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 (state == NULL)
 308                                 break;
 309                         nfs_remove_bad_delegation(state->inode);
 310                         ret = nfs4_schedule_stateid_recovery(server, state);
 311                         if (ret < 0)
 312                                 break;
 313                         goto wait_on_recovery;
 314                 case -NFS4ERR_EXPIRED:
 315                         if (state != NULL) {
 316                                 ret = nfs4_schedule_stateid_recovery(server, state);
 317                                 if (ret < 0)
 318                                         break;
 319                         }
 320                 case -NFS4ERR_STALE_STATEID:
 321                 case -NFS4ERR_STALE_CLIENTID:
 322                         nfs4_schedule_lease_recovery(clp);
 323                         goto wait_on_recovery;
 324 #if defined(CONFIG_NFS_V4_1)
 325                 case -NFS4ERR_BADSESSION:
 326                 case -NFS4ERR_BADSLOT:
 327                 case -NFS4ERR_BAD_HIGH_SLOT:
 328                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 329                 case -NFS4ERR_DEADSESSION:
 330                 case -NFS4ERR_SEQ_FALSE_RETRY:
 331                 case -NFS4ERR_SEQ_MISORDERED:
 332                         dprintk("%s ERROR: %d Reset session\n", __func__,
 333                                 errorcode);
 334                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 335                         goto wait_on_recovery;
 336 #endif /* defined(CONFIG_NFS_V4_1) */
 337                 case -NFS4ERR_FILE_OPEN:
 338                         if (exception->timeout > HZ) {
 339                                 /* We have retried a decent amount, time to
 340                                  * fail
 341                                  */
 342                                 ret = -EBUSY;
 343                                 break;
 344                         }
 345                 case -NFS4ERR_GRACE:
 346                 case -NFS4ERR_DELAY:
 347                         ret = nfs4_delay(server->client, &exception->timeout);
 348                         if (ret != 0)
 349                                 break;
 350                 case -NFS4ERR_RETRY_UNCACHED_REP:
 351                 case -NFS4ERR_OLD_STATEID:
 352                         exception->retry = 1;
 353                         break;
 354                 case -NFS4ERR_BADOWNER:
 355                         /* The following works around a Linux server bug! */
 356                 case -NFS4ERR_BADNAME:
 357                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 358                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 359                                 exception->retry = 1;
 360                                 printk(KERN_WARNING "NFS: v4 server %s "
 361                                                 "does not accept raw "
 362                                                 "uid/gids. "
 363                                                 "Reenabling the idmapper.\n",
 364                                                 server->nfs_client->cl_hostname);
 365                         }
 366         }
 367         /* We failed to handle the error */
 368         return nfs4_map_errors(ret);
 369 wait_on_recovery:
 370         ret = nfs4_wait_clnt_recover(clp);
 371         if (ret == 0)
 372                 exception->retry = 1;
 373         return ret;
 374 }
 375
 376
 377 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 378 {
 379         spin_lock(&clp->cl_lock);
 380         if (time_before(clp->cl_last_renewal,timestamp))
 381                 clp->cl_last_renewal = timestamp;
 382         spin_unlock(&clp->cl_lock);
 383 }
 384
 385 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 386 {
 387         do_renew_lease(server->nfs_client, timestamp);
 388 }
 389
 390 #if defined(CONFIG_NFS_V4_1)
 391
 392 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 393 {
 394         struct nfs4_session *session;
 395         struct nfs4_slot_table *tbl;
 396         bool send_new_highest_used_slotid = false;
 397
 398         if (!res->sr_slot) {
 399                 /* just wake up the next guy waiting since
 400                  * we may have not consumed a slot after all */
 401                 dprintk("%s: No slot\n", __func__);
 402                 return;
 403         }
 404         tbl = res->sr_slot->table;
 405         session = tbl->session;
 406
 407         spin_lock(&tbl->slot_tbl_lock);
 408         /* Be nice to the server: try to ensure that the last transmitted
 409          * value for highest_user_slotid <= target_highest_slotid
 410          */
 411         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 412                 send_new_highest_used_slotid = true;
 413
 414         if (nfs41_wake_and_assign_slot(tbl, res->sr_slot)) {
 415                 send_new_highest_used_slotid = false;
 416                 goto out_unlock;
 417         }
 418         nfs4_free_slot(tbl, res->sr_slot);
 419
 420         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 421                 send_new_highest_used_slotid = false;
 422 out_unlock:
 423         spin_unlock(&tbl->slot_tbl_lock);
 424         res->sr_slot = NULL;
 425         if (send_new_highest_used_slotid)
 426                 nfs41_server_notify_highest_slotid_update(session->clp);
 427 }
 428
 429 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 430 {
 431         struct nfs4_session *session;
 432         struct nfs4_slot *slot;
 433         struct nfs_client *clp;
 434         bool interrupted = false;
 435         int ret = 1;
 436
 437         /* don't increment the sequence number if the task wasn't sent */
 438         if (!RPC_WAS_SENT(task))
 439                 goto out;
 440
 441         slot = res->sr_slot;
 442         session = slot->table->session;
 443
 444         if (slot->interrupted) {
 445                 slot->interrupted = 0;
 446                 interrupted = true;
 447         }
 448
 449         /* Check the SEQUENCE operation status */
 450         switch (res->sr_status) {
 451         case 0:
 452                 /* Update the slot's sequence and clientid lease timer */
 453                 ++slot->seq_nr;
 454                 clp = session->clp;
 455                 do_renew_lease(clp, res->sr_timestamp);
 456                 /* Check sequence flags */
 457                 if (res->sr_status_flags != 0)
 458                         nfs4_schedule_lease_recovery(clp);
 459                 nfs41_update_target_slotid(slot->table, slot, res);
 460                 break;
 461         case 1:
 462                 /*
 463                  * sr_status remains 1 if an RPC level error occurred.
 464                  * The server may or may not have processed the sequence
 465                  * operation..
 466                  * Mark the slot as having hosted an interrupted RPC call.
 467                  */
 468                 slot->interrupted = 1;
 469                 goto out;
 470         case -NFS4ERR_DELAY:
 471                 /* The server detected a resend of the RPC call and
 472                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 473                  * of RFC5661.
 474                  */
 475                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 476                         __func__,
 477                         slot->slot_nr,
 478                         slot->seq_nr);
 479                 goto out_retry;
 480         case -NFS4ERR_BADSLOT:
 481                 /*
 482                  * The slot id we used was probably retired. Try again
 483                  * using a different slot id.
 484                  */
 485                 goto retry_nowait;
 486         case -NFS4ERR_SEQ_MISORDERED:
 487                 /*
 488                  * Was the last operation on this sequence interrupted?
 489                  * If so, retry after bumping the sequence number.
 490                  */
 491                 if (interrupted) {
 492                         ++slot->seq_nr;
 493                         goto retry_nowait;
 494                 }
 495                 /*
 496                  * Could this slot have been previously retired?
 497                  * If so, then the server may be expecting seq_nr = 1!
 498                  */
 499                 if (slot->seq_nr != 1) {
 500                         slot->seq_nr = 1;
 501                         goto retry_nowait;
 502                 }
 503                 break;
 504         case -NFS4ERR_SEQ_FALSE_RETRY:
 505                 ++slot->seq_nr;
 506                 goto retry_nowait;
 507         default:
 508                 /* Just update the slot sequence no. */
 509                 ++slot->seq_nr;
 510         }
 511 out:
 512         /* The session may be reset by one of the error handlers. */
 513         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 514         nfs41_sequence_free_slot(res);
 515         return ret;
 516 retry_nowait:
 517         if (rpc_restart_call_prepare(task)) {
 518                 task->tk_status = 0;
 519                 ret = 0;
 520         }
 521         goto out;
 522 out_retry:
 523         if (!rpc_restart_call(task))
 524                 goto out;
 525         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 526         return 0;
 527 }
 528
 529 static int nfs4_sequence_done(struct rpc_task *task,
 530                                struct nfs4_sequence_res *res)
 531 {
 532         if (res->sr_slot == NULL)
 533                 return 1;
 534         return nfs41_sequence_done(task, res);
 535 }
 536
 537 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 538                 struct nfs4_sequence_res *res, int cache_reply)
 539 {
 540         args->sa_slot = NULL;
 541         args->sa_cache_this = 0;
 542         args->sa_privileged = 0;
 543         if (cache_reply)
 544                 args->sa_cache_this = 1;
 545         res->sr_slot = NULL;
 546 }
 547
 548 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 549 {
 550         args->sa_privileged = 1;
 551 }
 552
 553 int nfs41_setup_sequence(struct nfs4_session *session,
 554                                 struct nfs4_sequence_args *args,
 555                                 struct nfs4_sequence_res *res,
 556                                 struct rpc_task *task)
 557 {
 558         struct nfs4_slot *slot;
 559         struct nfs4_slot_table *tbl;
 560
 561         dprintk("--> %s\n", __func__);
 562         /* slot already allocated? */
 563         if (res->sr_slot != NULL)
 564                 goto out_success;
 565
 566         tbl = &session->fc_slot_table;
 567
 568         task->tk_timeout = 0;
 569
 570         spin_lock(&tbl->slot_tbl_lock);
 571         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 572             !args->sa_privileged) {
 573                 /* The state manager will wait until the slot table is empty */
 574                 dprintk("%s session is draining\n", __func__);
 575                 goto out_sleep;
 576         }
 577
 578         slot = nfs4_alloc_slot(tbl);
 579         if (IS_ERR(slot)) {
 580                 /* If out of memory, try again in 1/4 second */
 581                 if (slot == ERR_PTR(-ENOMEM))
 582                         task->tk_timeout = HZ >> 2;
 583                 dprintk("<-- %s: no free slots\n", __func__);
 584                 goto out_sleep;
 585         }
 586         spin_unlock(&tbl->slot_tbl_lock);
 587
 588         args->sa_slot = slot;
 589
 590         dprintk("<-- %s slotid=%d seqid=%d\n", __func__,
 591                         slot->slot_nr, slot->seq_nr);
 592
 593         res->sr_slot = slot;
 594         res->sr_timestamp = jiffies;
 595         res->sr_status_flags = 0;
 596         /*
 597          * sr_status is only set in decode_sequence, and so will remain
 598          * set to 1 if an rpc level failure occurs.
 599          */
 600         res->sr_status = 1;
 601 out_success:
 602         rpc_call_start(task);
 603         return 0;
 604 out_sleep:
 605         /* Privileged tasks are queued with top priority */
 606         if (args->sa_privileged)
 607                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 608                                 NULL, RPC_PRIORITY_PRIVILEGED);
 609         else
 610                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 611         spin_unlock(&tbl->slot_tbl_lock);
 612         return -EAGAIN;
 613 }
 614 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 615
 616 int nfs4_setup_sequence(const struct nfs_server *server,
 617                         struct nfs4_sequence_args *args,
 618                         struct nfs4_sequence_res *res,
 619                         struct rpc_task *task)
 620 {
 621         struct nfs4_session *session = nfs4_get_session(server);
 622         int ret = 0;
 623
 624         if (session == NULL) {
 625                 rpc_call_start(task);
 626                 goto out;
 627         }
 628
 629         dprintk("--> %s clp %p session %p sr_slot %d\n",
 630                 __func__, session->clp, session, res->sr_slot ?
 631                         res->sr_slot->slot_nr : -1);
 632
 633         ret = nfs41_setup_sequence(session, args, res, task);
 634 out:
 635         dprintk("<-- %s status=%d\n", __func__, ret);
 636         return ret;
 637 }
 638
 639 struct nfs41_call_sync_data {
 640         const struct nfs_server *seq_server;
 641         struct nfs4_sequence_args *seq_args;
 642         struct nfs4_sequence_res *seq_res;
 643 };
 644
 645 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 646 {
 647         struct nfs41_call_sync_data *data = calldata;
 648         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 649
 650         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 651
 652         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 653 }
 654
 655 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 656 {
 657         struct nfs41_call_sync_data *data = calldata;
 658
 659         nfs41_sequence_done(task, data->seq_res);
 660 }
 661
 662 static const struct rpc_call_ops nfs41_call_sync_ops = {
 663         .rpc_call_prepare = nfs41_call_sync_prepare,
 664         .rpc_call_done = nfs41_call_sync_done,
 665 };
 666
 667 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 668                                    struct nfs_server *server,
 669                                    struct rpc_message *msg,
 670                                    struct nfs4_sequence_args *args,
 671                                    struct nfs4_sequence_res *res)
 672 {
 673         int ret;
 674         struct rpc_task *task;
 675         struct nfs41_call_sync_data data = {
 676                 .seq_server = server,
 677                 .seq_args = args,
 678                 .seq_res = res,
 679         };
 680         struct rpc_task_setup task_setup = {
 681                 .rpc_client = clnt,
 682                 .rpc_message = msg,
 683                 .callback_ops = &nfs41_call_sync_ops,
 684                 .callback_data = &data
 685         };
 686
 687         task = rpc_run_task(&task_setup);
 688         if (IS_ERR(task))
 689                 ret = PTR_ERR(task);
 690         else {
 691                 ret = task->tk_status;
 692                 rpc_put_task(task);
 693         }
 694         return ret;
 695 }
 696
 697 #else
 698 static
 699 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 700                 struct nfs4_sequence_res *res, int cache_reply)
 701 {
 702 }
 703
 704 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 705 {
 706 }
 707
 708
 709 static int nfs4_sequence_done(struct rpc_task *task,
 710                                struct nfs4_sequence_res *res)
 711 {
 712         return 1;
 713 }
 714 #endif /* CONFIG_NFS_V4_1 */
 715
 716 static
 717 int _nfs4_call_sync(struct rpc_clnt *clnt,
 718                     struct nfs_server *server,
 719                     struct rpc_message *msg,
 720                     struct nfs4_sequence_args *args,
 721                     struct nfs4_sequence_res *res)
 722 {
 723         return rpc_call_sync(clnt, msg, 0);
 724 }
 725
 726 static
 727 int nfs4_call_sync(struct rpc_clnt *clnt,
 728                    struct nfs_server *server,
 729                    struct rpc_message *msg,
 730                    struct nfs4_sequence_args *args,
 731                    struct nfs4_sequence_res *res,
 732                    int cache_reply)
 733 {
 734         nfs41_init_sequence(args, res, cache_reply);
 735         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 736                                                 args, res);
 737 }
 738
 739 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 740 {
 741         struct nfs_inode *nfsi = NFS_I(dir);
 742
 743         spin_lock(&dir->i_lock);
 744         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 745         if (!cinfo->atomic || cinfo->before != dir->i_version)
 746                 nfs_force_lookup_revalidate(dir);
 747         dir->i_version = cinfo->after;
 748         nfs_fscache_invalidate(dir);
 749         spin_unlock(&dir->i_lock);
 750 }
 751
 752 struct nfs4_opendata {
 753         struct kref kref;
 754         struct nfs_openargs o_arg;
 755         struct nfs_openres o_res;
 756         struct nfs_open_confirmargs c_arg;
 757         struct nfs_open_confirmres c_res;
 758         struct nfs4_string owner_name;
 759         struct nfs4_string group_name;
 760         struct nfs_fattr f_attr;
 761         struct dentry *dir;
 762         struct dentry *dentry;
 763         struct nfs4_state_owner *owner;
 764         struct nfs4_state *state;
 765         struct iattr attrs;
 766         unsigned long timestamp;
 767         unsigned int rpc_done : 1;
 768         int rpc_status;
 769         int cancelled;
 770 };
 771
 772 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 773                 int err, struct nfs4_exception *exception)
 774 {
 775         if (err != -EINVAL)
 776                 return false;
 777         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 778                 return false;
 779         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 780         exception->retry = 1;
 781         return true;
 782 }
 783
 784 static enum open_claim_type4
 785 nfs4_map_atomic_open_claim(struct nfs_server *server,
 786                 enum open_claim_type4 claim)
 787 {
 788         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 789                 return claim;
 790         switch (claim) {
 791         default:
 792                 return claim;
 793         case NFS4_OPEN_CLAIM_FH:
 794                 return NFS4_OPEN_CLAIM_NULL;
 795         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 796                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 797         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 798                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 799         }
 800 }
 801
 802 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 803 {
 804         p->o_res.f_attr = &p->f_attr;
 805         p->o_res.seqid = p->o_arg.seqid;
 806         p->c_res.seqid = p->c_arg.seqid;
 807         p->o_res.server = p->o_arg.server;
 808         p->o_res.access_request = p->o_arg.access;
 809         nfs_fattr_init(&p->f_attr);
 810         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 811 }
 812
 813 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 814                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 815                 const struct iattr *attrs,
 816                 enum open_claim_type4 claim,
 817                 gfp_t gfp_mask)
 818 {
 819         struct dentry *parent = dget_parent(dentry);
 820         struct inode *dir = parent->d_inode;
 821         struct nfs_server *server = NFS_SERVER(dir);
 822         struct nfs4_opendata *p;
 823
 824         p = kzalloc(sizeof(*p), gfp_mask);
 825         if (p == NULL)
 826                 goto err;
 827         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 828         if (p->o_arg.seqid == NULL)
 829                 goto err_free;
 830         nfs_sb_active(dentry->d_sb);
 831         p->dentry = dget(dentry);
 832         p->dir = parent;
 833         p->owner = sp;
 834         atomic_inc(&sp->so_count);
 835         p->o_arg.open_flags = flags;
 836         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 837         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
 838          * will return permission denied for all bits until close */
 839         if (!(flags & O_EXCL)) {
 840                 /* ask server to check for all possible rights as results
 841                  * are cached */
 842                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
 843                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
 844         }
 845         p->o_arg.clientid = server->nfs_client->cl_clientid;
 846         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
 847         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 848         p->o_arg.name = &dentry->d_name;
 849         p->o_arg.server = server;
 850         p->o_arg.bitmask = server->attr_bitmask;
 851         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
 852         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
 853         switch (p->o_arg.claim) {
 854         case NFS4_OPEN_CLAIM_NULL:
 855         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
 856         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
 857                 p->o_arg.fh = NFS_FH(dir);
 858                 break;
 859         case NFS4_OPEN_CLAIM_PREVIOUS:
 860         case NFS4_OPEN_CLAIM_FH:
 861         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 862         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 863                 p->o_arg.fh = NFS_FH(dentry->d_inode);
 864         }
 865         if (attrs != NULL && attrs->ia_valid != 0) {
 866                 __be32 verf[2];
 867
 868                 p->o_arg.u.attrs = &p->attrs;
 869                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 870
 871                 verf[0] = jiffies;
 872                 verf[1] = current->pid;
 873                 memcpy(p->o_arg.u.verifier.data, verf,
 874                                 sizeof(p->o_arg.u.verifier.data));
 875         }
 876         p->c_arg.fh = &p->o_res.fh;
 877         p->c_arg.stateid = &p->o_res.stateid;
 878         p->c_arg.seqid = p->o_arg.seqid;
 879         nfs4_init_opendata_res(p);
 880         kref_init(&p->kref);
 881         return p;
 882 err_free:
 883         kfree(p);
 884 err:
 885         dput(parent);
 886         return NULL;
 887 }
 888
 889 static void nfs4_opendata_free(struct kref *kref)
 890 {
 891         struct nfs4_opendata *p = container_of(kref,
 892                         struct nfs4_opendata, kref);
 893         struct super_block *sb = p->dentry->d_sb;
 894
 895         nfs_free_seqid(p->o_arg.seqid);
 896         if (p->state != NULL)
 897                 nfs4_put_open_state(p->state);
 898         nfs4_put_state_owner(p->owner);
 899         dput(p->dir);
 900         dput(p->dentry);
 901         nfs_sb_deactive(sb);
 902         nfs_fattr_free_names(&p->f_attr);
 903         kfree(p);
 904 }
 905
 906 static void nfs4_opendata_put(struct nfs4_opendata *p)
 907 {
 908         if (p != NULL)
 909                 kref_put(&p->kref, nfs4_opendata_free);
 910 }
 911
 912 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 913 {
 914         int ret;
 915
 916         ret = rpc_wait_for_completion_task(task);
 917         return ret;
 918 }
 919
 920 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 921 {
 922         int ret = 0;
 923
 924         if (open_mode & (O_EXCL|O_TRUNC))
 925                 goto out;
 926         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 927                 case FMODE_READ:
 928                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 929                                 && state->n_rdonly != 0;
 930                         break;
 931                 case FMODE_WRITE:
 932                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 933                                 && state->n_wronly != 0;
 934                         break;
 935                 case FMODE_READ|FMODE_WRITE:
 936                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 937                                 && state->n_rdwr != 0;
 938         }
 939 out:
 940         return ret;
 941 }
 942
 943 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 944 {
 945         if (delegation == NULL)
 946                 return 0;
 947         if ((delegation->type & fmode) != fmode)
 948                 return 0;
 949         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 950                 return 0;
 951         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
 952                 return 0;
 953         nfs_mark_delegation_referenced(delegation);
 954         return 1;
 955 }
 956
 957 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 958 {
 959         switch (fmode) {
 960                 case FMODE_WRITE:
 961                         state->n_wronly++;
 962                         break;
 963                 case FMODE_READ:
 964                         state->n_rdonly++;
 965                         break;
 966                 case FMODE_READ|FMODE_WRITE:
 967                         state->n_rdwr++;
 968         }
 969         nfs4_state_set_mode_locked(state, state->state | fmode);
 970 }
 971
 972 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 973 {
 974         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 975                 nfs4_stateid_copy(&state->stateid, stateid);
 976         nfs4_stateid_copy(&state->open_stateid, stateid);
 977         switch (fmode) {
 978                 case FMODE_READ:
 979                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 980                         break;
 981                 case FMODE_WRITE:
 982                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 983                         break;
 984                 case FMODE_READ|FMODE_WRITE:
 985                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 986         }
 987 }
 988
 989 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 990 {
 991         write_seqlock(&state->seqlock);
 992         nfs_set_open_stateid_locked(state, stateid, fmode);
 993         write_sequnlock(&state->seqlock);
 994 }
 995
 996 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 997 {
 998         /*
 999          * Protect the call to nfs4_state_set_mode_locked and
1000          * serialise the stateid update
1001          */
1002         write_seqlock(&state->seqlock);
1003         if (deleg_stateid != NULL) {
1004                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1005                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1006         }
1007         if (open_stateid != NULL)
1008                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1009         write_sequnlock(&state->seqlock);
1010         spin_lock(&state->owner->so_lock);
1011         update_open_stateflags(state, fmode);
1012         spin_unlock(&state->owner->so_lock);
1013 }
1014
1015 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1016 {
1017         struct nfs_inode *nfsi = NFS_I(state->inode);
1018         struct nfs_delegation *deleg_cur;
1019         int ret = 0;
1020
1021         fmode &= (FMODE_READ|FMODE_WRITE);
1022
1023         rcu_read_lock();
1024         deleg_cur = rcu_dereference(nfsi->delegation);
1025         if (deleg_cur == NULL)
1026                 goto no_delegation;
1027
1028         spin_lock(&deleg_cur->lock);
1029         if (nfsi->delegation != deleg_cur ||
1030            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1031             (deleg_cur->type & fmode) != fmode)
1032                 goto no_delegation_unlock;
1033
1034         if (delegation == NULL)
1035                 delegation = &deleg_cur->stateid;
1036         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1037                 goto no_delegation_unlock;
1038
1039         nfs_mark_delegation_referenced(deleg_cur);
1040         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1041         ret = 1;
1042 no_delegation_unlock:
1043         spin_unlock(&deleg_cur->lock);
1044 no_delegation:
1045         rcu_read_unlock();
1046
1047         if (!ret && open_stateid != NULL) {
1048                 __update_open_stateid(state, open_stateid, NULL, fmode);
1049                 ret = 1;
1050         }
1051
1052         return ret;
1053 }
1054
1055
1056 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1057 {
1058         struct nfs_delegation *delegation;
1059
1060         rcu_read_lock();
1061         delegation = rcu_dereference(NFS_I(inode)->delegation);
1062         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1063                 rcu_read_unlock();
1064                 return;
1065         }
1066         rcu_read_unlock();
1067         nfs4_inode_return_delegation(inode);
1068 }
1069
1070 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1071 {
1072         struct nfs4_state *state = opendata->state;
1073         struct nfs_inode *nfsi = NFS_I(state->inode);
1074         struct nfs_delegation *delegation;
1075         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1076         fmode_t fmode = opendata->o_arg.fmode;
1077         nfs4_stateid stateid;
1078         int ret = -EAGAIN;
1079
1080         for (;;) {
1081                 if (can_open_cached(state, fmode, open_mode)) {
1082                         spin_lock(&state->owner->so_lock);
1083                         if (can_open_cached(state, fmode, open_mode)) {
1084                                 update_open_stateflags(state, fmode);
1085                                 spin_unlock(&state->owner->so_lock);
1086                                 goto out_return_state;
1087                         }
1088                         spin_unlock(&state->owner->so_lock);
1089                 }
1090                 rcu_read_lock();
1091                 delegation = rcu_dereference(nfsi->delegation);
1092                 if (!can_open_delegated(delegation, fmode)) {
1093                         rcu_read_unlock();
1094                         break;
1095                 }
1096                 /* Save the delegation */
1097                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1098                 rcu_read_unlock();
1099                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1100                 if (ret != 0)
1101                         goto out;
1102                 ret = -EAGAIN;
1103
1104                 /* Try to update the stateid using the delegation */
1105                 if (update_open_stateid(state, NULL, &stateid, fmode))
1106                         goto out_return_state;
1107         }
1108 out:
1109         return ERR_PTR(ret);
1110 out_return_state:
1111         atomic_inc(&state->count);
1112         return state;
1113 }
1114
1115 static void
1116 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1117 {
1118         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1119         struct nfs_delegation *delegation;
1120         int delegation_flags = 0;
1121
1122         rcu_read_lock();
1123         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1124         if (delegation)
1125                 delegation_flags = delegation->flags;
1126         rcu_read_unlock();
1127         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1128                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1129                                    "returning a delegation for "
1130                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1131                                    clp->cl_hostname);
1132         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1133                 nfs_inode_set_delegation(state->inode,
1134                                          data->owner->so_cred,
1135                                          &data->o_res);
1136         else
1137                 nfs_inode_reclaim_delegation(state->inode,
1138                                              data->owner->so_cred,
1139                                              &data->o_res);
1140 }
1141
1142 /*
1143  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1144  * and update the nfs4_state.
1145  */
1146 static struct nfs4_state *
1147 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1148 {
1149         struct inode *inode = data->state->inode;
1150         struct nfs4_state *state = data->state;
1151         int ret;
1152
1153         if (!data->rpc_done) {
1154                 ret = data->rpc_status;
1155                 goto err;
1156         }
1157
1158         ret = -ESTALE;
1159         if (!(data->f_attr.valid & NFS_ATTR_FATTR_TYPE) ||
1160             !(data->f_attr.valid & NFS_ATTR_FATTR_FILEID) ||
1161             !(data->f_attr.valid & NFS_ATTR_FATTR_CHANGE))
1162                 goto err;
1163
1164         ret = -ENOMEM;
1165         state = nfs4_get_open_state(inode, data->owner);
1166         if (state == NULL)
1167                 goto err;
1168
1169         ret = nfs_refresh_inode(inode, &data->f_attr);
1170         if (ret)
1171                 goto err;
1172
1173         if (data->o_res.delegation_type != 0)
1174                 nfs4_opendata_check_deleg(data, state);
1175         update_open_stateid(state, &data->o_res.stateid, NULL,
1176                             data->o_arg.fmode);
1177
1178         return state;
1179 err:
1180         return ERR_PTR(ret);
1181
1182 }
1183
1184 static struct nfs4_state *
1185 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1186 {
1187         struct inode *inode;
1188         struct nfs4_state *state = NULL;
1189         int ret;
1190
1191         if (!data->rpc_done) {
1192                 state = nfs4_try_open_cached(data);
1193                 goto out;
1194         }
1195
1196         ret = -EAGAIN;
1197         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1198                 goto err;
1199         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1200         ret = PTR_ERR(inode);
1201         if (IS_ERR(inode))
1202                 goto err;
1203         ret = -ENOMEM;
1204         state = nfs4_get_open_state(inode, data->owner);
1205         if (state == NULL)
1206                 goto err_put_inode;
1207         if (data->o_res.delegation_type != 0)
1208                 nfs4_opendata_check_deleg(data, state);
1209         update_open_stateid(state, &data->o_res.stateid, NULL,
1210                         data->o_arg.fmode);
1211         iput(inode);
1212 out:
1213         nfs_release_seqid(data->o_arg.seqid);
1214         return state;
1215 err_put_inode:
1216         iput(inode);
1217 err:
1218         return ERR_PTR(ret);
1219 }
1220
1221 static struct nfs4_state *
1222 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1223 {
1224         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1225                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1226         return _nfs4_opendata_to_nfs4_state(data);
1227 }
1228
1229 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1230 {
1231         struct nfs_inode *nfsi = NFS_I(state->inode);
1232         struct nfs_open_context *ctx;
1233
1234         spin_lock(&state->inode->i_lock);
1235         list_for_each_entry(ctx, &nfsi->open_files, list) {
1236                 if (ctx->state != state)
1237                         continue;
1238                 get_nfs_open_context(ctx);
1239                 spin_unlock(&state->inode->i_lock);
1240                 return ctx;
1241         }
1242         spin_unlock(&state->inode->i_lock);
1243         return ERR_PTR(-ENOENT);
1244 }
1245
1246 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1247                 struct nfs4_state *state, enum open_claim_type4 claim)
1248 {
1249         struct nfs4_opendata *opendata;
1250
1251         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1252                         NULL, claim, GFP_NOFS);
1253         if (opendata == NULL)
1254                 return ERR_PTR(-ENOMEM);
1255         opendata->state = state;
1256         atomic_inc(&state->count);
1257         return opendata;
1258 }
1259
1260 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1261 {
1262         struct nfs4_state *newstate;
1263         int ret;
1264
1265         opendata->o_arg.open_flags = 0;
1266         opendata->o_arg.fmode = fmode;
1267         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1268         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1269         nfs4_init_opendata_res(opendata);
1270         ret = _nfs4_recover_proc_open(opendata);
1271         if (ret != 0)
1272                 return ret;
1273         newstate = nfs4_opendata_to_nfs4_state(opendata);
1274         if (IS_ERR(newstate))
1275                 return PTR_ERR(newstate);
1276         nfs4_close_state(newstate, fmode);
1277         *res = newstate;
1278         return 0;
1279 }
1280
1281 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1282 {
1283         struct nfs4_state *newstate;
1284         int ret;
1285
1286         /* memory barrier prior to reading state->n_* */
1287         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1288         smp_rmb();
1289         if (state->n_rdwr != 0) {
1290                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1291                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1292                 if (ret != 0)
1293                         return ret;
1294                 if (newstate != state)
1295                         return -ESTALE;
1296         }
1297         if (state->n_wronly != 0) {
1298                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1299                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1300                 if (ret != 0)
1301                         return ret;
1302                 if (newstate != state)
1303                         return -ESTALE;
1304         }
1305         if (state->n_rdonly != 0) {
1306                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1307                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1308                 if (ret != 0)
1309                         return ret;
1310                 if (newstate != state)
1311                         return -ESTALE;
1312         }
1313         /*
1314          * We may have performed cached opens for all three recoveries.
1315          * Check if we need to update the current stateid.
1316          */
1317         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1318             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1319                 write_seqlock(&state->seqlock);
1320                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1321                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1322                 write_sequnlock(&state->seqlock);
1323         }
1324         return 0;
1325 }
1326
1327 /*
1328  * OPEN_RECLAIM:
1329  *      reclaim state on the server after a reboot.
1330  */
1331 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1332 {
1333         struct nfs_delegation *delegation;
1334         struct nfs4_opendata *opendata;
1335         fmode_t delegation_type = 0;
1336         int status;
1337
1338         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1339                         NFS4_OPEN_CLAIM_PREVIOUS);
1340         if (IS_ERR(opendata))
1341                 return PTR_ERR(opendata);
1342         rcu_read_lock();
1343         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1344         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1345                 delegation_type = delegation->type;
1346         rcu_read_unlock();
1347         opendata->o_arg.u.delegation_type = delegation_type;
1348         status = nfs4_open_recover(opendata, state);
1349         nfs4_opendata_put(opendata);
1350         return status;
1351 }
1352
1353 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1354 {
1355         struct nfs_server *server = NFS_SERVER(state->inode);
1356         struct nfs4_exception exception = { };
1357         int err;
1358         do {
1359                 err = _nfs4_do_open_reclaim(ctx, state);
1360                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1361                         continue;
1362                 if (err != -NFS4ERR_DELAY)
1363                         break;
1364                 nfs4_handle_exception(server, err, &exception);
1365         } while (exception.retry);
1366         return err;
1367 }
1368
1369 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1370 {
1371         struct nfs_open_context *ctx;
1372         int ret;
1373
1374         ctx = nfs4_state_find_open_context(state);
1375         if (IS_ERR(ctx))
1376                 return PTR_ERR(ctx);
1377         ret = nfs4_do_open_reclaim(ctx, state);
1378         put_nfs_open_context(ctx);
1379         return ret;
1380 }
1381
1382 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1383 {
1384         struct nfs4_opendata *opendata;
1385         int ret;
1386
1387         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1388                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1389         if (IS_ERR(opendata))
1390                 return PTR_ERR(opendata);
1391         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1392         ret = nfs4_open_recover(opendata, state);
1393         nfs4_opendata_put(opendata);
1394         return ret;
1395 }
1396
1397 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1398 {
1399         struct nfs4_exception exception = { };
1400         struct nfs_server *server = NFS_SERVER(state->inode);
1401         int err;
1402         do {
1403                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1404                 switch (err) {
1405                         case 0:
1406                         case -ENOENT:
1407                         case -ESTALE:
1408                                 goto out;
1409                         case -NFS4ERR_BADSESSION:
1410                         case -NFS4ERR_BADSLOT:
1411                         case -NFS4ERR_BAD_HIGH_SLOT:
1412                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1413                         case -NFS4ERR_DEADSESSION:
1414                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1415                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1416                                 err = -EAGAIN;
1417                                 goto out;
1418                         case -NFS4ERR_STALE_CLIENTID:
1419                         case -NFS4ERR_STALE_STATEID:
1420                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1421                         case -NFS4ERR_EXPIRED:
1422                                 /* Don't recall a delegation if it was lost */
1423                                 nfs4_schedule_lease_recovery(server->nfs_client);
1424                                 err = -EAGAIN;
1425                                 goto out;
1426                         case -NFS4ERR_DELEG_REVOKED:
1427                         case -NFS4ERR_ADMIN_REVOKED:
1428                         case -NFS4ERR_BAD_STATEID:
1429                                 nfs_inode_find_state_and_recover(state->inode,
1430                                                 stateid);
1431                                 nfs4_schedule_stateid_recovery(server, state);
1432                         case -ENOMEM:
1433                                 err = 0;
1434                                 goto out;
1435                 }
1436                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1437                 err = nfs4_handle_exception(server, err, &exception);
1438         } while (exception.retry);
1439 out:
1440         return err;
1441 }
1442
1443 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1444 {
1445         struct nfs4_opendata *data = calldata;
1446
1447         data->rpc_status = task->tk_status;
1448         if (data->rpc_status == 0) {
1449                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1450                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1451                 renew_lease(data->o_res.server, data->timestamp);
1452                 data->rpc_done = 1;
1453         }
1454 }
1455
1456 static void nfs4_open_confirm_release(void *calldata)
1457 {
1458         struct nfs4_opendata *data = calldata;
1459         struct nfs4_state *state = NULL;
1460
1461         /* If this request hasn't been cancelled, do nothing */
1462         if (data->cancelled == 0)
1463                 goto out_free;
1464         /* In case of error, no cleanup! */
1465         if (!data->rpc_done)
1466                 goto out_free;
1467         state = nfs4_opendata_to_nfs4_state(data);
1468         if (!IS_ERR(state))
1469                 nfs4_close_state(state, data->o_arg.fmode);
1470 out_free:
1471         nfs4_opendata_put(data);
1472 }
1473
1474 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1475         .rpc_call_done = nfs4_open_confirm_done,
1476         .rpc_release = nfs4_open_confirm_release,
1477 };
1478
1479 /*
1480  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1481  */
1482 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1483 {
1484         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1485         struct rpc_task *task;
1486         struct  rpc_message msg = {
1487                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1488                 .rpc_argp = &data->c_arg,
1489                 .rpc_resp = &data->c_res,
1490                 .rpc_cred = data->owner->so_cred,
1491         };
1492         struct rpc_task_setup task_setup_data = {
1493                 .rpc_client = server->client,
1494                 .rpc_message = &msg,
1495                 .callback_ops = &nfs4_open_confirm_ops,
1496                 .callback_data = data,
1497                 .workqueue = nfsiod_workqueue,
1498                 .flags = RPC_TASK_ASYNC,
1499         };
1500         int status;
1501
1502         kref_get(&data->kref);
1503         data->rpc_done = 0;
1504         data->rpc_status = 0;
1505         data->timestamp = jiffies;
1506         task = rpc_run_task(&task_setup_data);
1507         if (IS_ERR(task))
1508                 return PTR_ERR(task);
1509         status = nfs4_wait_for_completion_rpc_task(task);
1510         if (status != 0) {
1511                 data->cancelled = 1;
1512                 smp_wmb();
1513         } else
1514                 status = data->rpc_status;
1515         rpc_put_task(task);
1516         return status;
1517 }
1518
1519 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1520 {
1521         struct nfs4_opendata *data = calldata;
1522         struct nfs4_state_owner *sp = data->owner;
1523
1524         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1525                 goto out_wait;
1526         /*
1527          * Check if we still need to send an OPEN call, or if we can use
1528          * a delegation instead.
1529          */
1530         if (data->state != NULL) {
1531                 struct nfs_delegation *delegation;
1532
1533                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1534                         goto out_no_action;
1535                 rcu_read_lock();
1536                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1537                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1538                     can_open_delegated(delegation, data->o_arg.fmode))
1539                         goto unlock_no_action;
1540                 rcu_read_unlock();
1541         }
1542         /* Update client id. */
1543         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1544         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1545                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1546                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1547                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1548         }
1549         data->timestamp = jiffies;
1550         if (nfs4_setup_sequence(data->o_arg.server,
1551                                 &data->o_arg.seq_args,
1552                                 &data->o_res.seq_res,
1553                                 task) != 0)
1554                 nfs_release_seqid(data->o_arg.seqid);
1555         return;
1556 unlock_no_action:
1557         rcu_read_unlock();
1558 out_no_action:
1559         task->tk_action = NULL;
1560 out_wait:
1561         nfs4_sequence_done(task, &data->o_res.seq_res);
1562 }
1563
1564 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1565 {
1566         struct nfs4_opendata *data = calldata;
1567
1568         data->rpc_status = task->tk_status;
1569
1570         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1571                 return;
1572
1573         if (task->tk_status == 0) {
1574                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1575                         switch (data->o_res.f_attr->mode & S_IFMT) {
1576                         case S_IFREG:
1577                                 break;
1578                         case S_IFLNK:
1579                                 data->rpc_status = -ELOOP;
1580                                 break;
1581                         case S_IFDIR:
1582                                 data->rpc_status = -EISDIR;
1583                                 break;
1584                         default:
1585                                 data->rpc_status = -ENOTDIR;
1586                         }
1587                 }
1588                 renew_lease(data->o_res.server, data->timestamp);
1589                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1590                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1591         }
1592         data->rpc_done = 1;
1593 }
1594
1595 static void nfs4_open_release(void *calldata)
1596 {
1597         struct nfs4_opendata *data = calldata;
1598         struct nfs4_state *state = NULL;
1599
1600         /* If this request hasn't been cancelled, do nothing */
1601         if (data->cancelled == 0)
1602                 goto out_free;
1603         /* In case of error, no cleanup! */
1604         if (data->rpc_status != 0 || !data->rpc_done)
1605                 goto out_free;
1606         /* In case we need an open_confirm, no cleanup! */
1607         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1608                 goto out_free;
1609         state = nfs4_opendata_to_nfs4_state(data);
1610         if (!IS_ERR(state))
1611                 nfs4_close_state(state, data->o_arg.fmode);
1612 out_free:
1613         nfs4_opendata_put(data);
1614 }
1615
1616 static const struct rpc_call_ops nfs4_open_ops = {
1617         .rpc_call_prepare = nfs4_open_prepare,
1618         .rpc_call_done = nfs4_open_done,
1619         .rpc_release = nfs4_open_release,
1620 };
1621
1622 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1623 {
1624         struct inode *dir = data->dir->d_inode;
1625         struct nfs_server *server = NFS_SERVER(dir);
1626         struct nfs_openargs *o_arg = &data->o_arg;
1627         struct nfs_openres *o_res = &data->o_res;
1628         struct rpc_task *task;
1629         struct rpc_message msg = {
1630                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1631                 .rpc_argp = o_arg,
1632                 .rpc_resp = o_res,
1633                 .rpc_cred = data->owner->so_cred,
1634         };
1635         struct rpc_task_setup task_setup_data = {
1636                 .rpc_client = server->client,
1637                 .rpc_message = &msg,
1638                 .callback_ops = &nfs4_open_ops,
1639                 .callback_data = data,
1640                 .workqueue = nfsiod_workqueue,
1641                 .flags = RPC_TASK_ASYNC,
1642         };
1643         int status;
1644
1645         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1646         kref_get(&data->kref);
1647         data->rpc_done = 0;
1648         data->rpc_status = 0;
1649         data->cancelled = 0;
1650         if (isrecover)
1651                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1652         task = rpc_run_task(&task_setup_data);
1653         if (IS_ERR(task))
1654                 return PTR_ERR(task);
1655         status = nfs4_wait_for_completion_rpc_task(task);
1656         if (status != 0) {
1657                 data->cancelled = 1;
1658                 smp_wmb();
1659         } else
1660                 status = data->rpc_status;
1661         rpc_put_task(task);
1662
1663         return status;
1664 }
1665
1666 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1667 {
1668         struct inode *dir = data->dir->d_inode;
1669         struct nfs_openres *o_res = &data->o_res;
1670         int status;
1671
1672         status = nfs4_run_open_task(data, 1);
1673         if (status != 0 || !data->rpc_done)
1674                 return status;
1675
1676         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1677
1678         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1679                 status = _nfs4_proc_open_confirm(data);
1680                 if (status != 0)
1681                         return status;
1682         }
1683
1684         return status;
1685 }
1686
1687 static int nfs4_opendata_access(struct rpc_cred *cred,
1688                                 struct nfs4_opendata *opendata,
1689                                 struct nfs4_state *state, fmode_t fmode,
1690                                 int openflags)
1691 {
1692         struct nfs_access_entry cache;
1693         u32 mask;
1694
1695         /* access call failed or for some reason the server doesn't
1696          * support any access modes -- defer access call until later */
1697         if (opendata->o_res.access_supported == 0)
1698                 return 0;
1699
1700         mask = 0;
1701         /* don't check MAY_WRITE - a newly created file may not have
1702          * write mode bits, but POSIX allows the creating process to write.
1703          * use openflags to check for exec, because fmode won't
1704          * always have FMODE_EXEC set when file open for exec. */
1705         if (openflags & __FMODE_EXEC) {
1706                 /* ONLY check for exec rights */
1707                 mask = MAY_EXEC;
1708         } else if (fmode & FMODE_READ)
1709                 mask = MAY_READ;
1710
1711         cache.cred = cred;
1712         cache.jiffies = jiffies;
1713         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1714         nfs_access_add_cache(state->inode, &cache);
1715
1716         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1717                 return 0;
1718
1719         /* even though OPEN succeeded, access is denied. Close the file */
1720         nfs4_close_state(state, fmode);
1721         return -EACCES;
1722 }
1723
1724 /*
1725  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1726  */
1727 static int _nfs4_proc_open(struct nfs4_opendata *data)
1728 {
1729         struct inode *dir = data->dir->d_inode;
1730         struct nfs_server *server = NFS_SERVER(dir);
1731         struct nfs_openargs *o_arg = &data->o_arg;
1732         struct nfs_openres *o_res = &data->o_res;
1733         int status;
1734
1735         status = nfs4_run_open_task(data, 0);
1736         if (!data->rpc_done)
1737                 return status;
1738         if (status != 0) {
1739                 if (status == -NFS4ERR_BADNAME &&
1740                                 !(o_arg->open_flags & O_CREAT))
1741                         return -ENOENT;
1742                 return status;
1743         }
1744
1745         nfs_fattr_map_and_free_names(server, &data->f_attr);
1746
1747         if (o_arg->open_flags & O_CREAT)
1748                 update_changeattr(dir, &o_res->cinfo);
1749         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1750                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1751         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1752                 status = _nfs4_proc_open_confirm(data);
1753                 if (status != 0)
1754                         return status;
1755         }
1756         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1757                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1758         return 0;
1759 }
1760
1761 static int nfs4_recover_expired_lease(struct nfs_server *server)
1762 {
1763         return nfs4_client_recover_expired_lease(server->nfs_client);
1764 }
1765
1766 /*
1767  * OPEN_EXPIRED:
1768  *      reclaim state on the server after a network partition.
1769  *      Assumes caller holds the appropriate lock
1770  */
1771 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1772 {
1773         struct nfs4_opendata *opendata;
1774         int ret;
1775
1776         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1777                         NFS4_OPEN_CLAIM_FH);
1778         if (IS_ERR(opendata))
1779                 return PTR_ERR(opendata);
1780         ret = nfs4_open_recover(opendata, state);
1781         if (ret == -ESTALE)
1782                 d_drop(ctx->dentry);
1783         nfs4_opendata_put(opendata);
1784         return ret;
1785 }
1786
1787 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1788 {
1789         struct nfs_server *server = NFS_SERVER(state->inode);
1790         struct nfs4_exception exception = { };
1791         int err;
1792
1793         do {
1794                 err = _nfs4_open_expired(ctx, state);
1795                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1796                         continue;
1797                 switch (err) {
1798                 default:
1799                         goto out;
1800                 case -NFS4ERR_GRACE:
1801                 case -NFS4ERR_DELAY:
1802                         nfs4_handle_exception(server, err, &exception);
1803                         err = 0;
1804                 }
1805         } while (exception.retry);
1806 out:
1807         return err;
1808 }
1809
1810 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1811 {
1812         struct nfs_open_context *ctx;
1813         int ret;
1814
1815         ctx = nfs4_state_find_open_context(state);
1816         if (IS_ERR(ctx))
1817                 return PTR_ERR(ctx);
1818         ret = nfs4_do_open_expired(ctx, state);
1819         put_nfs_open_context(ctx);
1820         return ret;
1821 }
1822
1823 #if defined(CONFIG_NFS_V4_1)
1824 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1825 {
1826         struct nfs_server *server = NFS_SERVER(state->inode);
1827         nfs4_stateid *stateid = &state->stateid;
1828         int status;
1829
1830         /* If a state reset has been done, test_stateid is unneeded */
1831         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1832                 return;
1833
1834         status = nfs41_test_stateid(server, stateid);
1835         if (status != NFS_OK) {
1836                 /* Free the stateid unless the server explicitly
1837                  * informs us the stateid is unrecognized. */
1838                 if (status != -NFS4ERR_BAD_STATEID)
1839                         nfs41_free_stateid(server, stateid);
1840                 nfs_remove_bad_delegation(state->inode);
1841
1842                 write_seqlock(&state->seqlock);
1843                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1844                 write_sequnlock(&state->seqlock);
1845                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1846         }
1847 }
1848
1849 /**
1850  * nfs41_check_open_stateid - possibly free an open stateid
1851  *
1852  * @state: NFSv4 state for an inode
1853  *
1854  * Returns NFS_OK if recovery for this stateid is now finished.
1855  * Otherwise a negative NFS4ERR value is returned.
1856  */
1857 static int nfs41_check_open_stateid(struct nfs4_state *state)
1858 {
1859         struct nfs_server *server = NFS_SERVER(state->inode);
1860         nfs4_stateid *stateid = &state->open_stateid;
1861         int status;
1862
1863         /* If a state reset has been done, test_stateid is unneeded */
1864         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1865             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1866             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1867                 return -NFS4ERR_BAD_STATEID;
1868
1869         status = nfs41_test_stateid(server, stateid);
1870         if (status != NFS_OK) {
1871                 /* Free the stateid unless the server explicitly
1872                  * informs us the stateid is unrecognized. */
1873                 if (status != -NFS4ERR_BAD_STATEID)
1874                         nfs41_free_stateid(server, stateid);
1875
1876                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1877                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1878                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1879         }
1880         return status;
1881 }
1882
1883 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1884 {
1885         int status;
1886
1887         nfs41_clear_delegation_stateid(state);
1888         status = nfs41_check_open_stateid(state);
1889         if (status != NFS_OK)
1890                 status = nfs4_open_expired(sp, state);
1891         return status;
1892 }
1893 #endif
1894
1895 /*
1896  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1897  * fields corresponding to attributes that were used to store the verifier.
1898  * Make sure we clobber those fields in the later setattr call
1899  */
1900 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1901 {
1902         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1903             !(sattr->ia_valid & ATTR_ATIME_SET))
1904                 sattr->ia_valid |= ATTR_ATIME;
1905
1906         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1907             !(sattr->ia_valid & ATTR_MTIME_SET))
1908                 sattr->ia_valid |= ATTR_MTIME;
1909 }
1910
1911 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
1912                 fmode_t fmode,
1913                 int flags,
1914                 struct nfs4_state **res)
1915 {
1916         struct nfs4_state_owner *sp = opendata->owner;
1917         struct nfs_server *server = sp->so_server;
1918         struct nfs4_state *state;
1919         unsigned int seq;
1920         int ret;
1921
1922         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
1923
1924         ret = _nfs4_proc_open(opendata);
1925         if (ret != 0)
1926                 goto out;
1927
1928         state = nfs4_opendata_to_nfs4_state(opendata);
1929         ret = PTR_ERR(state);
1930         if (IS_ERR(state))
1931                 goto out;
1932         if (server->caps & NFS_CAP_POSIX_LOCK)
1933                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1934
1935         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
1936         if (ret != 0)
1937                 goto out;
1938
1939         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
1940                 nfs4_schedule_stateid_recovery(server, state);
1941                 nfs4_wait_clnt_recover(server->nfs_client);
1942         }
1943         *res = state;
1944 out:
1945         return ret;
1946 }
1947
1948 /*
1949  * Returns a referenced nfs4_state
1950  */
1951 static int _nfs4_do_open(struct inode *dir,
1952                         struct dentry *dentry,
1953                         fmode_t fmode,
1954                         int flags,
1955                         struct iattr *sattr,
1956                         struct rpc_cred *cred,
1957                         struct nfs4_state **res,
1958                         struct nfs4_threshold **ctx_th)
1959 {
1960         struct nfs4_state_owner  *sp;
1961         struct nfs4_state     *state = NULL;
1962         struct nfs_server       *server = NFS_SERVER(dir);
1963         struct nfs4_opendata *opendata;
1964         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
1965         int status;
1966
1967         /* Protect against reboot recovery conflicts */
1968         status = -ENOMEM;
1969         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1970         if (sp == NULL) {
1971                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1972                 goto out_err;
1973         }
1974         status = nfs4_recover_expired_lease(server);
1975         if (status != 0)
1976                 goto err_put_state_owner;
1977         if (dentry->d_inode != NULL)
1978                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1979         status = -ENOMEM;
1980         if (dentry->d_inode)
1981                 claim = NFS4_OPEN_CLAIM_FH;
1982         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
1983                         claim, GFP_KERNEL);
1984         if (opendata == NULL)
1985                 goto err_put_state_owner;
1986
1987         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1988                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1989                 if (!opendata->f_attr.mdsthreshold)
1990                         goto err_opendata_put;
1991                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
1992         }
1993         if (dentry->d_inode != NULL)
1994                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1995
1996         status = _nfs4_open_and_get_state(opendata, fmode, flags, &state);
1997         if (status != 0)
1998                 goto err_opendata_put;
1999
2000         if (opendata->o_arg.open_flags & O_EXCL) {
2001                 nfs4_exclusive_attrset(opendata, sattr);
2002
2003                 nfs_fattr_init(opendata->o_res.f_attr);
2004                 status = nfs4_do_setattr(state->inode, cred,
2005                                 opendata->o_res.f_attr, sattr,
2006                                 state);
2007                 if (status == 0)
2008                         nfs_setattr_update_inode(state->inode, sattr);
2009                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2010         }
2011
2012         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2013                 *ctx_th = opendata->f_attr.mdsthreshold;
2014         else
2015                 kfree(opendata->f_attr.mdsthreshold);
2016         opendata->f_attr.mdsthreshold = NULL;
2017
2018         nfs4_opendata_put(opendata);
2019         nfs4_put_state_owner(sp);
2020         *res = state;
2021         return 0;
2022 err_opendata_put:
2023         kfree(opendata->f_attr.mdsthreshold);
2024         nfs4_opendata_put(opendata);
2025 err_put_state_owner:
2026         nfs4_put_state_owner(sp);
2027 out_err:
2028         *res = NULL;
2029         return status;
2030 }
2031
2032
2033 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2034                                         struct dentry *dentry,
2035                                         fmode_t fmode,
2036                                         int flags,
2037                                         struct iattr *sattr,
2038                                         struct rpc_cred *cred,
2039                                         struct nfs4_threshold **ctx_th)
2040 {
2041         struct nfs_server *server = NFS_SERVER(dir);
2042         struct nfs4_exception exception = { };
2043         struct nfs4_state *res;
2044         int status;
2045
2046         fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
2047         do {
2048                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
2049                                        &res, ctx_th);
2050                 if (status == 0)
2051                         break;
2052                 /* NOTE: BAD_SEQID means the server and client disagree about the
2053                  * book-keeping w.r.t. state-changing operations
2054                  * (OPEN/CLOSE/LOCK/LOCKU...)
2055                  * It is actually a sign of a bug on the client or on the server.
2056                  *
2057                  * If we receive a BAD_SEQID error in the particular case of
2058                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2059                  * have unhashed the old state_owner for us, and that we can
2060                  * therefore safely retry using a new one. We should still warn
2061                  * the user though...
2062                  */
2063                 if (status == -NFS4ERR_BAD_SEQID) {
2064                         pr_warn_ratelimited("NFS: v4 server %s "
2065                                         " returned a bad sequence-id error!\n",
2066                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2067                         exception.retry = 1;
2068                         continue;
2069                 }
2070                 /*
2071                  * BAD_STATEID on OPEN means that the server cancelled our
2072                  * state before it received the OPEN_CONFIRM.
2073                  * Recover by retrying the request as per the discussion
2074                  * on Page 181 of RFC3530.
2075                  */
2076                 if (status == -NFS4ERR_BAD_STATEID) {
2077                         exception.retry = 1;
2078                         continue;
2079                 }
2080                 if (status == -EAGAIN) {
2081                         /* We must have found a delegation */
2082                         exception.retry = 1;
2083                         continue;
2084                 }
2085                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2086                         continue;
2087                 res = ERR_PTR(nfs4_handle_exception(server,
2088                                         status, &exception));
2089         } while (exception.retry);
2090         return res;
2091 }
2092
2093 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2094                             struct nfs_fattr *fattr, struct iattr *sattr,
2095                             struct nfs4_state *state)
2096 {
2097         struct nfs_server *server = NFS_SERVER(inode);
2098         struct nfs_setattrargs  arg = {
2099                 .fh             = NFS_FH(inode),
2100                 .iap            = sattr,
2101                 .server         = server,
2102                 .bitmask = server->attr_bitmask,
2103         };
2104         struct nfs_setattrres  res = {
2105                 .fattr          = fattr,
2106                 .server         = server,
2107         };
2108         struct rpc_message msg = {
2109                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2110                 .rpc_argp       = &arg,
2111                 .rpc_resp       = &res,
2112                 .rpc_cred       = cred,
2113         };
2114         unsigned long timestamp = jiffies;
2115         int status;
2116
2117         nfs_fattr_init(fattr);
2118
2119         if (state != NULL && nfs4_valid_open_stateid(state)) {
2120                 struct nfs_lockowner lockowner = {
2121                         .l_owner = current->files,
2122                         .l_pid = current->tgid,
2123                 };
2124                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2125                                 &lockowner);
2126         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
2127                                 FMODE_WRITE)) {
2128                 /* Use that stateid */
2129         } else
2130                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2131
2132         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2133         if (status == 0 && state != NULL)
2134                 renew_lease(server, timestamp);
2135         return status;
2136 }
2137
2138 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2139                            struct nfs_fattr *fattr, struct iattr *sattr,
2140                            struct nfs4_state *state)
2141 {
2142         struct nfs_server *server = NFS_SERVER(inode);
2143         struct nfs4_exception exception = {
2144                 .state = state,
2145                 .inode = inode,
2146         };
2147         int err;
2148         do {
2149                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2150                 switch (err) {
2151                 case -NFS4ERR_OPENMODE:
2152                         if (state && !(state->state & FMODE_WRITE)) {
2153                                 err = -EBADF;
2154                                 if (sattr->ia_valid & ATTR_OPEN)
2155                                         err = -EACCES;
2156                                 goto out;
2157                         }
2158                 }
2159                 err = nfs4_handle_exception(server, err, &exception);
2160         } while (exception.retry);
2161 out:
2162         return err;
2163 }
2164
2165 struct nfs4_closedata {
2166         struct inode *inode;
2167         struct nfs4_state *state;
2168         struct nfs_closeargs arg;
2169         struct nfs_closeres res;
2170         struct nfs_fattr fattr;
2171         unsigned long timestamp;
2172         bool roc;
2173         u32 roc_barrier;
2174 };
2175
2176 static void nfs4_free_closedata(void *data)
2177 {
2178         struct nfs4_closedata *calldata = data;
2179         struct nfs4_state_owner *sp = calldata->state->owner;
2180         struct super_block *sb = calldata->state->inode->i_sb;
2181
2182         if (calldata->roc)
2183                 pnfs_roc_release(calldata->state->inode);
2184         nfs4_put_open_state(calldata->state);
2185         nfs_free_seqid(calldata->arg.seqid);
2186         nfs4_put_state_owner(sp);
2187         nfs_sb_deactive(sb);
2188         kfree(calldata);
2189 }
2190
2191 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2192                 fmode_t fmode)
2193 {
2194         spin_lock(&state->owner->so_lock);
2195         if (!(fmode & FMODE_READ))
2196                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2197         if (!(fmode & FMODE_WRITE))
2198                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2199         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2200         spin_unlock(&state->owner->so_lock);
2201 }
2202
2203 static void nfs4_close_done(struct rpc_task *task, void *data)
2204 {
2205         struct nfs4_closedata *calldata = data;
2206         struct nfs4_state *state = calldata->state;
2207         struct nfs_server *server = NFS_SERVER(calldata->inode);
2208
2209         dprintk("%s: begin!\n", __func__);
2210         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2211                 return;
2212         /* hmm. we are done with the inode, and in the process of freeing
2213          * the state_owner. we keep this around to process errors
2214          */
2215         switch (task->tk_status) {
2216                 case 0:
2217                         if (calldata->roc)
2218                                 pnfs_roc_set_barrier(state->inode,
2219                                                      calldata->roc_barrier);
2220                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2221                         renew_lease(server, calldata->timestamp);
2222                         nfs4_close_clear_stateid_flags(state,
2223                                         calldata->arg.fmode);
2224                         break;
2225                 case -NFS4ERR_STALE_STATEID:
2226                 case -NFS4ERR_OLD_STATEID:
2227                 case -NFS4ERR_BAD_STATEID:
2228                 case -NFS4ERR_EXPIRED:
2229                         if (calldata->arg.fmode == 0)
2230                                 break;
2231                 default:
2232                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2233                                 rpc_restart_call_prepare(task);
2234         }
2235         nfs_release_seqid(calldata->arg.seqid);
2236         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2237         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2238 }
2239
2240 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2241 {
2242         struct nfs4_closedata *calldata = data;
2243         struct nfs4_state *state = calldata->state;
2244         struct inode *inode = calldata->inode;
2245         int call_close = 0;
2246
2247         dprintk("%s: begin!\n", __func__);
2248         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2249                 goto out_wait;
2250
2251         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2252         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2253         spin_lock(&state->owner->so_lock);
2254         /* Calculate the change in open mode */
2255         if (state->n_rdwr == 0) {
2256                 if (state->n_rdonly == 0) {
2257                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2258                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2259                         calldata->arg.fmode &= ~FMODE_READ;
2260                 }
2261                 if (state->n_wronly == 0) {
2262                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2263                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2264                         calldata->arg.fmode &= ~FMODE_WRITE;
2265                 }
2266         }
2267         if (!nfs4_valid_open_stateid(state))
2268                 call_close = 0;
2269         spin_unlock(&state->owner->so_lock);
2270
2271         if (!call_close) {
2272                 /* Note: exit _without_ calling nfs4_close_done */
2273                 goto out_no_action;
2274         }
2275
2276         if (calldata->arg.fmode == 0) {
2277                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2278                 if (calldata->roc &&
2279                     pnfs_roc_drain(inode, &calldata->roc_barrier, task))
2280                         goto out_wait;
2281         }
2282
2283         nfs_fattr_init(calldata->res.fattr);
2284         calldata->timestamp = jiffies;
2285         if (nfs4_setup_sequence(NFS_SERVER(inode),
2286                                 &calldata->arg.seq_args,
2287                                 &calldata->res.seq_res,
2288                                 task) != 0)
2289                 nfs_release_seqid(calldata->arg.seqid);
2290         dprintk("%s: done!\n", __func__);
2291         return;
2292 out_no_action:
2293         task->tk_action = NULL;
2294 out_wait:
2295         nfs4_sequence_done(task, &calldata->res.seq_res);
2296 }
2297
2298 static const struct rpc_call_ops nfs4_close_ops = {
2299         .rpc_call_prepare = nfs4_close_prepare,
2300         .rpc_call_done = nfs4_close_done,
2301         .rpc_release = nfs4_free_closedata,
2302 };
2303
2304 /*
2305  * It is possible for data to be read/written from a mem-mapped file
2306  * after the sys_close call (which hits the vfs layer as a flush).
2307  * This means that we can't safely call nfsv4 close on a file until
2308  * the inode is cleared. This in turn means that we are not good
2309  * NFSv4 citizens - we do not indicate to the server to update the file's
2310  * share state even when we are done with one of the three share
2311  * stateid's in the inode.
2312  *
2313  * NOTE: Caller must be holding the sp->so_owner semaphore!
2314  */
2315 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2316 {
2317         struct nfs_server *server = NFS_SERVER(state->inode);
2318         struct nfs4_closedata *calldata;
2319         struct nfs4_state_owner *sp = state->owner;
2320         struct rpc_task *task;
2321         struct rpc_message msg = {
2322                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2323                 .rpc_cred = state->owner->so_cred,
2324         };
2325         struct rpc_task_setup task_setup_data = {
2326                 .rpc_client = server->client,
2327                 .rpc_message = &msg,
2328                 .callback_ops = &nfs4_close_ops,
2329                 .workqueue = nfsiod_workqueue,
2330                 .flags = RPC_TASK_ASYNC,
2331         };
2332         int status = -ENOMEM;
2333
2334         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2335         if (calldata == NULL)
2336                 goto out;
2337         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2338         calldata->inode = state->inode;
2339         calldata->state = state;
2340         calldata->arg.fh = NFS_FH(state->inode);
2341         calldata->arg.stateid = &state->open_stateid;
2342         /* Serialization for the sequence id */
2343         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2344         if (calldata->arg.seqid == NULL)
2345                 goto out_free_calldata;
2346         calldata->arg.fmode = 0;
2347         calldata->arg.bitmask = server->cache_consistency_bitmask;
2348         calldata->res.fattr = &calldata->fattr;
2349         calldata->res.seqid = calldata->arg.seqid;
2350         calldata->res.server = server;
2351         calldata->roc = pnfs_roc(state->inode);
2352         nfs_sb_active(calldata->inode->i_sb);
2353
2354         msg.rpc_argp = &calldata->arg;
2355         msg.rpc_resp = &calldata->res;
2356         task_setup_data.callback_data = calldata;
2357         task = rpc_run_task(&task_setup_data);
2358         if (IS_ERR(task))
2359                 return PTR_ERR(task);
2360         status = 0;
2361         if (wait)
2362                 status = rpc_wait_for_completion_task(task);
2363         rpc_put_task(task);
2364         return status;
2365 out_free_calldata:
2366         kfree(calldata);
2367 out:
2368         nfs4_put_open_state(state);
2369         nfs4_put_state_owner(sp);
2370         return status;
2371 }
2372
2373 static struct inode *
2374 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2375 {
2376         struct nfs4_state *state;
2377
2378         /* Protect against concurrent sillydeletes */
2379         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2380                              ctx->cred, &ctx->mdsthreshold);
2381         if (IS_ERR(state))
2382                 return ERR_CAST(state);
2383         ctx->state = state;
2384         return igrab(state->inode);
2385 }
2386
2387 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2388 {
2389         if (ctx->state == NULL)
2390                 return;
2391         if (is_sync)
2392                 nfs4_close_sync(ctx->state, ctx->mode);
2393         else
2394                 nfs4_close_state(ctx->state, ctx->mode);
2395 }
2396
2397 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2398 {
2399         struct nfs4_server_caps_arg args = {
2400                 .fhandle = fhandle,
2401         };
2402         struct nfs4_server_caps_res res = {};
2403         struct rpc_message msg = {
2404                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2405                 .rpc_argp = &args,
2406                 .rpc_resp = &res,
2407         };
2408         int status;
2409
2410         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2411         if (status == 0) {
2412                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2413                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2414                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2415                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2416                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2417                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2418                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2419                         server->caps |= NFS_CAP_ACLS;
2420                 if (res.has_links != 0)
2421                         server->caps |= NFS_CAP_HARDLINKS;
2422                 if (res.has_symlinks != 0)
2423                         server->caps |= NFS_CAP_SYMLINKS;
2424                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2425                         server->caps |= NFS_CAP_FILEID;
2426                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2427                         server->caps |= NFS_CAP_MODE;
2428                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2429                         server->caps |= NFS_CAP_NLINK;
2430                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2431                         server->caps |= NFS_CAP_OWNER;
2432                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2433                         server->caps |= NFS_CAP_OWNER_GROUP;
2434                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2435                         server->caps |= NFS_CAP_ATIME;
2436                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2437                         server->caps |= NFS_CAP_CTIME;
2438                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2439                         server->caps |= NFS_CAP_MTIME;
2440
2441                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2442                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2443                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2444                 server->acl_bitmask = res.acl_bitmask;
2445                 server->fh_expire_type = res.fh_expire_type;
2446         }
2447
2448         return status;
2449 }
2450
2451 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2452 {
2453         struct nfs4_exception exception = { };
2454         int err;
2455         do {
2456                 err = nfs4_handle_exception(server,
2457                                 _nfs4_server_capabilities(server, fhandle),
2458                                 &exception);
2459         } while (exception.retry);
2460         return err;
2461 }
2462
2463 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2464                 struct nfs_fsinfo *info)
2465 {
2466         struct nfs4_lookup_root_arg args = {
2467                 .bitmask = nfs4_fattr_bitmap,
2468         };
2469         struct nfs4_lookup_res res = {
2470                 .server = server,
2471                 .fattr = info->fattr,
2472                 .fh = fhandle,
2473         };
2474         struct rpc_message msg = {
2475                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2476                 .rpc_argp = &args,
2477                 .rpc_resp = &res,
2478         };
2479
2480         nfs_fattr_init(info->fattr);
2481         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2482 }
2483
2484 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2485                 struct nfs_fsinfo *info)
2486 {
2487         struct nfs4_exception exception = { };
2488         int err;
2489         do {
2490                 err = _nfs4_lookup_root(server, fhandle, info);
2491                 switch (err) {
2492                 case 0:
2493                 case -NFS4ERR_WRONGSEC:
2494                         goto out;
2495                 default:
2496                         err = nfs4_handle_exception(server, err, &exception);
2497                 }
2498         } while (exception.retry);
2499 out:
2500         return err;
2501 }
2502
2503 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2504                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2505 {
2506         struct rpc_auth *auth;
2507         int ret;
2508
2509         auth = rpcauth_create(flavor, server->client);
2510         if (IS_ERR(auth)) {
2511                 ret = -EIO;
2512                 goto out;
2513         }
2514         ret = nfs4_lookup_root(server, fhandle, info);
2515 out:
2516         return ret;
2517 }
2518
2519 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2520                               struct nfs_fsinfo *info)
2521 {
2522         int i, len, status = 0;
2523         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2524
2525         len = rpcauth_list_flavors(flav_array, ARRAY_SIZE(flav_array));
2526         if (len < 0)
2527                 return len;
2528
2529         for (i = 0; i < len; i++) {
2530                 /* AUTH_UNIX is the default flavor if none was specified,
2531                  * thus has already been tried. */
2532                 if (flav_array[i] == RPC_AUTH_UNIX)
2533                         continue;
2534
2535                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2536                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2537                         continue;
2538                 break;
2539         }
2540         /*
2541          * -EACCESS could mean that the user doesn't have correct permissions
2542          * to access the mount.  It could also mean that we tried to mount
2543          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2544          * existing mount programs don't handle -EACCES very well so it should
2545          * be mapped to -EPERM instead.
2546          */
2547         if (status == -EACCES)
2548                 status = -EPERM;
2549         return status;
2550 }
2551
2552 /*
2553  * get the file handle for the "/" directory on the server
2554  */
2555 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2556                          struct nfs_fsinfo *info)
2557 {
2558         int minor_version = server->nfs_client->cl_minorversion;
2559         int status = nfs4_lookup_root(server, fhandle, info);
2560         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2561                 /*
2562                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2563                  * by nfs4_map_errors() as this function exits.
2564                  */
2565                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2566         if (status == 0)
2567                 status = nfs4_server_capabilities(server, fhandle);
2568         if (status == 0)
2569                 status = nfs4_do_fsinfo(server, fhandle, info);
2570         return nfs4_map_errors(status);
2571 }
2572
2573 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2574                               struct nfs_fsinfo *info)
2575 {
2576         int error;
2577         struct nfs_fattr *fattr = info->fattr;
2578
2579         error = nfs4_server_capabilities(server, mntfh);
2580         if (error < 0) {
2581                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2582                 return error;
2583         }
2584
2585         error = nfs4_proc_getattr(server, mntfh, fattr);
2586         if (error < 0) {
2587                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2588                 return error;
2589         }
2590
2591         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2592             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2593                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2594
2595         return error;
2596 }
2597
2598 /*
2599  * Get locations and (maybe) other attributes of a referral.
2600  * Note that we'll actually follow the referral later when
2601  * we detect fsid mismatch in inode revalidation
2602  */
2603 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2604                              const struct qstr *name, struct nfs_fattr *fattr,
2605                              struct nfs_fh *fhandle)
2606 {
2607         int status = -ENOMEM;
2608         struct page *page = NULL;
2609         struct nfs4_fs_locations *locations = NULL;
2610
2611         page = alloc_page(GFP_KERNEL);
2612         if (page == NULL)
2613                 goto out;
2614         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2615         if (locations == NULL)
2616                 goto out;
2617
2618         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2619         if (status != 0)
2620                 goto out;
2621         /* Make sure server returned a different fsid for the referral */
2622         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2623                 dprintk("%s: server did not return a different fsid for"
2624                         " a referral at %s\n", __func__, name->name);
2625                 status = -EIO;
2626                 goto out;
2627         }
2628         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2629         nfs_fixup_referral_attributes(&locations->fattr);
2630
2631         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2632         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2633         memset(fhandle, 0, sizeof(struct nfs_fh));
2634 out:
2635         if (page)
2636                 __free_page(page);
2637         kfree(locations);
2638         return status;
2639 }
2640
2641 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2642 {
2643         struct nfs4_getattr_arg args = {
2644                 .fh = fhandle,
2645                 .bitmask = server->attr_bitmask,
2646         };
2647         struct nfs4_getattr_res res = {
2648                 .fattr = fattr,
2649                 .server = server,
2650         };
2651         struct rpc_message msg = {
2652                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2653                 .rpc_argp = &args,
2654                 .rpc_resp = &res,
2655         };
2656
2657         nfs_fattr_init(fattr);
2658         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2659 }
2660
2661 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2662 {
2663         struct nfs4_exception exception = { };
2664         int err;
2665         do {
2666                 err = nfs4_handle_exception(server,
2667                                 _nfs4_proc_getattr(server, fhandle, fattr),
2668                                 &exception);
2669         } while (exception.retry);
2670         return err;
2671 }
2672
2673 /*
2674  * The file is not closed if it is opened due to the a request to change
2675  * the size of the file. The open call will not be needed once the
2676  * VFS layer lookup-intents are implemented.
2677  *
2678  * Close is called when the inode is destroyed.
2679  * If we haven't opened the file for O_WRONLY, we
2680  * need to in the size_change case to obtain a stateid.
2681  *
2682  * Got race?
2683  * Because OPEN is always done by name in nfsv4, it is
2684  * possible that we opened a different file by the same
2685  * name.  We can recognize this race condition, but we
2686  * can't do anything about it besides returning an error.
2687  *
2688  * This will be fixed with VFS changes (lookup-intent).
2689  */
2690 static int
2691 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2692                   struct iattr *sattr)
2693 {
2694         struct inode *inode = dentry->d_inode;
2695         struct rpc_cred *cred = NULL;
2696         struct nfs4_state *state = NULL;
2697         int status;
2698
2699         if (pnfs_ld_layoutret_on_setattr(inode))
2700                 pnfs_commit_and_return_layout(inode);
2701
2702         nfs_fattr_init(fattr);
2703
2704         /* Deal with open(O_TRUNC) */
2705         if (sattr->ia_valid & ATTR_OPEN)
2706                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2707
2708         /* Optimization: if the end result is no change, don't RPC */
2709         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2710                 return 0;
2711
2712         /* Search for an existing open(O_WRITE) file */
2713         if (sattr->ia_valid & ATTR_FILE) {
2714                 struct nfs_open_context *ctx;
2715
2716                 ctx = nfs_file_open_context(sattr->ia_file);
2717                 if (ctx) {
2718                         cred = ctx->cred;
2719                         state = ctx->state;
2720                 }
2721         }
2722
2723         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2724         if (status == 0)
2725                 nfs_setattr_update_inode(inode, sattr);
2726         return status;
2727 }
2728
2729 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2730                 const struct qstr *name, struct nfs_fh *fhandle,
2731                 struct nfs_fattr *fattr)
2732 {
2733         struct nfs_server *server = NFS_SERVER(dir);
2734         int                    status;
2735         struct nfs4_lookup_arg args = {
2736                 .bitmask = server->attr_bitmask,
2737                 .dir_fh = NFS_FH(dir),
2738                 .name = name,
2739         };
2740         struct nfs4_lookup_res res = {
2741                 .server = server,
2742                 .fattr = fattr,
2743                 .fh = fhandle,
2744         };
2745         struct rpc_message msg = {
2746                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2747                 .rpc_argp = &args,
2748                 .rpc_resp = &res,
2749         };
2750
2751         nfs_fattr_init(fattr);
2752
2753         dprintk("NFS call  lookup %s\n", name->name);
2754         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2755         dprintk("NFS reply lookup: %d\n", status);
2756         return status;
2757 }
2758
2759 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2760 {
2761         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2762                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2763         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2764         fattr->nlink = 2;
2765 }
2766
2767 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2768                                    struct qstr *name, struct nfs_fh *fhandle,
2769                                    struct nfs_fattr *fattr)
2770 {
2771         struct nfs4_exception exception = { };
2772         struct rpc_clnt *client = *clnt;
2773         int err;
2774         do {
2775                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2776                 switch (err) {
2777                 case -NFS4ERR_BADNAME:
2778                         err = -ENOENT;
2779                         goto out;
2780                 case -NFS4ERR_MOVED:
2781                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2782                         goto out;
2783                 case -NFS4ERR_WRONGSEC:
2784                         err = -EPERM;
2785                         if (client != *clnt)
2786                                 goto out;
2787
2788                         client = nfs4_create_sec_client(client, dir, name);
2789                         if (IS_ERR(client))
2790                                 return PTR_ERR(client);
2791
2792                         exception.retry = 1;
2793                         break;
2794                 default:
2795                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2796                 }
2797         } while (exception.retry);
2798
2799 out:
2800         if (err == 0)
2801                 *clnt = client;
2802         else if (client != *clnt)
2803                 rpc_shutdown_client(client);
2804
2805         return err;
2806 }
2807
2808 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2809                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2810 {
2811         int status;
2812         struct rpc_clnt *client = NFS_CLIENT(dir);
2813
2814         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2815         if (client != NFS_CLIENT(dir)) {
2816                 rpc_shutdown_client(client);
2817                 nfs_fixup_secinfo_attributes(fattr);
2818         }
2819         return status;
2820 }
2821
2822 struct rpc_clnt *
2823 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2824                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2825 {
2826         int status;
2827         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2828
2829         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2830         if (status < 0) {
2831                 rpc_shutdown_client(client);
2832                 return ERR_PTR(status);
2833         }
2834         return client;
2835 }
2836
2837 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2838 {
2839         struct nfs_server *server = NFS_SERVER(inode);
2840         struct nfs4_accessargs args = {
2841                 .fh = NFS_FH(inode),
2842                 .bitmask = server->cache_consistency_bitmask,
2843         };
2844         struct nfs4_accessres res = {
2845                 .server = server,
2846         };
2847         struct rpc_message msg = {
2848                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2849                 .rpc_argp = &args,
2850                 .rpc_resp = &res,
2851                 .rpc_cred = entry->cred,
2852         };
2853         int mode = entry->mask;
2854         int status;
2855
2856         /*
2857          * Determine which access bits we want to ask for...
2858          */
2859         if (mode & MAY_READ)
2860                 args.access |= NFS4_ACCESS_READ;
2861         if (S_ISDIR(inode->i_mode)) {
2862                 if (mode & MAY_WRITE)
2863                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2864                 if (mode & MAY_EXEC)
2865                         args.access |= NFS4_ACCESS_LOOKUP;
2866         } else {
2867                 if (mode & MAY_WRITE)
2868                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2869                 if (mode & MAY_EXEC)
2870                         args.access |= NFS4_ACCESS_EXECUTE;
2871         }
2872
2873         res.fattr = nfs_alloc_fattr();
2874         if (res.fattr == NULL)
2875                 return -ENOMEM;
2876
2877         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2878         if (!status) {
2879                 nfs_access_set_mask(entry, res.access);
2880                 nfs_refresh_inode(inode, res.fattr);
2881         }
2882         nfs_free_fattr(res.fattr);
2883         return status;
2884 }
2885
2886 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2887 {
2888         struct nfs4_exception exception = { };
2889         int err;
2890         do {
2891                 err = nfs4_handle_exception(NFS_SERVER(inode),
2892                                 _nfs4_proc_access(inode, entry),
2893                                 &exception);
2894         } while (exception.retry);
2895         return err;
2896 }
2897
2898 /*
2899  * TODO: For the time being, we don't try to get any attributes
2900  * along with any of the zero-copy operations READ, READDIR,
2901  * READLINK, WRITE.
2902  *
2903  * In the case of the first three, we want to put the GETATTR
2904  * after the read-type operation -- this is because it is hard
2905  * to predict the length of a GETATTR response in v4, and thus
2906  * align the READ data correctly.  This means that the GETATTR
2907  * may end up partially falling into the page cache, and we should
2908  * shift it into the 'tail' of the xdr_buf before processing.
2909  * To do this efficiently, we need to know the total length
2910  * of data received, which doesn't seem to be available outside
2911  * of the RPC layer.
2912  *
2913  * In the case of WRITE, we also want to put the GETATTR after
2914  * the operation -- in this case because we want to make sure
2915  * we get the post-operation mtime and size.
2916  *
2917  * Both of these changes to the XDR layer would in fact be quite
2918  * minor, but I decided to leave them for a subsequent patch.
2919  */
2920 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2921                 unsigned int pgbase, unsigned int pglen)
2922 {
2923         struct nfs4_readlink args = {
2924                 .fh       = NFS_FH(inode),
2925                 .pgbase   = pgbase,
2926                 .pglen    = pglen,
2927                 .pages    = &page,
2928         };
2929         struct nfs4_readlink_res res;
2930         struct rpc_message msg = {
2931                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2932                 .rpc_argp = &args,
2933                 .rpc_resp = &res,
2934         };
2935
2936         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2937 }
2938
2939 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2940                 unsigned int pgbase, unsigned int pglen)
2941 {
2942         struct nfs4_exception exception = { };
2943         int err;
2944         do {
2945                 err = nfs4_handle_exception(NFS_SERVER(inode),
2946                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2947                                 &exception);
2948         } while (exception.retry);
2949         return err;
2950 }
2951
2952 /*
2953  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
2954  */
2955 static int
2956 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2957                  int flags)
2958 {
2959         struct nfs_open_context *ctx;
2960         struct nfs4_state *state;
2961         int status = 0;
2962
2963         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
2964         if (IS_ERR(ctx))
2965                 return PTR_ERR(ctx);
2966
2967         sattr->ia_mode &= ~current_umask();
2968         state = nfs4_do_open(dir, dentry, ctx->mode,
2969                         flags, sattr, ctx->cred,
2970                         &ctx->mdsthreshold);
2971         d_drop(dentry);
2972         if (IS_ERR(state)) {
2973                 status = PTR_ERR(state);
2974                 goto out;
2975         }
2976         d_add(dentry, igrab(state->inode));
2977         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2978         ctx->state = state;
2979 out:
2980         put_nfs_open_context(ctx);
2981         return status;
2982 }
2983
2984 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2985 {
2986         struct nfs_server *server = NFS_SERVER(dir);
2987         struct nfs_removeargs args = {
2988                 .fh = NFS_FH(dir),
2989                 .name = *name,
2990         };
2991         struct nfs_removeres res = {
2992                 .server = server,
2993         };
2994         struct rpc_message msg = {
2995                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2996                 .rpc_argp = &args,
2997                 .rpc_resp = &res,
2998         };
2999         int status;
3000
3001         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3002         if (status == 0)
3003                 update_changeattr(dir, &res.cinfo);
3004         return status;
3005 }
3006
3007 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3008 {
3009         struct nfs4_exception exception = { };
3010         int err;
3011         do {
3012                 err = nfs4_handle_exception(NFS_SERVER(dir),
3013                                 _nfs4_proc_remove(dir, name),
3014                                 &exception);
3015         } while (exception.retry);
3016         return err;
3017 }
3018
3019 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3020 {
3021         struct nfs_server *server = NFS_SERVER(dir);
3022         struct nfs_removeargs *args = msg->rpc_argp;
3023         struct nfs_removeres *res = msg->rpc_resp;
3024
3025         res->server = server;
3026         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3027         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3028 }
3029
3030 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3031 {
3032         nfs4_setup_sequence(NFS_SERVER(data->dir),
3033                         &data->args.seq_args,
3034                         &data->res.seq_res,
3035                         task);
3036 }
3037
3038 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3039 {
3040         struct nfs_removeres *res = task->tk_msg.rpc_resp;
3041
3042         if (!nfs4_sequence_done(task, &res->seq_res))
3043                 return 0;
3044         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3045                 return 0;
3046         update_changeattr(dir, &res->cinfo);
3047         return 1;
3048 }
3049
3050 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3051 {
3052         struct nfs_server *server = NFS_SERVER(dir);
3053         struct nfs_renameargs *arg = msg->rpc_argp;
3054         struct nfs_renameres *res = msg->rpc_resp;
3055
3056         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3057         res->server = server;
3058         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3059 }
3060
3061 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3062 {
3063         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3064                         &data->args.seq_args,
3065                         &data->res.seq_res,
3066                         task);
3067 }
3068
3069 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3070                                  struct inode *new_dir)
3071 {
3072         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3073
3074         if (!nfs4_sequence_done(task, &res->seq_res))
3075                 return 0;
3076         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3077                 return 0;
3078
3079         update_changeattr(old_dir, &res->old_cinfo);
3080         update_changeattr(new_dir, &res->new_cinfo);
3081         return 1;
3082 }
3083
3084 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3085                 struct inode *new_dir, struct qstr *new_name)
3086 {
3087         struct nfs_server *server = NFS_SERVER(old_dir);
3088         struct nfs_renameargs arg = {
3089                 .old_dir = NFS_FH(old_dir),
3090                 .new_dir = NFS_FH(new_dir),
3091                 .old_name = old_name,
3092                 .new_name = new_name,
3093         };
3094         struct nfs_renameres res = {
3095                 .server = server,
3096         };
3097         struct rpc_message msg = {
3098                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3099                 .rpc_argp = &arg,
3100                 .rpc_resp = &res,
3101         };
3102         int status = -ENOMEM;
3103
3104         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3105         if (!status) {
3106                 update_changeattr(old_dir, &res.old_cinfo);
3107                 update_changeattr(new_dir, &res.new_cinfo);
3108         }
3109         return status;
3110 }
3111
3112 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3113                 struct inode *new_dir, struct qstr *new_name)
3114 {
3115         struct nfs4_exception exception = { };
3116         int err;
3117         do {
3118                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3119                                 _nfs4_proc_rename(old_dir, old_name,
3120                                         new_dir, new_name),
3121                                 &exception);
3122         } while (exception.retry);
3123         return err;
3124 }
3125
3126 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3127 {
3128         struct nfs_server *server = NFS_SERVER(inode);
3129         struct nfs4_link_arg arg = {
3130                 .fh     = NFS_FH(inode),
3131                 .dir_fh = NFS_FH(dir),
3132                 .name   = name,
3133                 .bitmask = server->attr_bitmask,
3134         };
3135         struct nfs4_link_res res = {
3136                 .server = server,
3137         };
3138         struct rpc_message msg = {
3139                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3140                 .rpc_argp = &arg,
3141                 .rpc_resp = &res,
3142         };
3143         int status = -ENOMEM;
3144
3145         res.fattr = nfs_alloc_fattr();
3146         if (res.fattr == NULL)
3147                 goto out;
3148
3149         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3150         if (!status) {
3151                 update_changeattr(dir, &res.cinfo);
3152                 nfs_post_op_update_inode(inode, res.fattr);
3153         }
3154 out:
3155         nfs_free_fattr(res.fattr);
3156         return status;
3157 }
3158
3159 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3160 {
3161         struct nfs4_exception exception = { };
3162         int err;
3163         do {
3164                 err = nfs4_handle_exception(NFS_SERVER(inode),
3165                                 _nfs4_proc_link(inode, dir, name),
3166                                 &exception);
3167         } while (exception.retry);
3168         return err;
3169 }
3170
3171 struct nfs4_createdata {
3172         struct rpc_message msg;
3173         struct nfs4_create_arg arg;
3174         struct nfs4_create_res res;
3175         struct nfs_fh fh;
3176         struct nfs_fattr fattr;
3177 };
3178
3179 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3180                 struct qstr *name, struct iattr *sattr, u32 ftype)
3181 {
3182         struct nfs4_createdata *data;
3183
3184         data = kzalloc(sizeof(*data), GFP_KERNEL);
3185         if (data != NULL) {
3186                 struct nfs_server *server = NFS_SERVER(dir);
3187
3188                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3189                 data->msg.rpc_argp = &data->arg;
3190                 data->msg.rpc_resp = &data->res;
3191                 data->arg.dir_fh = NFS_FH(dir);
3192                 data->arg.server = server;
3193                 data->arg.name = name;
3194                 data->arg.attrs = sattr;
3195                 data->arg.ftype = ftype;
3196                 data->arg.bitmask = server->attr_bitmask;
3197                 data->res.server = server;
3198                 data->res.fh = &data->fh;
3199                 data->res.fattr = &data->fattr;
3200                 nfs_fattr_init(data->res.fattr);
3201         }
3202         return data;
3203 }
3204
3205 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3206 {
3207         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3208                                     &data->arg.seq_args, &data->res.seq_res, 1);
3209         if (status == 0) {
3210                 update_changeattr(dir, &data->res.dir_cinfo);
3211                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3212         }
3213         return status;
3214 }
3215
3216 static void nfs4_free_createdata(struct nfs4_createdata *data)
3217 {
3218         kfree(data);
3219 }
3220
3221 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3222                 struct page *page, unsigned int len, struct iattr *sattr)
3223 {
3224         struct nfs4_createdata *data;
3225         int status = -ENAMETOOLONG;
3226
3227         if (len > NFS4_MAXPATHLEN)
3228                 goto out;
3229
3230         status = -ENOMEM;
3231         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3232         if (data == NULL)
3233                 goto out;
3234
3235         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3236         data->arg.u.symlink.pages = &page;
3237         data->arg.u.symlink.len = len;
3238
3239         status = nfs4_do_create(dir, dentry, data);
3240
3241         nfs4_free_createdata(data);
3242 out:
3243         return status;
3244 }
3245
3246 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3247                 struct page *page, unsigned int len, struct iattr *sattr)
3248 {
3249         struct nfs4_exception exception = { };
3250         int err;
3251         do {
3252                 err = nfs4_handle_exception(NFS_SERVER(dir),
3253                                 _nfs4_proc_symlink(dir, dentry, page,
3254                                                         len, sattr),
3255                                 &exception);
3256         } while (exception.retry);
3257         return err;
3258 }
3259
3260 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3261                 struct iattr *sattr)
3262 {
3263         struct nfs4_createdata *data;
3264         int status = -ENOMEM;
3265
3266         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3267         if (data == NULL)
3268                 goto out;
3269
3270         status = nfs4_do_create(dir, dentry, data);
3271
3272         nfs4_free_createdata(data);
3273 out:
3274         return status;
3275 }
3276
3277 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3278                 struct iattr *sattr)
3279 {
3280         struct nfs4_exception exception = { };
3281         int err;
3282
3283         sattr->ia_mode &= ~current_umask();
3284         do {
3285                 err = nfs4_handle_exception(NFS_SERVER(dir),
3286                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3287                                 &exception);
3288         } while (exception.retry);
3289         return err;
3290 }
3291
3292 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3293                 u64 cookie, struct page **pages, unsigned int count, int plus)
3294 {
3295         struct inode            *dir = dentry->d_inode;
3296         struct nfs4_readdir_arg args = {
3297                 .fh = NFS_FH(dir),
3298                 .pages = pages,
3299                 .pgbase = 0,
3300                 .count = count,
3301                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3302                 .plus = plus,
3303         };
3304         struct nfs4_readdir_res res;
3305         struct rpc_message msg = {
3306                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3307                 .rpc_argp = &args,
3308                 .rpc_resp = &res,
3309                 .rpc_cred = cred,
3310         };
3311         int                     status;
3312
3313         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3314                         dentry->d_parent->d_name.name,
3315                         dentry->d_name.name,
3316                         (unsigned long long)cookie);
3317         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3318         res.pgbase = args.pgbase;
3319         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3320         if (status >= 0) {
3321                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3322                 status += args.pgbase;
3323         }
3324
3325         nfs_invalidate_atime(dir);
3326
3327         dprintk("%s: returns %d\n", __func__, status);
3328         return status;
3329 }
3330
3331 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3332                 u64 cookie, struct page **pages, unsigned int count, int plus)
3333 {
3334         struct nfs4_exception exception = { };
3335         int err;
3336         do {
3337                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3338                                 _nfs4_proc_readdir(dentry, cred, cookie,
3339                                         pages, count, plus),
3340                                 &exception);
3341         } while (exception.retry);
3342         return err;
3343 }
3344
3345 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3346                 struct iattr *sattr, dev_t rdev)
3347 {
3348         struct nfs4_createdata *data;
3349         int mode = sattr->ia_mode;
3350         int status = -ENOMEM;
3351
3352         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3353         if (data == NULL)
3354                 goto out;
3355
3356         if (S_ISFIFO(mode))
3357                 data->arg.ftype = NF4FIFO;
3358         else if (S_ISBLK(mode)) {
3359                 data->arg.ftype = NF4BLK;
3360                 data->arg.u.device.specdata1 = MAJOR(rdev);
3361                 data->arg.u.device.specdata2 = MINOR(rdev);
3362         }
3363         else if (S_ISCHR(mode)) {
3364                 data->arg.ftype = NF4CHR;
3365                 data->arg.u.device.specdata1 = MAJOR(rdev);
3366                 data->arg.u.device.specdata2 = MINOR(rdev);
3367         } else if (!S_ISSOCK(mode)) {
3368                 status = -EINVAL;
3369                 goto out_free;
3370         }
3371
3372         status = nfs4_do_create(dir, dentry, data);
3373 out_free:
3374         nfs4_free_createdata(data);
3375 out:
3376         return status;
3377 }
3378
3379 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3380                 struct iattr *sattr, dev_t rdev)
3381 {
3382         struct nfs4_exception exception = { };
3383         int err;
3384
3385         sattr->ia_mode &= ~current_umask();
3386         do {
3387                 err = nfs4_handle_exception(NFS_SERVER(dir),
3388                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3389                                 &exception);
3390         } while (exception.retry);
3391         return err;
3392 }
3393
3394 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3395                  struct nfs_fsstat *fsstat)
3396 {
3397         struct nfs4_statfs_arg args = {
3398                 .fh = fhandle,
3399                 .bitmask = server->attr_bitmask,
3400         };
3401         struct nfs4_statfs_res res = {
3402                 .fsstat = fsstat,
3403         };
3404         struct rpc_message msg = {
3405                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3406                 .rpc_argp = &args,
3407                 .rpc_resp = &res,
3408         };
3409
3410         nfs_fattr_init(fsstat->fattr);
3411         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3412 }
3413
3414 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3415 {
3416         struct nfs4_exception exception = { };
3417         int err;
3418         do {
3419                 err = nfs4_handle_exception(server,
3420                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3421                                 &exception);
3422         } while (exception.retry);
3423         return err;
3424 }
3425
3426 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3427                 struct nfs_fsinfo *fsinfo)
3428 {
3429         struct nfs4_fsinfo_arg args = {
3430                 .fh = fhandle,
3431                 .bitmask = server->attr_bitmask,
3432         };
3433         struct nfs4_fsinfo_res res = {
3434                 .fsinfo = fsinfo,
3435         };
3436         struct rpc_message msg = {
3437                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3438                 .rpc_argp = &args,
3439                 .rpc_resp = &res,
3440         };
3441
3442         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3443 }
3444
3445 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3446 {
3447         struct nfs4_exception exception = { };
3448         int err;
3449
3450         do {
3451                 err = nfs4_handle_exception(server,
3452                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3453                                 &exception);
3454         } while (exception.retry);
3455         return err;
3456 }
3457
3458 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3459 {
3460         int error;
3461
3462         nfs_fattr_init(fsinfo->fattr);
3463         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3464         if (error == 0) {
3465                 /* block layout checks this! */
3466                 server->pnfs_blksize = fsinfo->blksize;
3467                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3468         }
3469
3470         return error;
3471 }
3472
3473 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3474                 struct nfs_pathconf *pathconf)
3475 {
3476         struct nfs4_pathconf_arg args = {
3477                 .fh = fhandle,
3478                 .bitmask = server->attr_bitmask,
3479         };
3480         struct nfs4_pathconf_res res = {
3481                 .pathconf = pathconf,
3482         };
3483         struct rpc_message msg = {
3484                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3485                 .rpc_argp = &args,
3486                 .rpc_resp = &res,
3487         };
3488
3489         /* None of the pathconf attributes are mandatory to implement */
3490         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3491                 memset(pathconf, 0, sizeof(*pathconf));
3492                 return 0;
3493         }
3494
3495         nfs_fattr_init(pathconf->fattr);
3496         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3497 }
3498
3499 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3500                 struct nfs_pathconf *pathconf)
3501 {
3502         struct nfs4_exception exception = { };
3503         int err;
3504
3505         do {
3506                 err = nfs4_handle_exception(server,
3507                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3508                                 &exception);
3509         } while (exception.retry);
3510         return err;
3511 }
3512
3513 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
3514                 const struct nfs_open_context *ctx,
3515                 const struct nfs_lock_context *l_ctx,
3516                 fmode_t fmode)
3517 {
3518         const struct nfs_lockowner *lockowner = NULL;
3519
3520         if (l_ctx != NULL)
3521                 lockowner = &l_ctx->lockowner;
3522         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
3523 }
3524 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
3525
3526 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
3527                 const struct nfs_open_context *ctx,
3528                 const struct nfs_lock_context *l_ctx,
3529                 fmode_t fmode)
3530 {
3531         nfs4_stateid current_stateid;
3532
3533         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode))
3534                 return false;
3535         return nfs4_stateid_match(stateid, &current_stateid);
3536 }
3537
3538 static bool nfs4_error_stateid_expired(int err)
3539 {
3540         switch (err) {
3541         case -NFS4ERR_DELEG_REVOKED:
3542         case -NFS4ERR_ADMIN_REVOKED:
3543         case -NFS4ERR_BAD_STATEID:
3544         case -NFS4ERR_STALE_STATEID:
3545         case -NFS4ERR_OLD_STATEID:
3546         case -NFS4ERR_OPENMODE:
3547         case -NFS4ERR_EXPIRED:
3548                 return true;
3549         }
3550         return false;
3551 }
3552
3553 void __nfs4_read_done_cb(struct nfs_read_data *data)
3554 {
3555         nfs_invalidate_atime(data->header->inode);
3556 }
3557
3558 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3559 {
3560         struct nfs_server *server = NFS_SERVER(data->header->inode);
3561
3562         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3563                 rpc_restart_call_prepare(task);
3564                 return -EAGAIN;
3565         }
3566
3567         __nfs4_read_done_cb(data);
3568         if (task->tk_status > 0)
3569                 renew_lease(server, data->timestamp);
3570         return 0;
3571 }
3572
3573 static bool nfs4_read_stateid_changed(struct rpc_task *task,
3574                 struct nfs_readargs *args)
3575 {
3576
3577         if (!nfs4_error_stateid_expired(task->tk_status) ||
3578                 nfs4_stateid_is_current(&args->stateid,
3579                                 args->context,
3580                                 args->lock_context,
3581                                 FMODE_READ))
3582                 return false;
3583         rpc_restart_call_prepare(task);
3584         return true;
3585 }
3586
3587 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3588 {
3589
3590         dprintk("--> %s\n", __func__);
3591
3592         if (!nfs4_sequence_done(task, &data->res.seq_res))
3593                 return -EAGAIN;
3594         if (nfs4_read_stateid_changed(task, &data->args))
3595                 return -EAGAIN;
3596         return data->read_done_cb ? data->read_done_cb(task, data) :
3597                                     nfs4_read_done_cb(task, data);
3598 }
3599
3600 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3601 {
3602         data->timestamp   = jiffies;
3603         data->read_done_cb = nfs4_read_done_cb;
3604         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3605         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3606 }
3607
3608 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3609 {
3610         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3611                         &data->args.seq_args,
3612                         &data->res.seq_res,
3613                         task))
3614                 return;
3615         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3616                         data->args.lock_context, FMODE_READ);
3617 }
3618
3619 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3620 {
3621         struct inode *inode = data->header->inode;
3622
3623         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3624                 rpc_restart_call_prepare(task);
3625                 return -EAGAIN;
3626         }
3627         if (task->tk_status >= 0) {
3628                 renew_lease(NFS_SERVER(inode), data->timestamp);
3629                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3630         }
3631         return 0;
3632 }
3633
3634 static bool nfs4_write_stateid_changed(struct rpc_task *task,
3635                 struct nfs_writeargs *args)
3636 {
3637
3638         if (!nfs4_error_stateid_expired(task->tk_status) ||
3639                 nfs4_stateid_is_current(&args->stateid,
3640                                 args->context,
3641                                 args->lock_context,
3642                                 FMODE_WRITE))
3643                 return false;
3644         rpc_restart_call_prepare(task);
3645         return true;
3646 }
3647
3648 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3649 {
3650         if (!nfs4_sequence_done(task, &data->res.seq_res))
3651                 return -EAGAIN;
3652         if (nfs4_write_stateid_changed(task, &data->args))
3653                 return -EAGAIN;
3654         return data->write_done_cb ? data->write_done_cb(task, data) :
3655                 nfs4_write_done_cb(task, data);
3656 }
3657
3658 static
3659 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3660 {
3661         const struct nfs_pgio_header *hdr = data->header;
3662
3663         /* Don't request attributes for pNFS or O_DIRECT writes */
3664         if (data->ds_clp != NULL || hdr->dreq != NULL)
3665                 return false;
3666         /* Otherwise, request attributes if and only if we don't hold
3667          * a delegation
3668          */
3669         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3670 }
3671
3672 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3673 {
3674         struct nfs_server *server = NFS_SERVER(data->header->inode);
3675
3676         if (!nfs4_write_need_cache_consistency_data(data)) {
3677                 data->args.bitmask = NULL;
3678                 data->res.fattr = NULL;
3679         } else
3680                 data->args.bitmask = server->cache_consistency_bitmask;
3681
3682         if (!data->write_done_cb)
3683                 data->write_done_cb = nfs4_write_done_cb;
3684         data->res.server = server;
3685         data->timestamp   = jiffies;
3686
3687         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3688         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3689 }
3690
3691 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3692 {
3693         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3694                         &data->args.seq_args,
3695                         &data->res.seq_res,
3696                         task))
3697                 return;
3698         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3699                         data->args.lock_context, FMODE_WRITE);
3700 }
3701
3702 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3703 {
3704         nfs4_setup_sequence(NFS_SERVER(data->inode),
3705                         &data->args.seq_args,
3706                         &data->res.seq_res,
3707                         task);
3708 }
3709
3710 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3711 {
3712         struct inode *inode = data->inode;
3713
3714         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3715                 rpc_restart_call_prepare(task);
3716                 return -EAGAIN;
3717         }
3718         return 0;
3719 }
3720
3721 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3722 {
3723         if (!nfs4_sequence_done(task, &data->res.seq_res))
3724                 return -EAGAIN;
3725         return data->commit_done_cb(task, data);
3726 }
3727
3728 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3729 {
3730         struct nfs_server *server = NFS_SERVER(data->inode);
3731
3732         if (data->commit_done_cb == NULL)
3733                 data->commit_done_cb = nfs4_commit_done_cb;
3734         data->res.server = server;
3735         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3736         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3737 }
3738
3739 struct nfs4_renewdata {
3740         struct nfs_client       *client;
3741         unsigned long           timestamp;
3742 };
3743
3744 /*
3745  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3746  * standalone procedure for queueing an asynchronous RENEW.
3747  */
3748 static void nfs4_renew_release(void *calldata)
3749 {
3750         struct nfs4_renewdata *data = calldata;
3751         struct nfs_client *clp = data->client;
3752
3753         if (atomic_read(&clp->cl_count) > 1)
3754                 nfs4_schedule_state_renewal(clp);
3755         nfs_put_client(clp);
3756         kfree(data);
3757 }
3758
3759 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3760 {
3761         struct nfs4_renewdata *data = calldata;
3762         struct nfs_client *clp = data->client;
3763         unsigned long timestamp = data->timestamp;
3764
3765         if (task->tk_status < 0) {
3766                 /* Unless we're shutting down, schedule state recovery! */
3767                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3768                         return;
3769                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3770                         nfs4_schedule_lease_recovery(clp);
3771                         return;
3772                 }
3773                 nfs4_schedule_path_down_recovery(clp);
3774         }
3775         do_renew_lease(clp, timestamp);
3776 }
3777
3778 static const struct rpc_call_ops nfs4_renew_ops = {
3779         .rpc_call_done = nfs4_renew_done,
3780         .rpc_release = nfs4_renew_release,
3781 };
3782
3783 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3784 {
3785         struct rpc_message msg = {
3786                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3787                 .rpc_argp       = clp,
3788                 .rpc_cred       = cred,
3789         };
3790         struct nfs4_renewdata *data;
3791
3792         if (renew_flags == 0)
3793                 return 0;
3794         if (!atomic_inc_not_zero(&clp->cl_count))
3795                 return -EIO;
3796         data = kmalloc(sizeof(*data), GFP_NOFS);
3797         if (data == NULL)
3798                 return -ENOMEM;
3799         data->client = clp;
3800         data->timestamp = jiffies;
3801         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3802                         &nfs4_renew_ops, data);
3803 }
3804
3805 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3806 {
3807         struct rpc_message msg = {
3808                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3809                 .rpc_argp       = clp,
3810                 .rpc_cred       = cred,
3811         };
3812         unsigned long now = jiffies;
3813         int status;
3814
3815         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3816         if (status < 0)
3817                 return status;
3818         do_renew_lease(clp, now);
3819         return 0;
3820 }
3821
3822 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3823 {
3824         return (server->caps & NFS_CAP_ACLS)
3825                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3826                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3827 }
3828
3829 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3830  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3831  * the stack.
3832  */
3833 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3834
3835 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3836                 struct page **pages, unsigned int *pgbase)
3837 {
3838         struct page *newpage, **spages;
3839         int rc = 0;
3840         size_t len;
3841         spages = pages;
3842
3843         do {
3844                 len = min_t(size_t, PAGE_SIZE, buflen);
3845                 newpage = alloc_page(GFP_KERNEL);
3846
3847                 if (newpage == NULL)
3848                         goto unwind;
3849                 memcpy(page_address(newpage), buf, len);
3850                 buf += len;
3851                 buflen -= len;
3852                 *pages++ = newpage;
3853                 rc++;
3854         } while (buflen != 0);
3855
3856         return rc;
3857
3858 unwind:
3859         for(; rc > 0; rc--)
3860                 __free_page(spages[rc-1]);
3861         return -ENOMEM;
3862 }
3863
3864 struct nfs4_cached_acl {
3865         int cached;
3866         size_t len;
3867         char data[0];
3868 };
3869
3870 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3871 {
3872         struct nfs_inode *nfsi = NFS_I(inode);
3873
3874         spin_lock(&inode->i_lock);
3875         kfree(nfsi->nfs4_acl);
3876         nfsi->nfs4_acl = acl;
3877         spin_unlock(&inode->i_lock);
3878 }
3879
3880 static void nfs4_zap_acl_attr(struct inode *inode)
3881 {
3882         nfs4_set_cached_acl(inode, NULL);
3883 }
3884
3885 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3886 {
3887         struct nfs_inode *nfsi = NFS_I(inode);
3888         struct nfs4_cached_acl *acl;
3889         int ret = -ENOENT;
3890
3891         spin_lock(&inode->i_lock);
3892         acl = nfsi->nfs4_acl;
3893         if (acl == NULL)
3894                 goto out;
3895         if (buf == NULL) /* user is just asking for length */
3896                 goto out_len;
3897         if (acl->cached == 0)
3898                 goto out;
3899         ret = -ERANGE; /* see getxattr(2) man page */
3900         if (acl->len > buflen)
3901                 goto out;
3902         memcpy(buf, acl->data, acl->len);
3903 out_len:
3904         ret = acl->len;
3905 out:
3906         spin_unlock(&inode->i_lock);
3907         return ret;
3908 }
3909
3910 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3911 {
3912         struct nfs4_cached_acl *acl;
3913         size_t buflen = sizeof(*acl) + acl_len;
3914
3915         if (buflen <= PAGE_SIZE) {
3916                 acl = kmalloc(buflen, GFP_KERNEL);
3917                 if (acl == NULL)
3918                         goto out;
3919                 acl->cached = 1;
3920                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3921         } else {
3922                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3923                 if (acl == NULL)
3924                         goto out;
3925                 acl->cached = 0;
3926         }
3927         acl->len = acl_len;
3928 out:
3929         nfs4_set_cached_acl(inode, acl);
3930 }
3931
3932 /*
3933  * The getxattr API returns the required buffer length when called with a
3934  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3935  * the required buf.  On a NULL buf, we send a page of data to the server
3936  * guessing that the ACL request can be serviced by a page. If so, we cache
3937  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3938  * the cache. If not so, we throw away the page, and cache the required
3939  * length. The next getxattr call will then produce another round trip to
3940  * the server, this time with the input buf of the required size.
3941  */
3942 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3943 {
3944         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3945         struct nfs_getaclargs args = {
3946                 .fh = NFS_FH(inode),
3947                 .acl_pages = pages,
3948                 .acl_len = buflen,
3949         };
3950         struct nfs_getaclres res = {
3951                 .acl_len = buflen,
3952         };
3953         struct rpc_message msg = {
3954                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3955                 .rpc_argp = &args,
3956                 .rpc_resp = &res,
3957         };
3958         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3959         int ret = -ENOMEM, i;
3960
3961         /* As long as we're doing a round trip to the server anyway,
3962          * let's be prepared for a page of acl data. */
3963         if (npages == 0)
3964                 npages = 1;
3965         if (npages > ARRAY_SIZE(pages))
3966                 return -ERANGE;
3967
3968         for (i = 0; i < npages; i++) {
3969                 pages[i] = alloc_page(GFP_KERNEL);
3970                 if (!pages[i])
3971                         goto out_free;
3972         }
3973
3974         /* for decoding across pages */
3975         res.acl_scratch = alloc_page(GFP_KERNEL);
3976         if (!res.acl_scratch)
3977                 goto out_free;
3978
3979         args.acl_len = npages * PAGE_SIZE;
3980         args.acl_pgbase = 0;
3981
3982         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3983                 __func__, buf, buflen, npages, args.acl_len);
3984         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3985                              &msg, &args.seq_args, &res.seq_res, 0);
3986         if (ret)
3987                 goto out_free;
3988
3989         /* Handle the case where the passed-in buffer is too short */
3990         if (res.acl_flags & NFS4_ACL_TRUNC) {
3991                 /* Did the user only issue a request for the acl length? */
3992                 if (buf == NULL)
3993                         goto out_ok;
3994                 ret = -ERANGE;
3995                 goto out_free;
3996         }
3997         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
3998         if (buf) {
3999                 if (res.acl_len > buflen) {
4000                         ret = -ERANGE;
4001                         goto out_free;
4002                 }
4003                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4004         }
4005 out_ok:
4006         ret = res.acl_len;
4007 out_free:
4008         for (i = 0; i < npages; i++)
4009                 if (pages[i])
4010                         __free_page(pages[i]);
4011         if (res.acl_scratch)
4012                 __free_page(res.acl_scratch);
4013         return ret;
4014 }
4015
4016 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4017 {
4018         struct nfs4_exception exception = { };
4019         ssize_t ret;
4020         do {
4021                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4022                 if (ret >= 0)
4023                         break;
4024                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4025         } while (exception.retry);
4026         return ret;
4027 }
4028
4029 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4030 {
4031         struct nfs_server *server = NFS_SERVER(inode);
4032         int ret;
4033
4034         if (!nfs4_server_supports_acls(server))
4035                 return -EOPNOTSUPP;
4036         ret = nfs_revalidate_inode(server, inode);
4037         if (ret < 0)
4038                 return ret;
4039         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4040                 nfs_zap_acl_cache(inode);
4041         ret = nfs4_read_cached_acl(inode, buf, buflen);
4042         if (ret != -ENOENT)
4043                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4044                  * but no cached acl data, just the acl length */
4045                 return ret;
4046         return nfs4_get_acl_uncached(inode, buf, buflen);
4047 }
4048
4049 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4050 {
4051         struct nfs_server *server = NFS_SERVER(inode);
4052         struct page *pages[NFS4ACL_MAXPAGES];
4053         struct nfs_setaclargs arg = {
4054                 .fh             = NFS_FH(inode),
4055                 .acl_pages      = pages,
4056                 .acl_len        = buflen,
4057         };
4058         struct nfs_setaclres res;
4059         struct rpc_message msg = {
4060                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4061                 .rpc_argp       = &arg,
4062                 .rpc_resp       = &res,
4063         };
4064         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4065         int ret, i;
4066
4067         if (!nfs4_server_supports_acls(server))
4068                 return -EOPNOTSUPP;
4069         if (npages > ARRAY_SIZE(pages))
4070                 return -ERANGE;
4071         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4072         if (i < 0)
4073                 return i;
4074         nfs4_inode_return_delegation(inode);
4075         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4076
4077         /*
4078          * Free each page after tx, so the only ref left is
4079          * held by the network stack
4080          */
4081         for (; i > 0; i--)
4082                 put_page(pages[i-1]);
4083
4084         /*
4085          * Acl update can result in inode attribute update.
4086          * so mark the attribute cache invalid.
4087          */
4088         spin_lock(&inode->i_lock);
4089         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4090         spin_unlock(&inode->i_lock);
4091         nfs_access_zap_cache(inode);
4092         nfs_zap_acl_cache(inode);
4093         return ret;
4094 }
4095
4096 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4097 {
4098         struct nfs4_exception exception = { };
4099         int err;
4100         do {
4101                 err = nfs4_handle_exception(NFS_SERVER(inode),
4102                                 __nfs4_proc_set_acl(inode, buf, buflen),
4103                                 &exception);
4104         } while (exception.retry);
4105         return err;
4106 }
4107
4108 static int
4109 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4110 {
4111         struct nfs_client *clp = server->nfs_client;
4112
4113         if (task->tk_status >= 0)
4114                 return 0;
4115         switch(task->tk_status) {
4116                 case -NFS4ERR_DELEG_REVOKED:
4117                 case -NFS4ERR_ADMIN_REVOKED:
4118                 case -NFS4ERR_BAD_STATEID:
4119                         if (state == NULL)
4120                                 break;
4121                         nfs_remove_bad_delegation(state->inode);
4122                 case -NFS4ERR_OPENMODE:
4123                         if (state == NULL)
4124                                 break;
4125                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4126                                 goto stateid_invalid;
4127                         goto wait_on_recovery;
4128                 case -NFS4ERR_EXPIRED:
4129                         if (state != NULL) {
4130                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4131                                         goto stateid_invalid;
4132                         }
4133                 case -NFS4ERR_STALE_STATEID:
4134                 case -NFS4ERR_STALE_CLIENTID:
4135                         nfs4_schedule_lease_recovery(clp);
4136                         goto wait_on_recovery;
4137 #if defined(CONFIG_NFS_V4_1)
4138                 case -NFS4ERR_BADSESSION:
4139                 case -NFS4ERR_BADSLOT:
4140                 case -NFS4ERR_BAD_HIGH_SLOT:
4141                 case -NFS4ERR_DEADSESSION:
4142                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4143                 case -NFS4ERR_SEQ_FALSE_RETRY:
4144                 case -NFS4ERR_SEQ_MISORDERED:
4145                         dprintk("%s ERROR %d, Reset session\n", __func__,
4146                                 task->tk_status);
4147                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4148                         task->tk_status = 0;
4149                         return -EAGAIN;
4150 #endif /* CONFIG_NFS_V4_1 */
4151                 case -NFS4ERR_DELAY:
4152                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4153                 case -NFS4ERR_GRACE:
4154                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4155                         task->tk_status = 0;
4156                         return -EAGAIN;
4157                 case -NFS4ERR_RETRY_UNCACHED_REP:
4158                 case -NFS4ERR_OLD_STATEID:
4159                         task->tk_status = 0;
4160                         return -EAGAIN;
4161         }
4162         task->tk_status = nfs4_map_errors(task->tk_status);
4163         return 0;
4164 stateid_invalid:
4165         task->tk_status = -EIO;
4166         return 0;
4167 wait_on_recovery:
4168         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4169         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4170                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4171         task->tk_status = 0;
4172         return -EAGAIN;
4173 }
4174
4175 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4176                                     nfs4_verifier *bootverf)
4177 {
4178         __be32 verf[2];
4179
4180         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4181                 /* An impossible timestamp guarantees this value
4182                  * will never match a generated boot time. */
4183                 verf[0] = 0;
4184                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4185         } else {
4186                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4187                 verf[0] = (__be32)nn->boot_time.tv_sec;
4188                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4189         }
4190         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4191 }
4192
4193 static unsigned int
4194 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4195                                    char *buf, size_t len)
4196 {
4197         unsigned int result;
4198
4199         rcu_read_lock();
4200         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4201                                 clp->cl_ipaddr,
4202                                 rpc_peeraddr2str(clp->cl_rpcclient,
4203                                                         RPC_DISPLAY_ADDR),
4204                                 rpc_peeraddr2str(clp->cl_rpcclient,
4205                                                         RPC_DISPLAY_PROTO));
4206         rcu_read_unlock();
4207         return result;
4208 }
4209
4210 static unsigned int
4211 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4212                                 char *buf, size_t len)
4213 {
4214         char *nodename = clp->cl_rpcclient->cl_nodename;
4215
4216         if (nfs4_client_id_uniquifier[0] != '\0')
4217                 nodename = nfs4_client_id_uniquifier;
4218         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4219                                 clp->rpc_ops->version, clp->cl_minorversion,
4220                                 nodename);
4221 }
4222
4223 /**
4224  * nfs4_proc_setclientid - Negotiate client ID
4225  * @clp: state data structure
4226  * @program: RPC program for NFSv4 callback service
4227  * @port: IP port number for NFS4 callback service
4228  * @cred: RPC credential to use for this call
4229  * @res: where to place the result
4230  *
4231  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4232  */
4233 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4234                 unsigned short port, struct rpc_cred *cred,
4235                 struct nfs4_setclientid_res *res)
4236 {
4237         nfs4_verifier sc_verifier;
4238         struct nfs4_setclientid setclientid = {
4239                 .sc_verifier = &sc_verifier,
4240                 .sc_prog = program,
4241                 .sc_cb_ident = clp->cl_cb_ident,
4242         };
4243         struct rpc_message msg = {
4244                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4245                 .rpc_argp = &setclientid,
4246                 .rpc_resp = res,
4247                 .rpc_cred = cred,
4248         };
4249         int status;
4250
4251         /* nfs_client_id4 */
4252         nfs4_init_boot_verifier(clp, &sc_verifier);
4253         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4254                 setclientid.sc_name_len =
4255                                 nfs4_init_uniform_client_string(clp,
4256                                                 setclientid.sc_name,
4257                                                 sizeof(setclientid.sc_name));
4258         else
4259                 setclientid.sc_name_len =
4260                                 nfs4_init_nonuniform_client_string(clp,
4261                                                 setclientid.sc_name,
4262                                                 sizeof(setclientid.sc_name));
4263         /* cb_client4 */
4264         rcu_read_lock();
4265         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4266                                 sizeof(setclientid.sc_netid),
4267                                 rpc_peeraddr2str(clp->cl_rpcclient,
4268                                                         RPC_DISPLAY_NETID));
4269         rcu_read_unlock();
4270         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4271                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4272                                 clp->cl_ipaddr, port >> 8, port & 255);
4273
4274         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4275                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4276                 setclientid.sc_name_len, setclientid.sc_name);
4277         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4278         dprintk("NFS reply setclientid: %d\n", status);
4279         return status;
4280 }
4281
4282 /**
4283  * nfs4_proc_setclientid_confirm - Confirm client ID
4284  * @clp: state data structure
4285  * @res: result of a previous SETCLIENTID
4286  * @cred: RPC credential to use for this call
4287  *
4288  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4289  */
4290 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4291                 struct nfs4_setclientid_res *arg,
4292                 struct rpc_cred *cred)
4293 {
4294         struct nfs_fsinfo fsinfo;
4295         struct rpc_message msg = {
4296                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4297                 .rpc_argp = arg,
4298                 .rpc_resp = &fsinfo,
4299                 .rpc_cred = cred,
4300         };
4301         unsigned long now;
4302         int status;
4303
4304         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4305                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4306                 clp->cl_clientid);
4307         now = jiffies;
4308         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4309         if (status == 0) {
4310                 spin_lock(&clp->cl_lock);
4311                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4312                 clp->cl_last_renewal = now;
4313                 spin_unlock(&clp->cl_lock);
4314         }
4315         dprintk("NFS reply setclientid_confirm: %d\n", status);
4316         return status;
4317 }
4318
4319 struct nfs4_delegreturndata {
4320         struct nfs4_delegreturnargs args;
4321         struct nfs4_delegreturnres res;
4322         struct nfs_fh fh;
4323         nfs4_stateid stateid;
4324         unsigned long timestamp;
4325         struct nfs_fattr fattr;
4326         int rpc_status;
4327 };
4328
4329 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4330 {
4331         struct nfs4_delegreturndata *data = calldata;
4332
4333         if (!nfs4_sequence_done(task, &data->res.seq_res))
4334                 return;
4335
4336         switch (task->tk_status) {
4337         case -NFS4ERR_STALE_STATEID:
4338         case -NFS4ERR_EXPIRED:
4339         case 0:
4340                 renew_lease(data->res.server, data->timestamp);
4341                 break;
4342         default:
4343                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4344                                 -EAGAIN) {
4345                         rpc_restart_call_prepare(task);
4346                         return;
4347                 }
4348         }
4349         data->rpc_status = task->tk_status;
4350 }
4351
4352 static void nfs4_delegreturn_release(void *calldata)
4353 {
4354         kfree(calldata);
4355 }
4356
4357 #if defined(CONFIG_NFS_V4_1)
4358 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4359 {
4360         struct nfs4_delegreturndata *d_data;
4361
4362         d_data = (struct nfs4_delegreturndata *)data;
4363
4364         nfs4_setup_sequence(d_data->res.server,
4365                         &d_data->args.seq_args,
4366                         &d_data->res.seq_res,
4367                         task);
4368 }
4369 #endif /* CONFIG_NFS_V4_1 */
4370
4371 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4372 #if defined(CONFIG_NFS_V4_1)
4373         .rpc_call_prepare = nfs4_delegreturn_prepare,
4374 #endif /* CONFIG_NFS_V4_1 */
4375         .rpc_call_done = nfs4_delegreturn_done,
4376         .rpc_release = nfs4_delegreturn_release,
4377 };
4378
4379 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4380 {
4381         struct nfs4_delegreturndata *data;
4382         struct nfs_server *server = NFS_SERVER(inode);
4383         struct rpc_task *task;
4384         struct rpc_message msg = {
4385                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4386                 .rpc_cred = cred,
4387         };
4388         struct rpc_task_setup task_setup_data = {
4389                 .rpc_client = server->client,
4390                 .rpc_message = &msg,
4391                 .callback_ops = &nfs4_delegreturn_ops,
4392                 .flags = RPC_TASK_ASYNC,
4393         };
4394         int status = 0;
4395
4396         data = kzalloc(sizeof(*data), GFP_NOFS);
4397         if (data == NULL)
4398                 return -ENOMEM;
4399         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4400         data->args.fhandle = &data->fh;
4401         data->args.stateid = &data->stateid;
4402         data->args.bitmask = server->cache_consistency_bitmask;
4403         nfs_copy_fh(&data->fh, NFS_FH(inode));
4404         nfs4_stateid_copy(&data->stateid, stateid);
4405         data->res.fattr = &data->fattr;
4406         data->res.server = server;
4407         nfs_fattr_init(data->res.fattr);
4408         data->timestamp = jiffies;
4409         data->rpc_status = 0;
4410
4411         task_setup_data.callback_data = data;
4412         msg.rpc_argp = &data->args;
4413         msg.rpc_resp = &data->res;
4414         task = rpc_run_task(&task_setup_data);
4415         if (IS_ERR(task))
4416                 return PTR_ERR(task);
4417         if (!issync)
4418                 goto out;
4419         status = nfs4_wait_for_completion_rpc_task(task);
4420         if (status != 0)
4421                 goto out;
4422         status = data->rpc_status;
4423         if (status == 0)
4424                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4425         else
4426                 nfs_refresh_inode(inode, &data->fattr);
4427 out:
4428         rpc_put_task(task);
4429         return status;
4430 }
4431
4432 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4433 {
4434         struct nfs_server *server = NFS_SERVER(inode);
4435         struct nfs4_exception exception = { };
4436         int err;
4437         do {
4438                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4439                 switch (err) {
4440                         case -NFS4ERR_STALE_STATEID:
4441                         case -NFS4ERR_EXPIRED:
4442                         case 0:
4443                                 return 0;
4444                 }
4445                 err = nfs4_handle_exception(server, err, &exception);
4446         } while (exception.retry);
4447         return err;
4448 }
4449
4450 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4451 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4452
4453 /*
4454  * sleep, with exponential backoff, and retry the LOCK operation.
4455  */
4456 static unsigned long
4457 nfs4_set_lock_task_retry(unsigned long timeout)
4458 {
4459         freezable_schedule_timeout_killable(timeout);
4460         timeout <<= 1;
4461         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4462                 return NFS4_LOCK_MAXTIMEOUT;
4463         return timeout;
4464 }
4465
4466 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4467 {
4468         struct inode *inode = state->inode;
4469         struct nfs_server *server = NFS_SERVER(inode);
4470         struct nfs_client *clp = server->nfs_client;
4471         struct nfs_lockt_args arg = {
4472                 .fh = NFS_FH(inode),
4473                 .fl = request,
4474         };
4475         struct nfs_lockt_res res = {
4476                 .denied = request,
4477         };
4478         struct rpc_message msg = {
4479                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4480                 .rpc_argp       = &arg,
4481                 .rpc_resp       = &res,
4482                 .rpc_cred       = state->owner->so_cred,
4483         };
4484         struct nfs4_lock_state *lsp;
4485         int status;
4486
4487         arg.lock_owner.clientid = clp->cl_clientid;
4488         status = nfs4_set_lock_state(state, request);
4489         if (status != 0)
4490                 goto out;
4491         lsp = request->fl_u.nfs4_fl.owner;
4492         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4493         arg.lock_owner.s_dev = server->s_dev;
4494         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4495         switch (status) {
4496                 case 0:
4497                         request->fl_type = F_UNLCK;
4498                         break;
4499                 case -NFS4ERR_DENIED:
4500                         status = 0;
4501         }
4502         request->fl_ops->fl_release_private(request);
4503 out:
4504         return status;
4505 }
4506
4507 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4508 {
4509         struct nfs4_exception exception = { };
4510         int err;
4511
4512         do {
4513                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4514                                 _nfs4_proc_getlk(state, cmd, request),
4515                                 &exception);
4516         } while (exception.retry);
4517         return err;
4518 }
4519
4520 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4521 {
4522         int res = 0;
4523         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4524                 case FL_POSIX:
4525                         res = posix_lock_file_wait(file, fl);
4526                         break;
4527                 case FL_FLOCK:
4528                         res = flock_lock_file_wait(file, fl);
4529                         break;
4530                 default:
4531                         BUG();
4532         }
4533         return res;
4534 }
4535
4536 struct nfs4_unlockdata {
4537         struct nfs_locku_args arg;
4538         struct nfs_locku_res res;
4539         struct nfs4_lock_state *lsp;
4540         struct nfs_open_context *ctx;
4541         struct file_lock fl;
4542         const struct nfs_server *server;
4543         unsigned long timestamp;
4544 };
4545
4546 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4547                 struct nfs_open_context *ctx,
4548                 struct nfs4_lock_state *lsp,
4549                 struct nfs_seqid *seqid)
4550 {
4551         struct nfs4_unlockdata *p;
4552         struct inode *inode = lsp->ls_state->inode;
4553
4554         p = kzalloc(sizeof(*p), GFP_NOFS);
4555         if (p == NULL)
4556                 return NULL;
4557         p->arg.fh = NFS_FH(inode);
4558         p->arg.fl = &p->fl;
4559         p->arg.seqid = seqid;
4560         p->res.seqid = seqid;
4561         p->arg.stateid = &lsp->ls_stateid;
4562         p->lsp = lsp;
4563         atomic_inc(&lsp->ls_count);
4564         /* Ensure we don't close file until we're done freeing locks! */
4565         p->ctx = get_nfs_open_context(ctx);
4566         memcpy(&p->fl, fl, sizeof(p->fl));
4567         p->server = NFS_SERVER(inode);
4568         return p;
4569 }
4570
4571 static void nfs4_locku_release_calldata(void *data)
4572 {
4573         struct nfs4_unlockdata *calldata = data;
4574         nfs_free_seqid(calldata->arg.seqid);
4575         nfs4_put_lock_state(calldata->lsp);
4576         put_nfs_open_context(calldata->ctx);
4577         kfree(calldata);
4578 }
4579
4580 static void nfs4_locku_done(struct rpc_task *task, void *data)
4581 {
4582         struct nfs4_unlockdata *calldata = data;
4583
4584         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4585                 return;
4586         switch (task->tk_status) {
4587                 case 0:
4588                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4589                                         &calldata->res.stateid);
4590                         renew_lease(calldata->server, calldata->timestamp);
4591                         break;
4592                 case -NFS4ERR_BAD_STATEID:
4593                 case -NFS4ERR_OLD_STATEID:
4594                 case -NFS4ERR_STALE_STATEID:
4595                 case -NFS4ERR_EXPIRED:
4596                         break;
4597                 default:
4598                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4599                                 rpc_restart_call_prepare(task);
4600         }
4601         nfs_release_seqid(calldata->arg.seqid);
4602 }
4603
4604 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4605 {
4606         struct nfs4_unlockdata *calldata = data;
4607
4608         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4609                 goto out_wait;
4610         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4611                 /* Note: exit _without_ running nfs4_locku_done */
4612                 goto out_no_action;
4613         }
4614         calldata->timestamp = jiffies;
4615         if (nfs4_setup_sequence(calldata->server,
4616                                 &calldata->arg.seq_args,
4617                                 &calldata->res.seq_res,
4618                                 task) != 0)
4619                 nfs_release_seqid(calldata->arg.seqid);
4620         return;
4621 out_no_action:
4622         task->tk_action = NULL;
4623 out_wait:
4624         nfs4_sequence_done(task, &calldata->res.seq_res);
4625 }
4626
4627 static const struct rpc_call_ops nfs4_locku_ops = {
4628         .rpc_call_prepare = nfs4_locku_prepare,
4629         .rpc_call_done = nfs4_locku_done,
4630         .rpc_release = nfs4_locku_release_calldata,
4631 };
4632
4633 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4634                 struct nfs_open_context *ctx,
4635                 struct nfs4_lock_state *lsp,
4636                 struct nfs_seqid *seqid)
4637 {
4638         struct nfs4_unlockdata *data;
4639         struct rpc_message msg = {
4640                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4641                 .rpc_cred = ctx->cred,
4642         };
4643         struct rpc_task_setup task_setup_data = {
4644                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4645                 .rpc_message = &msg,
4646                 .callback_ops = &nfs4_locku_ops,
4647                 .workqueue = nfsiod_workqueue,
4648                 .flags = RPC_TASK_ASYNC,
4649         };
4650
4651         /* Ensure this is an unlock - when canceling a lock, the
4652          * canceled lock is passed in, and it won't be an unlock.
4653          */
4654         fl->fl_type = F_UNLCK;
4655
4656         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4657         if (data == NULL) {
4658                 nfs_free_seqid(seqid);
4659                 return ERR_PTR(-ENOMEM);
4660         }
4661
4662         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4663         msg.rpc_argp = &data->arg;
4664         msg.rpc_resp = &data->res;
4665         task_setup_data.callback_data = data;
4666         return rpc_run_task(&task_setup_data);
4667 }
4668
4669 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4670 {
4671         struct inode *inode = state->inode;
4672         struct nfs4_state_owner *sp = state->owner;
4673         struct nfs_inode *nfsi = NFS_I(inode);
4674         struct nfs_seqid *seqid;
4675         struct nfs4_lock_state *lsp;
4676         struct rpc_task *task;
4677         int status = 0;
4678         unsigned char fl_flags = request->fl_flags;
4679
4680         status = nfs4_set_lock_state(state, request);
4681         /* Unlock _before_ we do the RPC call */
4682         request->fl_flags |= FL_EXISTS;
4683         /* Exclude nfs_delegation_claim_locks() */
4684         mutex_lock(&sp->so_delegreturn_mutex);
4685         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
4686         down_read(&nfsi->rwsem);
4687         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4688                 up_read(&nfsi->rwsem);
4689                 mutex_unlock(&sp->so_delegreturn_mutex);
4690                 goto out;
4691         }
4692         up_read(&nfsi->rwsem);
4693         mutex_unlock(&sp->so_delegreturn_mutex);
4694         if (status != 0)
4695                 goto out;
4696         /* Is this a delegated lock? */
4697         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4698                 goto out;
4699         lsp = request->fl_u.nfs4_fl.owner;
4700         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4701         status = -ENOMEM;
4702         if (seqid == NULL)
4703                 goto out;
4704         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4705         status = PTR_ERR(task);
4706         if (IS_ERR(task))
4707                 goto out;
4708         status = nfs4_wait_for_completion_rpc_task(task);
4709         rpc_put_task(task);
4710 out:
4711         request->fl_flags = fl_flags;
4712         return status;
4713 }
4714
4715 struct nfs4_lockdata {
4716         struct nfs_lock_args arg;
4717         struct nfs_lock_res res;
4718         struct nfs4_lock_state *lsp;
4719         struct nfs_open_context *ctx;
4720         struct file_lock fl;
4721         unsigned long timestamp;
4722         int rpc_status;
4723         int cancelled;
4724         struct nfs_server *server;
4725 };
4726
4727 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4728                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4729                 gfp_t gfp_mask)
4730 {
4731         struct nfs4_lockdata *p;
4732         struct inode *inode = lsp->ls_state->inode;
4733         struct nfs_server *server = NFS_SERVER(inode);
4734
4735         p = kzalloc(sizeof(*p), gfp_mask);
4736         if (p == NULL)
4737                 return NULL;
4738
4739         p->arg.fh = NFS_FH(inode);
4740         p->arg.fl = &p->fl;
4741         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4742         if (p->arg.open_seqid == NULL)
4743                 goto out_free;
4744         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4745         if (p->arg.lock_seqid == NULL)
4746                 goto out_free_seqid;
4747         p->arg.lock_stateid = &lsp->ls_stateid;
4748         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4749         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4750         p->arg.lock_owner.s_dev = server->s_dev;
4751         p->res.lock_seqid = p->arg.lock_seqid;
4752         p->lsp = lsp;
4753         p->server = server;
4754         atomic_inc(&lsp->ls_count);
4755         p->ctx = get_nfs_open_context(ctx);
4756         memcpy(&p->fl, fl, sizeof(p->fl));
4757         return p;
4758 out_free_seqid:
4759         nfs_free_seqid(p->arg.open_seqid);
4760 out_free:
4761         kfree(p);
4762         return NULL;
4763 }
4764
4765 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4766 {
4767         struct nfs4_lockdata *data = calldata;
4768         struct nfs4_state *state = data->lsp->ls_state;
4769
4770         dprintk("%s: begin!\n", __func__);
4771         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4772                 goto out_wait;
4773         /* Do we need to do an open_to_lock_owner? */
4774         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4775                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
4776                         goto out_release_lock_seqid;
4777                 }
4778                 data->arg.open_stateid = &state->stateid;
4779                 data->arg.new_lock_owner = 1;
4780                 data->res.open_seqid = data->arg.open_seqid;
4781         } else
4782                 data->arg.new_lock_owner = 0;
4783         if (!nfs4_valid_open_stateid(state)) {
4784                 data->rpc_status = -EBADF;
4785                 task->tk_action = NULL;
4786                 goto out_release_open_seqid;
4787         }
4788         data->timestamp = jiffies;
4789         if (nfs4_setup_sequence(data->server,
4790                                 &data->arg.seq_args,
4791                                 &data->res.seq_res,
4792                                 task) == 0)
4793                 return;
4794 out_release_open_seqid:
4795         nfs_release_seqid(data->arg.open_seqid);
4796 out_release_lock_seqid:
4797         nfs_release_seqid(data->arg.lock_seqid);
4798 out_wait:
4799         nfs4_sequence_done(task, &data->res.seq_res);
4800         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4801 }
4802
4803 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4804 {
4805         struct nfs4_lockdata *data = calldata;
4806
4807         dprintk("%s: begin!\n", __func__);
4808
4809         if (!nfs4_sequence_done(task, &data->res.seq_res))
4810                 return;
4811
4812         data->rpc_status = task->tk_status;
4813         if (data->arg.new_lock_owner != 0) {
4814                 if (data->rpc_status == 0)
4815                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4816                 else
4817                         goto out;
4818         }
4819         if (data->rpc_status == 0) {
4820                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4821                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4822                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4823         }
4824 out:
4825         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4826 }
4827
4828 static void nfs4_lock_release(void *calldata)
4829 {
4830         struct nfs4_lockdata *data = calldata;
4831
4832         dprintk("%s: begin!\n", __func__);
4833         nfs_free_seqid(data->arg.open_seqid);
4834         if (data->cancelled != 0) {
4835                 struct rpc_task *task;
4836                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4837                                 data->arg.lock_seqid);
4838                 if (!IS_ERR(task))
4839                         rpc_put_task_async(task);
4840                 dprintk("%s: cancelling lock!\n", __func__);
4841         } else
4842                 nfs_free_seqid(data->arg.lock_seqid);
4843         nfs4_put_lock_state(data->lsp);
4844         put_nfs_open_context(data->ctx);
4845         kfree(data);
4846         dprintk("%s: done!\n", __func__);
4847 }
4848
4849 static const struct rpc_call_ops nfs4_lock_ops = {
4850         .rpc_call_prepare = nfs4_lock_prepare,
4851         .rpc_call_done = nfs4_lock_done,
4852         .rpc_release = nfs4_lock_release,
4853 };
4854
4855 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4856 {
4857         switch (error) {
4858         case -NFS4ERR_ADMIN_REVOKED:
4859         case -NFS4ERR_BAD_STATEID:
4860                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4861                 if (new_lock_owner != 0 ||
4862                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4863                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4864                 break;
4865         case -NFS4ERR_STALE_STATEID:
4866                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4867         case -NFS4ERR_EXPIRED:
4868                 nfs4_schedule_lease_recovery(server->nfs_client);
4869         };
4870 }
4871
4872 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4873 {
4874         struct nfs4_lockdata *data;
4875         struct rpc_task *task;
4876         struct rpc_message msg = {
4877                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4878                 .rpc_cred = state->owner->so_cred,
4879         };
4880         struct rpc_task_setup task_setup_data = {
4881                 .rpc_client = NFS_CLIENT(state->inode),
4882                 .rpc_message = &msg,
4883                 .callback_ops = &nfs4_lock_ops,
4884                 .workqueue = nfsiod_workqueue,
4885                 .flags = RPC_TASK_ASYNC,
4886         };
4887         int ret;
4888
4889         dprintk("%s: begin!\n", __func__);
4890         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4891                         fl->fl_u.nfs4_fl.owner,
4892                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4893         if (data == NULL)
4894                 return -ENOMEM;
4895         if (IS_SETLKW(cmd))
4896                 data->arg.block = 1;
4897         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4898         msg.rpc_argp = &data->arg;
4899         msg.rpc_resp = &data->res;
4900         task_setup_data.callback_data = data;
4901         if (recovery_type > NFS_LOCK_NEW) {
4902                 if (recovery_type == NFS_LOCK_RECLAIM)
4903                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4904                 nfs4_set_sequence_privileged(&data->arg.seq_args);
4905         }
4906         task = rpc_run_task(&task_setup_data);
4907         if (IS_ERR(task))
4908                 return PTR_ERR(task);
4909         ret = nfs4_wait_for_completion_rpc_task(task);
4910         if (ret == 0) {
4911                 ret = data->rpc_status;
4912                 if (ret)
4913                         nfs4_handle_setlk_error(data->server, data->lsp,
4914                                         data->arg.new_lock_owner, ret);
4915         } else
4916                 data->cancelled = 1;
4917         rpc_put_task(task);
4918         dprintk("%s: done, ret = %d!\n", __func__, ret);
4919         return ret;
4920 }
4921
4922 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4923 {
4924         struct nfs_server *server = NFS_SERVER(state->inode);
4925         struct nfs4_exception exception = {
4926                 .inode = state->inode,
4927         };
4928         int err;
4929
4930         do {
4931                 /* Cache the lock if possible... */
4932                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4933                         return 0;
4934                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4935                 if (err != -NFS4ERR_DELAY)
4936                         break;
4937                 nfs4_handle_exception(server, err, &exception);
4938         } while (exception.retry);
4939         return err;
4940 }
4941
4942 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4943 {
4944         struct nfs_server *server = NFS_SERVER(state->inode);
4945         struct nfs4_exception exception = {
4946                 .inode = state->inode,
4947         };
4948         int err;
4949
4950         err = nfs4_set_lock_state(state, request);
4951         if (err != 0)
4952                 return err;
4953         do {
4954                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4955                         return 0;
4956                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4957                 switch (err) {
4958                 default:
4959                         goto out;
4960                 case -NFS4ERR_GRACE:
4961                 case -NFS4ERR_DELAY:
4962                         nfs4_handle_exception(server, err, &exception);
4963                         err = 0;
4964                 }
4965         } while (exception.retry);
4966 out:
4967         return err;
4968 }
4969
4970 #if defined(CONFIG_NFS_V4_1)
4971 /**
4972  * nfs41_check_expired_locks - possibly free a lock stateid
4973  *
4974  * @state: NFSv4 state for an inode
4975  *
4976  * Returns NFS_OK if recovery for this stateid is now finished.
4977  * Otherwise a negative NFS4ERR value is returned.
4978  */
4979 static int nfs41_check_expired_locks(struct nfs4_state *state)
4980 {
4981         int status, ret = -NFS4ERR_BAD_STATEID;
4982         struct nfs4_lock_state *lsp;
4983         struct nfs_server *server = NFS_SERVER(state->inode);
4984
4985         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4986                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
4987                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4988                         if (status != NFS_OK) {
4989                                 /* Free the stateid unless the server
4990                                  * informs us the stateid is unrecognized. */
4991                                 if (status != -NFS4ERR_BAD_STATEID)
4992                                         nfs41_free_stateid(server,
4993                                                         &lsp->ls_stateid);
4994                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
4995                                 ret = status;
4996                         }
4997                 }
4998         };
4999
5000         return ret;
5001 }
5002
5003 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5004 {
5005         int status = NFS_OK;
5006
5007         if (test_bit(LK_STATE_IN_USE, &state->flags))
5008                 status = nfs41_check_expired_locks(state);
5009         if (status != NFS_OK)
5010                 status = nfs4_lock_expired(state, request);
5011         return status;
5012 }
5013 #endif
5014
5015 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5016 {
5017         struct nfs4_state_owner *sp = state->owner;
5018         struct nfs_inode *nfsi = NFS_I(state->inode);
5019         unsigned char fl_flags = request->fl_flags;
5020         unsigned int seq;
5021         int status = -ENOLCK;
5022
5023         if ((fl_flags & FL_POSIX) &&
5024                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5025                 goto out;
5026         /* Is this a delegated open? */
5027         status = nfs4_set_lock_state(state, request);
5028         if (status != 0)
5029                 goto out;
5030         request->fl_flags |= FL_ACCESS;
5031         status = do_vfs_lock(request->fl_file, request);
5032         if (status < 0)
5033                 goto out;
5034         down_read(&nfsi->rwsem);
5035         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5036                 /* Yes: cache locks! */
5037                 /* ...but avoid races with delegation recall... */
5038                 request->fl_flags = fl_flags & ~FL_SLEEP;
5039                 status = do_vfs_lock(request->fl_file, request);
5040                 goto out_unlock;
5041         }
5042         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5043         up_read(&nfsi->rwsem);
5044         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5045         if (status != 0)
5046                 goto out;
5047         down_read(&nfsi->rwsem);
5048         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5049                 status = -NFS4ERR_DELAY;
5050                 goto out_unlock;
5051         }
5052         /* Note: we always want to sleep here! */
5053         request->fl_flags = fl_flags | FL_SLEEP;
5054         if (do_vfs_lock(request->fl_file, request) < 0)
5055                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5056                         "manager!\n", __func__);
5057 out_unlock:
5058         up_read(&nfsi->rwsem);
5059 out:
5060         request->fl_flags = fl_flags;
5061         return status;
5062 }
5063
5064 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5065 {
5066         struct nfs4_exception exception = {
5067                 .state = state,
5068                 .inode = state->inode,
5069         };
5070         int err;
5071
5072         do {
5073                 err = _nfs4_proc_setlk(state, cmd, request);
5074                 if (err == -NFS4ERR_DENIED)
5075                         err = -EAGAIN;
5076                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5077                                 err, &exception);
5078         } while (exception.retry);
5079         return err;
5080 }
5081
5082 static int
5083 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5084 {
5085         struct nfs_open_context *ctx;
5086         struct nfs4_state *state;
5087         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5088         int status;
5089
5090         /* verify open state */
5091         ctx = nfs_file_open_context(filp);
5092         state = ctx->state;
5093
5094         if (request->fl_start < 0 || request->fl_end < 0)
5095                 return -EINVAL;
5096
5097         if (IS_GETLK(cmd)) {
5098                 if (state != NULL)
5099                         return nfs4_proc_getlk(state, F_GETLK, request);
5100                 return 0;
5101         }
5102
5103         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5104                 return -EINVAL;
5105
5106         if (request->fl_type == F_UNLCK) {
5107                 if (state != NULL)
5108                         return nfs4_proc_unlck(state, cmd, request);
5109                 return 0;
5110         }
5111
5112         if (state == NULL)
5113                 return -ENOLCK;
5114         /*
5115          * Don't rely on the VFS having checked the file open mode,
5116          * since it won't do this for flock() locks.
5117          */
5118         switch (request->fl_type) {
5119         case F_RDLCK:
5120                 if (!(filp->f_mode & FMODE_READ))
5121                         return -EBADF;
5122                 break;
5123         case F_WRLCK:
5124                 if (!(filp->f_mode & FMODE_WRITE))
5125                         return -EBADF;
5126         }
5127
5128         do {
5129                 status = nfs4_proc_setlk(state, cmd, request);
5130                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5131                         break;
5132                 timeout = nfs4_set_lock_task_retry(timeout);
5133                 status = -ERESTARTSYS;
5134                 if (signalled())
5135                         break;
5136         } while(status < 0);
5137         return status;
5138 }
5139
5140 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
5141 {
5142         struct nfs_server *server = NFS_SERVER(state->inode);
5143         struct nfs4_exception exception = { };
5144         int err;
5145
5146         err = nfs4_set_lock_state(state, fl);
5147         if (err != 0)
5148                 goto out;
5149         do {
5150                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5151                 switch (err) {
5152                         default:
5153                                 printk(KERN_ERR "NFS: %s: unhandled error "
5154                                         "%d.\n", __func__, err);
5155                         case 0:
5156                         case -ESTALE:
5157                                 goto out;
5158                         case -NFS4ERR_STALE_CLIENTID:
5159                         case -NFS4ERR_STALE_STATEID:
5160                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
5161                         case -NFS4ERR_EXPIRED:
5162                                 nfs4_schedule_lease_recovery(server->nfs_client);
5163                                 err = -EAGAIN;
5164                                 goto out;
5165                         case -NFS4ERR_BADSESSION:
5166                         case -NFS4ERR_BADSLOT:
5167                         case -NFS4ERR_BAD_HIGH_SLOT:
5168                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
5169                         case -NFS4ERR_DEADSESSION:
5170                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
5171                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
5172                                 err = -EAGAIN;
5173                                 goto out;
5174                         case -NFS4ERR_DELEG_REVOKED:
5175                         case -NFS4ERR_ADMIN_REVOKED:
5176                         case -NFS4ERR_BAD_STATEID:
5177                         case -NFS4ERR_OPENMODE:
5178                                 nfs4_schedule_stateid_recovery(server, state);
5179                                 err = 0;
5180                                 goto out;
5181                         case -ENOMEM:
5182                         case -NFS4ERR_DENIED:
5183                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
5184                                 err = 0;
5185                                 goto out;
5186                 }
5187                 set_bit(NFS_DELEGATED_STATE, &state->flags);
5188                 err = nfs4_handle_exception(server, err, &exception);
5189         } while (exception.retry);
5190 out:
5191         return err;
5192 }
5193
5194 struct nfs_release_lockowner_data {
5195         struct nfs4_lock_state *lsp;
5196         struct nfs_server *server;
5197         struct nfs_release_lockowner_args args;
5198 };
5199
5200 static void nfs4_release_lockowner_release(void *calldata)
5201 {
5202         struct nfs_release_lockowner_data *data = calldata;
5203         nfs4_free_lock_state(data->server, data->lsp);
5204         kfree(calldata);
5205 }
5206
5207 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5208         .rpc_release = nfs4_release_lockowner_release,
5209 };
5210
5211 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
5212 {
5213         struct nfs_server *server = lsp->ls_state->owner->so_server;
5214         struct nfs_release_lockowner_data *data;
5215         struct rpc_message msg = {
5216                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5217         };
5218
5219         if (server->nfs_client->cl_mvops->minor_version != 0)
5220                 return -EINVAL;
5221         data = kmalloc(sizeof(*data), GFP_NOFS);
5222         if (!data)
5223                 return -ENOMEM;
5224         data->lsp = lsp;
5225         data->server = server;
5226         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5227         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5228         data->args.lock_owner.s_dev = server->s_dev;
5229         msg.rpc_argp = &data->args;
5230         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5231         return 0;
5232 }
5233
5234 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5235
5236 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5237                                    const void *buf, size_t buflen,
5238                                    int flags, int type)
5239 {
5240         if (strcmp(key, "") != 0)
5241                 return -EINVAL;
5242
5243         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5244 }
5245
5246 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5247                                    void *buf, size_t buflen, int type)
5248 {
5249         if (strcmp(key, "") != 0)
5250                 return -EINVAL;
5251
5252         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5253 }
5254
5255 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5256                                        size_t list_len, const char *name,
5257                                        size_t name_len, int type)
5258 {
5259         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5260
5261         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5262                 return 0;
5263
5264         if (list && len <= list_len)
5265                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5266         return len;
5267 }
5268
5269 /*
5270  * nfs_fhget will use either the mounted_on_fileid or the fileid
5271  */
5272 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5273 {
5274         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5275                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5276               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5277               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5278                 return;
5279
5280         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5281                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5282         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5283         fattr->nlink = 2;
5284 }
5285
5286 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5287                                    const struct qstr *name,
5288                                    struct nfs4_fs_locations *fs_locations,
5289                                    struct page *page)
5290 {
5291         struct nfs_server *server = NFS_SERVER(dir);
5292         u32 bitmask[2] = {
5293                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5294         };
5295         struct nfs4_fs_locations_arg args = {
5296                 .dir_fh = NFS_FH(dir),
5297                 .name = name,
5298                 .page = page,
5299                 .bitmask = bitmask,
5300         };
5301         struct nfs4_fs_locations_res res = {
5302                 .fs_locations = fs_locations,
5303         };
5304         struct rpc_message msg = {
5305                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5306                 .rpc_argp = &args,
5307                 .rpc_resp = &res,
5308         };
5309         int status;
5310
5311         dprintk("%s: start\n", __func__);
5312
5313         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5314          * is not supported */
5315         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5316                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5317         else
5318                 bitmask[0] |= FATTR4_WORD0_FILEID;
5319
5320         nfs_fattr_init(&fs_locations->fattr);
5321         fs_locations->server = server;
5322         fs_locations->nlocations = 0;
5323         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5324         dprintk("%s: returned status = %d\n", __func__, status);
5325         return status;
5326 }
5327
5328 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5329                            const struct qstr *name,
5330                            struct nfs4_fs_locations *fs_locations,
5331                            struct page *page)
5332 {
5333         struct nfs4_exception exception = { };
5334         int err;
5335         do {
5336                 err = nfs4_handle_exception(NFS_SERVER(dir),
5337                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5338                                 &exception);
5339         } while (exception.retry);
5340         return err;
5341 }
5342
5343 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5344 {
5345         int status;
5346         struct nfs4_secinfo_arg args = {
5347                 .dir_fh = NFS_FH(dir),
5348                 .name   = name,
5349         };
5350         struct nfs4_secinfo_res res = {
5351                 .flavors     = flavors,
5352         };
5353         struct rpc_message msg = {
5354                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5355                 .rpc_argp = &args,
5356                 .rpc_resp = &res,
5357         };
5358
5359         dprintk("NFS call  secinfo %s\n", name->name);
5360         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5361         dprintk("NFS reply  secinfo: %d\n", status);
5362         return status;
5363 }
5364
5365 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5366                       struct nfs4_secinfo_flavors *flavors)
5367 {
5368         struct nfs4_exception exception = { };
5369         int err;
5370         do {
5371                 err = nfs4_handle_exception(NFS_SERVER(dir),
5372                                 _nfs4_proc_secinfo(dir, name, flavors),
5373                                 &exception);
5374         } while (exception.retry);
5375         return err;
5376 }
5377
5378 #ifdef CONFIG_NFS_V4_1
5379 /*
5380  * Check the exchange flags returned by the server for invalid flags, having
5381  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5382  * DS flags set.
5383  */
5384 static int nfs4_check_cl_exchange_flags(u32 flags)
5385 {
5386         if (flags & ~EXCHGID4_FLAG_MASK_R)
5387                 goto out_inval;
5388         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5389             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5390                 goto out_inval;
5391         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5392                 goto out_inval;
5393         return NFS_OK;
5394 out_inval:
5395         return -NFS4ERR_INVAL;
5396 }
5397
5398 static bool
5399 nfs41_same_server_scope(struct nfs41_server_scope *a,
5400                         struct nfs41_server_scope *b)
5401 {
5402         if (a->server_scope_sz == b->server_scope_sz &&
5403             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5404                 return true;
5405
5406         return false;
5407 }
5408
5409 /*
5410  * nfs4_proc_bind_conn_to_session()
5411  *
5412  * The 4.1 client currently uses the same TCP connection for the
5413  * fore and backchannel.
5414  */
5415 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5416 {
5417         int status;
5418         struct nfs41_bind_conn_to_session_res res;
5419         struct rpc_message msg = {
5420                 .rpc_proc =
5421                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5422                 .rpc_argp = clp,
5423                 .rpc_resp = &res,
5424                 .rpc_cred = cred,
5425         };
5426
5427         dprintk("--> %s\n", __func__);
5428
5429         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5430         if (unlikely(res.session == NULL)) {
5431                 status = -ENOMEM;
5432                 goto out;
5433         }
5434
5435         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5436         if (status == 0) {
5437                 if (memcmp(res.session->sess_id.data,
5438                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5439                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5440                         status = -EIO;
5441                         goto out_session;
5442                 }
5443                 if (res.dir != NFS4_CDFS4_BOTH) {
5444                         dprintk("NFS: %s: Unexpected direction from server\n",
5445                                 __func__);
5446                         status = -EIO;
5447                         goto out_session;
5448                 }
5449                 if (res.use_conn_in_rdma_mode) {
5450                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5451                                 __func__);
5452                         status = -EIO;
5453                         goto out_session;
5454                 }
5455         }
5456 out_session:
5457         kfree(res.session);
5458 out:
5459         dprintk("<-- %s status= %d\n", __func__, status);
5460         return status;
5461 }
5462
5463 /*
5464  * nfs4_proc_exchange_id()
5465  *
5466  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5467  *
5468  * Since the clientid has expired, all compounds using sessions
5469  * associated with the stale clientid will be returning
5470  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5471  * be in some phase of session reset.
5472  */
5473 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5474 {
5475         nfs4_verifier verifier;
5476         struct nfs41_exchange_id_args args = {
5477                 .verifier = &verifier,
5478                 .client = clp,
5479                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5480         };
5481         struct nfs41_exchange_id_res res = {
5482                 0
5483         };
5484         int status;
5485         struct rpc_message msg = {
5486                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5487                 .rpc_argp = &args,
5488                 .rpc_resp = &res,
5489                 .rpc_cred = cred,
5490         };
5491
5492         nfs4_init_boot_verifier(clp, &verifier);
5493         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5494                                                         sizeof(args.id));
5495         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5496                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5497                 args.id_len, args.id);
5498
5499         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5500                                         GFP_NOFS);
5501         if (unlikely(res.server_owner == NULL)) {
5502                 status = -ENOMEM;
5503                 goto out;
5504         }
5505
5506         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5507                                         GFP_NOFS);
5508         if (unlikely(res.server_scope == NULL)) {
5509                 status = -ENOMEM;
5510                 goto out_server_owner;
5511         }
5512
5513         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5514         if (unlikely(res.impl_id == NULL)) {
5515                 status = -ENOMEM;
5516                 goto out_server_scope;
5517         }
5518
5519         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5520         if (status == 0)
5521                 status = nfs4_check_cl_exchange_flags(res.flags);
5522
5523         if (status == 0) {
5524                 clp->cl_clientid = res.clientid;
5525                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5526                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5527                         clp->cl_seqid = res.seqid;
5528
5529                 kfree(clp->cl_serverowner);
5530                 clp->cl_serverowner = res.server_owner;
5531                 res.server_owner = NULL;
5532
5533                 /* use the most recent implementation id */
5534                 kfree(clp->cl_implid);
5535                 clp->cl_implid = res.impl_id;
5536
5537                 if (clp->cl_serverscope != NULL &&
5538                     !nfs41_same_server_scope(clp->cl_serverscope,
5539                                              res.server_scope)) {
5540                         dprintk("%s: server_scope mismatch detected\n",
5541                                 __func__);
5542                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5543                         kfree(clp->cl_serverscope);
5544                         clp->cl_serverscope = NULL;
5545                 }
5546
5547                 if (clp->cl_serverscope == NULL) {
5548                         clp->cl_serverscope = res.server_scope;
5549                         goto out;
5550                 }
5551         } else
5552                 kfree(res.impl_id);
5553
5554 out_server_owner:
5555         kfree(res.server_owner);
5556 out_server_scope:
5557         kfree(res.server_scope);
5558 out:
5559         if (clp->cl_implid != NULL)
5560                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5561                         "domain: %s, name: %s, date: %llu,%u\n",
5562                         clp->cl_implid->domain, clp->cl_implid->name,
5563                         clp->cl_implid->date.seconds,
5564                         clp->cl_implid->date.nseconds);
5565         dprintk("NFS reply exchange_id: %d\n", status);
5566         return status;
5567 }
5568
5569 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5570                 struct rpc_cred *cred)
5571 {
5572         struct rpc_message msg = {
5573                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5574                 .rpc_argp = clp,
5575                 .rpc_cred = cred,
5576         };
5577         int status;
5578
5579         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5580         if (status)
5581                 dprintk("NFS: Got error %d from the server %s on "
5582                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5583         return status;
5584 }
5585
5586 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5587                 struct rpc_cred *cred)
5588 {
5589         unsigned int loop;
5590         int ret;
5591
5592         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5593                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5594                 switch (ret) {
5595                 case -NFS4ERR_DELAY:
5596                 case -NFS4ERR_CLIENTID_BUSY:
5597                         ssleep(1);
5598                         break;
5599                 default:
5600                         return ret;
5601                 }
5602         }
5603         return 0;
5604 }
5605
5606 int nfs4_destroy_clientid(struct nfs_client *clp)
5607 {
5608         struct rpc_cred *cred;
5609         int ret = 0;
5610
5611         if (clp->cl_mvops->minor_version < 1)
5612                 goto out;
5613         if (clp->cl_exchange_flags == 0)
5614                 goto out;
5615         if (clp->cl_preserve_clid)
5616                 goto out;
5617         cred = nfs4_get_exchange_id_cred(clp);
5618         ret = nfs4_proc_destroy_clientid(clp, cred);
5619         if (cred)
5620                 put_rpccred(cred);
5621         switch (ret) {
5622         case 0:
5623         case -NFS4ERR_STALE_CLIENTID:
5624                 clp->cl_exchange_flags = 0;
5625         }
5626 out:
5627         return ret;
5628 }
5629
5630 struct nfs4_get_lease_time_data {
5631         struct nfs4_get_lease_time_args *args;
5632         struct nfs4_get_lease_time_res *res;
5633         struct nfs_client *clp;
5634 };
5635
5636 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5637                                         void *calldata)
5638 {
5639         struct nfs4_get_lease_time_data *data =
5640                         (struct nfs4_get_lease_time_data *)calldata;
5641
5642         dprintk("--> %s\n", __func__);
5643         /* just setup sequence, do not trigger session recovery
5644            since we're invoked within one */
5645         nfs41_setup_sequence(data->clp->cl_session,
5646                         &data->args->la_seq_args,
5647                         &data->res->lr_seq_res,
5648                         task);
5649         dprintk("<-- %s\n", __func__);
5650 }
5651
5652 /*
5653  * Called from nfs4_state_manager thread for session setup, so don't recover
5654  * from sequence operation or clientid errors.
5655  */
5656 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5657 {
5658         struct nfs4_get_lease_time_data *data =
5659                         (struct nfs4_get_lease_time_data *)calldata;
5660
5661         dprintk("--> %s\n", __func__);
5662         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5663                 return;
5664         switch (task->tk_status) {
5665         case -NFS4ERR_DELAY:
5666         case -NFS4ERR_GRACE:
5667                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5668                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5669                 task->tk_status = 0;
5670                 /* fall through */
5671         case -NFS4ERR_RETRY_UNCACHED_REP:
5672                 rpc_restart_call_prepare(task);
5673                 return;
5674         }
5675         dprintk("<-- %s\n", __func__);
5676 }
5677
5678 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5679         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5680         .rpc_call_done = nfs4_get_lease_time_done,
5681 };
5682
5683 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5684 {
5685         struct rpc_task *task;
5686         struct nfs4_get_lease_time_args args;
5687         struct nfs4_get_lease_time_res res = {
5688                 .lr_fsinfo = fsinfo,
5689         };
5690         struct nfs4_get_lease_time_data data = {
5691                 .args = &args,
5692                 .res = &res,
5693                 .clp = clp,
5694         };
5695         struct rpc_message msg = {
5696                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5697                 .rpc_argp = &args,
5698                 .rpc_resp = &res,
5699         };
5700         struct rpc_task_setup task_setup = {
5701                 .rpc_client = clp->cl_rpcclient,
5702                 .rpc_message = &msg,
5703                 .callback_ops = &nfs4_get_lease_time_ops,
5704                 .callback_data = &data,
5705                 .flags = RPC_TASK_TIMEOUT,
5706         };
5707         int status;
5708
5709         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5710         nfs4_set_sequence_privileged(&args.la_seq_args);
5711         dprintk("--> %s\n", __func__);
5712         task = rpc_run_task(&task_setup);
5713
5714         if (IS_ERR(task))
5715                 status = PTR_ERR(task);
5716         else {
5717                 status = task->tk_status;
5718                 rpc_put_task(task);
5719         }
5720         dprintk("<-- %s return %d\n", __func__, status);
5721
5722         return status;
5723 }
5724
5725 /*
5726  * Initialize the values to be used by the client in CREATE_SESSION
5727  * If nfs4_init_session set the fore channel request and response sizes,
5728  * use them.
5729  *
5730  * Set the back channel max_resp_sz_cached to zero to force the client to
5731  * always set csa_cachethis to FALSE because the current implementation
5732  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5733  */
5734 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5735 {
5736         struct nfs4_session *session = args->client->cl_session;
5737         unsigned int mxrqst_sz = session->fc_target_max_rqst_sz,
5738                      mxresp_sz = session->fc_target_max_resp_sz;
5739
5740         if (mxrqst_sz == 0)
5741                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5742         if (mxresp_sz == 0)
5743                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5744         /* Fore channel attributes */
5745         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5746         args->fc_attrs.max_resp_sz = mxresp_sz;
5747         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5748         args->fc_attrs.max_reqs = max_session_slots;
5749
5750         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5751                 "max_ops=%u max_reqs=%u\n",
5752                 __func__,
5753                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5754                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5755
5756         /* Back channel attributes */
5757         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5758         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5759         args->bc_attrs.max_resp_sz_cached = 0;
5760         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5761         args->bc_attrs.max_reqs = 1;
5762
5763         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5764                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5765                 __func__,
5766                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5767                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5768                 args->bc_attrs.max_reqs);
5769 }
5770
5771 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5772 {
5773         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5774         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5775
5776         if (rcvd->max_resp_sz > sent->max_resp_sz)
5777                 return -EINVAL;
5778         /*
5779          * Our requested max_ops is the minimum we need; we're not
5780          * prepared to break up compounds into smaller pieces than that.
5781          * So, no point even trying to continue if the server won't
5782          * cooperate:
5783          */
5784         if (rcvd->max_ops < sent->max_ops)
5785                 return -EINVAL;
5786         if (rcvd->max_reqs == 0)
5787                 return -EINVAL;
5788         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5789                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5790         return 0;
5791 }
5792
5793 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5794 {
5795         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5796         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5797
5798         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5799                 return -EINVAL;
5800         if (rcvd->max_resp_sz < sent->max_resp_sz)
5801                 return -EINVAL;
5802         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5803                 return -EINVAL;
5804         /* These would render the backchannel useless: */
5805         if (rcvd->max_ops != sent->max_ops)
5806                 return -EINVAL;
5807         if (rcvd->max_reqs != sent->max_reqs)
5808                 return -EINVAL;
5809         return 0;
5810 }
5811
5812 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5813                                      struct nfs4_session *session)
5814 {
5815         int ret;
5816
5817         ret = nfs4_verify_fore_channel_attrs(args, session);
5818         if (ret)
5819                 return ret;
5820         return nfs4_verify_back_channel_attrs(args, session);
5821 }
5822
5823 static int _nfs4_proc_create_session(struct nfs_client *clp,
5824                 struct rpc_cred *cred)
5825 {
5826         struct nfs4_session *session = clp->cl_session;
5827         struct nfs41_create_session_args args = {
5828                 .client = clp,
5829                 .cb_program = NFS4_CALLBACK,
5830         };
5831         struct nfs41_create_session_res res = {
5832                 .client = clp,
5833         };
5834         struct rpc_message msg = {
5835                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5836                 .rpc_argp = &args,
5837                 .rpc_resp = &res,
5838                 .rpc_cred = cred,
5839         };
5840         int status;
5841
5842         nfs4_init_channel_attrs(&args);
5843         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5844
5845         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5846
5847         if (!status) {
5848                 /* Verify the session's negotiated channel_attrs values */
5849                 status = nfs4_verify_channel_attrs(&args, session);
5850                 /* Increment the clientid slot sequence id */
5851                 clp->cl_seqid++;
5852         }
5853
5854         return status;
5855 }
5856
5857 /*
5858  * Issues a CREATE_SESSION operation to the server.
5859  * It is the responsibility of the caller to verify the session is
5860  * expired before calling this routine.
5861  */
5862 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5863 {
5864         int status;
5865         unsigned *ptr;
5866         struct nfs4_session *session = clp->cl_session;
5867
5868         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5869
5870         status = _nfs4_proc_create_session(clp, cred);
5871         if (status)
5872                 goto out;
5873
5874         /* Init or reset the session slot tables */
5875         status = nfs4_setup_session_slot_tables(session);
5876         dprintk("slot table setup returned %d\n", status);
5877         if (status)
5878                 goto out;
5879
5880         ptr = (unsigned *)&session->sess_id.data[0];
5881         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5882                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5883 out:
5884         dprintk("<-- %s\n", __func__);
5885         return status;
5886 }
5887
5888 /*
5889  * Issue the over-the-wire RPC DESTROY_SESSION.
5890  * The caller must serialize access to this routine.
5891  */
5892 int nfs4_proc_destroy_session(struct nfs4_session *session,
5893                 struct rpc_cred *cred)
5894 {
5895         struct rpc_message msg = {
5896                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5897                 .rpc_argp = session,
5898                 .rpc_cred = cred,
5899         };
5900         int status = 0;
5901
5902         dprintk("--> nfs4_proc_destroy_session\n");
5903
5904         /* session is still being setup */
5905         if (session->clp->cl_cons_state != NFS_CS_READY)
5906                 return status;
5907
5908         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5909
5910         if (status)
5911                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5912                         "Session has been destroyed regardless...\n", status);
5913
5914         dprintk("<-- nfs4_proc_destroy_session\n");
5915         return status;
5916 }
5917
5918 /*
5919  * Renew the cl_session lease.
5920  */
5921 struct nfs4_sequence_data {
5922         struct nfs_client *clp;
5923         struct nfs4_sequence_args args;
5924         struct nfs4_sequence_res res;
5925 };
5926
5927 static void nfs41_sequence_release(void *data)
5928 {
5929         struct nfs4_sequence_data *calldata = data;
5930         struct nfs_client *clp = calldata->clp;
5931
5932         if (atomic_read(&clp->cl_count) > 1)
5933                 nfs4_schedule_state_renewal(clp);
5934         nfs_put_client(clp);
5935         kfree(calldata);
5936 }
5937
5938 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5939 {
5940         switch(task->tk_status) {
5941         case -NFS4ERR_DELAY:
5942                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5943                 return -EAGAIN;
5944         default:
5945                 nfs4_schedule_lease_recovery(clp);
5946         }
5947         return 0;
5948 }
5949
5950 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5951 {
5952         struct nfs4_sequence_data *calldata = data;
5953         struct nfs_client *clp = calldata->clp;
5954
5955         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5956                 return;
5957
5958         if (task->tk_status < 0) {
5959                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5960                 if (atomic_read(&clp->cl_count) == 1)
5961                         goto out;
5962
5963                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5964                         rpc_restart_call_prepare(task);
5965                         return;
5966                 }
5967         }
5968         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5969 out:
5970         dprintk("<-- %s\n", __func__);
5971 }
5972
5973 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5974 {
5975         struct nfs4_sequence_data *calldata = data;
5976         struct nfs_client *clp = calldata->clp;
5977         struct nfs4_sequence_args *args;
5978         struct nfs4_sequence_res *res;
5979
5980         args = task->tk_msg.rpc_argp;
5981         res = task->tk_msg.rpc_resp;
5982
5983         nfs41_setup_sequence(clp->cl_session, args, res, task);
5984 }
5985
5986 static const struct rpc_call_ops nfs41_sequence_ops = {
5987         .rpc_call_done = nfs41_sequence_call_done,
5988         .rpc_call_prepare = nfs41_sequence_prepare,
5989         .rpc_release = nfs41_sequence_release,
5990 };
5991
5992 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
5993                 struct rpc_cred *cred,
5994                 bool is_privileged)
5995 {
5996         struct nfs4_sequence_data *calldata;
5997         struct rpc_message msg = {
5998                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5999                 .rpc_cred = cred,
6000         };
6001         struct rpc_task_setup task_setup_data = {
6002                 .rpc_client = clp->cl_rpcclient,
6003                 .rpc_message = &msg,
6004                 .callback_ops = &nfs41_sequence_ops,
6005                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
6006         };
6007
6008         if (!atomic_inc_not_zero(&clp->cl_count))
6009                 return ERR_PTR(-EIO);
6010         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6011         if (calldata == NULL) {
6012                 nfs_put_client(clp);
6013                 return ERR_PTR(-ENOMEM);
6014         }
6015         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6016         if (is_privileged)
6017                 nfs4_set_sequence_privileged(&calldata->args);
6018         msg.rpc_argp = &calldata->args;
6019         msg.rpc_resp = &calldata->res;
6020         calldata->clp = clp;
6021         task_setup_data.callback_data = calldata;
6022
6023         return rpc_run_task(&task_setup_data);
6024 }
6025
6026 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6027 {
6028         struct rpc_task *task;
6029         int ret = 0;
6030
6031         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6032                 return 0;
6033         task = _nfs41_proc_sequence(clp, cred, false);
6034         if (IS_ERR(task))
6035                 ret = PTR_ERR(task);
6036         else
6037                 rpc_put_task_async(task);
6038         dprintk("<-- %s status=%d\n", __func__, ret);
6039         return ret;
6040 }
6041
6042 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6043 {
6044         struct rpc_task *task;
6045         int ret;
6046
6047         task = _nfs41_proc_sequence(clp, cred, true);
6048         if (IS_ERR(task)) {
6049                 ret = PTR_ERR(task);
6050                 goto out;
6051         }
6052         ret = rpc_wait_for_completion_task(task);
6053         if (!ret) {
6054                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6055
6056                 if (task->tk_status == 0)
6057                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6058                 ret = task->tk_status;
6059         }
6060         rpc_put_task(task);
6061 out:
6062         dprintk("<-- %s status=%d\n", __func__, ret);
6063         return ret;
6064 }
6065
6066 struct nfs4_reclaim_complete_data {
6067         struct nfs_client *clp;
6068         struct nfs41_reclaim_complete_args arg;
6069         struct nfs41_reclaim_complete_res res;
6070 };
6071
6072 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6073 {
6074         struct nfs4_reclaim_complete_data *calldata = data;
6075
6076         nfs41_setup_sequence(calldata->clp->cl_session,
6077                         &calldata->arg.seq_args,
6078                         &calldata->res.seq_res,
6079                         task);
6080 }
6081
6082 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6083 {
6084         switch(task->tk_status) {
6085         case 0:
6086         case -NFS4ERR_COMPLETE_ALREADY:
6087         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6088                 break;
6089         case -NFS4ERR_DELAY:
6090                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6091                 /* fall through */
6092         case -NFS4ERR_RETRY_UNCACHED_REP:
6093                 return -EAGAIN;
6094         default:
6095                 nfs4_schedule_lease_recovery(clp);
6096         }
6097         return 0;
6098 }
6099
6100 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6101 {
6102         struct nfs4_reclaim_complete_data *calldata = data;
6103         struct nfs_client *clp = calldata->clp;
6104         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6105
6106         dprintk("--> %s\n", __func__);
6107         if (!nfs41_sequence_done(task, res))
6108                 return;
6109
6110         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6111                 rpc_restart_call_prepare(task);
6112                 return;
6113         }
6114         dprintk("<-- %s\n", __func__);
6115 }
6116
6117 static void nfs4_free_reclaim_complete_data(void *data)
6118 {
6119         struct nfs4_reclaim_complete_data *calldata = data;
6120
6121         kfree(calldata);
6122 }
6123
6124 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6125         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6126         .rpc_call_done = nfs4_reclaim_complete_done,
6127         .rpc_release = nfs4_free_reclaim_complete_data,
6128 };
6129
6130 /*
6131  * Issue a global reclaim complete.
6132  */
6133 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6134 {
6135         struct nfs4_reclaim_complete_data *calldata;
6136         struct rpc_task *task;
6137         struct rpc_message msg = {
6138                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6139         };
6140         struct rpc_task_setup task_setup_data = {
6141                 .rpc_client = clp->cl_rpcclient,
6142                 .rpc_message = &msg,
6143                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6144                 .flags = RPC_TASK_ASYNC,
6145         };
6146         int status = -ENOMEM;
6147
6148         dprintk("--> %s\n", __func__);
6149         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6150         if (calldata == NULL)
6151                 goto out;
6152         calldata->clp = clp;
6153         calldata->arg.one_fs = 0;
6154
6155         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6156         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
6157         msg.rpc_argp = &calldata->arg;
6158         msg.rpc_resp = &calldata->res;
6159         task_setup_data.callback_data = calldata;
6160         task = rpc_run_task(&task_setup_data);
6161         if (IS_ERR(task)) {
6162                 status = PTR_ERR(task);
6163                 goto out;
6164         }
6165         status = nfs4_wait_for_completion_rpc_task(task);
6166         if (status == 0)
6167                 status = task->tk_status;
6168         rpc_put_task(task);
6169         return 0;
6170 out:
6171         dprintk("<-- %s status=%d\n", __func__, status);
6172         return status;
6173 }
6174
6175 static void
6176 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6177 {
6178         struct nfs4_layoutget *lgp = calldata;
6179         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6180         struct nfs4_session *session = nfs4_get_session(server);
6181
6182         dprintk("--> %s\n", __func__);
6183         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6184          * right now covering the LAYOUTGET we are about to send.
6185          * However, that is not so catastrophic, and there seems
6186          * to be no way to prevent it completely.
6187          */
6188         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
6189                                 &lgp->res.seq_res, task))
6190                 return;
6191         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6192                                           NFS_I(lgp->args.inode)->layout,
6193                                           lgp->args.ctx->state)) {
6194                 rpc_exit(task, NFS4_OK);
6195         }
6196 }
6197
6198 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6199 {
6200         struct nfs4_layoutget *lgp = calldata;
6201         struct inode *inode = lgp->args.inode;
6202         struct nfs_server *server = NFS_SERVER(inode);
6203         struct pnfs_layout_hdr *lo;
6204         struct nfs4_state *state = NULL;
6205         unsigned long timeo, giveup;
6206
6207         dprintk("--> %s\n", __func__);
6208
6209         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
6210                 goto out;
6211
6212         switch (task->tk_status) {
6213         case 0:
6214                 goto out;
6215         case -NFS4ERR_LAYOUTTRYLATER:
6216         case -NFS4ERR_RECALLCONFLICT:
6217                 timeo = rpc_get_timeout(task->tk_client);
6218                 giveup = lgp->args.timestamp + timeo;
6219                 if (time_after(giveup, jiffies))
6220                         task->tk_status = -NFS4ERR_DELAY;
6221                 break;
6222         case -NFS4ERR_EXPIRED:
6223         case -NFS4ERR_BAD_STATEID:
6224                 spin_lock(&inode->i_lock);
6225                 lo = NFS_I(inode)->layout;
6226                 if (!lo || list_empty(&lo->plh_segs)) {
6227                         spin_unlock(&inode->i_lock);
6228                         /* If the open stateid was bad, then recover it. */
6229                         state = lgp->args.ctx->state;
6230                 } else {
6231                         LIST_HEAD(head);
6232
6233                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6234                         spin_unlock(&inode->i_lock);
6235                         /* Mark the bad layout state as invalid, then
6236                          * retry using the open stateid. */
6237                         pnfs_free_lseg_list(&head);
6238                 }
6239         }
6240         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6241                 rpc_restart_call_prepare(task);
6242 out:
6243         dprintk("<-- %s\n", __func__);
6244 }
6245
6246 static size_t max_response_pages(struct nfs_server *server)
6247 {
6248         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6249         return nfs_page_array_len(0, max_resp_sz);
6250 }
6251
6252 static void nfs4_free_pages(struct page **pages, size_t size)
6253 {
6254         int i;
6255
6256         if (!pages)
6257                 return;
6258
6259         for (i = 0; i < size; i++) {
6260                 if (!pages[i])
6261                         break;
6262                 __free_page(pages[i]);
6263         }
6264         kfree(pages);
6265 }
6266
6267 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6268 {
6269         struct page **pages;
6270         int i;
6271
6272         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6273         if (!pages) {
6274                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6275                 return NULL;
6276         }
6277
6278         for (i = 0; i < size; i++) {
6279                 pages[i] = alloc_page(gfp_flags);
6280                 if (!pages[i]) {
6281                         dprintk("%s: failed to allocate page\n", __func__);
6282                         nfs4_free_pages(pages, size);
6283                         return NULL;
6284                 }
6285         }
6286
6287         return pages;
6288 }
6289
6290 static void nfs4_layoutget_release(void *calldata)
6291 {
6292         struct nfs4_layoutget *lgp = calldata;
6293         struct inode *inode = lgp->args.inode;
6294         struct nfs_server *server = NFS_SERVER(inode);
6295         size_t max_pages = max_response_pages(server);
6296
6297         dprintk("--> %s\n", __func__);
6298         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6299         pnfs_put_layout_hdr(NFS_I(inode)->layout);
6300         put_nfs_open_context(lgp->args.ctx);
6301         kfree(calldata);
6302         dprintk("<-- %s\n", __func__);
6303 }
6304
6305 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6306         .rpc_call_prepare = nfs4_layoutget_prepare,
6307         .rpc_call_done = nfs4_layoutget_done,
6308         .rpc_release = nfs4_layoutget_release,
6309 };
6310
6311 struct pnfs_layout_segment *
6312 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6313 {
6314         struct inode *inode = lgp->args.inode;
6315         struct nfs_server *server = NFS_SERVER(inode);
6316         size_t max_pages = max_response_pages(server);
6317         struct rpc_task *task;
6318         struct rpc_message msg = {
6319                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6320                 .rpc_argp = &lgp->args,
6321                 .rpc_resp = &lgp->res,
6322         };
6323         struct rpc_task_setup task_setup_data = {
6324                 .rpc_client = server->client,
6325                 .rpc_message = &msg,
6326                 .callback_ops = &nfs4_layoutget_call_ops,
6327                 .callback_data = lgp,
6328                 .flags = RPC_TASK_ASYNC,
6329         };
6330         struct pnfs_layout_segment *lseg = NULL;
6331         int status = 0;
6332
6333         dprintk("--> %s\n", __func__);
6334
6335         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6336         if (!lgp->args.layout.pages) {
6337                 nfs4_layoutget_release(lgp);
6338                 return ERR_PTR(-ENOMEM);
6339         }
6340         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6341         lgp->args.timestamp = jiffies;
6342
6343         lgp->res.layoutp = &lgp->args.layout;
6344         lgp->res.seq_res.sr_slot = NULL;
6345         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6346
6347         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
6348         pnfs_get_layout_hdr(NFS_I(inode)->layout);
6349
6350         task = rpc_run_task(&task_setup_data);
6351         if (IS_ERR(task))
6352                 return ERR_CAST(task);
6353         status = nfs4_wait_for_completion_rpc_task(task);
6354         if (status == 0)
6355                 status = task->tk_status;
6356         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
6357         if (status == 0 && lgp->res.layoutp->len)
6358                 lseg = pnfs_layout_process(lgp);
6359         rpc_put_task(task);
6360         dprintk("<-- %s status=%d\n", __func__, status);
6361         if (status)
6362                 return ERR_PTR(status);
6363         return lseg;
6364 }
6365
6366 static void
6367 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6368 {
6369         struct nfs4_layoutreturn *lrp = calldata;
6370
6371         dprintk("--> %s\n", __func__);
6372         nfs41_setup_sequence(lrp->clp->cl_session,
6373                         &lrp->args.seq_args,
6374                         &lrp->res.seq_res,
6375                         task);
6376 }
6377
6378 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6379 {
6380         struct nfs4_layoutreturn *lrp = calldata;
6381         struct nfs_server *server;
6382
6383         dprintk("--> %s\n", __func__);
6384
6385         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
6386                 return;
6387
6388         server = NFS_SERVER(lrp->args.inode);
6389         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6390                 rpc_restart_call_prepare(task);
6391                 return;
6392         }
6393         dprintk("<-- %s\n", __func__);
6394 }
6395
6396 static void nfs4_layoutreturn_release(void *calldata)
6397 {
6398         struct nfs4_layoutreturn *lrp = calldata;
6399         struct pnfs_layout_hdr *lo = lrp->args.layout;
6400
6401         dprintk("--> %s\n", __func__);
6402         spin_lock(&lo->plh_inode->i_lock);
6403         if (lrp->res.lrs_present)
6404                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6405         lo->plh_block_lgets--;
6406         spin_unlock(&lo->plh_inode->i_lock);
6407         pnfs_put_layout_hdr(lrp->args.layout);
6408         kfree(calldata);
6409         dprintk("<-- %s\n", __func__);
6410 }
6411
6412 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6413         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6414         .rpc_call_done = nfs4_layoutreturn_done,
6415         .rpc_release = nfs4_layoutreturn_release,
6416 };
6417
6418 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6419 {
6420         struct rpc_task *task;
6421         struct rpc_message msg = {
6422                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6423                 .rpc_argp = &lrp->args,
6424                 .rpc_resp = &lrp->res,
6425         };
6426         struct rpc_task_setup task_setup_data = {
6427                 .rpc_client = lrp->clp->cl_rpcclient,
6428                 .rpc_message = &msg,
6429                 .callback_ops = &nfs4_layoutreturn_call_ops,
6430                 .callback_data = lrp,
6431         };
6432         int status;
6433
6434         dprintk("--> %s\n", __func__);
6435         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6436         task = rpc_run_task(&task_setup_data);
6437         if (IS_ERR(task))
6438                 return PTR_ERR(task);
6439         status = task->tk_status;
6440         dprintk("<-- %s status=%d\n", __func__, status);
6441         rpc_put_task(task);
6442         return status;
6443 }
6444
6445 /*
6446  * Retrieve the list of Data Server devices from the MDS.
6447  */
6448 static int _nfs4_getdevicelist(struct nfs_server *server,
6449                                     const struct nfs_fh *fh,
6450                                     struct pnfs_devicelist *devlist)
6451 {
6452         struct nfs4_getdevicelist_args args = {
6453                 .fh = fh,
6454                 .layoutclass = server->pnfs_curr_ld->id,
6455         };
6456         struct nfs4_getdevicelist_res res = {
6457                 .devlist = devlist,
6458         };
6459         struct rpc_message msg = {
6460                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6461                 .rpc_argp = &args,
6462                 .rpc_resp = &res,
6463         };
6464         int status;
6465
6466         dprintk("--> %s\n", __func__);
6467         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6468                                 &res.seq_res, 0);
6469         dprintk("<-- %s status=%d\n", __func__, status);
6470         return status;
6471 }
6472
6473 int nfs4_proc_getdevicelist(struct nfs_server *server,
6474                             const struct nfs_fh *fh,
6475                             struct pnfs_devicelist *devlist)
6476 {
6477         struct nfs4_exception exception = { };
6478         int err;
6479
6480         do {
6481                 err = nfs4_handle_exception(server,
6482                                 _nfs4_getdevicelist(server, fh, devlist),
6483                                 &exception);
6484         } while (exception.retry);
6485
6486         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6487                 err, devlist->num_devs);
6488
6489         return err;
6490 }
6491 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6492
6493 static int
6494 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6495 {
6496         struct nfs4_getdeviceinfo_args args = {
6497                 .pdev = pdev,
6498         };
6499         struct nfs4_getdeviceinfo_res res = {
6500                 .pdev = pdev,
6501         };
6502         struct rpc_message msg = {
6503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6504                 .rpc_argp = &args,
6505                 .rpc_resp = &res,
6506         };
6507         int status;
6508
6509         dprintk("--> %s\n", __func__);
6510         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6511         dprintk("<-- %s status=%d\n", __func__, status);
6512
6513         return status;
6514 }
6515
6516 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6517 {
6518         struct nfs4_exception exception = { };
6519         int err;
6520
6521         do {
6522                 err = nfs4_handle_exception(server,
6523                                         _nfs4_proc_getdeviceinfo(server, pdev),
6524                                         &exception);
6525         } while (exception.retry);
6526         return err;
6527 }
6528 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6529
6530 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6531 {
6532         struct nfs4_layoutcommit_data *data = calldata;
6533         struct nfs_server *server = NFS_SERVER(data->args.inode);
6534         struct nfs4_session *session = nfs4_get_session(server);
6535
6536         nfs41_setup_sequence(session,
6537                         &data->args.seq_args,
6538                         &data->res.seq_res,
6539                         task);
6540 }
6541
6542 static void
6543 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6544 {
6545         struct nfs4_layoutcommit_data *data = calldata;
6546         struct nfs_server *server = NFS_SERVER(data->args.inode);
6547
6548         if (!nfs41_sequence_done(task, &data->res.seq_res))
6549                 return;
6550
6551         switch (task->tk_status) { /* Just ignore these failures */
6552         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6553         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6554         case -NFS4ERR_BADLAYOUT:     /* no layout */
6555         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6556                 task->tk_status = 0;
6557                 break;
6558         case 0:
6559                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6560                                                    data->res.fattr);
6561                 break;
6562         default:
6563                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6564                         rpc_restart_call_prepare(task);
6565                         return;
6566                 }
6567         }
6568 }
6569
6570 static void nfs4_layoutcommit_release(void *calldata)
6571 {
6572         struct nfs4_layoutcommit_data *data = calldata;
6573
6574         pnfs_cleanup_layoutcommit(data);
6575         put_rpccred(data->cred);
6576         kfree(data);
6577 }
6578
6579 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6580         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6581         .rpc_call_done = nfs4_layoutcommit_done,
6582         .rpc_release = nfs4_layoutcommit_release,
6583 };
6584
6585 int
6586 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6587 {
6588         struct rpc_message msg = {
6589                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6590                 .rpc_argp = &data->args,
6591                 .rpc_resp = &data->res,
6592                 .rpc_cred = data->cred,
6593         };
6594         struct rpc_task_setup task_setup_data = {
6595                 .task = &data->task,
6596                 .rpc_client = NFS_CLIENT(data->args.inode),
6597                 .rpc_message = &msg,
6598                 .callback_ops = &nfs4_layoutcommit_ops,
6599                 .callback_data = data,
6600                 .flags = RPC_TASK_ASYNC,
6601         };
6602         struct rpc_task *task;
6603         int status = 0;
6604
6605         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6606                 "lbw: %llu inode %lu\n",
6607                 data->task.tk_pid, sync,
6608                 data->args.lastbytewritten,
6609                 data->args.inode->i_ino);
6610
6611         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6612         task = rpc_run_task(&task_setup_data);
6613         if (IS_ERR(task))
6614                 return PTR_ERR(task);
6615         if (sync == false)
6616                 goto out;
6617         status = nfs4_wait_for_completion_rpc_task(task);
6618         if (status != 0)
6619                 goto out;
6620         status = task->tk_status;
6621 out:
6622         dprintk("%s: status %d\n", __func__, status);
6623         rpc_put_task(task);
6624         return status;
6625 }
6626
6627 static int
6628 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6629                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6630 {
6631         struct nfs41_secinfo_no_name_args args = {
6632                 .style = SECINFO_STYLE_CURRENT_FH,
6633         };
6634         struct nfs4_secinfo_res res = {
6635                 .flavors = flavors,
6636         };
6637         struct rpc_message msg = {
6638                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6639                 .rpc_argp = &args,
6640                 .rpc_resp = &res,
6641         };
6642         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6643 }
6644
6645 static int
6646 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6647                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6648 {
6649         struct nfs4_exception exception = { };
6650         int err;
6651         do {
6652                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6653                 switch (err) {
6654                 case 0:
6655                 case -NFS4ERR_WRONGSEC:
6656                 case -NFS4ERR_NOTSUPP:
6657                         goto out;
6658                 default:
6659                         err = nfs4_handle_exception(server, err, &exception);
6660                 }
6661         } while (exception.retry);
6662 out:
6663         return err;
6664 }
6665
6666 static int
6667 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6668                     struct nfs_fsinfo *info)
6669 {
6670         int err;
6671         struct page *page;
6672         rpc_authflavor_t flavor;
6673         struct nfs4_secinfo_flavors *flavors;
6674
6675         page = alloc_page(GFP_KERNEL);
6676         if (!page) {
6677                 err = -ENOMEM;
6678                 goto out;
6679         }
6680
6681         flavors = page_address(page);
6682         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6683
6684         /*
6685          * Fall back on "guess and check" method if
6686          * the server doesn't support SECINFO_NO_NAME
6687          */
6688         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6689                 err = nfs4_find_root_sec(server, fhandle, info);
6690                 goto out_freepage;
6691         }
6692         if (err)
6693                 goto out_freepage;
6694
6695         flavor = nfs_find_best_sec(flavors);
6696         if (err == 0)
6697                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6698
6699 out_freepage:
6700         put_page(page);
6701         if (err == -EACCES)
6702                 return -EPERM;
6703 out:
6704         return err;
6705 }
6706
6707 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6708 {
6709         int status;
6710         struct nfs41_test_stateid_args args = {
6711                 .stateid = stateid,
6712         };
6713         struct nfs41_test_stateid_res res;
6714         struct rpc_message msg = {
6715                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6716                 .rpc_argp = &args,
6717                 .rpc_resp = &res,
6718         };
6719
6720         dprintk("NFS call  test_stateid %p\n", stateid);
6721         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6722         nfs4_set_sequence_privileged(&args.seq_args);
6723         status = nfs4_call_sync_sequence(server->client, server, &msg,
6724                         &args.seq_args, &res.seq_res);
6725         if (status != NFS_OK) {
6726                 dprintk("NFS reply test_stateid: failed, %d\n", status);
6727                 return status;
6728         }
6729         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
6730         return -res.status;
6731 }
6732
6733 /**
6734  * nfs41_test_stateid - perform a TEST_STATEID operation
6735  *
6736  * @server: server / transport on which to perform the operation
6737  * @stateid: state ID to test
6738  *
6739  * Returns NFS_OK if the server recognizes that "stateid" is valid.
6740  * Otherwise a negative NFS4ERR value is returned if the operation
6741  * failed or the state ID is not currently valid.
6742  */
6743 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6744 {
6745         struct nfs4_exception exception = { };
6746         int err;
6747         do {
6748                 err = _nfs41_test_stateid(server, stateid);
6749                 if (err != -NFS4ERR_DELAY)
6750                         break;
6751                 nfs4_handle_exception(server, err, &exception);
6752         } while (exception.retry);
6753         return err;
6754 }
6755
6756 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6757 {
6758         struct nfs41_free_stateid_args args = {
6759                 .stateid = stateid,
6760         };
6761         struct nfs41_free_stateid_res res;
6762         struct rpc_message msg = {
6763                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6764                 .rpc_argp = &args,
6765                 .rpc_resp = &res,
6766         };
6767         int status;
6768
6769         dprintk("NFS call  free_stateid %p\n", stateid);
6770         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6771         nfs4_set_sequence_privileged(&args.seq_args);
6772         status = nfs4_call_sync_sequence(server->client, server, &msg,
6773                         &args.seq_args, &res.seq_res);
6774         dprintk("NFS reply free_stateid: %d\n", status);
6775         return status;
6776 }
6777
6778 /**
6779  * nfs41_free_stateid - perform a FREE_STATEID operation
6780  *
6781  * @server: server / transport on which to perform the operation
6782  * @stateid: state ID to release
6783  *
6784  * Returns NFS_OK if the server freed "stateid".  Otherwise a
6785  * negative NFS4ERR value is returned.
6786  */
6787 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6788 {
6789         struct nfs4_exception exception = { };
6790         int err;
6791         do {
6792                 err = _nfs4_free_stateid(server, stateid);
6793                 if (err != -NFS4ERR_DELAY)
6794                         break;
6795                 nfs4_handle_exception(server, err, &exception);
6796         } while (exception.retry);
6797         return err;
6798 }
6799
6800 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6801                 const nfs4_stateid *s2)
6802 {
6803         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6804                 return false;
6805
6806         if (s1->seqid == s2->seqid)
6807                 return true;
6808         if (s1->seqid == 0 || s2->seqid == 0)
6809                 return true;
6810
6811         return false;
6812 }
6813
6814 #endif /* CONFIG_NFS_V4_1 */
6815
6816 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6817                 const nfs4_stateid *s2)
6818 {
6819         return nfs4_stateid_match(s1, s2);
6820 }
6821
6822
6823 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6824         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6825         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6826         .recover_open   = nfs4_open_reclaim,
6827         .recover_lock   = nfs4_lock_reclaim,
6828         .establish_clid = nfs4_init_clientid,
6829         .get_clid_cred  = nfs4_get_setclientid_cred,
6830         .detect_trunking = nfs40_discover_server_trunking,
6831 };
6832
6833 #if defined(CONFIG_NFS_V4_1)
6834 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6835         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6836         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6837         .recover_open   = nfs4_open_reclaim,
6838         .recover_lock   = nfs4_lock_reclaim,
6839         .establish_clid = nfs41_init_clientid,
6840         .get_clid_cred  = nfs4_get_exchange_id_cred,
6841         .reclaim_complete = nfs41_proc_reclaim_complete,
6842         .detect_trunking = nfs41_discover_server_trunking,
6843 };
6844 #endif /* CONFIG_NFS_V4_1 */
6845
6846 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6847         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6848         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6849         .recover_open   = nfs4_open_expired,
6850         .recover_lock   = nfs4_lock_expired,
6851         .establish_clid = nfs4_init_clientid,
6852         .get_clid_cred  = nfs4_get_setclientid_cred,
6853 };
6854
6855 #if defined(CONFIG_NFS_V4_1)
6856 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6857         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6858         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6859         .recover_open   = nfs41_open_expired,
6860         .recover_lock   = nfs41_lock_expired,
6861         .establish_clid = nfs41_init_clientid,
6862         .get_clid_cred  = nfs4_get_exchange_id_cred,
6863 };
6864 #endif /* CONFIG_NFS_V4_1 */
6865
6866 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6867         .sched_state_renewal = nfs4_proc_async_renew,
6868         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6869         .renew_lease = nfs4_proc_renew,
6870 };
6871
6872 #if defined(CONFIG_NFS_V4_1)
6873 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6874         .sched_state_renewal = nfs41_proc_async_sequence,
6875         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6876         .renew_lease = nfs4_proc_sequence,
6877 };
6878 #endif
6879
6880 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6881         .minor_version = 0,
6882         .init_caps = NFS_CAP_READDIRPLUS
6883                 | NFS_CAP_ATOMIC_OPEN
6884                 | NFS_CAP_CHANGE_ATTR
6885                 | NFS_CAP_POSIX_LOCK,
6886         .call_sync = _nfs4_call_sync,
6887         .match_stateid = nfs4_match_stateid,
6888         .find_root_sec = nfs4_find_root_sec,
6889         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6890         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6891         .state_renewal_ops = &nfs40_state_renewal_ops,
6892 };
6893
6894 #if defined(CONFIG_NFS_V4_1)
6895 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6896         .minor_version = 1,
6897         .init_caps = NFS_CAP_READDIRPLUS
6898                 | NFS_CAP_ATOMIC_OPEN
6899                 | NFS_CAP_CHANGE_ATTR
6900                 | NFS_CAP_POSIX_LOCK
6901                 | NFS_CAP_STATEID_NFSV41
6902                 | NFS_CAP_ATOMIC_OPEN_V1,
6903         .call_sync = nfs4_call_sync_sequence,
6904         .match_stateid = nfs41_match_stateid,
6905         .find_root_sec = nfs41_find_root_sec,
6906         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6907         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6908         .state_renewal_ops = &nfs41_state_renewal_ops,
6909 };
6910 #endif
6911
6912 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6913         [0] = &nfs_v4_0_minor_ops,
6914 #if defined(CONFIG_NFS_V4_1)
6915         [1] = &nfs_v4_1_minor_ops,
6916 #endif
6917 };
6918
6919 const struct inode_operations nfs4_dir_inode_operations = {
6920         .create         = nfs_create,
6921         .lookup         = nfs_lookup,
6922         .atomic_open    = nfs_atomic_open,
6923         .link           = nfs_link,
6924         .unlink         = nfs_unlink,
6925         .symlink        = nfs_symlink,
6926         .mkdir          = nfs_mkdir,
6927         .rmdir          = nfs_rmdir,
6928         .mknod          = nfs_mknod,
6929         .rename         = nfs_rename,
6930         .permission     = nfs_permission,
6931         .getattr        = nfs_getattr,
6932         .setattr        = nfs_setattr,
6933         .getxattr       = generic_getxattr,
6934         .setxattr       = generic_setxattr,
6935         .listxattr      = generic_listxattr,
6936         .removexattr    = generic_removexattr,
6937 };
6938
6939 static const struct inode_operations nfs4_file_inode_operations = {
6940         .permission     = nfs_permission,
6941         .getattr        = nfs_getattr,
6942         .setattr        = nfs_setattr,
6943         .getxattr       = generic_getxattr,
6944         .setxattr       = generic_setxattr,
6945         .listxattr      = generic_listxattr,
6946         .removexattr    = generic_removexattr,
6947 };
6948
6949 const struct nfs_rpc_ops nfs_v4_clientops = {
6950         .version        = 4,                    /* protocol version */
6951         .dentry_ops     = &nfs4_dentry_operations,
6952         .dir_inode_ops  = &nfs4_dir_inode_operations,
6953         .file_inode_ops = &nfs4_file_inode_operations,
6954         .file_ops       = &nfs4_file_operations,
6955         .getroot        = nfs4_proc_get_root,
6956         .submount       = nfs4_submount,
6957         .try_mount      = nfs4_try_mount,
6958         .getattr        = nfs4_proc_getattr,
6959         .setattr        = nfs4_proc_setattr,
6960         .lookup         = nfs4_proc_lookup,
6961         .access         = nfs4_proc_access,
6962         .readlink       = nfs4_proc_readlink,
6963         .create         = nfs4_proc_create,
6964         .remove         = nfs4_proc_remove,
6965         .unlink_setup   = nfs4_proc_unlink_setup,
6966         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
6967         .unlink_done    = nfs4_proc_unlink_done,
6968         .rename         = nfs4_proc_rename,
6969         .rename_setup   = nfs4_proc_rename_setup,
6970         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
6971         .rename_done    = nfs4_proc_rename_done,
6972         .link           = nfs4_proc_link,
6973         .symlink        = nfs4_proc_symlink,
6974         .mkdir          = nfs4_proc_mkdir,
6975         .rmdir          = nfs4_proc_remove,
6976         .readdir        = nfs4_proc_readdir,
6977         .mknod          = nfs4_proc_mknod,
6978         .statfs         = nfs4_proc_statfs,
6979         .fsinfo         = nfs4_proc_fsinfo,
6980         .pathconf       = nfs4_proc_pathconf,
6981         .set_capabilities = nfs4_server_capabilities,
6982         .decode_dirent  = nfs4_decode_dirent,
6983         .read_setup     = nfs4_proc_read_setup,
6984         .read_pageio_init = pnfs_pageio_init_read,
6985         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
6986         .read_done      = nfs4_read_done,
6987         .write_setup    = nfs4_proc_write_setup,
6988         .write_pageio_init = pnfs_pageio_init_write,
6989         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
6990         .write_done     = nfs4_write_done,
6991         .commit_setup   = nfs4_proc_commit_setup,
6992         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
6993         .commit_done    = nfs4_commit_done,
6994         .lock           = nfs4_proc_lock,
6995         .clear_acl_cache = nfs4_zap_acl_attr,
6996         .close_context  = nfs4_close_context,
6997         .open_context   = nfs4_atomic_open,
6998         .have_delegation = nfs4_have_delegation,
6999         .return_delegation = nfs4_inode_return_delegation,
7000         .alloc_client   = nfs4_alloc_client,
7001         .init_client    = nfs4_init_client,
7002         .free_client    = nfs4_free_client,
7003         .create_server  = nfs4_create_server,
7004         .clone_server   = nfs_clone_server,
7005 };
7006
7007 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7008         .prefix = XATTR_NAME_NFSV4_ACL,
7009         .list   = nfs4_xattr_list_nfs4_acl,
7010         .get    = nfs4_xattr_get_nfs4_acl,
7011         .set    = nfs4_xattr_set_nfs4_acl,
7012 };
7013
7014 const struct xattr_handler *nfs4_xattr_handlers[] = {
7015         &nfs4_xattr_nfs4_acl_handler,
7016         NULL
7017 };
7018
7019 /*
7020  * Local variables:
7021  *  c-basic-offset: 8
7022  * End:
7023  */