for invalid flags, having
5419  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5420  * DS flags set.
5421  */
5422 static int nfs4_check_cl_exchange_flags(u32 flags)
5423 {
5424         if (flags & ~EXCHGID4_FLAG_MASK_R)
5425                 goto out_inval;
5426         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5427             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5428                 goto out_inval;
5429         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5430                 goto out_inval;
5431         return NFS_OK;
5432 out_inval:
5433         return -NFS4ERR_INVAL;
5434 }
5435
5436 static bool
5437 nfs41_same_server_scope(struct nfs41_server_scope *a,
5438                         struct nfs41_server_scope *b)
5439 {
5440         if (a->server_scope_sz == b->server_scope_sz &&
5441             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5442                 return true;
5443
5444         return false;
5445 }
5446
5447 /*
5448  * nfs4_proc_bind_conn_to_session()
5449  *
5450  * The 4.1 client currently uses the same TCP connection for the
5451  * fore and backchannel.
5452  */
5453 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5454 {
5455         int status;
5456         struct nfs41_bind_conn_to_session_res res;
5457         struct rpc_message msg = {
5458                 .rpc_proc =
5459                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5460                 .rpc_argp = clp,
5461                 .rpc_resp = &res,
5462                 .rpc_cred = cred,
5463         };
5464
5465         dprintk("--> %s\n", __func__);
5466
5467         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5468         if (unlikely(res.session == NULL)) {
5469                 status = -ENOMEM;
5470                 goto out;
5471         }
5472
5473         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5474         if (status == 0) {
5475                 if (memcmp(res.session->sess_id.data,
5476                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5477                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5478                         status = -EIO;
5479                         goto out_session;
5480                 }
5481                 if (res.dir != NFS4_CDFS4_BOTH) {
5482                         dprintk("NFS: %s: Unexpected direction from server\n",
5483                                 __func__);
5484                         status = -EIO;
5485                         goto out_session;
5486                 }
5487                 if (res.use_conn_in_rdma_mode) {
5488                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5489                                 __func__);
5490                         status = -EIO;
5491                         goto out_session;
5492                 }
5493         }
5494 out_session:
5495         kfree(res.session);
5496 out:
5497         dprintk("<-- %s status= %d\n", __func__, status);
5498         return status;
5499 }
5500
5501 /*
5502  * nfs4_proc_exchange_id()
5503  *
5504  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5505  *
5506  * Since the clientid has expired, all compounds using sessions
5507  * associated with the stale clientid will be returning
5508  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5509  * be in some phase of session reset.
5510  */
5511 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5512 {
5513         nfs4_verifier verifier;
5514         struct nfs41_exchange_id_args args = {
5515                 .verifier = &verifier,
5516                 .client = clp,
5517                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5518         };
5519         struct nfs41_exchange_id_res res = {
5520                 0
5521         };
5522         int status;
5523         struct rpc_message msg = {
5524                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5525                 .rpc_argp = &args,
5526                 .rpc_resp = &res,
5527                 .rpc_cred = cred,
5528         };
5529
5530         nfs4_init_boot_verifier(clp, &verifier);
5531         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5532                                                         sizeof(args.id));
5533         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5534                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5535                 args.id_len, args.id);
5536
5537         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5538                                         GFP_NOFS);
5539         if (unlikely(res.server_owner == NULL)) {
5540                 status = -ENOMEM;
5541                 goto out;
5542         }
5543
5544         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5545                                         GFP_NOFS);
5546         if (unlikely(res.server_scope == NULL)) {
5547                 status = -ENOMEM;
5548                 goto out_server_owner;
5549         }
5550
5551         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5552         if (unlikely(res.impl_id == NULL)) {
5553                 status = -ENOMEM;
5554                 goto out_server_scope;
5555         }
5556
5557         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5558         if (status == 0)
5559                 status = nfs4_check_cl_exchange_flags(res.flags);
5560
5561         if (status == 0) {
5562                 clp->cl_clientid = res.clientid;
5563                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5564                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5565                         clp->cl_seqid = res.seqid;
5566
5567                 kfree(clp->cl_serverowner);
5568                 clp->cl_serverowner = res.server_owner;
5569                 res.server_owner = NULL;
5570
5571                 /* use the most recent implementation id */
5572                 kfree(clp->cl_implid);
5573                 clp->cl_implid = res.impl_id;
5574
5575                 if (clp->cl_serverscope != NULL &&
5576                     !nfs41_same_server_scope(clp->cl_serverscope,
5577                                              res.server_scope)) {
5578                         dprintk("%s: server_scope mismatch detected\n",
5579                                 __func__);
5580                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5581                         kfree(clp->cl_serverscope);
5582                         clp->cl_serverscope = NULL;
5583                 }
5584
5585                 if (clp->cl_serverscope == NULL) {
5586                         clp->cl_serverscope = res.server_scope;
5587                         goto out;
5588                 }
5589         } else
5590                 kfree(res.impl_id);
5591
5592 out_server_owner:
5593         kfree(res.server_owner);
5594 out_server_scope:
5595         kfree(res.server_scope);
5596 out:
5597         if (clp->cl_implid != NULL)
5598                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5599                         "domain: %s, name: %s, date: %llu,%u\n",
5600                         clp->cl_implid->domain, clp->cl_implid->name,
5601                         clp->cl_implid->date.seconds,
5602                         clp->cl_implid->date.nseconds);
5603         dprintk("NFS reply exchange_id: %d\n", status);
5604         return status;
5605 }
5606
5607 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5608                 struct rpc_cred *cred)
5609 {
5610         struct rpc_message msg = {
5611                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5612                 .rpc_argp = clp,
5613                 .rpc_cred = cred,
5614         };
5615         int status;
5616
5617         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5618         if (status)
5619                 dprintk("NFS: Got error %d from the server %s on "
5620                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5621         return status;
5622 }
5623
5624 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5625                 struct rpc_cred *cred)
5626 {
5627         unsigned int loop;
5628         int ret;
5629
5630         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5631                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5632                 switch (ret) {
5633                 case -NFS4ERR_DELAY:
5634                 case -NFS4ERR_CLIENTID_BUSY:
5635                         ssleep(1);
5636                         break;
5637                 default:
5638                         return ret;
5639                 }
5640         }
5641         return 0;
5642 }
5643
5644 int nfs4_destroy_clientid(struct nfs_client *clp)
5645 {
5646         struct rpc_cred *cred;
5647         int ret = 0;
5648
5649         if (clp->cl_mvops->minor_version < 1)
5650                 goto out;
5651         if (clp->cl_exchange_flags == 0)
5652                 goto out;
5653         if (clp->cl_preserve_clid)
5654                 goto out;
5655         cred = nfs4_get_exchange_id_cred(clp);
5656         ret = nfs4_proc_destroy_clientid(clp, cred);
5657         if (cred)
5658                 put_rpccred(cred);
5659         switch (ret) {
5660         case 0:
5661         case -NFS4ERR_STALE_CLIENTID:
5662                 clp->cl_exchange_flags = 0;
5663         }
5664 out:
5665         return ret;
5666 }
5667
5668 struct nfs4_get_lease_time_data {
5669         struct nfs4_get_lease_time_args *args;
5670         struct nfs4_get_lease_time_res *res;
5671         struct nfs_client *clp;
5672 };
5673
5674 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5675                                         void *calldata)
5676 {
5677         struct nfs4_get_lease_time_data *data =
5678                         (struct nfs4_get_lease_time_data *)calldata;
5679
5680         dprintk("--> %s\n", __func__);
5681         /* just setup sequence, do not trigger session recovery
5682            since we're invoked within one */
5683         nfs41_setup_sequence(data->clp->cl_session,
5684                         &data->args->la_seq_args,
5685                         &data->res->lr_seq_res,
5686                         task);
5687         dprintk("<-- %s\n", __func__);
5688 }
5689
5690 /*
5691  * Called from nfs4_state_manager thread for session setup, so don't recover
5692  * from sequence operation or clientid errors.
5693  */
5694 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5695 {
5696         struct nfs4_get_lease_time_data *data =
5697                         (struct nfs4_get_lease_time_data *)calldata;
5698
5699         dprintk("--> %s\n", __func__);
5700         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5701                 return;
5702         switch (task->tk_status) {
5703         case -NFS4ERR_DELAY:
5704         case -NFS4ERR_GRACE:
5705                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5706                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5707                 task->tk_status = 0;
5708                 /* fall through */
5709         case -NFS4ERR_RETRY_UNCACHED_REP:
5710                 rpc_restart_call_prepare(task);
5711                 return;
5712         }
5713         dprintk("<-- %s\n", __func__);
5714 }
5715
5716 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5717         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5718         .rpc_call_done = nfs4_get_lease_time_done,
5719 };
5720
5721 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5722 {
5723         struct rpc_task *task;
5724         struct nfs4_get_lease_time_args args;
5725         struct nfs4_get_lease_time_res res = {
5726                 .lr_fsinfo = fsinfo,
5727         };
5728         struct nfs4_get_lease_time_data data = {
5729                 .args = &args,
5730                 .res = &res,
5731                 .clp = clp,
5732         };
5733         struct rpc_message msg = {
5734                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5735                 .rpc_argp = &args,
5736                 .rpc_resp = &res,
5737         };
5738         struct rpc_task_setup task_setup = {
5739                 .rpc_client = clp->cl_rpcclient,
5740                 .rpc_message = &msg,
5741                 .callback_ops = &nfs4_get_lease_time_ops,
5742                 .callback_data = &data,
5743                 .flags = RPC_TASK_TIMEOUT,
5744         };
5745         int status;
5746
5747         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5748         nfs4_set_sequence_privileged(&args.la_seq_args);
5749         dprintk("--> %s\n", __func__);
5750         task = rpc_run_task(&task_setup);
5751
5752         if (IS_ERR(task))
5753                 status = PTR_ERR(task);
5754         else {
5755                 status = task->tk_status;
5756                 rpc_put_task(task);
5757         }
5758         dprintk("<-- %s return %d\n", __func__, status);
5759
5760         return status;
5761 }
5762
5763 /*
5764  * Initialize the values to be used by the client in CREATE_SESSION
5765  * If nfs4_init_session set the fore channel request and response sizes,
5766  * use them.
5767  *
5768  * Set the back channel max_resp_sz_cached to zero to force the client to
5769  * always set csa_cachethis to FALSE because the current implementation
5770  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5771  */
5772 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5773 {
5774         struct nfs4_session *session = args->client->cl_session;
5775         unsigned int mxrqst_sz = session->fc_target_max_rqst_sz,
5776                      mxresp_sz = session->fc_target_max_resp_sz;
5777
5778         if (mxrqst_sz == 0)
5779                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5780         if (mxresp_sz == 0)
5781                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5782         /* Fore channel attributes */
5783         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5784         args->fc_attrs.max_resp_sz = mxresp_sz;
5785         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5786         args->fc_attrs.max_reqs = max_session_slots;
5787
5788         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5789                 "max_ops=%u max_reqs=%u\n",
5790                 __func__,
5791                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5792                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5793
5794         /* Back channel attributes */
5795         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5796         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5797         args->bc_attrs.max_resp_sz_cached = 0;
5798         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5799         args->bc_attrs.max_reqs = 1;
5800
5801         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5802                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5803                 __func__,
5804                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5805                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5806                 args->bc_attrs.max_reqs);
5807 }
5808
5809 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5810 {
5811         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5812         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5813
5814         if (rcvd->max_resp_sz > sent->max_resp_sz)
5815                 return -EINVAL;
5816         /*
5817          * Our requested max_ops is the minimum we need; we're not
5818          * prepared to break up compounds into smaller pieces than that.
5819          * So, no point even trying to continue if the server won't
5820          * cooperate:
5821          */
5822         if (rcvd->max_ops < sent->max_ops)
5823                 return -EINVAL;
5824         if (rcvd->max_reqs == 0)
5825                 return -EINVAL;
5826         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5827                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5828         return 0;
5829 }
5830
5831 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5832 {
5833         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5834         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5835
5836         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5837                 return -EINVAL;
5838         if (rcvd->max_resp_sz < sent->max_resp_sz)
5839                 return -EINVAL;
5840         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5841                 return -EINVAL;
5842         /* These would render the backchannel useless: */
5843         if (rcvd->max_ops != sent->max_ops)
5844                 return -EINVAL;
5845         if (rcvd->max_reqs != sent->max_reqs)
5846                 return -EINVAL;
5847         return 0;
5848 }
5849
5850 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5851                                      struct nfs4_session *session)
5852 {
5853         int ret;
5854
5855         ret = nfs4_verify_fore_channel_attrs(args, session);
5856         if (ret)
5857                 return ret;
5858         return nfs4_verify_back_channel_attrs(args, session);
5859 }
5860
5861 static int _nfs4_proc_create_session(struct nfs_client *clp,
5862                 struct rpc_cred *cred)
5863 {
5864         struct nfs4_session *session = clp->cl_session;
5865         struct nfs41_create_session_args args = {
5866                 .client = clp,
5867                 .cb_program = NFS4_CALLBACK,
5868         };
5869         struct nfs41_create_session_res res = {
5870                 .client = clp,
5871         };
5872         struct rpc_message msg = {
5873                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5874                 .rpc_argp = &args,
5875                 .rpc_resp = &res,
5876                 .rpc_cred = cred,
5877         };
5878         int status;
5879
5880         nfs4_init_channel_attrs(&args);
5881         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5882
5883         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5884
5885         if (!status) {
5886                 /* Verify the session's negotiated channel_attrs values */
5887                 status = nfs4_verify_channel_attrs(&args, session);
5888                 /* Increment the clientid slot sequence id */
5889                 clp->cl_seqid++;
5890         }
5891
5892         return status;
5893 }
5894
5895 /*
5896  * Issues a CREATE_SESSION operation to the server.
5897  * It is the responsibility of the caller to verify the session is
5898  * expired before calling this routine.
5899  */
5900 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5901 {
5902         int status;
5903         unsigned *ptr;
5904         struct nfs4_session *session = clp->cl_session;
5905
5906         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5907
5908         status = _nfs4_proc_create_session(clp, cred);
5909         if (status)
5910                 goto out;
5911
5912         /* Init or reset the session slot tables */
5913         status = nfs4_setup_session_slot_tables(session);
5914         dprintk("slot table setup returned %d\n", status);
5915         if (status)
5916                 goto out;
5917
5918         ptr = (unsigned *)&session->sess_id.data[0];
5919         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5920                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5921 out:
5922         dprintk("<-- %s\n", __func__);
5923         return status;
5924 }
5925
5926 /*
5927  * Issue the over-the-wire RPC DESTROY_SESSION.
5928  * The caller must serialize access to this routine.
5929  */
5930 int nfs4_proc_destroy_session(struct nfs4_session *session,
5931                 struct rpc_cred *cred)
5932 {
5933         struct rpc_message msg = {
5934                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5935                 .rpc_argp = session,
5936                 .rpc_cred = cred,
5937         };
5938         int status = 0;
5939
5940         dprintk("--> nfs4_proc_destroy_session\n");
5941
5942         /* session is still being setup */
5943         if (session->clp->cl_cons_state != NFS_CS_READY)
5944                 return status;
5945
5946         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5947
5948         if (status)
5949                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5950                         "Session has been destroyed regardless...\n", status);
5951
5952         dprintk("<-- nfs4_proc_destroy_session\n");
5953         return status;
5954 }
5955
5956 /*
5957  * Renew the cl_session lease.
5958  */
5959 struct nfs4_sequence_data {
5960         struct nfs_client *clp;
5961         struct nfs4_sequence_args args;
5962         struct nfs4_sequence_res res;
5963 };
5964
5965 static void nfs41_sequence_release(void *data)
5966 {
5967         struct nfs4_sequence_data *calldata = data;
5968         struct nfs_client *clp = calldata->clp;
5969
5970         if (atomic_read(&clp->cl_count) > 1)
5971                 nfs4_schedule_state_renewal(clp);
5972         nfs_put_client(clp);
5973         kfree(calldata);
5974 }
5975
5976 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5977 {
5978         switch(task->tk_status) {
5979         case -NFS4ERR_DELAY:
5980                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5981                 return -EAGAIN;
5982         default:
5983                 nfs4_schedule_lease_recovery(clp);
5984         }
5985         return 0;
5986 }
5987
5988 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5989 {
5990         struct nfs4_sequence_data *calldata = data;
5991         struct nfs_client *clp = calldata->clp;
5992
5993         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5994                 return;
5995
5996         if (task->tk_status < 0) {
5997                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5998                 if (atomic_read(&clp->cl_count) == 1)
5999                         goto out;
6000
6001                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
6002                         rpc_restart_call_prepare(task);
6003                         return;
6004                 }
6005         }
6006         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
6007 out:
6008         dprintk("<-- %s\n", __func__);
6009 }
6010
6011 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
6012 {
6013         struct nfs4_sequence_data *calldata = data;
6014         struct nfs_client *clp = calldata->clp;
6015         struct nfs4_sequence_args *args;
6016         struct nfs4_sequence_res *res;
6017
6018         args = task->tk_msg.rpc_argp;
6019         res = task->tk_msg.rpc_resp;
6020
6021         nfs41_setup_sequence(clp->cl_session, args, res, task);
6022 }
6023
6024 static const struct rpc_call_ops nfs41_sequence_ops = {
6025         .rpc_call_done = nfs41_sequence_call_done,
6026         .rpc_call_prepare = nfs41_sequence_prepare,
6027         .rpc_release = nfs41_sequence_release,
6028 };
6029
6030 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
6031                 struct rpc_cred *cred,
6032                 bool is_privileged)
6033 {
6034         struct nfs4_sequence_data *calldata;
6035         struct rpc_message msg = {
6036                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
6037                 .rpc_cred = cred,
6038         };
6039         struct rpc_task_setup task_setup_data = {
6040                 .rpc_client = clp->cl_rpcclient,
6041                 .rpc_message = &msg,
6042                 .callback_ops = &nfs41_sequence_ops,
6043                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6044         };
6045
6046         if (!atomic_inc_not_zero(&clp->cl_count))
6047                 return ERR_PTR(-EIO);
6048         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6049         if (calldata == NULL) {
6050                 nfs_put_client(clp);
6051                 return ERR_PTR(-ENOMEM);
6052         }
6053         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6054         if (is_privileged)
6055                 nfs4_set_sequence_privileged(&calldata->args);
6056         msg.rpc_argp = &calldata->args;
6057         msg.rpc_resp = &calldata->res;
6058         calldata->clp = clp;
6059         task_setup_data.callback_data = calldata;
6060
6061         return rpc_run_task(&task_setup_data);
6062 }
6063
6064 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6065 {
6066         struct rpc_task *task;
6067         int ret = 0;
6068
6069         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6070                 return -EAGAIN;
6071         task = _nfs41_proc_sequence(clp, cred, false);
6072         if (IS_ERR(task))
6073                 ret = PTR_ERR(task);
6074         else
6075                 rpc_put_task_async(task);
6076         dprintk("<-- %s status=%d\n", __func__, ret);
6077         return ret;
6078 }
6079
6080 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6081 {
6082         struct rpc_task *task;
6083         int ret;
6084
6085         task = _nfs41_proc_sequence(clp, cred, true);
6086         if (IS_ERR(task)) {
6087                 ret = PTR_ERR(task);
6088                 goto out;
6089         }
6090         ret = rpc_wait_for_completion_task(task);
6091         if (!ret) {
6092                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6093
6094                 if (task->tk_status == 0)
6095                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6096                 ret = task->tk_status;
6097         }
6098         rpc_put_task(task);
6099 out:
6100         dprintk("<-- %s status=%d\n", __func__, ret);
6101         return ret;
6102 }
6103
6104 struct nfs4_reclaim_complete_data {
6105         struct nfs_client *clp;
6106         struct nfs41_reclaim_complete_args arg;
6107         struct nfs41_reclaim_complete_res res;
6108 };
6109
6110 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6111 {
6112         struct nfs4_reclaim_complete_data *calldata = data;
6113
6114         nfs41_setup_sequence(calldata->clp->cl_session,
6115                         &calldata->arg.seq_args,
6116                         &calldata->res.seq_res,
6117                         task);
6118 }
6119
6120 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6121 {
6122         switch(task->tk_status) {
6123         case 0:
6124         case -NFS4ERR_COMPLETE_ALREADY:
6125         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6126                 break;
6127         case -NFS4ERR_DELAY:
6128                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6129                 /* fall through */
6130         case -NFS4ERR_RETRY_UNCACHED_REP:
6131                 return -EAGAIN;
6132         default:
6133                 nfs4_schedule_lease_recovery(clp);
6134         }
6135         return 0;
6136 }
6137
6138 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6139 {
6140         struct nfs4_reclaim_complete_data *calldata = data;
6141         struct nfs_client *clp = calldata->clp;
6142         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6143
6144         dprintk("--> %s\n", __func__);
6145         if (!nfs41_sequence_done(task, res))
6146                 return;
6147
6148         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6149                 rpc_restart_call_prepare(task);
6150                 return;
6151         }
6152         dprintk("<-- %s\n", __func__);
6153 }
6154
6155 static void nfs4_free_reclaim_complete_data(void *data)
6156 {
6157         struct nfs4_reclaim_complete_data *calldata = data;
6158
6159         kfree(calldata);
6160 }
6161
6162 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6163         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6164         .rpc_call_done = nfs4_reclaim_complete_done,
6165         .rpc_release = nfs4_free_reclaim_complete_data,
6166 };
6167
6168 /*
6169  * Issue a global reclaim complete.
6170  */
6171 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6172 {
6173         struct nfs4_reclaim_complete_data *calldata;
6174         struct rpc_task *task;
6175         struct rpc_message msg = {
6176                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6177         };
6178         struct rpc_task_setup task_setup_data = {
6179                 .rpc_client = clp->cl_rpcclient,
6180                 .rpc_message = &msg,
6181                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6182                 .flags = RPC_TASK_ASYNC,
6183         };
6184         int status = -ENOMEM;
6185
6186         dprintk("--> %s\n", __func__);
6187         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6188         if (calldata == NULL)
6189                 goto out;
6190         calldata->clp = clp;
6191         calldata->arg.one_fs = 0;
6192
6193         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6194         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
6195         msg.rpc_argp = &calldata->arg;
6196         msg.rpc_resp = &calldata->res;
6197         task_setup_data.callback_data = calldata;
6198         task = rpc_run_task(&task_setup_data);
6199         if (IS_ERR(task)) {
6200                 status = PTR_ERR(task);
6201                 goto out;
6202         }
6203         status = nfs4_wait_for_completion_rpc_task(task);
6204         if (status == 0)
6205                 status = task->tk_status;
6206         rpc_put_task(task);
6207         return 0;
6208 out:
6209         dprintk("<-- %s status=%d\n", __func__, status);
6210         return status;
6211 }
6212
6213 static void
6214 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6215 {
6216         struct nfs4_layoutget *lgp = calldata;
6217         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6218         struct nfs4_session *session = nfs4_get_session(server);
6219
6220         dprintk("--> %s\n", __func__);
6221         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6222          * right now covering the LAYOUTGET we are about to send.
6223          * However, that is not so catastrophic, and there seems
6224          * to be no way to prevent it completely.
6225          */
6226         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
6227                                 &lgp->res.seq_res, task))
6228                 return;
6229         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6230                                           NFS_I(lgp->args.inode)->layout,
6231                                           lgp->args.ctx->state)) {
6232                 rpc_exit(task, NFS4_OK);
6233         }
6234 }
6235
6236 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6237 {
6238         struct nfs4_layoutget *lgp = calldata;
6239         struct inode *inode = lgp->args.inode;
6240         struct nfs_server *server = NFS_SERVER(inode);
6241         struct pnfs_layout_hdr *lo;
6242         struct nfs4_state *state = NULL;
6243         unsigned long timeo, now, giveup;
6244
6245         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
6246
6247         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
6248                 goto out;
6249
6250         switch (task->tk_status) {
6251         case 0:
6252                 goto out;
6253         /*
6254          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
6255          * (or clients) writing to the same RAID stripe
6256          */
6257         case -NFS4ERR_LAYOUTTRYLATER:
6258         /*
6259          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
6260          * existing layout before getting a new one).
6261          */
6262         case -NFS4ERR_RECALLCONFLICT:
6263                 timeo = rpc_get_timeout(task->tk_client);
6264                 giveup = lgp->args.timestamp + timeo;
6265                 now = jiffies;
6266                 if (time_after(giveup, now)) {
6267                         unsigned long delay;
6268
6269                         /* Delay for:
6270                          * - Not less then NFS4_POLL_RETRY_MIN.
6271                          * - One last time a jiffie before we give up
6272                          * - exponential backoff (time_now minus start_attempt)
6273                          */
6274                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
6275                                     min((giveup - now - 1),
6276                                         now - lgp->args.timestamp));
6277
6278                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
6279                                 __func__, delay);
6280                         rpc_delay(task, delay);
6281                         task->tk_status = 0;
6282                         rpc_restart_call_prepare(task);
6283                         goto out; /* Do not call nfs4_async_handle_error() */
6284                 }
6285                 break;
6286         case -NFS4ERR_EXPIRED:
6287         case -NFS4ERR_BAD_STATEID:
6288                 spin_lock(&inode->i_lock);
6289                 lo = NFS_I(inode)->layout;
6290                 if (!lo || list_empty(&lo->plh_segs)) {
6291                         spin_unlock(&inode->i_lock);
6292                         /* If the open stateid was bad, then recover it. */
6293                         state = lgp->args.ctx->state;
6294                 } else {
6295                         LIST_HEAD(head);
6296
6297                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6298                         spin_unlock(&inode->i_lock);
6299                         /* Mark the bad layout state as invalid, then
6300                          * retry using the open stateid. */
6301                         pnfs_free_lseg_list(&head);
6302                 }
6303         }
6304         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6305                 rpc_restart_call_prepare(task);
6306 out:
6307         dprintk("<-- %s\n", __func__);
6308 }
6309
6310 static size_t max_response_pages(struct nfs_server *server)
6311 {
6312         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6313         return nfs_page_array_len(0, max_resp_sz);
6314 }
6315
6316 static void nfs4_free_pages(struct page **pages, size_t size)
6317 {
6318         int i;
6319
6320         if (!pages)
6321                 return;
6322
6323         for (i = 0; i < size; i++) {
6324                 if (!pages[i])
6325                         break;
6326                 __free_page(pages[i]);
6327         }
6328         kfree(pages);
6329 }
6330
6331 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6332 {
6333         struct page **pages;
6334         int i;
6335
6336         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6337         if (!pages) {
6338                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6339                 return NULL;
6340         }
6341
6342         for (i = 0; i < size; i++) {
6343                 pages[i] = alloc_page(gfp_flags);
6344                 if (!pages[i]) {
6345                         dprintk("%s: failed to allocate page\n", __func__);
6346                         nfs4_free_pages(pages, size);
6347                         return NULL;
6348                 }
6349         }
6350
6351         return pages;
6352 }
6353
6354 static void nfs4_layoutget_release(void *calldata)
6355 {
6356         struct nfs4_layoutget *lgp = calldata;
6357         struct inode *inode = lgp->args.inode;
6358         struct nfs_server *server = NFS_SERVER(inode);
6359         size_t max_pages = max_response_pages(server);
6360
6361         dprintk("--> %s\n", __func__);
6362         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6363         pnfs_put_layout_hdr(NFS_I(inode)->layout);
6364         put_nfs_open_context(lgp->args.ctx);
6365         kfree(calldata);
6366         dprintk("<-- %s\n", __func__);
6367 }
6368
6369 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6370         .rpc_call_prepare = nfs4_layoutget_prepare,
6371         .rpc_call_done = nfs4_layoutget_done,
6372         .rpc_release = nfs4_layoutget_release,
6373 };
6374
6375 struct pnfs_layout_segment *
6376 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6377 {
6378         struct inode *inode = lgp->args.inode;
6379         struct nfs_server *server = NFS_SERVER(inode);
6380         size_t max_pages = max_response_pages(server);
6381         struct rpc_task *task;
6382         struct rpc_message msg = {
6383                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6384                 .rpc_argp = &lgp->args,
6385                 .rpc_resp = &lgp->res,
6386         };
6387         struct rpc_task_setup task_setup_data = {
6388                 .rpc_client = server->client,
6389                 .rpc_message = &msg,
6390                 .callback_ops = &nfs4_layoutget_call_ops,
6391                 .callback_data = lgp,
6392                 .flags = RPC_TASK_ASYNC,
6393         };
6394         struct pnfs_layout_segment *lseg = NULL;
6395         int status = 0;
6396
6397         dprintk("--> %s\n", __func__);
6398
6399         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6400         if (!lgp->args.layout.pages) {
6401                 nfs4_layoutget_release(lgp);
6402                 return ERR_PTR(-ENOMEM);
6403         }
6404         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6405         lgp->args.timestamp = jiffies;
6406
6407         lgp->res.layoutp = &lgp->args.layout;
6408         lgp->res.seq_res.sr_slot = NULL;
6409         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6410
6411         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
6412         pnfs_get_layout_hdr(NFS_I(inode)->layout);
6413
6414         task = rpc_run_task(&task_setup_data);
6415         if (IS_ERR(task))
6416                 return ERR_CAST(task);
6417         status = nfs4_wait_for_completion_rpc_task(task);
6418         if (status == 0)
6419                 status = task->tk_status;
6420         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
6421         if (status == 0 && lgp->res.layoutp->len)
6422                 lseg = pnfs_layout_process(lgp);
6423         rpc_put_task(task);
6424         dprintk("<-- %s status=%d\n", __func__, status);
6425         if (status)
6426                 return ERR_PTR(status);
6427         return lseg;
6428 }
6429
6430 static void
6431 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6432 {
6433         struct nfs4_layoutreturn *lrp = calldata;
6434
6435         dprintk("--> %s\n", __func__);
6436         nfs41_setup_sequence(lrp->clp->cl_session,
6437                         &lrp->args.seq_args,
6438                         &lrp->res.seq_res,
6439                         task);
6440 }
6441
6442 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6443 {
6444         struct nfs4_layoutreturn *lrp = calldata;
6445         struct nfs_server *server;
6446
6447         dprintk("--> %s\n", __func__);
6448
6449         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
6450                 return;
6451
6452         server = NFS_SERVER(lrp->args.inode);
6453         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6454                 rpc_restart_call_prepare(task);
6455                 return;
6456         }
6457         dprintk("<-- %s\n", __func__);
6458 }
6459
6460 static void nfs4_layoutreturn_release(void *calldata)
6461 {
6462         struct nfs4_layoutreturn *lrp = calldata;
6463         struct pnfs_layout_hdr *lo = lrp->args.layout;
6464
6465         dprintk("--> %s\n", __func__);
6466         spin_lock(&lo->plh_inode->i_lock);
6467         if (lrp->res.lrs_present)
6468                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6469         lo->plh_block_lgets--;
6470         spin_unlock(&lo->plh_inode->i_lock);
6471         pnfs_put_layout_hdr(lrp->args.layout);
6472         kfree(calldata);
6473         dprintk("<-- %s\n", __func__);
6474 }
6475
6476 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6477         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6478         .rpc_call_done = nfs4_layoutreturn_done,
6479         .rpc_release = nfs4_layoutreturn_release,
6480 };
6481
6482 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6483 {
6484         struct rpc_task *task;
6485         struct rpc_message msg = {
6486                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6487                 .rpc_argp = &lrp->args,
6488                 .rpc_resp = &lrp->res,
6489         };
6490         struct rpc_task_setup task_setup_data = {
6491                 .rpc_client = lrp->clp->cl_rpcclient,
6492                 .rpc_message = &msg,
6493                 .callback_ops = &nfs4_layoutreturn_call_ops,
6494                 .callback_data = lrp,
6495         };
6496         int status;
6497
6498         dprintk("--> %s\n", __func__);
6499         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6500         task = rpc_run_task(&task_setup_data);
6501         if (IS_ERR(task))
6502                 return PTR_ERR(task);
6503         status = task->tk_status;
6504         dprintk("<-- %s status=%d\n", __func__, status);
6505         rpc_put_task(task);
6506         return status;
6507 }
6508
6509 /*
6510  * Retrieve the list of Data Server devices from the MDS.
6511  */
6512 static int _nfs4_getdevicelist(struct nfs_server *server,
6513                                     const struct nfs_fh *fh,
6514                                     struct pnfs_devicelist *devlist)
6515 {
6516         struct nfs4_getdevicelist_args args = {
6517                 .fh = fh,
6518                 .layoutclass = server->pnfs_curr_ld->id,
6519         };
6520         struct nfs4_getdevicelist_res res = {
6521                 .devlist = devlist,
6522         };
6523         struct rpc_message msg = {
6524                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6525                 .rpc_argp = &args,
6526                 .rpc_resp = &res,
6527         };
6528         int status;
6529
6530         dprintk("--> %s\n", __func__);
6531         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6532                                 &res.seq_res, 0);
6533         dprintk("<-- %s status=%d\n", __func__, status);
6534         return status;
6535 }
6536
6537 int nfs4_proc_getdevicelist(struct nfs_server *server,
6538                             const struct nfs_fh *fh,
6539                             struct pnfs_devicelist *devlist)
6540 {
6541         struct nfs4_exception exception = { };
6542         int err;
6543
6544         do {
6545                 err = nfs4_handle_exception(server,
6546                                 _nfs4_getdevicelist(server, fh, devlist),
6547                                 &exception);
6548         } while (exception.retry);
6549
6550         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6551                 err, devlist->num_devs);
6552
6553         return err;
6554 }
6555 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6556
6557 static int
6558 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6559 {
6560         struct nfs4_getdeviceinfo_args args = {
6561                 .pdev = pdev,
6562         };
6563         struct nfs4_getdeviceinfo_res res = {
6564                 .pdev = pdev,
6565         };
6566         struct rpc_message msg = {
6567                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6568                 .rpc_argp = &args,
6569                 .rpc_resp = &res,
6570         };
6571         int status;
6572
6573         dprintk("--> %s\n", __func__);
6574         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6575         dprintk("<-- %s status=%d\n", __func__, status);
6576
6577         return status;
6578 }
6579
6580 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6581 {
6582         struct nfs4_exception exception = { };
6583         int err;
6584
6585         do {
6586                 err = nfs4_handle_exception(server,
6587                                         _nfs4_proc_getdeviceinfo(server, pdev),
6588                                         &exception);
6589         } while (exception.retry);
6590         return err;
6591 }
6592 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6593
6594 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6595 {
6596         struct nfs4_layoutcommit_data *data = calldata;
6597         struct nfs_server *server = NFS_SERVER(data->args.inode);
6598         struct nfs4_session *session = nfs4_get_session(server);
6599
6600         nfs41_setup_sequence(session,
6601                         &data->args.seq_args,
6602                         &data->res.seq_res,
6603                         task);
6604 }
6605
6606 static void
6607 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6608 {
6609         struct nfs4_layoutcommit_data *data = calldata;
6610         struct nfs_server *server = NFS_SERVER(data->args.inode);
6611
6612         if (!nfs41_sequence_done(task, &data->res.seq_res))
6613                 return;
6614
6615         switch (task->tk_status) { /* Just ignore these failures */
6616         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6617         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6618         case -NFS4ERR_BADLAYOUT:     /* no layout */
6619         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6620                 task->tk_status = 0;
6621                 break;
6622         case 0:
6623                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6624                                                    data->res.fattr);
6625                 break;
6626         default:
6627                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6628                         rpc_restart_call_prepare(task);
6629                         return;
6630                 }
6631         }
6632 }
6633
6634 static void nfs4_layoutcommit_release(void *calldata)
6635 {
6636         struct nfs4_layoutcommit_data *data = calldata;
6637
6638         pnfs_cleanup_layoutcommit(data);
6639         put_rpccred(data->cred);
6640         kfree(data);
6641 }
6642
6643 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6644         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6645         .rpc_call_done = nfs4_layoutcommit_done,
6646         .rpc_release = nfs4_layoutcommit_release,
6647 };
6648
6649 int
6650 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6651 {
6652         struct rpc_message msg = {
6653                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6654                 .rpc_argp = &data->args,
6655                 .rpc_resp = &data->res,
6656                 .rpc_cred = data->cred,
6657         };
6658         struct rpc_task_setup task_setup_data = {
6659                 .task = &data->task,
6660                 .rpc_client = NFS_CLIENT(data->args.inode),
6661                 .rpc_message = &msg,
6662                 .callback_ops = &nfs4_layoutcommit_ops,
6663                 .callback_data = data,
6664                 .flags = RPC_TASK_ASYNC,
6665         };
6666         struct rpc_task *task;
6667         int status = 0;
6668
6669         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6670                 "lbw: %llu inode %lu\n",
6671                 data->task.tk_pid, sync,
6672                 data->args.lastbytewritten,
6673                 data->args.inode->i_ino);
6674
6675         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6676         task = rpc_run_task(&task_setup_data);
6677         if (IS_ERR(task))
6678                 return PTR_ERR(task);
6679         if (sync == false)
6680                 goto out;
6681         status = nfs4_wait_for_completion_rpc_task(task);
6682         if (status != 0)
6683                 goto out;
6684         status = task->tk_status;
6685 out:
6686         dprintk("%s: status %d\n", __func__, status);
6687         rpc_put_task(task);
6688         return status;
6689 }
6690
6691 static int
6692 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6693                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6694 {
6695         struct nfs41_secinfo_no_name_args args = {
6696                 .style = SECINFO_STYLE_CURRENT_FH,
6697         };
6698         struct nfs4_secinfo_res res = {
6699                 .flavors = flavors,
6700         };
6701         struct rpc_message msg = {
6702                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6703                 .rpc_argp = &args,
6704                 .rpc_resp = &res,
6705         };
6706         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6707 }
6708
6709 static int
6710 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6711                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6712 {
6713         struct nfs4_exception exception = { };
6714         int err;
6715         do {
6716                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6717                 switch (err) {
6718                 case 0:
6719                 case -NFS4ERR_WRONGSEC:
6720                 case -ENOTSUPP:
6721                         goto out;
6722                 default:
6723                         err = nfs4_handle_exception(server, err, &exception);
6724                 }
6725         } while (exception.retry);
6726 out:
6727         return err;
6728 }
6729
6730 static int
6731 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6732                     struct nfs_fsinfo *info)
6733 {
6734         int err;
6735         struct page *page;
6736         rpc_authflavor_t flavor;
6737         struct nfs4_secinfo_flavors *flavors;
6738
6739         page = alloc_page(GFP_KERNEL);
6740         if (!page) {
6741                 err = -ENOMEM;
6742                 goto out;
6743         }
6744
6745         flavors = page_address(page);
6746         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6747
6748         /*
6749          * Fall back on "guess and check" method if
6750          * the server doesn't support SECINFO_NO_NAME
6751          */
6752         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
6753                 err = nfs4_find_root_sec(server, fhandle, info);
6754                 goto out_freepage;
6755         }
6756         if (err)
6757                 goto out_freepage;
6758
6759         flavor = nfs_find_best_sec(flavors);
6760         if (err == 0)
6761                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6762
6763 out_freepage:
6764         put_page(page);
6765         if (err == -EACCES)
6766                 return -EPERM;
6767 out:
6768         return err;
6769 }
6770
6771 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6772 {
6773         int status;
6774         struct nfs41_test_stateid_args args = {
6775                 .stateid = stateid,
6776         };
6777         struct nfs41_test_stateid_res res;
6778         struct rpc_message msg = {
6779                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6780                 .rpc_argp = &args,
6781                 .rpc_resp = &res,
6782         };
6783
6784         dprintk("NFS call  test_stateid %p\n", stateid);
6785         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6786         nfs4_set_sequence_privileged(&args.seq_args);
6787         status = nfs4_call_sync_sequence(server->client, server, &msg,
6788                         &args.seq_args, &res.seq_res);
6789         if (status != NFS_OK) {
6790                 dprintk("NFS reply test_stateid: failed, %d\n", status);
6791                 return status;
6792         }
6793         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
6794         return -res.status;
6795 }
6796
6797 /**
6798  * nfs41_test_stateid - perform a TEST_STATEID operation
6799  *
6800  * @server: server / transport on which to perform the operation
6801  * @stateid: state ID to test
6802  *
6803  * Returns NFS_OK if the server recognizes that "stateid" is valid.
6804  * Otherwise a negative NFS4ERR value is returned if the operation
6805  * failed or the state ID is not currently valid.
6806  */
6807 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6808 {
6809         struct nfs4_exception exception = { };
6810         int err;
6811         do {
6812                 err = _nfs41_test_stateid(server, stateid);
6813                 if (err != -NFS4ERR_DELAY)
6814                         break;
6815                 nfs4_handle_exception(server, err, &exception);
6816         } while (exception.retry);
6817         return err;
6818 }
6819
6820 struct nfs_free_stateid_data {
6821         struct nfs_server *server;
6822         struct nfs41_free_stateid_args args;
6823         struct nfs41_free_stateid_res res;
6824 };
6825
6826 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
6827 {
6828         struct nfs_free_stateid_data *data = calldata;
6829         nfs41_setup_sequence(nfs4_get_session(data->server),
6830                         &data->args.seq_args,
6831                         &data->res.seq_res,
6832                         task);
6833 }
6834
6835 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
6836 {
6837         struct nfs_free_stateid_data *data = calldata;
6838
6839         nfs41_sequence_done(task, &data->res.seq_res);
6840
6841         switch (task->tk_status) {
6842         case -NFS4ERR_DELAY:
6843                 if (nfs4_async_handle_error(task, data->server, NULL) == -EAGAIN)
6844                         rpc_restart_call_prepare(task);
6845         }
6846 }
6847
6848 static void nfs41_free_stateid_release(void *calldata)
6849 {
6850         kfree(calldata);
6851 }
6852
6853 const struct rpc_call_ops nfs41_free_stateid_ops = {
6854         .rpc_call_prepare = nfs41_free_stateid_prepare,
6855         .rpc_call_done = nfs41_free_stateid_done,
6856         .rpc_release = nfs41_free_stateid_release,
6857 };
6858
6859 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
6860                 nfs4_stateid *stateid,
6861                 bool privileged)
6862 {
6863         struct rpc_message msg = {
6864                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6865         };
6866         struct rpc_task_setup task_setup = {
6867                 .rpc_client = server->client,
6868                 .rpc_message = &msg,
6869                 .callback_ops = &nfs41_free_stateid_ops,
6870                 .flags = RPC_TASK_ASYNC,
6871         };
6872         struct nfs_free_stateid_data *data;
6873
6874         dprintk("NFS call  free_stateid %p\n", stateid);
6875         data = kmalloc(sizeof(*data), GFP_NOFS);
6876         if (!data)
6877                 return ERR_PTR(-ENOMEM);
6878         data->server = server;
6879         nfs4_stateid_copy(&data->args.stateid, stateid);
6880
6881         task_setup.callback_data = data;
6882
6883         msg.rpc_argp = &data->args;
6884         msg.rpc_resp = &data->res;
6885         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6886         if (privileged)
6887                 nfs4_set_sequence_privileged(&data->args.seq_args);
6888
6889         return rpc_run_task(&task_setup);
6890 }
6891
6892 /**
6893  * nfs41_free_stateid - perform a FREE_STATEID operation
6894  *
6895  * @server: server / transport on which to perform the operation
6896  * @stateid: state ID to release
6897  *
6898  * Returns NFS_OK if the server freed "stateid".  Otherwise a
6899  * negative NFS4ERR value is returned.
6900  */
6901 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6902 {
6903         struct rpc_task *task;
6904         int ret;
6905
6906         task = _nfs41_free_stateid(server, stateid, true);
6907         if (IS_ERR(task))
6908                 return PTR_ERR(task);
6909         ret = rpc_wait_for_completion_task(task);
6910         if (!ret)
6911                 ret = task->tk_status;
6912         rpc_put_task(task);
6913         return ret;
6914 }
6915
6916 static int nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
6917 {
6918         struct rpc_task *task;
6919
6920         task = _nfs41_free_stateid(server, &lsp->ls_stateid, false);
6921         nfs4_free_lock_state(server, lsp);
6922         if (IS_ERR(task))
6923                 return PTR_ERR(task);
6924         rpc_put_task(task);
6925         return 0;
6926 }
6927
6928 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6929                 const nfs4_stateid *s2)
6930 {
6931         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6932                 return false;
6933
6934         if (s1->seqid == s2->seqid)
6935                 return true;
6936         if (s1->seqid == 0 || s2->seqid == 0)
6937                 return true;
6938
6939         return false;
6940 }
6941
6942 #endif /* CONFIG_NFS_V4_1 */
6943
6944 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6945                 const nfs4_stateid *s2)
6946 {
6947         return nfs4_stateid_match(s1, s2);
6948 }
6949
6950
6951 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6952         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6953         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6954         .recover_open   = nfs4_open_reclaim,
6955         .recover_lock   = nfs4_lock_reclaim,
6956         .establish_clid = nfs4_init_clientid,
6957         .get_clid_cred  = nfs4_get_setclientid_cred,
6958         .detect_trunking = nfs40_discover_server_trunking,
6959 };
6960
6961 #if defined(CONFIG_NFS_V4_1)
6962 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6963         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6964         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6965         .recover_open   = nfs4_open_reclaim,
6966         .recover_lock   = nfs4_lock_reclaim,
6967         .establish_clid = nfs41_init_clientid,
6968         .get_clid_cred  = nfs4_get_exchange_id_cred,
6969         .reclaim_complete = nfs41_proc_reclaim_complete,
6970         .detect_trunking = nfs41_discover_server_trunking,
6971 };
6972 #endif /* CONFIG_NFS_V4_1 */
6973
6974 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6975         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6976         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6977         .recover_open   = nfs4_open_expired,
6978         .recover_lock   = nfs4_lock_expired,
6979         .establish_clid = nfs4_init_clientid,
6980         .get_clid_cred  = nfs4_get_setclientid_cred,
6981 };
6982
6983 #if defined(CONFIG_NFS_V4_1)
6984 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6985         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6986         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6987         .recover_open   = nfs41_open_expired,
6988         .recover_lock   = nfs41_lock_expired,
6989         .establish_clid = nfs41_init_clientid,
6990         .get_clid_cred  = nfs4_get_exchange_id_cred,
6991 };
6992 #endif /* CONFIG_NFS_V4_1 */
6993
6994 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6995         .sched_state_renewal = nfs4_proc_async_renew,
6996         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6997         .renew_lease = nfs4_proc_renew,
6998 };
6999
7000 #if defined(CONFIG_NFS_V4_1)
7001 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
7002         .sched_state_renewal = nfs41_proc_async_sequence,
7003         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
7004         .renew_lease = nfs4_proc_sequence,
7005 };
7006 #endif
7007
7008 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
7009         .minor_version = 0,
7010         .init_caps = NFS_CAP_READDIRPLUS
7011                 | NFS_CAP_ATOMIC_OPEN
7012                 | NFS_CAP_CHANGE_ATTR
7013                 | NFS_CAP_POSIX_LOCK,
7014         .call_sync = _nfs4_call_sync,
7015         .match_stateid = nfs4_match_stateid,
7016         .find_root_sec = nfs4_find_root_sec,
7017         .free_lock_state = nfs4_release_lockowner,
7018         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
7019         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
7020         .state_renewal_ops = &nfs40_state_renewal_ops,
7021 };
7022
7023 #if defined(CONFIG_NFS_V4_1)
7024 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
7025         .minor_version = 1,
7026         .init_caps = NFS_CAP_READDIRPLUS
7027                 | NFS_CAP_ATOMIC_OPEN
7028                 | NFS_CAP_CHANGE_ATTR
7029                 | NFS_CAP_POSIX_LOCK
7030                 | NFS_CAP_STATEID_NFSV41
7031                 | NFS_CAP_ATOMIC_OPEN_V1,
7032         .call_sync = nfs4_call_sync_sequence,
7033         .match_stateid = nfs41_match_stateid,
7034         .find_root_sec = nfs41_find_root_sec,
7035         .free_lock_state = nfs41_free_lock_state,
7036         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
7037         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
7038         .state_renewal_ops = &nfs41_state_renewal_ops,
7039 };
7040 #endif
7041
7042 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
7043         [0] = &nfs_v4_0_minor_ops,
7044 #if defined(CONFIG_NFS_V4_1)
7045         [1] = &nfs_v4_1_minor_ops,
7046 #endif
7047 };
7048
7049 const struct inode_operations nfs4_dir_inode_operations = {
7050         .create         = nfs_create,
7051         .lookup         = nfs_lookup,
7052         .atomic_open    = nfs_atomic_open,
7053         .link           = nfs_link,
7054         .unlink         = nfs_unlink,
7055         .symlink        = nfs_symlink,
7056         .mkdir          = nfs_mkdir,
7057         .rmdir          = nfs_rmdir,
7058         .mknod          = nfs_mknod,
7059         .rename         = nfs_rename,
7060         .permission     = nfs_permission,
7061         .getattr        = nfs_getattr,
7062         .setattr        = nfs_setattr,
7063         .getxattr       = generic_getxattr,
7064         .setxattr       = generic_setxattr,
7065         .listxattr      = generic_listxattr,
7066         .removexattr    = generic_removexattr,
7067 };
7068
7069 static const struct inode_operations nfs4_file_inode_operations = {
7070         .permission     = nfs_permission,
7071         .getattr        = nfs_getattr,
7072         .setattr        = nfs_setattr,
7073         .getxattr       = generic_getxattr,
7074         .setxattr       = generic_setxattr,
7075         .listxattr      = generic_listxattr,
7076         .removexattr    = generic_removexattr,
7077 };
7078
7079 const struct nfs_rpc_ops nfs_v4_clientops = {
7080         .version        = 4,                    /* protocol version */
7081         .dentry_ops     = &nfs4_dentry_operations,
7082         .dir_inode_ops  = &nfs4_dir_inode_operations,
7083         .file_inode_ops = &nfs4_file_inode_operations,
7084         .file_ops       = &nfs4_file_operations,
7085         .getroot        = nfs4_proc_get_root,
7086         .submount       = nfs4_submount,
7087         .try_mount      = nfs4_try_mount,
7088         .getattr        = nfs4_proc_getattr,
7089         .setattr        = nfs4_proc_setattr,
7090         .lookup         = nfs4_proc_lookup,
7091         .access         = nfs4_proc_access,
7092         .readlink       = nfs4_proc_readlink,
7093         .create         = nfs4_proc_create,
7094         .remove         = nfs4_proc_remove,
7095         .unlink_setup   = nfs4_proc_unlink_setup,
7096         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
7097         .unlink_done    = nfs4_proc_unlink_done,
7098         .rename         = nfs4_proc_rename,
7099         .rename_setup   = nfs4_proc_rename_setup,
7100         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
7101         .rename_done    = nfs4_proc_rename_done,
7102         .link           = nfs4_proc_link,
7103         .symlink        = nfs4_proc_symlink,
7104         .mkdir          = nfs4_proc_mkdir,
7105         .rmdir          = nfs4_proc_remove,
7106         .readdir        = nfs4_proc_readdir,
7107         .mknod          = nfs4_proc_mknod,
7108         .statfs         = nfs4_proc_statfs,
7109         .fsinfo         = nfs4_proc_fsinfo,
7110         .pathconf       = nfs4_proc_pathconf,
7111         .set_capabilities = nfs4_server_capabilities,
7112         .decode_dirent  = nfs4_decode_dirent,
7113         .read_setup     = nfs4_proc_read_setup,
7114         .read_pageio_init = pnfs_pageio_init_read,
7115         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
7116         .read_done      = nfs4_read_done,
7117         .write_setup    = nfs4_proc_write_setup,
7118         .write_pageio_init = pnfs_pageio_init_write,
7119         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
7120         .write_done     = nfs4_write_done,
7121         .commit_setup   = nfs4_proc_commit_setup,
7122         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
7123         .commit_done    = nfs4_commit_done,
7124         .lock           = nfs4_proc_lock,
7125         .clear_acl_cache = nfs4_zap_acl_attr,
7126         .close_context  = nfs4_close_context,
7127         .open_context   = nfs4_atomic_open,
7128         .have_delegation = nfs4_have_delegation,
7129         .return_delegation = nfs4_inode_return_delegation,
7130         .alloc_client   = nfs4_alloc_client,
7131         .init_client    = nfs4_init_client,
7132         .free_client    = nfs4_free_client,
7133         .create_server  = nfs4_create_server,
7134         .clone_server   = nfs_clone_server,
7135 };
7136
7137 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7138         .prefix = XATTR_NAME_NFSV4_ACL,
7139         .list   = nfs4_xattr_list_nfs4_acl,
7140         .get    = nfs4_xattr_get_nfs4_acl,
7141         .set    = nfs4_xattr_set_nfs4_acl,
7142 };
7143
7144 const struct xattr_handler *nfs4_xattr_handlers[] = {
7145         &nfs4_xattr_nfs4_acl_handler,
7146         NULL
7147 };
7148
7149 /*
7150  * Local variables:
7151  *  c-basic-offset: 8
7152  * End:
7153  */