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 -EAGAIN;
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 -EAGAIN;
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         *res = state;
1942 out:
1943         return ret;
1944 }
1945
1946 /*
1947  * Returns a referenced nfs4_state
1948  */
1949 static int _nfs4_do_open(struct inode *dir,
1950                         struct dentry *dentry,
1951                         fmode_t fmode,
1952                         int flags,
1953                         struct iattr *sattr,
1954                         struct rpc_cred *cred,
1955                         struct nfs4_state **res,
1956                         struct nfs4_threshold **ctx_th)
1957 {
1958         struct nfs4_state_owner  *sp;
1959         struct nfs4_state     *state = NULL;
1960         struct nfs_server       *server = NFS_SERVER(dir);
1961         struct nfs4_opendata *opendata;
1962         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
1963         int status;
1964
1965         /* Protect against reboot recovery conflicts */
1966         status = -ENOMEM;
1967         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1968         if (sp == NULL) {
1969                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1970                 goto out_err;
1971         }
1972         status = nfs4_recover_expired_lease(server);
1973         if (status != 0)
1974                 goto err_put_state_owner;
1975         if (dentry->d_inode != NULL)
1976                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1977         status = -ENOMEM;
1978         if (dentry->d_inode)
1979                 claim = NFS4_OPEN_CLAIM_FH;
1980         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
1981                         claim, GFP_KERNEL);
1982         if (opendata == NULL)
1983                 goto err_put_state_owner;
1984
1985         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1986                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1987                 if (!opendata->f_attr.mdsthreshold)
1988                         goto err_opendata_put;
1989                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
1990         }
1991         if (dentry->d_inode != NULL)
1992                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1993
1994         status = _nfs4_open_and_get_state(opendata, fmode, flags, &state);
1995         if (status != 0)
1996                 goto err_opendata_put;
1997
1998         if (opendata->o_arg.open_flags & O_EXCL) {
1999                 nfs4_exclusive_attrset(opendata, sattr);
2000
2001                 nfs_fattr_init(opendata->o_res.f_attr);
2002                 status = nfs4_do_setattr(state->inode, cred,
2003                                 opendata->o_res.f_attr, sattr,
2004                                 state);
2005                 if (status == 0)
2006                         nfs_setattr_update_inode(state->inode, sattr);
2007                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2008         }
2009
2010         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2011                 *ctx_th = opendata->f_attr.mdsthreshold;
2012         else
2013                 kfree(opendata->f_attr.mdsthreshold);
2014         opendata->f_attr.mdsthreshold = NULL;
2015
2016         nfs4_opendata_put(opendata);
2017         nfs4_put_state_owner(sp);
2018         *res = state;
2019         return 0;
2020 err_opendata_put:
2021         kfree(opendata->f_attr.mdsthreshold);
2022         nfs4_opendata_put(opendata);
2023 err_put_state_owner:
2024         nfs4_put_state_owner(sp);
2025 out_err:
2026         *res = NULL;
2027         return status;
2028 }
2029
2030
2031 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2032                                         struct dentry *dentry,
2033                                         fmode_t fmode,
2034                                         int flags,
2035                                         struct iattr *sattr,
2036                                         struct rpc_cred *cred,
2037                                         struct nfs4_threshold **ctx_th)
2038 {
2039         struct nfs_server *server = NFS_SERVER(dir);
2040         struct nfs4_exception exception = { };
2041         struct nfs4_state *res;
2042         int status;
2043
2044         fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
2045         do {
2046                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
2047                                        &res, ctx_th);
2048                 if (status == 0)
2049                         break;
2050                 /* NOTE: BAD_SEQID means the server and client disagree about the
2051                  * book-keeping w.r.t. state-changing operations
2052                  * (OPEN/CLOSE/LOCK/LOCKU...)
2053                  * It is actually a sign of a bug on the client or on the server.
2054                  *
2055                  * If we receive a BAD_SEQID error in the particular case of
2056                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2057                  * have unhashed the old state_owner for us, and that we can
2058                  * therefore safely retry using a new one. We should still warn
2059                  * the user though...
2060                  */
2061                 if (status == -NFS4ERR_BAD_SEQID) {
2062                         pr_warn_ratelimited("NFS: v4 server %s "
2063                                         " returned a bad sequence-id error!\n",
2064                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2065                         exception.retry = 1;
2066                         continue;
2067                 }
2068                 /*
2069                  * BAD_STATEID on OPEN means that the server cancelled our
2070                  * state before it received the OPEN_CONFIRM.
2071                  * Recover by retrying the request as per the discussion
2072                  * on Page 181 of RFC3530.
2073                  */
2074                 if (status == -NFS4ERR_BAD_STATEID) {
2075                         exception.retry = 1;
2076                         continue;
2077                 }
2078                 if (status == -EAGAIN) {
2079                         /* We must have found a delegation */
2080                         exception.retry = 1;
2081                         continue;
2082                 }
2083                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2084                         continue;
2085                 res = ERR_PTR(nfs4_handle_exception(server,
2086                                         status, &exception));
2087         } while (exception.retry);
2088         return res;
2089 }
2090
2091 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2092                             struct nfs_fattr *fattr, struct iattr *sattr,
2093                             struct nfs4_state *state)
2094 {
2095         struct nfs_server *server = NFS_SERVER(inode);
2096         struct nfs_setattrargs  arg = {
2097                 .fh             = NFS_FH(inode),
2098                 .iap            = sattr,
2099                 .server         = server,
2100                 .bitmask = server->attr_bitmask,
2101         };
2102         struct nfs_setattrres  res = {
2103                 .fattr          = fattr,
2104                 .server         = server,
2105         };
2106         struct rpc_message msg = {
2107                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2108                 .rpc_argp       = &arg,
2109                 .rpc_resp       = &res,
2110                 .rpc_cred       = cred,
2111         };
2112         unsigned long timestamp = jiffies;
2113         int status;
2114
2115         nfs_fattr_init(fattr);
2116
2117         if (state != NULL && nfs4_valid_open_stateid(state)) {
2118                 struct nfs_lockowner lockowner = {
2119                         .l_owner = current->files,
2120                         .l_pid = current->tgid,
2121                 };
2122                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2123                                 &lockowner);
2124         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
2125                                 FMODE_WRITE)) {
2126                 /* Use that stateid */
2127         } else
2128                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2129
2130         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2131         if (status == 0 && state != NULL)
2132                 renew_lease(server, timestamp);
2133         return status;
2134 }
2135
2136 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2137                            struct nfs_fattr *fattr, struct iattr *sattr,
2138                            struct nfs4_state *state)
2139 {
2140         struct nfs_server *server = NFS_SERVER(inode);
2141         struct nfs4_exception exception = {
2142                 .state = state,
2143                 .inode = inode,
2144         };
2145         int err;
2146         do {
2147                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2148                 switch (err) {
2149                 case -NFS4ERR_OPENMODE:
2150                         if (state && !(state->state & FMODE_WRITE)) {
2151                                 err = -EBADF;
2152                                 if (sattr->ia_valid & ATTR_OPEN)
2153                                         err = -EACCES;
2154                                 goto out;
2155                         }
2156                 }
2157                 err = nfs4_handle_exception(server, err, &exception);
2158         } while (exception.retry);
2159 out:
2160         return err;
2161 }
2162
2163 struct nfs4_closedata {
2164         struct inode *inode;
2165         struct nfs4_state *state;
2166         struct nfs_closeargs arg;
2167         struct nfs_closeres res;
2168         struct nfs_fattr fattr;
2169         unsigned long timestamp;
2170         bool roc;
2171         u32 roc_barrier;
2172 };
2173
2174 static void nfs4_free_closedata(void *data)
2175 {
2176         struct nfs4_closedata *calldata = data;
2177         struct nfs4_state_owner *sp = calldata->state->owner;
2178         struct super_block *sb = calldata->state->inode->i_sb;
2179
2180         if (calldata->roc)
2181                 pnfs_roc_release(calldata->state->inode);
2182         nfs4_put_open_state(calldata->state);
2183         nfs_free_seqid(calldata->arg.seqid);
2184         nfs4_put_state_owner(sp);
2185         nfs_sb_deactive(sb);
2186         kfree(calldata);
2187 }
2188
2189 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2190                 fmode_t fmode)
2191 {
2192         spin_lock(&state->owner->so_lock);
2193         if (!(fmode & FMODE_READ))
2194                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2195         if (!(fmode & FMODE_WRITE))
2196                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2197         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2198         spin_unlock(&state->owner->so_lock);
2199 }
2200
2201 static void nfs4_close_done(struct rpc_task *task, void *data)
2202 {
2203         struct nfs4_closedata *calldata = data;
2204         struct nfs4_state *state = calldata->state;
2205         struct nfs_server *server = NFS_SERVER(calldata->inode);
2206
2207         dprintk("%s: begin!\n", __func__);
2208         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2209                 return;
2210         /* hmm. we are done with the inode, and in the process of freeing
2211          * the state_owner. we keep this around to process errors
2212          */
2213         switch (task->tk_status) {
2214                 case 0:
2215                         if (calldata->roc)
2216                                 pnfs_roc_set_barrier(state->inode,
2217                                                      calldata->roc_barrier);
2218                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2219                         renew_lease(server, calldata->timestamp);
2220                         nfs4_close_clear_stateid_flags(state,
2221                                         calldata->arg.fmode);
2222                         break;
2223                 case -NFS4ERR_STALE_STATEID:
2224                 case -NFS4ERR_OLD_STATEID:
2225                 case -NFS4ERR_BAD_STATEID:
2226                 case -NFS4ERR_EXPIRED:
2227                         if (calldata->arg.fmode == 0)
2228                                 break;
2229                 default:
2230                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2231                                 rpc_restart_call_prepare(task);
2232         }
2233         nfs_release_seqid(calldata->arg.seqid);
2234         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2235         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2236 }
2237
2238 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2239 {
2240         struct nfs4_closedata *calldata = data;
2241         struct nfs4_state *state = calldata->state;
2242         struct inode *inode = calldata->inode;
2243         int call_close = 0;
2244
2245         dprintk("%s: begin!\n", __func__);
2246         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2247                 goto out_wait;
2248
2249         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2250         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2251         spin_lock(&state->owner->so_lock);
2252         /* Calculate the change in open mode */
2253         if (state->n_rdwr == 0) {
2254                 if (state->n_rdonly == 0) {
2255                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2256                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2257                         calldata->arg.fmode &= ~FMODE_READ;
2258                 }
2259                 if (state->n_wronly == 0) {
2260                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2261                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2262                         calldata->arg.fmode &= ~FMODE_WRITE;
2263                 }
2264         }
2265         if (!nfs4_valid_open_stateid(state))
2266                 call_close = 0;
2267         spin_unlock(&state->owner->so_lock);
2268
2269         if (!call_close) {
2270                 /* Note: exit _without_ calling nfs4_close_done */
2271                 goto out_no_action;
2272         }
2273
2274         if (calldata->arg.fmode == 0) {
2275                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2276                 if (calldata->roc &&
2277                     pnfs_roc_drain(inode, &calldata->roc_barrier, task))
2278                         goto out_wait;
2279         }
2280
2281         nfs_fattr_init(calldata->res.fattr);
2282         calldata->timestamp = jiffies;
2283         if (nfs4_setup_sequence(NFS_SERVER(inode),
2284                                 &calldata->arg.seq_args,
2285                                 &calldata->res.seq_res,
2286                                 task) != 0)
2287                 nfs_release_seqid(calldata->arg.seqid);
2288         dprintk("%s: done!\n", __func__);
2289         return;
2290 out_no_action:
2291         task->tk_action = NULL;
2292 out_wait:
2293         nfs4_sequence_done(task, &calldata->res.seq_res);
2294 }
2295
2296 static const struct rpc_call_ops nfs4_close_ops = {
2297         .rpc_call_prepare = nfs4_close_prepare,
2298         .rpc_call_done = nfs4_close_done,
2299         .rpc_release = nfs4_free_closedata,
2300 };
2301
2302 /*
2303  * It is possible for data to be read/written from a mem-mapped file
2304  * after the sys_close call (which hits the vfs layer as a flush).
2305  * This means that we can't safely call nfsv4 close on a file until
2306  * the inode is cleared. This in turn means that we are not good
2307  * NFSv4 citizens - we do not indicate to the server to update the file's
2308  * share state even when we are done with one of the three share
2309  * stateid's in the inode.
2310  *
2311  * NOTE: Caller must be holding the sp->so_owner semaphore!
2312  */
2313 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2314 {
2315         struct nfs_server *server = NFS_SERVER(state->inode);
2316         struct nfs4_closedata *calldata;
2317         struct nfs4_state_owner *sp = state->owner;
2318         struct rpc_task *task;
2319         struct rpc_message msg = {
2320                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2321                 .rpc_cred = state->owner->so_cred,
2322         };
2323         struct rpc_task_setup task_setup_data = {
2324                 .rpc_client = server->client,
2325                 .rpc_message = &msg,
2326                 .callback_ops = &nfs4_close_ops,
2327                 .workqueue = nfsiod_workqueue,
2328                 .flags = RPC_TASK_ASYNC,
2329         };
2330         int status = -ENOMEM;
2331
2332         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2333         if (calldata == NULL)
2334                 goto out;
2335         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2336         calldata->inode = state->inode;
2337         calldata->state = state;
2338         calldata->arg.fh = NFS_FH(state->inode);
2339         calldata->arg.stateid = &state->open_stateid;
2340         /* Serialization for the sequence id */
2341         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2342         if (calldata->arg.seqid == NULL)
2343                 goto out_free_calldata;
2344         calldata->arg.fmode = 0;
2345         calldata->arg.bitmask = server->cache_consistency_bitmask;
2346         calldata->res.fattr = &calldata->fattr;
2347         calldata->res.seqid = calldata->arg.seqid;
2348         calldata->res.server = server;
2349         calldata->roc = pnfs_roc(state->inode);
2350         nfs_sb_active(calldata->inode->i_sb);
2351
2352         msg.rpc_argp = &calldata->arg;
2353         msg.rpc_resp = &calldata->res;
2354         task_setup_data.callback_data = calldata;
2355         task = rpc_run_task(&task_setup_data);
2356         if (IS_ERR(task))
2357                 return PTR_ERR(task);
2358         status = 0;
2359         if (wait)
2360                 status = rpc_wait_for_completion_task(task);
2361         rpc_put_task(task);
2362         return status;
2363 out_free_calldata:
2364         kfree(calldata);
2365 out:
2366         nfs4_put_open_state(state);
2367         nfs4_put_state_owner(sp);
2368         return status;
2369 }
2370
2371 static struct inode *
2372 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2373 {
2374         struct nfs4_state *state;
2375
2376         /* Protect against concurrent sillydeletes */
2377         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2378                              ctx->cred, &ctx->mdsthreshold);
2379         if (IS_ERR(state))
2380                 return ERR_CAST(state);
2381         ctx->state = state;
2382         return igrab(state->inode);
2383 }
2384
2385 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2386 {
2387         if (ctx->state == NULL)
2388                 return;
2389         if (is_sync)
2390                 nfs4_close_sync(ctx->state, ctx->mode);
2391         else
2392                 nfs4_close_state(ctx->state, ctx->mode);
2393 }
2394
2395 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2396 {
2397         struct nfs4_server_caps_arg args = {
2398                 .fhandle = fhandle,
2399         };
2400         struct nfs4_server_caps_res res = {};
2401         struct rpc_message msg = {
2402                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2403                 .rpc_argp = &args,
2404                 .rpc_resp = &res,
2405         };
2406         int status;
2407
2408         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2409         if (status == 0) {
2410                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2411                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2412                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2413                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2414                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2415                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2416                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2417                         server->caps |= NFS_CAP_ACLS;
2418                 if (res.has_links != 0)
2419                         server->caps |= NFS_CAP_HARDLINKS;
2420                 if (res.has_symlinks != 0)
2421                         server->caps |= NFS_CAP_SYMLINKS;
2422                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2423                         server->caps |= NFS_CAP_FILEID;
2424                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2425                         server->caps |= NFS_CAP_MODE;
2426                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2427                         server->caps |= NFS_CAP_NLINK;
2428                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2429                         server->caps |= NFS_CAP_OWNER;
2430                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2431                         server->caps |= NFS_CAP_OWNER_GROUP;
2432                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2433                         server->caps |= NFS_CAP_ATIME;
2434                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2435                         server->caps |= NFS_CAP_CTIME;
2436                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2437                         server->caps |= NFS_CAP_MTIME;
2438
2439                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2440                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2441                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2442                 server->acl_bitmask = res.acl_bitmask;
2443                 server->fh_expire_type = res.fh_expire_type;
2444         }
2445
2446         return status;
2447 }
2448
2449 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2450 {
2451         struct nfs4_exception exception = { };
2452         int err;
2453         do {
2454                 err = nfs4_handle_exception(server,
2455                                 _nfs4_server_capabilities(server, fhandle),
2456                                 &exception);
2457         } while (exception.retry);
2458         return err;
2459 }
2460
2461 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2462                 struct nfs_fsinfo *info)
2463 {
2464         struct nfs4_lookup_root_arg args = {
2465                 .bitmask = nfs4_fattr_bitmap,
2466         };
2467         struct nfs4_lookup_res res = {
2468                 .server = server,
2469                 .fattr = info->fattr,
2470                 .fh = fhandle,
2471         };
2472         struct rpc_message msg = {
2473                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2474                 .rpc_argp = &args,
2475                 .rpc_resp = &res,
2476         };
2477
2478         nfs_fattr_init(info->fattr);
2479         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2480 }
2481
2482 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2483                 struct nfs_fsinfo *info)
2484 {
2485         struct nfs4_exception exception = { };
2486         int err;
2487         do {
2488                 err = _nfs4_lookup_root(server, fhandle, info);
2489                 switch (err) {
2490                 case 0:
2491                 case -NFS4ERR_WRONGSEC:
2492                         goto out;
2493                 default:
2494                         err = nfs4_handle_exception(server, err, &exception);
2495                 }
2496         } while (exception.retry);
2497 out:
2498         return err;
2499 }
2500
2501 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2502                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2503 {
2504         struct rpc_auth *auth;
2505         int ret;
2506
2507         auth = rpcauth_create(flavor, server->client);
2508         if (IS_ERR(auth)) {
2509                 ret = -EACCES;
2510                 goto out;
2511         }
2512         ret = nfs4_lookup_root(server, fhandle, info);
2513 out:
2514         return ret;
2515 }
2516
2517 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2518                               struct nfs_fsinfo *info)
2519 {
2520         int i, len, status = 0;
2521         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2522
2523         len = rpcauth_list_flavors(flav_array, ARRAY_SIZE(flav_array));
2524         if (len < 0)
2525                 return len;
2526
2527         for (i = 0; i < len; i++) {
2528                 /* AUTH_UNIX is the default flavor if none was specified,
2529                  * thus has already been tried. */
2530                 if (flav_array[i] == RPC_AUTH_UNIX)
2531                         continue;
2532
2533                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2534                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2535                         continue;
2536                 break;
2537         }
2538         /*
2539          * -EACCESS could mean that the user doesn't have correct permissions
2540          * to access the mount.  It could also mean that we tried to mount
2541          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2542          * existing mount programs don't handle -EACCES very well so it should
2543          * be mapped to -EPERM instead.
2544          */
2545         if (status == -EACCES)
2546                 status = -EPERM;
2547         return status;
2548 }
2549
2550 static int nfs4_do_find_root_sec(struct nfs_server *server,
2551                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
2552 {
2553         int mv = server->nfs_client->cl_minorversion;
2554         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
2555 }
2556
2557 /**
2558  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
2559  * @server: initialized nfs_server handle
2560  * @fhandle: we fill in the pseudo-fs root file handle
2561  * @info: we fill in an FSINFO struct
2562  *
2563  * Returns zero on success, or a negative errno.
2564  */
2565 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2566                          struct nfs_fsinfo *info)
2567 {
2568         int status;
2569
2570         status = nfs4_lookup_root(server, fhandle, info);
2571         if ((status == -NFS4ERR_WRONGSEC) &&
2572             !(server->flags & NFS_MOUNT_SECFLAVOUR))
2573                 status = nfs4_do_find_root_sec(server, fhandle, info);
2574
2575         if (status == 0)
2576                 status = nfs4_server_capabilities(server, fhandle);
2577         if (status == 0)
2578                 status = nfs4_do_fsinfo(server, fhandle, info);
2579
2580         return nfs4_map_errors(status);
2581 }
2582
2583 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2584                               struct nfs_fsinfo *info)
2585 {
2586         int error;
2587         struct nfs_fattr *fattr = info->fattr;
2588
2589         error = nfs4_server_capabilities(server, mntfh);
2590         if (error < 0) {
2591                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2592                 return error;
2593         }
2594
2595         error = nfs4_proc_getattr(server, mntfh, fattr);
2596         if (error < 0) {
2597                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2598                 return error;
2599         }
2600
2601         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2602             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2603                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2604
2605         return error;
2606 }
2607
2608 /*
2609  * Get locations and (maybe) other attributes of a referral.
2610  * Note that we'll actually follow the referral later when
2611  * we detect fsid mismatch in inode revalidation
2612  */
2613 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2614                              const struct qstr *name, struct nfs_fattr *fattr,
2615                              struct nfs_fh *fhandle)
2616 {
2617         int status = -ENOMEM;
2618         struct page *page = NULL;
2619         struct nfs4_fs_locations *locations = NULL;
2620
2621         page = alloc_page(GFP_KERNEL);
2622         if (page == NULL)
2623                 goto out;
2624         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2625         if (locations == NULL)
2626                 goto out;
2627
2628         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2629         if (status != 0)
2630                 goto out;
2631         /* Make sure server returned a different fsid for the referral */
2632         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2633                 dprintk("%s: server did not return a different fsid for"
2634                         " a referral at %s\n", __func__, name->name);
2635                 status = -EIO;
2636                 goto out;
2637         }
2638         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2639         nfs_fixup_referral_attributes(&locations->fattr);
2640
2641         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2642         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2643         memset(fhandle, 0, sizeof(struct nfs_fh));
2644 out:
2645         if (page)
2646                 __free_page(page);
2647         kfree(locations);
2648         return status;
2649 }
2650
2651 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2652 {
2653         struct nfs4_getattr_arg args = {
2654                 .fh = fhandle,
2655                 .bitmask = server->attr_bitmask,
2656         };
2657         struct nfs4_getattr_res res = {
2658                 .fattr = fattr,
2659                 .server = server,
2660         };
2661         struct rpc_message msg = {
2662                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2663                 .rpc_argp = &args,
2664                 .rpc_resp = &res,
2665         };
2666
2667         nfs_fattr_init(fattr);
2668         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2669 }
2670
2671 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2672 {
2673         struct nfs4_exception exception = { };
2674         int err;
2675         do {
2676                 err = nfs4_handle_exception(server,
2677                                 _nfs4_proc_getattr(server, fhandle, fattr),
2678                                 &exception);
2679         } while (exception.retry);
2680         return err;
2681 }
2682
2683 /*
2684  * The file is not closed if it is opened due to the a request to change
2685  * the size of the file. The open call will not be needed once the
2686  * VFS layer lookup-intents are implemented.
2687  *
2688  * Close is called when the inode is destroyed.
2689  * If we haven't opened the file for O_WRONLY, we
2690  * need to in the size_change case to obtain a stateid.
2691  *
2692  * Got race?
2693  * Because OPEN is always done by name in nfsv4, it is
2694  * possible that we opened a different file by the same
2695  * name.  We can recognize this race condition, but we
2696  * can't do anything about it besides returning an error.
2697  *
2698  * This will be fixed with VFS changes (lookup-intent).
2699  */
2700 static int
2701 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2702                   struct iattr *sattr)
2703 {
2704         struct inode *inode = dentry->d_inode;
2705         struct rpc_cred *cred = NULL;
2706         struct nfs4_state *state = NULL;
2707         int status;
2708
2709         if (pnfs_ld_layoutret_on_setattr(inode))
2710                 pnfs_commit_and_return_layout(inode);
2711
2712         nfs_fattr_init(fattr);
2713
2714         /* Deal with open(O_TRUNC) */
2715         if (sattr->ia_valid & ATTR_OPEN)
2716                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2717
2718         /* Optimization: if the end result is no change, don't RPC */
2719         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2720                 return 0;
2721
2722         /* Search for an existing open(O_WRITE) file */
2723         if (sattr->ia_valid & ATTR_FILE) {
2724                 struct nfs_open_context *ctx;
2725
2726                 ctx = nfs_file_open_context(sattr->ia_file);
2727                 if (ctx) {
2728                         cred = ctx->cred;
2729                         state = ctx->state;
2730                 }
2731         }
2732
2733         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2734         if (status == 0)
2735                 nfs_setattr_update_inode(inode, sattr);
2736         return status;
2737 }
2738
2739 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2740                 const struct qstr *name, struct nfs_fh *fhandle,
2741                 struct nfs_fattr *fattr)
2742 {
2743         struct nfs_server *server = NFS_SERVER(dir);
2744         int                    status;
2745         struct nfs4_lookup_arg args = {
2746                 .bitmask = server->attr_bitmask,
2747                 .dir_fh = NFS_FH(dir),
2748                 .name = name,
2749         };
2750         struct nfs4_lookup_res res = {
2751                 .server = server,
2752                 .fattr = fattr,
2753                 .fh = fhandle,
2754         };
2755         struct rpc_message msg = {
2756                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2757                 .rpc_argp = &args,
2758                 .rpc_resp = &res,
2759         };
2760
2761         nfs_fattr_init(fattr);
2762
2763         dprintk("NFS call  lookup %s\n", name->name);
2764         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2765         dprintk("NFS reply lookup: %d\n", status);
2766         return status;
2767 }
2768
2769 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2770 {
2771         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2772                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2773         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2774         fattr->nlink = 2;
2775 }
2776
2777 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2778                                    struct qstr *name, struct nfs_fh *fhandle,
2779                                    struct nfs_fattr *fattr)
2780 {
2781         struct nfs4_exception exception = { };
2782         struct rpc_clnt *client = *clnt;
2783         int err;
2784         do {
2785                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2786                 switch (err) {
2787                 case -NFS4ERR_BADNAME:
2788                         err = -ENOENT;
2789                         goto out;
2790                 case -NFS4ERR_MOVED:
2791                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2792                         goto out;
2793                 case -NFS4ERR_WRONGSEC:
2794                         err = -EPERM;
2795                         if (client != *clnt)
2796                                 goto out;
2797
2798                         client = nfs4_create_sec_client(client, dir, name);
2799                         if (IS_ERR(client))
2800                                 return PTR_ERR(client);
2801
2802                         exception.retry = 1;
2803                         break;
2804                 default:
2805                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2806                 }
2807         } while (exception.retry);
2808
2809 out:
2810         if (err == 0)
2811                 *clnt = client;
2812         else if (client != *clnt)
2813                 rpc_shutdown_client(client);
2814
2815         return err;
2816 }
2817
2818 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2819                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2820 {
2821         int status;
2822         struct rpc_clnt *client = NFS_CLIENT(dir);
2823
2824         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2825         if (client != NFS_CLIENT(dir)) {
2826                 rpc_shutdown_client(client);
2827                 nfs_fixup_secinfo_attributes(fattr);
2828         }
2829         return status;
2830 }
2831
2832 struct rpc_clnt *
2833 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2834                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2835 {
2836         int status;
2837         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2838
2839         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2840         if (status < 0) {
2841                 rpc_shutdown_client(client);
2842                 return ERR_PTR(status);
2843         }
2844         return client;
2845 }
2846
2847 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2848 {
2849         struct nfs_server *server = NFS_SERVER(inode);
2850         struct nfs4_accessargs args = {
2851                 .fh = NFS_FH(inode),
2852                 .bitmask = server->cache_consistency_bitmask,
2853         };
2854         struct nfs4_accessres res = {
2855                 .server = server,
2856         };
2857         struct rpc_message msg = {
2858                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2859                 .rpc_argp = &args,
2860                 .rpc_resp = &res,
2861                 .rpc_cred = entry->cred,
2862         };
2863         int mode = entry->mask;
2864         int status;
2865
2866         /*
2867          * Determine which access bits we want to ask for...
2868          */
2869         if (mode & MAY_READ)
2870                 args.access |= NFS4_ACCESS_READ;
2871         if (S_ISDIR(inode->i_mode)) {
2872                 if (mode & MAY_WRITE)
2873                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2874                 if (mode & MAY_EXEC)
2875                         args.access |= NFS4_ACCESS_LOOKUP;
2876         } else {
2877                 if (mode & MAY_WRITE)
2878                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2879                 if (mode & MAY_EXEC)
2880                         args.access |= NFS4_ACCESS_EXECUTE;
2881         }
2882
2883         res.fattr = nfs_alloc_fattr();
2884         if (res.fattr == NULL)
2885                 return -ENOMEM;
2886
2887         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2888         if (!status) {
2889                 nfs_access_set_mask(entry, res.access);
2890                 nfs_refresh_inode(inode, res.fattr);
2891         }
2892         nfs_free_fattr(res.fattr);
2893         return status;
2894 }
2895
2896 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2897 {
2898         struct nfs4_exception exception = { };
2899         int err;
2900         do {
2901                 err = nfs4_handle_exception(NFS_SERVER(inode),
2902                                 _nfs4_proc_access(inode, entry),
2903                                 &exception);
2904         } while (exception.retry);
2905         return err;
2906 }
2907
2908 /*
2909  * TODO: For the time being, we don't try to get any attributes
2910  * along with any of the zero-copy operations READ, READDIR,
2911  * READLINK, WRITE.
2912  *
2913  * In the case of the first three, we want to put the GETATTR
2914  * after the read-type operation -- this is because it is hard
2915  * to predict the length of a GETATTR response in v4, and thus
2916  * align the READ data correctly.  This means that the GETATTR
2917  * may end up partially falling into the page cache, and we should
2918  * shift it into the 'tail' of the xdr_buf before processing.
2919  * To do this efficiently, we need to know the total length
2920  * of data received, which doesn't seem to be available outside
2921  * of the RPC layer.
2922  *
2923  * In the case of WRITE, we also want to put the GETATTR after
2924  * the operation -- in this case because we want to make sure
2925  * we get the post-operation mtime and size.
2926  *
2927  * Both of these changes to the XDR layer would in fact be quite
2928  * minor, but I decided to leave them for a subsequent patch.
2929  */
2930 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2931                 unsigned int pgbase, unsigned int pglen)
2932 {
2933         struct nfs4_readlink args = {
2934                 .fh       = NFS_FH(inode),
2935                 .pgbase   = pgbase,
2936                 .pglen    = pglen,
2937                 .pages    = &page,
2938         };
2939         struct nfs4_readlink_res res;
2940         struct rpc_message msg = {
2941                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2942                 .rpc_argp = &args,
2943                 .rpc_resp = &res,
2944         };
2945
2946         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2947 }
2948
2949 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2950                 unsigned int pgbase, unsigned int pglen)
2951 {
2952         struct nfs4_exception exception = { };
2953         int err;
2954         do {
2955                 err = nfs4_handle_exception(NFS_SERVER(inode),
2956                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2957                                 &exception);
2958         } while (exception.retry);
2959         return err;
2960 }
2961
2962 /*
2963  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
2964  */
2965 static int
2966 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2967                  int flags)
2968 {
2969         struct nfs_open_context *ctx;
2970         struct nfs4_state *state;
2971         int status = 0;
2972
2973         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
2974         if (IS_ERR(ctx))
2975                 return PTR_ERR(ctx);
2976
2977         sattr->ia_mode &= ~current_umask();
2978         state = nfs4_do_open(dir, dentry, ctx->mode,
2979                         flags, sattr, ctx->cred,
2980                         &ctx->mdsthreshold);
2981         d_drop(dentry);
2982         if (IS_ERR(state)) {
2983                 status = PTR_ERR(state);
2984                 goto out;
2985         }
2986         d_add(dentry, igrab(state->inode));
2987         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2988         ctx->state = state;
2989 out:
2990         put_nfs_open_context(ctx);
2991         return status;
2992 }
2993
2994 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2995 {
2996         struct nfs_server *server = NFS_SERVER(dir);
2997         struct nfs_removeargs args = {
2998                 .fh = NFS_FH(dir),
2999                 .name = *name,
3000         };
3001         struct nfs_removeres res = {
3002                 .server = server,
3003         };
3004         struct rpc_message msg = {
3005                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3006                 .rpc_argp = &args,
3007                 .rpc_resp = &res,
3008         };
3009         int status;
3010
3011         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3012         if (status == 0)
3013                 update_changeattr(dir, &res.cinfo);
3014         return status;
3015 }
3016
3017 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3018 {
3019         struct nfs4_exception exception = { };
3020         int err;
3021         do {
3022                 err = nfs4_handle_exception(NFS_SERVER(dir),
3023                                 _nfs4_proc_remove(dir, name),
3024                                 &exception);
3025         } while (exception.retry);
3026         return err;
3027 }
3028
3029 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3030 {
3031         struct nfs_server *server = NFS_SERVER(dir);
3032         struct nfs_removeargs *args = msg->rpc_argp;
3033         struct nfs_removeres *res = msg->rpc_resp;
3034
3035         res->server = server;
3036         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3037         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3038 }
3039
3040 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3041 {
3042         nfs4_setup_sequence(NFS_SERVER(data->dir),
3043                         &data->args.seq_args,
3044                         &data->res.seq_res,
3045                         task);
3046 }
3047
3048 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3049 {
3050         struct nfs_removeres *res = task->tk_msg.rpc_resp;
3051
3052         if (!nfs4_sequence_done(task, &res->seq_res))
3053                 return 0;
3054         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3055                 return 0;
3056         update_changeattr(dir, &res->cinfo);
3057         return 1;
3058 }
3059
3060 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3061 {
3062         struct nfs_server *server = NFS_SERVER(dir);
3063         struct nfs_renameargs *arg = msg->rpc_argp;
3064         struct nfs_renameres *res = msg->rpc_resp;
3065
3066         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3067         res->server = server;
3068         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3069 }
3070
3071 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3072 {
3073         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3074                         &data->args.seq_args,
3075                         &data->res.seq_res,
3076                         task);
3077 }
3078
3079 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3080                                  struct inode *new_dir)
3081 {
3082         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3083
3084         if (!nfs4_sequence_done(task, &res->seq_res))
3085                 return 0;
3086         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3087                 return 0;
3088
3089         update_changeattr(old_dir, &res->old_cinfo);
3090         update_changeattr(new_dir, &res->new_cinfo);
3091         return 1;
3092 }
3093
3094 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3095                 struct inode *new_dir, struct qstr *new_name)
3096 {
3097         struct nfs_server *server = NFS_SERVER(old_dir);
3098         struct nfs_renameargs arg = {
3099                 .old_dir = NFS_FH(old_dir),
3100                 .new_dir = NFS_FH(new_dir),
3101                 .old_name = old_name,
3102                 .new_name = new_name,
3103         };
3104         struct nfs_renameres res = {
3105                 .server = server,
3106         };
3107         struct rpc_message msg = {
3108                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3109                 .rpc_argp = &arg,
3110                 .rpc_resp = &res,
3111         };
3112         int status = -ENOMEM;
3113
3114         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3115         if (!status) {
3116                 update_changeattr(old_dir, &res.old_cinfo);
3117                 update_changeattr(new_dir, &res.new_cinfo);
3118         }
3119         return status;
3120 }
3121
3122 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3123                 struct inode *new_dir, struct qstr *new_name)
3124 {
3125         struct nfs4_exception exception = { };
3126         int err;
3127         do {
3128                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3129                                 _nfs4_proc_rename(old_dir, old_name,
3130                                         new_dir, new_name),
3131                                 &exception);
3132         } while (exception.retry);
3133         return err;
3134 }
3135
3136 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3137 {
3138         struct nfs_server *server = NFS_SERVER(inode);
3139         struct nfs4_link_arg arg = {
3140                 .fh     = NFS_FH(inode),
3141                 .dir_fh = NFS_FH(dir),
3142                 .name   = name,
3143                 .bitmask = server->attr_bitmask,
3144         };
3145         struct nfs4_link_res res = {
3146                 .server = server,
3147         };
3148         struct rpc_message msg = {
3149                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3150                 .rpc_argp = &arg,
3151                 .rpc_resp = &res,
3152         };
3153         int status = -ENOMEM;
3154
3155         res.fattr = nfs_alloc_fattr();
3156         if (res.fattr == NULL)
3157                 goto out;
3158
3159         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3160         if (!status) {
3161                 update_changeattr(dir, &res.cinfo);
3162                 nfs_post_op_update_inode(inode, res.fattr);
3163         }
3164 out:
3165         nfs_free_fattr(res.fattr);
3166         return status;
3167 }
3168
3169 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3170 {
3171         struct nfs4_exception exception = { };
3172         int err;
3173         do {
3174                 err = nfs4_handle_exception(NFS_SERVER(inode),
3175                                 _nfs4_proc_link(inode, dir, name),
3176                                 &exception);
3177         } while (exception.retry);
3178         return err;
3179 }
3180
3181 struct nfs4_createdata {
3182         struct rpc_message msg;
3183         struct nfs4_create_arg arg;
3184         struct nfs4_create_res res;
3185         struct nfs_fh fh;
3186         struct nfs_fattr fattr;
3187 };
3188
3189 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3190                 struct qstr *name, struct iattr *sattr, u32 ftype)
3191 {
3192         struct nfs4_createdata *data;
3193
3194         data = kzalloc(sizeof(*data), GFP_KERNEL);
3195         if (data != NULL) {
3196                 struct nfs_server *server = NFS_SERVER(dir);
3197
3198                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3199                 data->msg.rpc_argp = &data->arg;
3200                 data->msg.rpc_resp = &data->res;
3201                 data->arg.dir_fh = NFS_FH(dir);
3202                 data->arg.server = server;
3203                 data->arg.name = name;
3204                 data->arg.attrs = sattr;
3205                 data->arg.ftype = ftype;
3206                 data->arg.bitmask = server->attr_bitmask;
3207                 data->res.server = server;
3208                 data->res.fh = &data->fh;
3209                 data->res.fattr = &data->fattr;
3210                 nfs_fattr_init(data->res.fattr);
3211         }
3212         return data;
3213 }
3214
3215 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3216 {
3217         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3218                                     &data->arg.seq_args, &data->res.seq_res, 1);
3219         if (status == 0) {
3220                 update_changeattr(dir, &data->res.dir_cinfo);
3221                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3222         }
3223         return status;
3224 }
3225
3226 static void nfs4_free_createdata(struct nfs4_createdata *data)
3227 {
3228         kfree(data);
3229 }
3230
3231 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3232                 struct page *page, unsigned int len, struct iattr *sattr)
3233 {
3234         struct nfs4_createdata *data;
3235         int status = -ENAMETOOLONG;
3236
3237         if (len > NFS4_MAXPATHLEN)
3238                 goto out;
3239
3240         status = -ENOMEM;
3241         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3242         if (data == NULL)
3243                 goto out;
3244
3245         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3246         data->arg.u.symlink.pages = &page;
3247         data->arg.u.symlink.len = len;
3248
3249         status = nfs4_do_create(dir, dentry, data);
3250
3251         nfs4_free_createdata(data);
3252 out:
3253         return status;
3254 }
3255
3256 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3257                 struct page *page, unsigned int len, struct iattr *sattr)
3258 {
3259         struct nfs4_exception exception = { };
3260         int err;
3261         do {
3262                 err = nfs4_handle_exception(NFS_SERVER(dir),
3263                                 _nfs4_proc_symlink(dir, dentry, page,
3264                                                         len, sattr),
3265                                 &exception);
3266         } while (exception.retry);
3267         return err;
3268 }
3269
3270 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3271                 struct iattr *sattr)
3272 {
3273         struct nfs4_createdata *data;
3274         int status = -ENOMEM;
3275
3276         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3277         if (data == NULL)
3278                 goto out;
3279
3280         status = nfs4_do_create(dir, dentry, data);
3281
3282         nfs4_free_createdata(data);
3283 out:
3284         return status;
3285 }
3286
3287 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3288                 struct iattr *sattr)
3289 {
3290         struct nfs4_exception exception = { };
3291         int err;
3292
3293         sattr->ia_mode &= ~current_umask();
3294         do {
3295                 err = nfs4_handle_exception(NFS_SERVER(dir),
3296                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3297                                 &exception);
3298         } while (exception.retry);
3299         return err;
3300 }
3301
3302 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3303                 u64 cookie, struct page **pages, unsigned int count, int plus)
3304 {
3305         struct inode            *dir = dentry->d_inode;
3306         struct nfs4_readdir_arg args = {
3307                 .fh = NFS_FH(dir),
3308                 .pages = pages,
3309                 .pgbase = 0,
3310                 .count = count,
3311                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3312                 .plus = plus,
3313         };
3314         struct nfs4_readdir_res res;
3315         struct rpc_message msg = {
3316                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3317                 .rpc_argp = &args,
3318                 .rpc_resp = &res,
3319                 .rpc_cred = cred,
3320         };
3321         int                     status;
3322
3323         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3324                         dentry->d_parent->d_name.name,
3325                         dentry->d_name.name,
3326                         (unsigned long long)cookie);
3327         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3328         res.pgbase = args.pgbase;
3329         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3330         if (status >= 0) {
3331                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3332                 status += args.pgbase;
3333         }
3334
3335         nfs_invalidate_atime(dir);
3336
3337         dprintk("%s: returns %d\n", __func__, status);
3338         return status;
3339 }
3340
3341 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3342                 u64 cookie, struct page **pages, unsigned int count, int plus)
3343 {
3344         struct nfs4_exception exception = { };
3345         int err;
3346         do {
3347                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3348                                 _nfs4_proc_readdir(dentry, cred, cookie,
3349                                         pages, count, plus),
3350                                 &exception);
3351         } while (exception.retry);
3352         return err;
3353 }
3354
3355 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3356                 struct iattr *sattr, dev_t rdev)
3357 {
3358         struct nfs4_createdata *data;
3359         int mode = sattr->ia_mode;
3360         int status = -ENOMEM;
3361
3362         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3363         if (data == NULL)
3364                 goto out;
3365
3366         if (S_ISFIFO(mode))
3367                 data->arg.ftype = NF4FIFO;
3368         else if (S_ISBLK(mode)) {
3369                 data->arg.ftype = NF4BLK;
3370                 data->arg.u.device.specdata1 = MAJOR(rdev);
3371                 data->arg.u.device.specdata2 = MINOR(rdev);
3372         }
3373         else if (S_ISCHR(mode)) {
3374                 data->arg.ftype = NF4CHR;
3375                 data->arg.u.device.specdata1 = MAJOR(rdev);
3376                 data->arg.u.device.specdata2 = MINOR(rdev);
3377         } else if (!S_ISSOCK(mode)) {
3378                 status = -EINVAL;
3379                 goto out_free;
3380         }
3381
3382         status = nfs4_do_create(dir, dentry, data);
3383 out_free:
3384         nfs4_free_createdata(data);
3385 out:
3386         return status;
3387 }
3388
3389 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3390                 struct iattr *sattr, dev_t rdev)
3391 {
3392         struct nfs4_exception exception = { };
3393         int err;
3394
3395         sattr->ia_mode &= ~current_umask();
3396         do {
3397                 err = nfs4_handle_exception(NFS_SERVER(dir),
3398                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3399                                 &exception);
3400         } while (exception.retry);
3401         return err;
3402 }
3403
3404 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3405                  struct nfs_fsstat *fsstat)
3406 {
3407         struct nfs4_statfs_arg args = {
3408                 .fh = fhandle,
3409                 .bitmask = server->attr_bitmask,
3410         };
3411         struct nfs4_statfs_res res = {
3412                 .fsstat = fsstat,
3413         };
3414         struct rpc_message msg = {
3415                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3416                 .rpc_argp = &args,
3417                 .rpc_resp = &res,
3418         };
3419
3420         nfs_fattr_init(fsstat->fattr);
3421         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3422 }
3423
3424 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3425 {
3426         struct nfs4_exception exception = { };
3427         int err;
3428         do {
3429                 err = nfs4_handle_exception(server,
3430                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3431                                 &exception);
3432         } while (exception.retry);
3433         return err;
3434 }
3435
3436 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3437                 struct nfs_fsinfo *fsinfo)
3438 {
3439         struct nfs4_fsinfo_arg args = {
3440                 .fh = fhandle,
3441                 .bitmask = server->attr_bitmask,
3442         };
3443         struct nfs4_fsinfo_res res = {
3444                 .fsinfo = fsinfo,
3445         };
3446         struct rpc_message msg = {
3447                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3448                 .rpc_argp = &args,
3449                 .rpc_resp = &res,
3450         };
3451
3452         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3453 }
3454
3455 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3456 {
3457         struct nfs4_exception exception = { };
3458         int err;
3459
3460         do {
3461                 err = nfs4_handle_exception(server,
3462                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3463                                 &exception);
3464         } while (exception.retry);
3465         return err;
3466 }
3467
3468 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3469 {
3470         int error;
3471
3472         nfs_fattr_init(fsinfo->fattr);
3473         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3474         if (error == 0) {
3475                 /* block layout checks this! */
3476                 server->pnfs_blksize = fsinfo->blksize;
3477                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3478         }
3479
3480         return error;
3481 }
3482
3483 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3484                 struct nfs_pathconf *pathconf)
3485 {
3486         struct nfs4_pathconf_arg args = {
3487                 .fh = fhandle,
3488                 .bitmask = server->attr_bitmask,
3489         };
3490         struct nfs4_pathconf_res res = {
3491                 .pathconf = pathconf,
3492         };
3493         struct rpc_message msg = {
3494                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3495                 .rpc_argp = &args,
3496                 .rpc_resp = &res,
3497         };
3498
3499         /* None of the pathconf attributes are mandatory to implement */
3500         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3501                 memset(pathconf, 0, sizeof(*pathconf));
3502                 return 0;
3503         }
3504
3505         nfs_fattr_init(pathconf->fattr);
3506         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3507 }
3508
3509 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3510                 struct nfs_pathconf *pathconf)
3511 {
3512         struct nfs4_exception exception = { };
3513         int err;
3514
3515         do {
3516                 err = nfs4_handle_exception(server,
3517                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3518                                 &exception);
3519         } while (exception.retry);
3520         return err;
3521 }
3522
3523 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
3524                 const struct nfs_open_context *ctx,
3525                 const struct nfs_lock_context *l_ctx,
3526                 fmode_t fmode)
3527 {
3528         const struct nfs_lockowner *lockowner = NULL;
3529
3530         if (l_ctx != NULL)
3531                 lockowner = &l_ctx->lockowner;
3532         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
3533 }
3534 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
3535
3536 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
3537                 const struct nfs_open_context *ctx,
3538                 const struct nfs_lock_context *l_ctx,
3539                 fmode_t fmode)
3540 {
3541         nfs4_stateid current_stateid;
3542
3543         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode))
3544                 return false;
3545         return nfs4_stateid_match(stateid, &current_stateid);
3546 }
3547
3548 static bool nfs4_error_stateid_expired(int err)
3549 {
3550         switch (err) {
3551         case -NFS4ERR_DELEG_REVOKED:
3552         case -NFS4ERR_ADMIN_REVOKED:
3553         case -NFS4ERR_BAD_STATEID:
3554         case -NFS4ERR_STALE_STATEID:
3555         case -NFS4ERR_OLD_STATEID:
3556         case -NFS4ERR_OPENMODE:
3557         case -NFS4ERR_EXPIRED:
3558                 return true;
3559         }
3560         return false;
3561 }
3562
3563 void __nfs4_read_done_cb(struct nfs_read_data *data)
3564 {
3565         nfs_invalidate_atime(data->header->inode);
3566 }
3567
3568 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3569 {
3570         struct nfs_server *server = NFS_SERVER(data->header->inode);
3571
3572         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3573                 rpc_restart_call_prepare(task);
3574                 return -EAGAIN;
3575         }
3576
3577         __nfs4_read_done_cb(data);
3578         if (task->tk_status > 0)
3579                 renew_lease(server, data->timestamp);
3580         return 0;
3581 }
3582
3583 static bool nfs4_read_stateid_changed(struct rpc_task *task,
3584                 struct nfs_readargs *args)
3585 {
3586
3587         if (!nfs4_error_stateid_expired(task->tk_status) ||
3588                 nfs4_stateid_is_current(&args->stateid,
3589                                 args->context,
3590                                 args->lock_context,
3591                                 FMODE_READ))
3592                 return false;
3593         rpc_restart_call_prepare(task);
3594         return true;
3595 }
3596
3597 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3598 {
3599
3600         dprintk("--> %s\n", __func__);
3601
3602         if (!nfs4_sequence_done(task, &data->res.seq_res))
3603                 return -EAGAIN;
3604         if (nfs4_read_stateid_changed(task, &data->args))
3605                 return -EAGAIN;
3606         return data->read_done_cb ? data->read_done_cb(task, data) :
3607                                     nfs4_read_done_cb(task, data);
3608 }
3609
3610 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3611 {
3612         data->timestamp   = jiffies;
3613         data->read_done_cb = nfs4_read_done_cb;
3614         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3615         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3616 }
3617
3618 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3619 {
3620         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3621                         &data->args.seq_args,
3622                         &data->res.seq_res,
3623                         task))
3624                 return;
3625         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3626                         data->args.lock_context, FMODE_READ);
3627 }
3628
3629 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3630 {
3631         struct inode *inode = data->header->inode;
3632
3633         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3634                 rpc_restart_call_prepare(task);
3635                 return -EAGAIN;
3636         }
3637         if (task->tk_status >= 0) {
3638                 renew_lease(NFS_SERVER(inode), data->timestamp);
3639                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3640         }
3641         return 0;
3642 }
3643
3644 static bool nfs4_write_stateid_changed(struct rpc_task *task,
3645                 struct nfs_writeargs *args)
3646 {
3647
3648         if (!nfs4_error_stateid_expired(task->tk_status) ||
3649                 nfs4_stateid_is_current(&args->stateid,
3650                                 args->context,
3651                                 args->lock_context,
3652                                 FMODE_WRITE))
3653                 return false;
3654         rpc_restart_call_prepare(task);
3655         return true;
3656 }
3657
3658 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3659 {
3660         if (!nfs4_sequence_done(task, &data->res.seq_res))
3661                 return -EAGAIN;
3662         if (nfs4_write_stateid_changed(task, &data->args))
3663                 return -EAGAIN;
3664         return data->write_done_cb ? data->write_done_cb(task, data) :
3665                 nfs4_write_done_cb(task, data);
3666 }
3667
3668 static
3669 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3670 {
3671         const struct nfs_pgio_header *hdr = data->header;
3672
3673         /* Don't request attributes for pNFS or O_DIRECT writes */
3674         if (data->ds_clp != NULL || hdr->dreq != NULL)
3675                 return false;
3676         /* Otherwise, request attributes if and only if we don't hold
3677          * a delegation
3678          */
3679         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3680 }
3681
3682 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3683 {
3684         struct nfs_server *server = NFS_SERVER(data->header->inode);
3685
3686         if (!nfs4_write_need_cache_consistency_data(data)) {
3687                 data->args.bitmask = NULL;
3688                 data->res.fattr = NULL;
3689         } else
3690                 data->args.bitmask = server->cache_consistency_bitmask;
3691
3692         if (!data->write_done_cb)
3693                 data->write_done_cb = nfs4_write_done_cb;
3694         data->res.server = server;
3695         data->timestamp   = jiffies;
3696
3697         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3698         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3699 }
3700
3701 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3702 {
3703         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3704                         &data->args.seq_args,
3705                         &data->res.seq_res,
3706                         task))
3707                 return;
3708         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3709                         data->args.lock_context, FMODE_WRITE);
3710 }
3711
3712 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3713 {
3714         nfs4_setup_sequence(NFS_SERVER(data->inode),
3715                         &data->args.seq_args,
3716                         &data->res.seq_res,
3717                         task);
3718 }
3719
3720 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3721 {
3722         struct inode *inode = data->inode;
3723
3724         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3725                 rpc_restart_call_prepare(task);
3726                 return -EAGAIN;
3727         }
3728         return 0;
3729 }
3730
3731 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3732 {
3733         if (!nfs4_sequence_done(task, &data->res.seq_res))
3734                 return -EAGAIN;
3735         return data->commit_done_cb(task, data);
3736 }
3737
3738 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3739 {
3740         struct nfs_server *server = NFS_SERVER(data->inode);
3741
3742         if (data->commit_done_cb == NULL)
3743                 data->commit_done_cb = nfs4_commit_done_cb;
3744         data->res.server = server;
3745         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3746         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3747 }
3748
3749 struct nfs4_renewdata {
3750         struct nfs_client       *client;
3751         unsigned long           timestamp;
3752 };
3753
3754 /*
3755  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3756  * standalone procedure for queueing an asynchronous RENEW.
3757  */
3758 static void nfs4_renew_release(void *calldata)
3759 {
3760         struct nfs4_renewdata *data = calldata;
3761         struct nfs_client *clp = data->client;
3762
3763         if (atomic_read(&clp->cl_count) > 1)
3764                 nfs4_schedule_state_renewal(clp);
3765         nfs_put_client(clp);
3766         kfree(data);
3767 }
3768
3769 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3770 {
3771         struct nfs4_renewdata *data = calldata;
3772         struct nfs_client *clp = data->client;
3773         unsigned long timestamp = data->timestamp;
3774
3775         if (task->tk_status < 0) {
3776                 /* Unless we're shutting down, schedule state recovery! */
3777                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3778                         return;
3779                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3780                         nfs4_schedule_lease_recovery(clp);
3781                         return;
3782                 }
3783                 nfs4_schedule_path_down_recovery(clp);
3784         }
3785         do_renew_lease(clp, timestamp);
3786 }
3787
3788 static const struct rpc_call_ops nfs4_renew_ops = {
3789         .rpc_call_done = nfs4_renew_done,
3790         .rpc_release = nfs4_renew_release,
3791 };
3792
3793 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3794 {
3795         struct rpc_message msg = {
3796                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3797                 .rpc_argp       = clp,
3798                 .rpc_cred       = cred,
3799         };
3800         struct nfs4_renewdata *data;
3801
3802         if (renew_flags == 0)
3803                 return 0;
3804         if (!atomic_inc_not_zero(&clp->cl_count))
3805                 return -EIO;
3806         data = kmalloc(sizeof(*data), GFP_NOFS);
3807         if (data == NULL)
3808                 return -ENOMEM;
3809         data->client = clp;
3810         data->timestamp = jiffies;
3811         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3812                         &nfs4_renew_ops, data);
3813 }
3814
3815 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3816 {
3817         struct rpc_message msg = {
3818                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3819                 .rpc_argp       = clp,
3820                 .rpc_cred       = cred,
3821         };
3822         unsigned long now = jiffies;
3823         int status;
3824
3825         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3826         if (status < 0)
3827                 return status;
3828         do_renew_lease(clp, now);
3829         return 0;
3830 }
3831
3832 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3833 {
3834         return (server->caps & NFS_CAP_ACLS)
3835                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3836                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3837 }
3838
3839 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3840  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3841  * the stack.
3842  */
3843 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3844
3845 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3846                 struct page **pages, unsigned int *pgbase)
3847 {
3848         struct page *newpage, **spages;
3849         int rc = 0;
3850         size_t len;
3851         spages = pages;
3852
3853         do {
3854                 len = min_t(size_t, PAGE_SIZE, buflen);
3855                 newpage = alloc_page(GFP_KERNEL);
3856
3857                 if (newpage == NULL)
3858                         goto unwind;
3859                 memcpy(page_address(newpage), buf, len);
3860                 buf += len;
3861                 buflen -= len;
3862                 *pages++ = newpage;
3863                 rc++;
3864         } while (buflen != 0);
3865
3866         return rc;
3867
3868 unwind:
3869         for(; rc > 0; rc--)
3870                 __free_page(spages[rc-1]);
3871         return -ENOMEM;
3872 }
3873
3874 struct nfs4_cached_acl {
3875         int cached;
3876         size_t len;
3877         char data[0];
3878 };
3879
3880 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3881 {
3882         struct nfs_inode *nfsi = NFS_I(inode);
3883
3884         spin_lock(&inode->i_lock);
3885         kfree(nfsi->nfs4_acl);
3886         nfsi->nfs4_acl = acl;
3887         spin_unlock(&inode->i_lock);
3888 }
3889
3890 static void nfs4_zap_acl_attr(struct inode *inode)
3891 {
3892         nfs4_set_cached_acl(inode, NULL);
3893 }
3894
3895 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3896 {
3897         struct nfs_inode *nfsi = NFS_I(inode);
3898         struct nfs4_cached_acl *acl;
3899         int ret = -ENOENT;
3900
3901         spin_lock(&inode->i_lock);
3902         acl = nfsi->nfs4_acl;
3903         if (acl == NULL)
3904                 goto out;
3905         if (buf == NULL) /* user is just asking for length */
3906                 goto out_len;
3907         if (acl->cached == 0)
3908                 goto out;
3909         ret = -ERANGE; /* see getxattr(2) man page */
3910         if (acl->len > buflen)
3911                 goto out;
3912         memcpy(buf, acl->data, acl->len);
3913 out_len:
3914         ret = acl->len;
3915 out:
3916         spin_unlock(&inode->i_lock);
3917         return ret;
3918 }
3919
3920 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3921 {
3922         struct nfs4_cached_acl *acl;
3923         size_t buflen = sizeof(*acl) + acl_len;
3924
3925         if (buflen <= PAGE_SIZE) {
3926                 acl = kmalloc(buflen, GFP_KERNEL);
3927                 if (acl == NULL)
3928                         goto out;
3929                 acl->cached = 1;
3930                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3931         } else {
3932                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3933                 if (acl == NULL)
3934                         goto out;
3935                 acl->cached = 0;
3936         }
3937         acl->len = acl_len;
3938 out:
3939         nfs4_set_cached_acl(inode, acl);
3940 }
3941
3942 /*
3943  * The getxattr API returns the required buffer length when called with a
3944  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3945  * the required buf.  On a NULL buf, we send a page of data to the server
3946  * guessing that the ACL request can be serviced by a page. If so, we cache
3947  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3948  * the cache. If not so, we throw away the page, and cache the required
3949  * length. The next getxattr call will then produce another round trip to
3950  * the server, this time with the input buf of the required size.
3951  */
3952 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3953 {
3954         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3955         struct nfs_getaclargs args = {
3956                 .fh = NFS_FH(inode),
3957                 .acl_pages = pages,
3958                 .acl_len = buflen,
3959         };
3960         struct nfs_getaclres res = {
3961                 .acl_len = buflen,
3962         };
3963         struct rpc_message msg = {
3964                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3965                 .rpc_argp = &args,
3966                 .rpc_resp = &res,
3967         };
3968         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3969         int ret = -ENOMEM, i;
3970
3971         /* As long as we're doing a round trip to the server anyway,
3972          * let's be prepared for a page of acl data. */
3973         if (npages == 0)
3974                 npages = 1;
3975         if (npages > ARRAY_SIZE(pages))
3976                 return -ERANGE;
3977
3978         for (i = 0; i < npages; i++) {
3979                 pages[i] = alloc_page(GFP_KERNEL);
3980                 if (!pages[i])
3981                         goto out_free;
3982         }
3983
3984         /* for decoding across pages */
3985         res.acl_scratch = alloc_page(GFP_KERNEL);
3986         if (!res.acl_scratch)
3987                 goto out_free;
3988
3989         args.acl_len = npages * PAGE_SIZE;
3990         args.acl_pgbase = 0;
3991
3992         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3993                 __func__, buf, buflen, npages, args.acl_len);
3994         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3995                              &msg, &args.seq_args, &res.seq_res, 0);
3996         if (ret)
3997                 goto out_free;
3998
3999         /* Handle the case where the passed-in buffer is too short */
4000         if (res.acl_flags & NFS4_ACL_TRUNC) {
4001                 /* Did the user only issue a request for the acl length? */
4002                 if (buf == NULL)
4003                         goto out_ok;
4004                 ret = -ERANGE;
4005                 goto out_free;
4006         }
4007         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4008         if (buf) {
4009                 if (res.acl_len > buflen) {
4010                         ret = -ERANGE;
4011                         goto out_free;
4012                 }
4013                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4014         }
4015 out_ok:
4016         ret = res.acl_len;
4017 out_free:
4018         for (i = 0; i < npages; i++)
4019                 if (pages[i])
4020                         __free_page(pages[i]);
4021         if (res.acl_scratch)
4022                 __free_page(res.acl_scratch);
4023         return ret;
4024 }
4025
4026 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4027 {
4028         struct nfs4_exception exception = { };
4029         ssize_t ret;
4030         do {
4031                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4032                 if (ret >= 0)
4033                         break;
4034                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4035         } while (exception.retry);
4036         return ret;
4037 }
4038
4039 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4040 {
4041         struct nfs_server *server = NFS_SERVER(inode);
4042         int ret;
4043
4044         if (!nfs4_server_supports_acls(server))
4045                 return -EOPNOTSUPP;
4046         ret = nfs_revalidate_inode(server, inode);
4047         if (ret < 0)
4048                 return ret;
4049         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4050                 nfs_zap_acl_cache(inode);
4051         ret = nfs4_read_cached_acl(inode, buf, buflen);
4052         if (ret != -ENOENT)
4053                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4054                  * but no cached acl data, just the acl length */
4055                 return ret;
4056         return nfs4_get_acl_uncached(inode, buf, buflen);
4057 }
4058
4059 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4060 {
4061         struct nfs_server *server = NFS_SERVER(inode);
4062         struct page *pages[NFS4ACL_MAXPAGES];
4063         struct nfs_setaclargs arg = {
4064                 .fh             = NFS_FH(inode),
4065                 .acl_pages      = pages,
4066                 .acl_len        = buflen,
4067         };
4068         struct nfs_setaclres res;
4069         struct rpc_message msg = {
4070                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4071                 .rpc_argp       = &arg,
4072                 .rpc_resp       = &res,
4073         };
4074         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4075         int ret, i;
4076
4077         if (!nfs4_server_supports_acls(server))
4078                 return -EOPNOTSUPP;
4079         if (npages > ARRAY_SIZE(pages))
4080                 return -ERANGE;
4081         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4082         if (i < 0)
4083                 return i;
4084         nfs4_inode_return_delegation(inode);
4085         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4086
4087         /*
4088          * Free each page after tx, so the only ref left is
4089          * held by the network stack
4090          */
4091         for (; i > 0; i--)
4092                 put_page(pages[i-1]);
4093
4094         /*
4095          * Acl update can result in inode attribute update.
4096          * so mark the attribute cache invalid.
4097          */
4098         spin_lock(&inode->i_lock);
4099         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4100         spin_unlock(&inode->i_lock);
4101         nfs_access_zap_cache(inode);
4102         nfs_zap_acl_cache(inode);
4103         return ret;
4104 }
4105
4106 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4107 {
4108         struct nfs4_exception exception = { };
4109         int err;
4110         do {
4111                 err = nfs4_handle_exception(NFS_SERVER(inode),
4112                                 __nfs4_proc_set_acl(inode, buf, buflen),
4113                                 &exception);
4114         } while (exception.retry);
4115         return err;
4116 }
4117
4118 static int
4119 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4120 {
4121         struct nfs_client *clp = server->nfs_client;
4122
4123         if (task->tk_status >= 0)
4124                 return 0;
4125         switch(task->tk_status) {
4126                 case -NFS4ERR_DELEG_REVOKED:
4127                 case -NFS4ERR_ADMIN_REVOKED:
4128                 case -NFS4ERR_BAD_STATEID:
4129                         if (state == NULL)
4130                                 break;
4131                         nfs_remove_bad_delegation(state->inode);
4132                 case -NFS4ERR_OPENMODE:
4133                         if (state == NULL)
4134                                 break;
4135                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4136                                 goto stateid_invalid;
4137                         goto wait_on_recovery;
4138                 case -NFS4ERR_EXPIRED:
4139                         if (state != NULL) {
4140                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4141                                         goto stateid_invalid;
4142                         }
4143                 case -NFS4ERR_STALE_STATEID:
4144                 case -NFS4ERR_STALE_CLIENTID:
4145                         nfs4_schedule_lease_recovery(clp);
4146                         goto wait_on_recovery;
4147 #if defined(CONFIG_NFS_V4_1)
4148                 case -NFS4ERR_BADSESSION:
4149                 case -NFS4ERR_BADSLOT:
4150                 case -NFS4ERR_BAD_HIGH_SLOT:
4151                 case -NFS4ERR_DEADSESSION:
4152                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4153                 case -NFS4ERR_SEQ_FALSE_RETRY:
4154                 case -NFS4ERR_SEQ_MISORDERED:
4155                         dprintk("%s ERROR %d, Reset session\n", __func__,
4156                                 task->tk_status);
4157                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4158                         task->tk_status = 0;
4159                         return -EAGAIN;
4160 #endif /* CONFIG_NFS_V4_1 */
4161                 case -NFS4ERR_DELAY:
4162                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4163                 case -NFS4ERR_GRACE:
4164                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4165                         task->tk_status = 0;
4166                         return -EAGAIN;
4167                 case -NFS4ERR_RETRY_UNCACHED_REP:
4168                 case -NFS4ERR_OLD_STATEID:
4169                         task->tk_status = 0;
4170                         return -EAGAIN;
4171         }
4172         task->tk_status = nfs4_map_errors(task->tk_status);
4173         return 0;
4174 stateid_invalid:
4175         task->tk_status = -EIO;
4176         return 0;
4177 wait_on_recovery:
4178         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4179         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4180                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4181         task->tk_status = 0;
4182         return -EAGAIN;
4183 }
4184
4185 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4186                                     nfs4_verifier *bootverf)
4187 {
4188         __be32 verf[2];
4189
4190         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4191                 /* An impossible timestamp guarantees this value
4192                  * will never match a generated boot time. */
4193                 verf[0] = 0;
4194                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4195         } else {
4196                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4197                 verf[0] = (__be32)nn->boot_time.tv_sec;
4198                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4199         }
4200         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4201 }
4202
4203 static unsigned int
4204 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4205                                    char *buf, size_t len)
4206 {
4207         unsigned int result;
4208
4209         rcu_read_lock();
4210         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4211                                 clp->cl_ipaddr,
4212                                 rpc_peeraddr2str(clp->cl_rpcclient,
4213                                                         RPC_DISPLAY_ADDR),
4214                                 rpc_peeraddr2str(clp->cl_rpcclient,
4215                                                         RPC_DISPLAY_PROTO));
4216         rcu_read_unlock();
4217         return result;
4218 }
4219
4220 static unsigned int
4221 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4222                                 char *buf, size_t len)
4223 {
4224         char *nodename = clp->cl_rpcclient->cl_nodename;
4225
4226         if (nfs4_client_id_uniquifier[0] != '\0')
4227                 nodename = nfs4_client_id_uniquifier;
4228         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4229                                 clp->rpc_ops->version, clp->cl_minorversion,
4230                                 nodename);
4231 }
4232
4233 /**
4234  * nfs4_proc_setclientid - Negotiate client ID
4235  * @clp: state data structure
4236  * @program: RPC program for NFSv4 callback service
4237  * @port: IP port number for NFS4 callback service
4238  * @cred: RPC credential to use for this call
4239  * @res: where to place the result
4240  *
4241  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4242  */
4243 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4244                 unsigned short port, struct rpc_cred *cred,
4245                 struct nfs4_setclientid_res *res)
4246 {
4247         nfs4_verifier sc_verifier;
4248         struct nfs4_setclientid setclientid = {
4249                 .sc_verifier = &sc_verifier,
4250                 .sc_prog = program,
4251                 .sc_cb_ident = clp->cl_cb_ident,
4252         };
4253         struct rpc_message msg = {
4254                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4255                 .rpc_argp = &setclientid,
4256                 .rpc_resp = res,
4257                 .rpc_cred = cred,
4258         };
4259         int status;
4260
4261         /* nfs_client_id4 */
4262         nfs4_init_boot_verifier(clp, &sc_verifier);
4263         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4264                 setclientid.sc_name_len =
4265                                 nfs4_init_uniform_client_string(clp,
4266                                                 setclientid.sc_name,
4267                                                 sizeof(setclientid.sc_name));
4268         else
4269                 setclientid.sc_name_len =
4270                                 nfs4_init_nonuniform_client_string(clp,
4271                                                 setclientid.sc_name,
4272                                                 sizeof(setclientid.sc_name));
4273         /* cb_client4 */
4274         rcu_read_lock();
4275         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4276                                 sizeof(setclientid.sc_netid),
4277                                 rpc_peeraddr2str(clp->cl_rpcclient,
4278                                                         RPC_DISPLAY_NETID));
4279         rcu_read_unlock();
4280         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4281                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4282                                 clp->cl_ipaddr, port >> 8, port & 255);
4283
4284         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4285                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4286                 setclientid.sc_name_len, setclientid.sc_name);
4287         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4288         dprintk("NFS reply setclientid: %d\n", status);
4289         return status;
4290 }
4291
4292 /**
4293  * nfs4_proc_setclientid_confirm - Confirm client ID
4294  * @clp: state data structure
4295  * @res: result of a previous SETCLIENTID
4296  * @cred: RPC credential to use for this call
4297  *
4298  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4299  */
4300 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4301                 struct nfs4_setclientid_res *arg,
4302                 struct rpc_cred *cred)
4303 {
4304         struct nfs_fsinfo fsinfo;
4305         struct rpc_message msg = {
4306                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4307                 .rpc_argp = arg,
4308                 .rpc_resp = &fsinfo,
4309                 .rpc_cred = cred,
4310         };
4311         unsigned long now;
4312         int status;
4313
4314         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4315                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4316                 clp->cl_clientid);
4317         now = jiffies;
4318         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4319         if (status == 0) {
4320                 spin_lock(&clp->cl_lock);
4321                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4322                 clp->cl_last_renewal = now;
4323                 spin_unlock(&clp->cl_lock);
4324         }
4325         dprintk("NFS reply setclientid_confirm: %d\n", status);
4326         return status;
4327 }
4328
4329 struct nfs4_delegreturndata {
4330         struct nfs4_delegreturnargs args;
4331         struct nfs4_delegreturnres res;
4332         struct nfs_fh fh;
4333         nfs4_stateid stateid;
4334         unsigned long timestamp;
4335         struct nfs_fattr fattr;
4336         int rpc_status;
4337 };
4338
4339 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4340 {
4341         struct nfs4_delegreturndata *data = calldata;
4342
4343         if (!nfs4_sequence_done(task, &data->res.seq_res))
4344                 return;
4345
4346         switch (task->tk_status) {
4347         case -NFS4ERR_STALE_STATEID:
4348         case -NFS4ERR_EXPIRED:
4349         case 0:
4350                 renew_lease(data->res.server, data->timestamp);
4351                 break;
4352         default:
4353                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4354                                 -EAGAIN) {
4355                         rpc_restart_call_prepare(task);
4356                         return;
4357                 }
4358         }
4359         data->rpc_status = task->tk_status;
4360 }
4361
4362 static void nfs4_delegreturn_release(void *calldata)
4363 {
4364         kfree(calldata);
4365 }
4366
4367 #if defined(CONFIG_NFS_V4_1)
4368 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4369 {
4370         struct nfs4_delegreturndata *d_data;
4371
4372         d_data = (struct nfs4_delegreturndata *)data;
4373
4374         nfs4_setup_sequence(d_data->res.server,
4375                         &d_data->args.seq_args,
4376                         &d_data->res.seq_res,
4377                         task);
4378 }
4379 #endif /* CONFIG_NFS_V4_1 */
4380
4381 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4382 #if defined(CONFIG_NFS_V4_1)
4383         .rpc_call_prepare = nfs4_delegreturn_prepare,
4384 #endif /* CONFIG_NFS_V4_1 */
4385         .rpc_call_done = nfs4_delegreturn_done,
4386         .rpc_release = nfs4_delegreturn_release,
4387 };
4388
4389 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4390 {
4391         struct nfs4_delegreturndata *data;
4392         struct nfs_server *server = NFS_SERVER(inode);
4393         struct rpc_task *task;
4394         struct rpc_message msg = {
4395                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4396                 .rpc_cred = cred,
4397         };
4398         struct rpc_task_setup task_setup_data = {
4399                 .rpc_client = server->client,
4400                 .rpc_message = &msg,
4401                 .callback_ops = &nfs4_delegreturn_ops,
4402                 .flags = RPC_TASK_ASYNC,
4403         };
4404         int status = 0;
4405
4406         data = kzalloc(sizeof(*data), GFP_NOFS);
4407         if (data == NULL)
4408                 return -ENOMEM;
4409         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4410         data->args.fhandle = &data->fh;
4411         data->args.stateid = &data->stateid;
4412         data->args.bitmask = server->cache_consistency_bitmask;
4413         nfs_copy_fh(&data->fh, NFS_FH(inode));
4414         nfs4_stateid_copy(&data->stateid, stateid);
4415         data->res.fattr = &data->fattr;
4416         data->res.server = server;
4417         nfs_fattr_init(data->res.fattr);
4418         data->timestamp = jiffies;
4419         data->rpc_status = 0;
4420
4421         task_setup_data.callback_data = data;
4422         msg.rpc_argp = &data->args;
4423         msg.rpc_resp = &data->res;
4424         task = rpc_run_task(&task_setup_data);
4425         if (IS_ERR(task))
4426                 return PTR_ERR(task);
4427         if (!issync)
4428                 goto out;
4429         status = nfs4_wait_for_completion_rpc_task(task);
4430         if (status != 0)
4431                 goto out;
4432         status = data->rpc_status;
4433         if (status == 0)
4434                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4435         else
4436                 nfs_refresh_inode(inode, &data->fattr);
4437 out:
4438         rpc_put_task(task);
4439         return status;
4440 }
4441
4442 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4443 {
4444         struct nfs_server *server = NFS_SERVER(inode);
4445         struct nfs4_exception exception = { };
4446         int err;
4447         do {
4448                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4449                 switch (err) {
4450                         case -NFS4ERR_STALE_STATEID:
4451                         case -NFS4ERR_EXPIRED:
4452                         case 0:
4453                                 return 0;
4454                 }
4455                 err = nfs4_handle_exception(server, err, &exception);
4456         } while (exception.retry);
4457         return err;
4458 }
4459
4460 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4461 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4462
4463 /*
4464  * sleep, with exponential backoff, and retry the LOCK operation.
4465  */
4466 static unsigned long
4467 nfs4_set_lock_task_retry(unsigned long timeout)
4468 {
4469         freezable_schedule_timeout_killable(timeout);
4470         timeout <<= 1;
4471         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4472                 return NFS4_LOCK_MAXTIMEOUT;
4473         return timeout;
4474 }
4475
4476 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4477 {
4478         struct inode *inode = state->inode;
4479         struct nfs_server *server = NFS_SERVER(inode);
4480         struct nfs_client *clp = server->nfs_client;
4481         struct nfs_lockt_args arg = {
4482                 .fh = NFS_FH(inode),
4483                 .fl = request,
4484         };
4485         struct nfs_lockt_res res = {
4486                 .denied = request,
4487         };
4488         struct rpc_message msg = {
4489                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4490                 .rpc_argp       = &arg,
4491                 .rpc_resp       = &res,
4492                 .rpc_cred       = state->owner->so_cred,
4493         };
4494         struct nfs4_lock_state *lsp;
4495         int status;
4496
4497         arg.lock_owner.clientid = clp->cl_clientid;
4498         status = nfs4_set_lock_state(state, request);
4499         if (status != 0)
4500                 goto out;
4501         lsp = request->fl_u.nfs4_fl.owner;
4502         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4503         arg.lock_owner.s_dev = server->s_dev;
4504         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4505         switch (status) {
4506                 case 0:
4507                         request->fl_type = F_UNLCK;
4508                         break;
4509                 case -NFS4ERR_DENIED:
4510                         status = 0;
4511         }
4512         request->fl_ops->fl_release_private(request);
4513 out:
4514         return status;
4515 }
4516
4517 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4518 {
4519         struct nfs4_exception exception = { };
4520         int err;
4521
4522         do {
4523                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4524                                 _nfs4_proc_getlk(state, cmd, request),
4525                                 &exception);
4526         } while (exception.retry);
4527         return err;
4528 }
4529
4530 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4531 {
4532         int res = 0;
4533         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4534                 case FL_POSIX:
4535                         res = posix_lock_file_wait(file, fl);
4536                         break;
4537                 case FL_FLOCK:
4538                         res = flock_lock_file_wait(file, fl);
4539                         break;
4540                 default:
4541                         BUG();
4542         }
4543         return res;
4544 }
4545
4546 struct nfs4_unlockdata {
4547         struct nfs_locku_args arg;
4548         struct nfs_locku_res res;
4549         struct nfs4_lock_state *lsp;
4550         struct nfs_open_context *ctx;
4551         struct file_lock fl;
4552         const struct nfs_server *server;
4553         unsigned long timestamp;
4554 };
4555
4556 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4557                 struct nfs_open_context *ctx,
4558                 struct nfs4_lock_state *lsp,
4559                 struct nfs_seqid *seqid)
4560 {
4561         struct nfs4_unlockdata *p;
4562         struct inode *inode = lsp->ls_state->inode;
4563
4564         p = kzalloc(sizeof(*p), GFP_NOFS);
4565         if (p == NULL)
4566                 return NULL;
4567         p->arg.fh = NFS_FH(inode);
4568         p->arg.fl = &p->fl;
4569         p->arg.seqid = seqid;
4570         p->res.seqid = seqid;
4571         p->arg.stateid = &lsp->ls_stateid;
4572         p->lsp = lsp;
4573         atomic_inc(&lsp->ls_count);
4574         /* Ensure we don't close file until we're done freeing locks! */
4575         p->ctx = get_nfs_open_context(ctx);
4576         memcpy(&p->fl, fl, sizeof(p->fl));
4577         p->server = NFS_SERVER(inode);
4578         return p;
4579 }
4580
4581 static void nfs4_locku_release_calldata(void *data)
4582 {
4583         struct nfs4_unlockdata *calldata = data;
4584         nfs_free_seqid(calldata->arg.seqid);
4585         nfs4_put_lock_state(calldata->lsp);
4586         put_nfs_open_context(calldata->ctx);
4587         kfree(calldata);
4588 }
4589
4590 static void nfs4_locku_done(struct rpc_task *task, void *data)
4591 {
4592         struct nfs4_unlockdata *calldata = data;
4593
4594         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4595                 return;
4596         switch (task->tk_status) {
4597                 case 0:
4598                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4599                                         &calldata->res.stateid);
4600                         renew_lease(calldata->server, calldata->timestamp);
4601                         break;
4602                 case -NFS4ERR_BAD_STATEID:
4603                 case -NFS4ERR_OLD_STATEID:
4604                 case -NFS4ERR_STALE_STATEID:
4605                 case -NFS4ERR_EXPIRED:
4606                         break;
4607                 default:
4608                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4609                                 rpc_restart_call_prepare(task);
4610         }
4611         nfs_release_seqid(calldata->arg.seqid);
4612 }
4613
4614 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4615 {
4616         struct nfs4_unlockdata *calldata = data;
4617
4618         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4619                 goto out_wait;
4620         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4621                 /* Note: exit _without_ running nfs4_locku_done */
4622                 goto out_no_action;
4623         }
4624         calldata->timestamp = jiffies;
4625         if (nfs4_setup_sequence(calldata->server,
4626                                 &calldata->arg.seq_args,
4627                                 &calldata->res.seq_res,
4628                                 task) != 0)
4629                 nfs_release_seqid(calldata->arg.seqid);
4630         return;
4631 out_no_action:
4632         task->tk_action = NULL;
4633 out_wait:
4634         nfs4_sequence_done(task, &calldata->res.seq_res);
4635 }
4636
4637 static const struct rpc_call_ops nfs4_locku_ops = {
4638         .rpc_call_prepare = nfs4_locku_prepare,
4639         .rpc_call_done = nfs4_locku_done,
4640         .rpc_release = nfs4_locku_release_calldata,
4641 };
4642
4643 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4644                 struct nfs_open_context *ctx,
4645                 struct nfs4_lock_state *lsp,
4646                 struct nfs_seqid *seqid)
4647 {
4648         struct nfs4_unlockdata *data;
4649         struct rpc_message msg = {
4650                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4651                 .rpc_cred = ctx->cred,
4652         };
4653         struct rpc_task_setup task_setup_data = {
4654                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4655                 .rpc_message = &msg,
4656                 .callback_ops = &nfs4_locku_ops,
4657                 .workqueue = nfsiod_workqueue,
4658                 .flags = RPC_TASK_ASYNC,
4659         };
4660
4661         /* Ensure this is an unlock - when canceling a lock, the
4662          * canceled lock is passed in, and it won't be an unlock.
4663          */
4664         fl->fl_type = F_UNLCK;
4665
4666         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4667         if (data == NULL) {
4668                 nfs_free_seqid(seqid);
4669                 return ERR_PTR(-ENOMEM);
4670         }
4671
4672         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4673         msg.rpc_argp = &data->arg;
4674         msg.rpc_resp = &data->res;
4675         task_setup_data.callback_data = data;
4676         return rpc_run_task(&task_setup_data);
4677 }
4678
4679 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4680 {
4681         struct inode *inode = state->inode;
4682         struct nfs4_state_owner *sp = state->owner;
4683         struct nfs_inode *nfsi = NFS_I(inode);
4684         struct nfs_seqid *seqid;
4685         struct nfs4_lock_state *lsp;
4686         struct rpc_task *task;
4687         int status = 0;
4688         unsigned char fl_flags = request->fl_flags;
4689
4690         status = nfs4_set_lock_state(state, request);
4691         /* Unlock _before_ we do the RPC call */
4692         request->fl_flags |= FL_EXISTS;
4693         /* Exclude nfs_delegation_claim_locks() */
4694         mutex_lock(&sp->so_delegreturn_mutex);
4695         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
4696         down_read(&nfsi->rwsem);
4697         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4698                 up_read(&nfsi->rwsem);
4699                 mutex_unlock(&sp->so_delegreturn_mutex);
4700                 goto out;
4701         }
4702         up_read(&nfsi->rwsem);
4703         mutex_unlock(&sp->so_delegreturn_mutex);
4704         if (status != 0)
4705                 goto out;
4706         /* Is this a delegated lock? */
4707         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4708                 goto out;
4709         lsp = request->fl_u.nfs4_fl.owner;
4710         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4711         status = -ENOMEM;
4712         if (seqid == NULL)
4713                 goto out;
4714         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4715         status = PTR_ERR(task);
4716         if (IS_ERR(task))
4717                 goto out;
4718         status = nfs4_wait_for_completion_rpc_task(task);
4719         rpc_put_task(task);
4720 out:
4721         request->fl_flags = fl_flags;
4722         return status;
4723 }
4724
4725 struct nfs4_lockdata {
4726         struct nfs_lock_args arg;
4727         struct nfs_lock_res res;
4728         struct nfs4_lock_state *lsp;
4729         struct nfs_open_context *ctx;
4730         struct file_lock fl;
4731         unsigned long timestamp;
4732         int rpc_status;
4733         int cancelled;
4734         struct nfs_server *server;
4735 };
4736
4737 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4738                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4739                 gfp_t gfp_mask)
4740 {
4741         struct nfs4_lockdata *p;
4742         struct inode *inode = lsp->ls_state->inode;
4743         struct nfs_server *server = NFS_SERVER(inode);
4744
4745         p = kzalloc(sizeof(*p), gfp_mask);
4746         if (p == NULL)
4747                 return NULL;
4748
4749         p->arg.fh = NFS_FH(inode);
4750         p->arg.fl = &p->fl;
4751         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4752         if (p->arg.open_seqid == NULL)
4753                 goto out_free;
4754         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4755         if (p->arg.lock_seqid == NULL)
4756                 goto out_free_seqid;
4757         p->arg.lock_stateid = &lsp->ls_stateid;
4758         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4759         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4760         p->arg.lock_owner.s_dev = server->s_dev;
4761         p->res.lock_seqid = p->arg.lock_seqid;
4762         p->lsp = lsp;
4763         p->server = server;
4764         atomic_inc(&lsp->ls_count);
4765         p->ctx = get_nfs_open_context(ctx);
4766         memcpy(&p->fl, fl, sizeof(p->fl));
4767         return p;
4768 out_free_seqid:
4769         nfs_free_seqid(p->arg.open_seqid);
4770 out_free:
4771         kfree(p);
4772         return NULL;
4773 }
4774
4775 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4776 {
4777         struct nfs4_lockdata *data = calldata;
4778         struct nfs4_state *state = data->lsp->ls_state;
4779
4780         dprintk("%s: begin!\n", __func__);
4781         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4782                 goto out_wait;
4783         /* Do we need to do an open_to_lock_owner? */
4784         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4785                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
4786                         goto out_release_lock_seqid;
4787                 }
4788                 data->arg.open_stateid = &state->stateid;
4789                 data->arg.new_lock_owner = 1;
4790                 data->res.open_seqid = data->arg.open_seqid;
4791         } else
4792                 data->arg.new_lock_owner = 0;
4793         if (!nfs4_valid_open_stateid(state)) {
4794                 data->rpc_status = -EBADF;
4795                 task->tk_action = NULL;
4796                 goto out_release_open_seqid;
4797         }
4798         data->timestamp = jiffies;
4799         if (nfs4_setup_sequence(data->server,
4800                                 &data->arg.seq_args,
4801                                 &data->res.seq_res,
4802                                 task) == 0)
4803                 return;
4804 out_release_open_seqid:
4805         nfs_release_seqid(data->arg.open_seqid);
4806 out_release_lock_seqid:
4807         nfs_release_seqid(data->arg.lock_seqid);
4808 out_wait:
4809         nfs4_sequence_done(task, &data->res.seq_res);
4810         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4811 }
4812
4813 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4814 {
4815         struct nfs4_lockdata *data = calldata;
4816
4817         dprintk("%s: begin!\n", __func__);
4818
4819         if (!nfs4_sequence_done(task, &data->res.seq_res))
4820                 return;
4821
4822         data->rpc_status = task->tk_status;
4823         if (data->arg.new_lock_owner != 0) {
4824                 if (data->rpc_status == 0)
4825                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4826                 else
4827                         goto out;
4828         }
4829         if (data->rpc_status == 0) {
4830                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4831                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4832                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4833         }
4834 out:
4835         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4836 }
4837
4838 static void nfs4_lock_release(void *calldata)
4839 {
4840         struct nfs4_lockdata *data = calldata;
4841
4842         dprintk("%s: begin!\n", __func__);
4843         nfs_free_seqid(data->arg.open_seqid);
4844         if (data->cancelled != 0) {
4845                 struct rpc_task *task;
4846                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4847                                 data->arg.lock_seqid);
4848                 if (!IS_ERR(task))
4849                         rpc_put_task_async(task);
4850                 dprintk("%s: cancelling lock!\n", __func__);
4851         } else
4852                 nfs_free_seqid(data->arg.lock_seqid);
4853         nfs4_put_lock_state(data->lsp);
4854         put_nfs_open_context(data->ctx);
4855         kfree(data);
4856         dprintk("%s: done!\n", __func__);
4857 }
4858
4859 static const struct rpc_call_ops nfs4_lock_ops = {
4860         .rpc_call_prepare = nfs4_lock_prepare,
4861         .rpc_call_done = nfs4_lock_done,
4862         .rpc_release = nfs4_lock_release,
4863 };
4864
4865 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4866 {
4867         switch (error) {
4868         case -NFS4ERR_ADMIN_REVOKED:
4869         case -NFS4ERR_BAD_STATEID:
4870                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4871                 if (new_lock_owner != 0 ||
4872                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4873                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4874                 break;
4875         case -NFS4ERR_STALE_STATEID:
4876                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4877         case -NFS4ERR_EXPIRED:
4878                 nfs4_schedule_lease_recovery(server->nfs_client);
4879         };
4880 }
4881
4882 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4883 {
4884         struct nfs4_lockdata *data;
4885         struct rpc_task *task;
4886         struct rpc_message msg = {
4887                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4888                 .rpc_cred = state->owner->so_cred,
4889         };
4890         struct rpc_task_setup task_setup_data = {
4891                 .rpc_client = NFS_CLIENT(state->inode),
4892                 .rpc_message = &msg,
4893                 .callback_ops = &nfs4_lock_ops,
4894                 .workqueue = nfsiod_workqueue,
4895                 .flags = RPC_TASK_ASYNC,
4896         };
4897         int ret;
4898
4899         dprintk("%s: begin!\n", __func__);
4900         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4901                         fl->fl_u.nfs4_fl.owner,
4902                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4903         if (data == NULL)
4904                 return -ENOMEM;
4905         if (IS_SETLKW(cmd))
4906                 data->arg.block = 1;
4907         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4908         msg.rpc_argp = &data->arg;
4909         msg.rpc_resp = &data->res;
4910         task_setup_data.callback_data = data;
4911         if (recovery_type > NFS_LOCK_NEW) {
4912                 if (recovery_type == NFS_LOCK_RECLAIM)
4913                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4914                 nfs4_set_sequence_privileged(&data->arg.seq_args);
4915         }
4916         task = rpc_run_task(&task_setup_data);
4917         if (IS_ERR(task))
4918                 return PTR_ERR(task);
4919         ret = nfs4_wait_for_completion_rpc_task(task);
4920         if (ret == 0) {
4921                 ret = data->rpc_status;
4922                 if (ret)
4923                         nfs4_handle_setlk_error(data->server, data->lsp,
4924                                         data->arg.new_lock_owner, ret);
4925         } else
4926                 data->cancelled = 1;
4927         rpc_put_task(task);
4928         dprintk("%s: done, ret = %d!\n", __func__, ret);
4929         return ret;
4930 }
4931
4932 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4933 {
4934         struct nfs_server *server = NFS_SERVER(state->inode);
4935         struct nfs4_exception exception = {
4936                 .inode = state->inode,
4937         };
4938         int err;
4939
4940         do {
4941                 /* Cache the lock if possible... */
4942                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4943                         return 0;
4944                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4945                 if (err != -NFS4ERR_DELAY)
4946                         break;
4947                 nfs4_handle_exception(server, err, &exception);
4948         } while (exception.retry);
4949         return err;
4950 }
4951
4952 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4953 {
4954         struct nfs_server *server = NFS_SERVER(state->inode);
4955         struct nfs4_exception exception = {
4956                 .inode = state->inode,
4957         };
4958         int err;
4959
4960         err = nfs4_set_lock_state(state, request);
4961         if (err != 0)
4962                 return err;
4963         do {
4964                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4965                         return 0;
4966                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4967                 switch (err) {
4968                 default:
4969                         goto out;
4970                 case -NFS4ERR_GRACE:
4971                 case -NFS4ERR_DELAY:
4972                         nfs4_handle_exception(server, err, &exception);
4973                         err = 0;
4974                 }
4975         } while (exception.retry);
4976 out:
4977         return err;
4978 }
4979
4980 #if defined(CONFIG_NFS_V4_1)
4981 /**
4982  * nfs41_check_expired_locks - possibly free a lock stateid
4983  *
4984  * @state: NFSv4 state for an inode
4985  *
4986  * Returns NFS_OK if recovery for this stateid is now finished.
4987  * Otherwise a negative NFS4ERR value is returned.
4988  */
4989 static int nfs41_check_expired_locks(struct nfs4_state *state)
4990 {
4991         int status, ret = -NFS4ERR_BAD_STATEID;
4992         struct nfs4_lock_state *lsp;
4993         struct nfs_server *server = NFS_SERVER(state->inode);
4994
4995         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4996                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
4997                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4998                         if (status != NFS_OK) {
4999                                 /* Free the stateid unless the server
5000                                  * informs us the stateid is unrecognized. */
5001                                 if (status != -NFS4ERR_BAD_STATEID)
5002                                         nfs41_free_stateid(server,
5003                                                         &lsp->ls_stateid);
5004                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5005                                 ret = status;
5006                         }
5007                 }
5008         };
5009
5010         return ret;
5011 }
5012
5013 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5014 {
5015         int status = NFS_OK;
5016
5017         if (test_bit(LK_STATE_IN_USE, &state->flags))
5018                 status = nfs41_check_expired_locks(state);
5019         if (status != NFS_OK)
5020                 status = nfs4_lock_expired(state, request);
5021         return status;
5022 }
5023 #endif
5024
5025 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5026 {
5027         struct nfs4_state_owner *sp = state->owner;
5028         struct nfs_inode *nfsi = NFS_I(state->inode);
5029         unsigned char fl_flags = request->fl_flags;
5030         unsigned int seq;
5031         int status = -ENOLCK;
5032
5033         if ((fl_flags & FL_POSIX) &&
5034                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5035                 goto out;
5036         /* Is this a delegated open? */
5037         status = nfs4_set_lock_state(state, request);
5038         if (status != 0)
5039                 goto out;
5040         request->fl_flags |= FL_ACCESS;
5041         status = do_vfs_lock(request->fl_file, request);
5042         if (status < 0)
5043                 goto out;
5044         down_read(&nfsi->rwsem);
5045         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5046                 /* Yes: cache locks! */
5047                 /* ...but avoid races with delegation recall... */
5048                 request->fl_flags = fl_flags & ~FL_SLEEP;
5049                 status = do_vfs_lock(request->fl_file, request);
5050                 goto out_unlock;
5051         }
5052         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5053         up_read(&nfsi->rwsem);
5054         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5055         if (status != 0)
5056                 goto out;
5057         down_read(&nfsi->rwsem);
5058         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5059                 status = -NFS4ERR_DELAY;
5060                 goto out_unlock;
5061         }
5062         /* Note: we always want to sleep here! */
5063         request->fl_flags = fl_flags | FL_SLEEP;
5064         if (do_vfs_lock(request->fl_file, request) < 0)
5065                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5066                         "manager!\n", __func__);
5067 out_unlock:
5068         up_read(&nfsi->rwsem);
5069 out:
5070         request->fl_flags = fl_flags;
5071         return status;
5072 }
5073
5074 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5075 {
5076         struct nfs4_exception exception = {
5077                 .state = state,
5078                 .inode = state->inode,
5079         };
5080         int err;
5081
5082         do {
5083                 err = _nfs4_proc_setlk(state, cmd, request);
5084                 if (err == -NFS4ERR_DENIED)
5085                         err = -EAGAIN;
5086                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5087                                 err, &exception);
5088         } while (exception.retry);
5089         return err;
5090 }
5091
5092 static int
5093 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5094 {
5095         struct nfs_open_context *ctx;
5096         struct nfs4_state *state;
5097         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5098         int status;
5099
5100         /* verify open state */
5101         ctx = nfs_file_open_context(filp);
5102         state = ctx->state;
5103
5104         if (request->fl_start < 0 || request->fl_end < 0)
5105                 return -EINVAL;
5106
5107         if (IS_GETLK(cmd)) {
5108                 if (state != NULL)
5109                         return nfs4_proc_getlk(state, F_GETLK, request);
5110                 return 0;
5111         }
5112
5113         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5114                 return -EINVAL;
5115
5116         if (request->fl_type == F_UNLCK) {
5117                 if (state != NULL)
5118                         return nfs4_proc_unlck(state, cmd, request);
5119                 return 0;
5120         }
5121
5122         if (state == NULL)
5123                 return -ENOLCK;
5124         /*
5125          * Don't rely on the VFS having checked the file open mode,
5126          * since it won't do this for flock() locks.
5127          */
5128         switch (request->fl_type) {
5129         case F_RDLCK:
5130                 if (!(filp->f_mode & FMODE_READ))
5131                         return -EBADF;
5132                 break;
5133         case F_WRLCK:
5134                 if (!(filp->f_mode & FMODE_WRITE))
5135                         return -EBADF;
5136         }
5137
5138         do {
5139                 status = nfs4_proc_setlk(state, cmd, request);
5140                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5141                         break;
5142                 timeout = nfs4_set_lock_task_retry(timeout);
5143                 status = -ERESTARTSYS;
5144                 if (signalled())
5145                         break;
5146         } while(status < 0);
5147         return status;
5148 }
5149
5150 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
5151 {
5152         struct nfs_server *server = NFS_SERVER(state->inode);
5153         struct nfs4_exception exception = { };
5154         int err;
5155
5156         err = nfs4_set_lock_state(state, fl);
5157         if (err != 0)
5158                 goto out;
5159         do {
5160                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5161                 switch (err) {
5162                         default:
5163                                 printk(KERN_ERR "NFS: %s: unhandled error "
5164                                         "%d.\n", __func__, err);
5165                         case 0:
5166                         case -ESTALE:
5167                                 goto out;
5168                         case -NFS4ERR_STALE_CLIENTID:
5169                         case -NFS4ERR_STALE_STATEID:
5170                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
5171                         case -NFS4ERR_EXPIRED:
5172                                 nfs4_schedule_lease_recovery(server->nfs_client);
5173                                 err = -EAGAIN;
5174                                 goto out;
5175                         case -NFS4ERR_BADSESSION:
5176                         case -NFS4ERR_BADSLOT:
5177                         case -NFS4ERR_BAD_HIGH_SLOT:
5178                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
5179                         case -NFS4ERR_DEADSESSION:
5180                                 set_bit(NFS_DELEGATED_STATE, &state->flags);
5181                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
5182                                 err = -EAGAIN;
5183                                 goto out;
5184                         case -NFS4ERR_DELEG_REVOKED:
5185                         case -NFS4ERR_ADMIN_REVOKED:
5186                         case -NFS4ERR_BAD_STATEID:
5187                         case -NFS4ERR_OPENMODE:
5188                                 nfs4_schedule_stateid_recovery(server, state);
5189                                 err = 0;
5190                                 goto out;
5191                         case -ENOMEM:
5192                         case -NFS4ERR_DENIED:
5193                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
5194                                 err = 0;
5195                                 goto out;
5196                 }
5197                 set_bit(NFS_DELEGATED_STATE, &state->flags);
5198                 err = nfs4_handle_exception(server, err, &exception);
5199         } while (exception.retry);
5200 out:
5201         return err;
5202 }
5203
5204 struct nfs_release_lockowner_data {
5205         struct nfs4_lock_state *lsp;
5206         struct nfs_server *server;
5207         struct nfs_release_lockowner_args args;
5208 };
5209
5210 static void nfs4_release_lockowner_release(void *calldata)
5211 {
5212         struct nfs_release_lockowner_data *data = calldata;
5213         nfs4_free_lock_state(data->server, data->lsp);
5214         kfree(calldata);
5215 }
5216
5217 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5218         .rpc_release = nfs4_release_lockowner_release,
5219 };
5220
5221 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
5222 {
5223         struct nfs_server *server = lsp->ls_state->owner->so_server;
5224         struct nfs_release_lockowner_data *data;
5225         struct rpc_message msg = {
5226                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5227         };
5228
5229         if (server->nfs_client->cl_mvops->minor_version != 0)
5230                 return -EINVAL;
5231         data = kmalloc(sizeof(*data), GFP_NOFS);
5232         if (!data)
5233                 return -ENOMEM;
5234         data->lsp = lsp;
5235         data->server = server;
5236         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5237         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5238         data->args.lock_owner.s_dev = server->s_dev;
5239         msg.rpc_argp = &data->args;
5240         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5241         return 0;
5242 }
5243
5244 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5245
5246 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5247                                    const void *buf, size_t buflen,
5248                                    int flags, int type)
5249 {
5250         if (strcmp(key, "") != 0)
5251                 return -EINVAL;
5252
5253         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5254 }
5255
5256 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5257                                    void *buf, size_t buflen, int type)
5258 {
5259         if (strcmp(key, "") != 0)
5260                 return -EINVAL;
5261
5262         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5263 }
5264
5265 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5266                                        size_t list_len, const char *name,
5267                                        size_t name_len, int type)
5268 {
5269         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5270
5271         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5272                 return 0;
5273
5274         if (list && len <= list_len)
5275                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5276         return len;
5277 }
5278
5279 /*
5280  * nfs_fhget will use either the mounted_on_fileid or the fileid
5281  */
5282 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5283 {
5284         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5285                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5286               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5287               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5288                 return;
5289
5290         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5291                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5292         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5293         fattr->nlink = 2;
5294 }
5295
5296 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5297                                    const struct qstr *name,
5298                                    struct nfs4_fs_locations *fs_locations,
5299                                    struct page *page)
5300 {
5301         struct nfs_server *server = NFS_SERVER(dir);
5302         u32 bitmask[2] = {
5303                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5304         };
5305         struct nfs4_fs_locations_arg args = {
5306                 .dir_fh = NFS_FH(dir),
5307                 .name = name,
5308                 .page = page,
5309                 .bitmask = bitmask,
5310         };
5311         struct nfs4_fs_locations_res res = {
5312                 .fs_locations = fs_locations,
5313         };
5314         struct rpc_message msg = {
5315                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5316                 .rpc_argp = &args,
5317                 .rpc_resp = &res,
5318         };
5319         int status;
5320
5321         dprintk("%s: start\n", __func__);
5322
5323         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5324          * is not supported */
5325         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5326                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5327         else
5328                 bitmask[0] |= FATTR4_WORD0_FILEID;
5329
5330         nfs_fattr_init(&fs_locations->fattr);
5331         fs_locations->server = server;
5332         fs_locations->nlocations = 0;
5333         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5334         dprintk("%s: returned status = %d\n", __func__, status);
5335         return status;
5336 }
5337
5338 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5339                            const struct qstr *name,
5340                            struct nfs4_fs_locations *fs_locations,
5341                            struct page *page)
5342 {
5343         struct nfs4_exception exception = { };
5344         int err;
5345         do {
5346                 err = nfs4_handle_exception(NFS_SERVER(dir),
5347                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5348                                 &exception);
5349         } while (exception.retry);
5350         return err;
5351 }
5352
5353 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5354 {
5355         int status;
5356         struct nfs4_secinfo_arg args = {
5357                 .dir_fh = NFS_FH(dir),
5358                 .name   = name,
5359         };
5360         struct nfs4_secinfo_res res = {
5361                 .flavors     = flavors,
5362         };
5363         struct rpc_message msg = {
5364                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5365                 .rpc_argp = &args,
5366                 .rpc_resp = &res,
5367         };
5368
5369         dprintk("NFS call  secinfo %s\n", name->name);
5370         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5371         dprintk("NFS reply  secinfo: %d\n", status);
5372         return status;
5373 }
5374
5375 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5376                       struct nfs4_secinfo_flavors *flavors)
5377 {
5378         struct nfs4_exception exception = { };
5379         int err;
5380         do {
5381                 err = nfs4_handle_exception(NFS_SERVER(dir),
5382                                 _nfs4_proc_secinfo(dir, name, flavors),
5383                                 &exception);
5384         } while (exception.retry);
5385         return err;
5386 }
5387
5388 #ifdef CONFIG_NFS_V4_1
5389 /*
5390  * Check the exchange flags returned by the server for invalid flags, having
5391  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5392  * DS flags set.
5393  */
5394 static int nfs4_check_cl_exchange_flags(u32 flags)
5395 {
5396         if (flags & ~EXCHGID4_FLAG_MASK_R)
5397                 goto out_inval;
5398         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5399             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5400                 goto out_inval;
5401         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5402                 goto out_inval;
5403         return NFS_OK;
5404 out_inval:
5405         return -NFS4ERR_INVAL;
5406 }
5407
5408 static bool
5409 nfs41_same_server_scope(struct nfs41_server_scope *a,
5410                         struct nfs41_server_scope *b)
5411 {
5412         if (a->server_scope_sz == b->server_scope_sz &&
5413             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5414                 return true;
5415
5416         return false;
5417 }
5418
5419 /*
5420  * nfs4_proc_bind_conn_to_session()
5421  *
5422  * The 4.1 client currently uses the same TCP connection for the
5423  * fore and backchannel.
5424  */
5425 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5426 {
5427         int status;
5428         struct nfs41_bind_conn_to_session_res res;
5429         struct rpc_message msg = {
5430                 .rpc_proc =
5431                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5432                 .rpc_argp = clp,
5433                 .rpc_resp = &res,
5434                 .rpc_cred = cred,
5435         };
5436
5437         dprintk("--> %s\n", __func__);
5438
5439         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5440         if (unlikely(res.session == NULL)) {
5441                 status = -ENOMEM;
5442                 goto out;
5443         }
5444
5445         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5446         if (status == 0) {
5447                 if (memcmp(res.session->sess_id.data,
5448                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5449                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5450                         status = -EIO;
5451                         goto out_session;
5452                 }
5453                 if (res.dir != NFS4_CDFS4_BOTH) {
5454                         dprintk("NFS: %s: Unexpected direction from server\n",
5455                                 __func__);
5456                         status = -EIO;
5457                         goto out_session;
5458                 }
5459                 if (res.use_conn_in_rdma_mode) {
5460                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5461                                 __func__);
5462                         status = -EIO;
5463                         goto out_session;
5464                 }
5465         }
5466 out_session:
5467         kfree(res.session);
5468 out:
5469         dprintk("<-- %s status= %d\n", __func__, status);
5470         return status;
5471 }
5472
5473 /*
5474  * nfs4_proc_exchange_id()
5475  *
5476  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5477  *
5478  * Since the clientid has expired, all compounds using sessions
5479  * associated with the stale clientid will be returning
5480  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5481  * be in some phase of session reset.
5482  */
5483 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5484 {
5485         nfs4_verifier verifier;
5486         struct nfs41_exchange_id_args args = {
5487                 .verifier = &verifier,
5488                 .client = clp,
5489                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5490         };
5491         struct nfs41_exchange_id_res res = {
5492                 0
5493         };
5494         int status;
5495         struct rpc_message msg = {
5496                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5497                 .rpc_argp = &args,
5498                 .rpc_resp = &res,
5499                 .rpc_cred = cred,
5500         };
5501
5502         nfs4_init_boot_verifier(clp, &verifier);
5503         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5504                                                         sizeof(args.id));
5505         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5506                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5507                 args.id_len, args.id);
5508
5509         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5510                                         GFP_NOFS);
5511         if (unlikely(res.server_owner == NULL)) {
5512                 status = -ENOMEM;
5513                 goto out;
5514         }
5515
5516         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5517                                         GFP_NOFS);
5518         if (unlikely(res.server_scope == NULL)) {
5519                 status = -ENOMEM;
5520                 goto out_server_owner;
5521         }
5522
5523         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5524         if (unlikely(res.impl_id == NULL)) {
5525                 status = -ENOMEM;
5526                 goto out_server_scope;
5527         }
5528
5529         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5530         if (status == 0)
5531                 status = nfs4_check_cl_exchange_flags(res.flags);
5532
5533         if (status == 0) {
5534                 clp->cl_clientid = res.clientid;
5535                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5536                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5537                         clp->cl_seqid = res.seqid;
5538
5539                 kfree(clp->cl_serverowner);
5540                 clp->cl_serverowner = res.server_owner;
5541                 res.server_owner = NULL;
5542
5543                 /* use the most recent implementation id */
5544                 kfree(clp->cl_implid);
5545                 clp->cl_implid = res.impl_id;
5546
5547                 if (clp->cl_serverscope != NULL &&
5548                     !nfs41_same_server_scope(clp->cl_serverscope,
5549                                              res.server_scope)) {
5550                         dprintk("%s: server_scope mismatch detected\n",
5551                                 __func__);
5552                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5553                         kfree(clp->cl_serverscope);
5554                         clp->cl_serverscope = NULL;
5555                 }
5556
5557                 if (clp->cl_serverscope == NULL) {
5558                         clp->cl_serverscope = res.server_scope;
5559                         goto out;
5560                 }
5561         } else
5562                 kfree(res.impl_id);
5563
5564 out_server_owner:
5565         kfree(res.server_owner);
5566 out_server_scope:
5567         kfree(res.server_scope);
5568 out:
5569         if (clp->cl_implid != NULL)
5570                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5571                         "domain: %s, name: %s, date: %llu,%u\n",
5572                         clp->cl_implid->domain, clp->cl_implid->name,
5573                         clp->cl_implid->date.seconds,
5574                         clp->cl_implid->date.nseconds);
5575         dprintk("NFS reply exchange_id: %d\n", status);
5576         return status;
5577 }
5578
5579 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5580                 struct rpc_cred *cred)
5581 {
5582         struct rpc_message msg = {
5583                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5584                 .rpc_argp = clp,
5585                 .rpc_cred = cred,
5586         };
5587         int status;
5588
5589         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5590         if (status)
5591                 dprintk("NFS: Got error %d from the server %s on "
5592                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5593         return status;
5594 }
5595
5596 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5597                 struct rpc_cred *cred)
5598 {
5599         unsigned int loop;
5600         int ret;
5601
5602         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5603                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5604                 switch (ret) {
5605                 case -NFS4ERR_DELAY:
5606                 case -NFS4ERR_CLIENTID_BUSY:
5607                         ssleep(1);
5608                         break;
5609                 default:
5610                         return ret;
5611                 }
5612         }
5613         return 0;
5614 }
5615
5616 int nfs4_destroy_clientid(struct nfs_client *clp)
5617 {
5618         struct rpc_cred *cred;
5619         int ret = 0;
5620
5621         if (clp->cl_mvops->minor_version < 1)
5622                 goto out;
5623         if (clp->cl_exchange_flags == 0)
5624                 goto out;
5625         if (clp->cl_preserve_clid)
5626                 goto out;
5627         cred = nfs4_get_exchange_id_cred(clp);
5628         ret = nfs4_proc_destroy_clientid(clp, cred);
5629         if (cred)
5630                 put_rpccred(cred);
5631         switch (ret) {
5632         case 0:
5633         case -NFS4ERR_STALE_CLIENTID:
5634                 clp->cl_exchange_flags = 0;
5635         }
5636 out:
5637         return ret;
5638 }
5639
5640 struct nfs4_get_lease_time_data {
5641         struct nfs4_get_lease_time_args *args;
5642         struct nfs4_get_lease_time_res *res;
5643         struct nfs_client *clp;
5644 };
5645
5646 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5647                                         void *calldata)
5648 {
5649         struct nfs4_get_lease_time_data *data =
5650                         (struct nfs4_get_lease_time_data *)calldata;
5651
5652         dprintk("--> %s\n", __func__);
5653         /* just setup sequence, do not trigger session recovery
5654            since we're invoked within one */
5655         nfs41_setup_sequence(data->clp->cl_session,
5656                         &data->args->la_seq_args,
5657                         &data->res->lr_seq_res,
5658                         task);
5659         dprintk("<-- %s\n", __func__);
5660 }
5661
5662 /*
5663  * Called from nfs4_state_manager thread for session setup, so don't recover
5664  * from sequence operation or clientid errors.
5665  */
5666 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5667 {
5668         struct nfs4_get_lease_time_data *data =
5669                         (struct nfs4_get_lease_time_data *)calldata;
5670
5671         dprintk("--> %s\n", __func__);
5672         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5673                 return;
5674         switch (task->tk_status) {
5675         case -NFS4ERR_DELAY:
5676         case -NFS4ERR_GRACE:
5677                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5678                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5679                 task->tk_status = 0;
5680                 /* fall through */
5681         case -NFS4ERR_RETRY_UNCACHED_REP:
5682                 rpc_restart_call_prepare(task);
5683                 return;
5684         }
5685         dprintk("<-- %s\n", __func__);
5686 }
5687
5688 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5689         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5690         .rpc_call_done = nfs4_get_lease_time_done,
5691 };
5692
5693 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5694 {
5695         struct rpc_task *task;
5696         struct nfs4_get_lease_time_args args;
5697         struct nfs4_get_lease_time_res res = {
5698                 .lr_fsinfo = fsinfo,
5699         };
5700         struct nfs4_get_lease_time_data data = {
5701                 .args = &args,
5702                 .res = &res,
5703                 .clp = clp,
5704         };
5705         struct rpc_message msg = {
5706                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5707                 .rpc_argp = &args,
5708                 .rpc_resp = &res,
5709         };
5710         struct rpc_task_setup task_setup = {
5711                 .rpc_client = clp->cl_rpcclient,
5712                 .rpc_message = &msg,
5713                 .callback_ops = &nfs4_get_lease_time_ops,
5714                 .callback_data = &data,
5715                 .flags = RPC_TASK_TIMEOUT,
5716         };
5717         int status;
5718
5719         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5720         nfs4_set_sequence_privileged(&args.la_seq_args);
5721         dprintk("--> %s\n", __func__);
5722         task = rpc_run_task(&task_setup);
5723
5724         if (IS_ERR(task))
5725                 status = PTR_ERR(task);
5726         else {
5727                 status = task->tk_status;
5728                 rpc_put_task(task);
5729         }
5730         dprintk("<-- %s return %d\n", __func__, status);
5731
5732         return status;
5733 }
5734
5735 /*
5736  * Initialize the values to be used by the client in CREATE_SESSION
5737  * If nfs4_init_session set the fore channel request and response sizes,
5738  * use them.
5739  *
5740  * Set the back channel max_resp_sz_cached to zero to force the client to
5741  * always set csa_cachethis to FALSE because the current implementation
5742  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5743  */
5744 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5745 {
5746         struct nfs4_session *session = args->client->cl_session;
5747         unsigned int mxrqst_sz = session->fc_target_max_rqst_sz,
5748                      mxresp_sz = session->fc_target_max_resp_sz;
5749
5750         if (mxrqst_sz == 0)
5751                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5752         if (mxresp_sz == 0)
5753                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5754         /* Fore channel attributes */
5755         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5756         args->fc_attrs.max_resp_sz = mxresp_sz;
5757         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5758         args->fc_attrs.max_reqs = max_session_slots;
5759
5760         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5761                 "max_ops=%u max_reqs=%u\n",
5762                 __func__,
5763                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5764                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5765
5766         /* Back channel attributes */
5767         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5768         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5769         args->bc_attrs.max_resp_sz_cached = 0;
5770         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5771         args->bc_attrs.max_reqs = 1;
5772
5773         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5774                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5775                 __func__,
5776                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5777                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5778                 args->bc_attrs.max_reqs);
5779 }
5780
5781 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5782 {
5783         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5784         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5785
5786         if (rcvd->max_resp_sz > sent->max_resp_sz)
5787                 return -EINVAL;
5788         /*
5789          * Our requested max_ops is the minimum we need; we're not
5790          * prepared to break up compounds into smaller pieces than that.
5791          * So, no point even trying to continue if the server won't
5792          * cooperate:
5793          */
5794         if (rcvd->max_ops < sent->max_ops)
5795                 return -EINVAL;
5796         if (rcvd->max_reqs == 0)
5797                 return -EINVAL;
5798         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5799                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5800         return 0;
5801 }
5802
5803 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5804 {
5805         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5806         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5807
5808         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5809                 return -EINVAL;
5810         if (rcvd->max_resp_sz < sent->max_resp_sz)
5811                 return -EINVAL;
5812         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5813                 return -EINVAL;
5814         /* These would render the backchannel useless: */
5815         if (rcvd->max_ops != sent->max_ops)
5816                 return -EINVAL;
5817         if (rcvd->max_reqs != sent->max_reqs)
5818                 return -EINVAL;
5819         return 0;
5820 }
5821
5822 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5823                                      struct nfs4_session *session)
5824 {
5825         int ret;
5826
5827         ret = nfs4_verify_fore_channel_attrs(args, session);
5828         if (ret)
5829                 return ret;
5830         return nfs4_verify_back_channel_attrs(args, session);
5831 }
5832
5833 static int _nfs4_proc_create_session(struct nfs_client *clp,
5834                 struct rpc_cred *cred)
5835 {
5836         struct nfs4_session *session = clp->cl_session;
5837         struct nfs41_create_session_args args = {
5838                 .client = clp,
5839                 .cb_program = NFS4_CALLBACK,
5840         };
5841         struct nfs41_create_session_res res = {
5842                 .client = clp,
5843         };
5844         struct rpc_message msg = {
5845                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5846                 .rpc_argp = &args,
5847                 .rpc_resp = &res,
5848                 .rpc_cred = cred,
5849         };
5850         int status;
5851
5852         nfs4_init_channel_attrs(&args);
5853         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5854
5855         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5856
5857         if (!status) {
5858                 /* Verify the session's negotiated channel_attrs values */
5859                 status = nfs4_verify_channel_attrs(&args, session);
5860                 /* Increment the clientid slot sequence id */
5861                 clp->cl_seqid++;
5862         }
5863
5864         return status;
5865 }
5866
5867 /*
5868  * Issues a CREATE_SESSION operation to the server.
5869  * It is the responsibility of the caller to verify the session is
5870  * expired before calling this routine.
5871  */
5872 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5873 {
5874         int status;
5875         unsigned *ptr;
5876         struct nfs4_session *session = clp->cl_session;
5877
5878         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5879
5880         status = _nfs4_proc_create_session(clp, cred);
5881         if (status)
5882                 goto out;
5883
5884         /* Init or reset the session slot tables */
5885         status = nfs4_setup_session_slot_tables(session);
5886         dprintk("slot table setup returned %d\n", status);
5887         if (status)
5888                 goto out;
5889
5890         ptr = (unsigned *)&session->sess_id.data[0];
5891         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5892                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5893 out:
5894         dprintk("<-- %s\n", __func__);
5895         return status;
5896 }
5897
5898 /*
5899  * Issue the over-the-wire RPC DESTROY_SESSION.
5900  * The caller must serialize access to this routine.
5901  */
5902 int nfs4_proc_destroy_session(struct nfs4_session *session,
5903                 struct rpc_cred *cred)
5904 {
5905         struct rpc_message msg = {
5906                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5907                 .rpc_argp = session,
5908                 .rpc_cred = cred,
5909         };
5910         int status = 0;
5911
5912         dprintk("--> nfs4_proc_destroy_session\n");
5913
5914         /* session is still being setup */
5915         if (session->clp->cl_cons_state != NFS_CS_READY)
5916                 return status;
5917
5918         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5919
5920         if (status)
5921                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5922                         "Session has been destroyed regardless...\n", status);
5923
5924         dprintk("<-- nfs4_proc_destroy_session\n");
5925         return status;
5926 }
5927
5928 /*
5929  * Renew the cl_session lease.
5930  */
5931 struct nfs4_sequence_data {
5932         struct nfs_client *clp;
5933         struct nfs4_sequence_args args;
5934         struct nfs4_sequence_res res;
5935 };
5936
5937 static void nfs41_sequence_release(void *data)
5938 {
5939         struct nfs4_sequence_data *calldata = data;
5940         struct nfs_client *clp = calldata->clp;
5941
5942         if (atomic_read(&clp->cl_count) > 1)
5943                 nfs4_schedule_state_renewal(clp);
5944         nfs_put_client(clp);
5945         kfree(calldata);
5946 }
5947
5948 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5949 {
5950         switch(task->tk_status) {
5951         case -NFS4ERR_DELAY:
5952                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5953                 return -EAGAIN;
5954         default:
5955                 nfs4_schedule_lease_recovery(clp);
5956         }
5957         return 0;
5958 }
5959
5960 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5961 {
5962         struct nfs4_sequence_data *calldata = data;
5963         struct nfs_client *clp = calldata->clp;
5964
5965         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5966                 return;
5967
5968         if (task->tk_status < 0) {
5969                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5970                 if (atomic_read(&clp->cl_count) == 1)
5971                         goto out;
5972
5973                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5974                         rpc_restart_call_prepare(task);
5975                         return;
5976                 }
5977         }
5978         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5979 out:
5980         dprintk("<-- %s\n", __func__);
5981 }
5982
5983 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5984 {
5985         struct nfs4_sequence_data *calldata = data;
5986         struct nfs_client *clp = calldata->clp;
5987         struct nfs4_sequence_args *args;
5988         struct nfs4_sequence_res *res;
5989
5990         args = task->tk_msg.rpc_argp;
5991         res = task->tk_msg.rpc_resp;
5992
5993         nfs41_setup_sequence(clp->cl_session, args, res, task);
5994 }
5995
5996 static const struct rpc_call_ops nfs41_sequence_ops = {
5997         .rpc_call_done = nfs41_sequence_call_done,
5998         .rpc_call_prepare = nfs41_sequence_prepare,
5999         .rpc_release = nfs41_sequence_release,
6000 };
6001
6002 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
6003                 struct rpc_cred *cred,
6004                 bool is_privileged)
6005 {
6006         struct nfs4_sequence_data *calldata;
6007         struct rpc_message msg = {
6008                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
6009                 .rpc_cred = cred,
6010         };
6011         struct rpc_task_setup task_setup_data = {
6012                 .rpc_client = clp->cl_rpcclient,
6013                 .rpc_message = &msg,
6014                 .callback_ops = &nfs41_sequence_ops,
6015                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
6016         };
6017
6018         if (!atomic_inc_not_zero(&clp->cl_count))
6019                 return ERR_PTR(-EIO);
6020         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6021         if (calldata == NULL) {
6022                 nfs_put_client(clp);
6023                 return ERR_PTR(-ENOMEM);
6024         }
6025         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6026         if (is_privileged)
6027                 nfs4_set_sequence_privileged(&calldata->args);
6028         msg.rpc_argp = &calldata->args;
6029         msg.rpc_resp = &calldata->res;
6030         calldata->clp = clp;
6031         task_setup_data.callback_data = calldata;
6032
6033         return rpc_run_task(&task_setup_data);
6034 }
6035
6036 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6037 {
6038         struct rpc_task *task;
6039         int ret = 0;
6040
6041         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6042                 return 0;
6043         task = _nfs41_proc_sequence(clp, cred, false);
6044         if (IS_ERR(task))
6045                 ret = PTR_ERR(task);
6046         else
6047                 rpc_put_task_async(task);
6048         dprintk("<-- %s status=%d\n", __func__, ret);
6049         return ret;
6050 }
6051
6052 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6053 {
6054         struct rpc_task *task;
6055         int ret;
6056
6057         task = _nfs41_proc_sequence(clp, cred, true);
6058         if (IS_ERR(task)) {
6059                 ret = PTR_ERR(task);
6060                 goto out;
6061         }
6062         ret = rpc_wait_for_completion_task(task);
6063         if (!ret) {
6064                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6065
6066                 if (task->tk_status == 0)
6067                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6068                 ret = task->tk_status;
6069         }
6070         rpc_put_task(task);
6071 out:
6072         dprintk("<-- %s status=%d\n", __func__, ret);
6073         return ret;
6074 }
6075
6076 struct nfs4_reclaim_complete_data {
6077         struct nfs_client *clp;
6078         struct nfs41_reclaim_complete_args arg;
6079         struct nfs41_reclaim_complete_res res;
6080 };
6081
6082 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6083 {
6084         struct nfs4_reclaim_complete_data *calldata = data;
6085
6086         nfs41_setup_sequence(calldata->clp->cl_session,
6087                         &calldata->arg.seq_args,
6088                         &calldata->res.seq_res,
6089                         task);
6090 }
6091
6092 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6093 {
6094         switch(task->tk_status) {
6095         case 0:
6096         case -NFS4ERR_COMPLETE_ALREADY:
6097         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6098                 break;
6099         case -NFS4ERR_DELAY:
6100                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6101                 /* fall through */
6102         case -NFS4ERR_RETRY_UNCACHED_REP:
6103                 return -EAGAIN;
6104         default:
6105                 nfs4_schedule_lease_recovery(clp);
6106         }
6107         return 0;
6108 }
6109
6110 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6111 {
6112         struct nfs4_reclaim_complete_data *calldata = data;
6113         struct nfs_client *clp = calldata->clp;
6114         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6115
6116         dprintk("--> %s\n", __func__);
6117         if (!nfs41_sequence_done(task, res))
6118                 return;
6119
6120         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6121                 rpc_restart_call_prepare(task);
6122                 return;
6123         }
6124         dprintk("<-- %s\n", __func__);
6125 }
6126
6127 static void nfs4_free_reclaim_complete_data(void *data)
6128 {
6129         struct nfs4_reclaim_complete_data *calldata = data;
6130
6131         kfree(calldata);
6132 }
6133
6134 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6135         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6136         .rpc_call_done = nfs4_reclaim_complete_done,
6137         .rpc_release = nfs4_free_reclaim_complete_data,
6138 };
6139
6140 /*
6141  * Issue a global reclaim complete.
6142  */
6143 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6144 {
6145         struct nfs4_reclaim_complete_data *calldata;
6146         struct rpc_task *task;
6147         struct rpc_message msg = {
6148                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6149         };
6150         struct rpc_task_setup task_setup_data = {
6151                 .rpc_client = clp->cl_rpcclient,
6152                 .rpc_message = &msg,
6153                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6154                 .flags = RPC_TASK_ASYNC,
6155         };
6156         int status = -ENOMEM;
6157
6158         dprintk("--> %s\n", __func__);
6159         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6160         if (calldata == NULL)
6161                 goto out;
6162         calldata->clp = clp;
6163         calldata->arg.one_fs = 0;
6164
6165         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6166         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
6167         msg.rpc_argp = &calldata->arg;
6168         msg.rpc_resp = &calldata->res;
6169         task_setup_data.callback_data = calldata;
6170         task = rpc_run_task(&task_setup_data);
6171         if (IS_ERR(task)) {
6172                 status = PTR_ERR(task);
6173                 goto out;
6174         }
6175         status = nfs4_wait_for_completion_rpc_task(task);
6176         if (status == 0)
6177                 status = task->tk_status;
6178         rpc_put_task(task);
6179         return 0;
6180 out:
6181         dprintk("<-- %s status=%d\n", __func__, status);
6182         return status;
6183 }
6184
6185 static void
6186 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6187 {
6188         struct nfs4_layoutget *lgp = calldata;
6189         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6190         struct nfs4_session *session = nfs4_get_session(server);
6191
6192         dprintk("--> %s\n", __func__);
6193         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6194          * right now covering the LAYOUTGET we are about to send.
6195          * However, that is not so catastrophic, and there seems
6196          * to be no way to prevent it completely.
6197          */
6198         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
6199                                 &lgp->res.seq_res, task))
6200                 return;
6201         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6202                                           NFS_I(lgp->args.inode)->layout,
6203                                           lgp->args.ctx->state)) {
6204                 rpc_exit(task, NFS4_OK);
6205         }
6206 }
6207
6208 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6209 {
6210         struct nfs4_layoutget *lgp = calldata;
6211         struct inode *inode = lgp->args.inode;
6212         struct nfs_server *server = NFS_SERVER(inode);
6213         struct pnfs_layout_hdr *lo;
6214         struct nfs4_state *state = NULL;
6215         unsigned long timeo, giveup;
6216
6217         dprintk("--> %s\n", __func__);
6218
6219         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
6220                 goto out;
6221
6222         switch (task->tk_status) {
6223         case 0:
6224                 goto out;
6225         case -NFS4ERR_LAYOUTTRYLATER:
6226         case -NFS4ERR_RECALLCONFLICT:
6227                 timeo = rpc_get_timeout(task->tk_client);
6228                 giveup = lgp->args.timestamp + timeo;
6229                 if (time_after(giveup, jiffies))
6230                         task->tk_status = -NFS4ERR_DELAY;
6231                 break;
6232         case -NFS4ERR_EXPIRED:
6233         case -NFS4ERR_BAD_STATEID:
6234                 spin_lock(&inode->i_lock);
6235                 lo = NFS_I(inode)->layout;
6236                 if (!lo || list_empty(&lo->plh_segs)) {
6237                         spin_unlock(&inode->i_lock);
6238                         /* If the open stateid was bad, then recover it. */
6239                         state = lgp->args.ctx->state;
6240                 } else {
6241                         LIST_HEAD(head);
6242
6243                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6244                         spin_unlock(&inode->i_lock);
6245                         /* Mark the bad layout state as invalid, then
6246                          * retry using the open stateid. */
6247                         pnfs_free_lseg_list(&head);
6248                 }
6249         }
6250         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6251                 rpc_restart_call_prepare(task);
6252 out:
6253         dprintk("<-- %s\n", __func__);
6254 }
6255
6256 static size_t max_response_pages(struct nfs_server *server)
6257 {
6258         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6259         return nfs_page_array_len(0, max_resp_sz);
6260 }
6261
6262 static void nfs4_free_pages(struct page **pages, size_t size)
6263 {
6264         int i;
6265
6266         if (!pages)
6267                 return;
6268
6269         for (i = 0; i < size; i++) {
6270                 if (!pages[i])
6271                         break;
6272                 __free_page(pages[i]);
6273         }
6274         kfree(pages);
6275 }
6276
6277 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6278 {
6279         struct page **pages;
6280         int i;
6281
6282         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6283         if (!pages) {
6284                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6285                 return NULL;
6286         }
6287
6288         for (i = 0; i < size; i++) {
6289                 pages[i] = alloc_page(gfp_flags);
6290                 if (!pages[i]) {
6291                         dprintk("%s: failed to allocate page\n", __func__);
6292                         nfs4_free_pages(pages, size);
6293                         return NULL;
6294                 }
6295         }
6296
6297         return pages;
6298 }
6299
6300 static void nfs4_layoutget_release(void *calldata)
6301 {
6302         struct nfs4_layoutget *lgp = calldata;
6303         struct inode *inode = lgp->args.inode;
6304         struct nfs_server *server = NFS_SERVER(inode);
6305         size_t max_pages = max_response_pages(server);
6306
6307         dprintk("--> %s\n", __func__);
6308         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6309         pnfs_put_layout_hdr(NFS_I(inode)->layout);
6310         put_nfs_open_context(lgp->args.ctx);
6311         kfree(calldata);
6312         dprintk("<-- %s\n", __func__);
6313 }
6314
6315 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6316         .rpc_call_prepare = nfs4_layoutget_prepare,
6317         .rpc_call_done = nfs4_layoutget_done,
6318         .rpc_release = nfs4_layoutget_release,
6319 };
6320
6321 struct pnfs_layout_segment *
6322 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6323 {
6324         struct inode *inode = lgp->args.inode;
6325         struct nfs_server *server = NFS_SERVER(inode);
6326         size_t max_pages = max_response_pages(server);
6327         struct rpc_task *task;
6328         struct rpc_message msg = {
6329                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6330                 .rpc_argp = &lgp->args,
6331                 .rpc_resp = &lgp->res,
6332         };
6333         struct rpc_task_setup task_setup_data = {
6334                 .rpc_client = server->client,
6335                 .rpc_message = &msg,
6336                 .callback_ops = &nfs4_layoutget_call_ops,
6337                 .callback_data = lgp,
6338                 .flags = RPC_TASK_ASYNC,
6339         };
6340         struct pnfs_layout_segment *lseg = NULL;
6341         int status = 0;
6342
6343         dprintk("--> %s\n", __func__);
6344
6345         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6346         if (!lgp->args.layout.pages) {
6347                 nfs4_layoutget_release(lgp);
6348                 return ERR_PTR(-ENOMEM);
6349         }
6350         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6351         lgp->args.timestamp = jiffies;
6352
6353         lgp->res.layoutp = &lgp->args.layout;
6354         lgp->res.seq_res.sr_slot = NULL;
6355         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6356
6357         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
6358         pnfs_get_layout_hdr(NFS_I(inode)->layout);
6359
6360         task = rpc_run_task(&task_setup_data);
6361         if (IS_ERR(task))
6362                 return ERR_CAST(task);
6363         status = nfs4_wait_for_completion_rpc_task(task);
6364         if (status == 0)
6365                 status = task->tk_status;
6366         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
6367         if (status == 0 && lgp->res.layoutp->len)
6368                 lseg = pnfs_layout_process(lgp);
6369         rpc_put_task(task);
6370         dprintk("<-- %s status=%d\n", __func__, status);
6371         if (status)
6372                 return ERR_PTR(status);
6373         return lseg;
6374 }
6375
6376 static void
6377 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6378 {
6379         struct nfs4_layoutreturn *lrp = calldata;
6380
6381         dprintk("--> %s\n", __func__);
6382         nfs41_setup_sequence(lrp->clp->cl_session,
6383                         &lrp->args.seq_args,
6384                         &lrp->res.seq_res,
6385                         task);
6386 }
6387
6388 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6389 {
6390         struct nfs4_layoutreturn *lrp = calldata;
6391         struct nfs_server *server;
6392
6393         dprintk("--> %s\n", __func__);
6394
6395         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
6396                 return;
6397
6398         server = NFS_SERVER(lrp->args.inode);
6399         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6400                 rpc_restart_call_prepare(task);
6401                 return;
6402         }
6403         dprintk("<-- %s\n", __func__);
6404 }
6405
6406 static void nfs4_layoutreturn_release(void *calldata)
6407 {
6408         struct nfs4_layoutreturn *lrp = calldata;
6409         struct pnfs_layout_hdr *lo = lrp->args.layout;
6410
6411         dprintk("--> %s\n", __func__);
6412         spin_lock(&lo->plh_inode->i_lock);
6413         if (lrp->res.lrs_present)
6414                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6415         lo->plh_block_lgets--;
6416         spin_unlock(&lo->plh_inode->i_lock);
6417         pnfs_put_layout_hdr(lrp->args.layout);
6418         kfree(calldata);
6419         dprintk("<-- %s\n", __func__);
6420 }
6421
6422 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6423         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6424         .rpc_call_done = nfs4_layoutreturn_done,
6425         .rpc_release = nfs4_layoutreturn_release,
6426 };
6427
6428 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6429 {
6430         struct rpc_task *task;
6431         struct rpc_message msg = {
6432                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6433                 .rpc_argp = &lrp->args,
6434                 .rpc_resp = &lrp->res,
6435         };
6436         struct rpc_task_setup task_setup_data = {
6437                 .rpc_client = lrp->clp->cl_rpcclient,
6438                 .rpc_message = &msg,
6439                 .callback_ops = &nfs4_layoutreturn_call_ops,
6440                 .callback_data = lrp,
6441         };
6442         int status;
6443
6444         dprintk("--> %s\n", __func__);
6445         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6446         task = rpc_run_task(&task_setup_data);
6447         if (IS_ERR(task))
6448                 return PTR_ERR(task);
6449         status = task->tk_status;
6450         dprintk("<-- %s status=%d\n", __func__, status);
6451         rpc_put_task(task);
6452         return status;
6453 }
6454
6455 /*
6456  * Retrieve the list of Data Server devices from the MDS.
6457  */
6458 static int _nfs4_getdevicelist(struct nfs_server *server,
6459                                     const struct nfs_fh *fh,
6460                                     struct pnfs_devicelist *devlist)
6461 {
6462         struct nfs4_getdevicelist_args args = {
6463                 .fh = fh,
6464                 .layoutclass = server->pnfs_curr_ld->id,
6465         };
6466         struct nfs4_getdevicelist_res res = {
6467                 .devlist = devlist,
6468         };
6469         struct rpc_message msg = {
6470                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6471                 .rpc_argp = &args,
6472                 .rpc_resp = &res,
6473         };
6474         int status;
6475
6476         dprintk("--> %s\n", __func__);
6477         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6478                                 &res.seq_res, 0);
6479         dprintk("<-- %s status=%d\n", __func__, status);
6480         return status;
6481 }
6482
6483 int nfs4_proc_getdevicelist(struct nfs_server *server,
6484                             const struct nfs_fh *fh,
6485                             struct pnfs_devicelist *devlist)
6486 {
6487         struct nfs4_exception exception = { };
6488         int err;
6489
6490         do {
6491                 err = nfs4_handle_exception(server,
6492                                 _nfs4_getdevicelist(server, fh, devlist),
6493                                 &exception);
6494         } while (exception.retry);
6495
6496         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6497                 err, devlist->num_devs);
6498
6499         return err;
6500 }
6501 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6502
6503 static int
6504 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6505 {
6506         struct nfs4_getdeviceinfo_args args = {
6507                 .pdev = pdev,
6508         };
6509         struct nfs4_getdeviceinfo_res res = {
6510                 .pdev = pdev,
6511         };
6512         struct rpc_message msg = {
6513                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6514                 .rpc_argp = &args,
6515                 .rpc_resp = &res,
6516         };
6517         int status;
6518
6519         dprintk("--> %s\n", __func__);
6520         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6521         dprintk("<-- %s status=%d\n", __func__, status);
6522
6523         return status;
6524 }
6525
6526 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6527 {
6528         struct nfs4_exception exception = { };
6529         int err;
6530
6531         do {
6532                 err = nfs4_handle_exception(server,
6533                                         _nfs4_proc_getdeviceinfo(server, pdev),
6534                                         &exception);
6535         } while (exception.retry);
6536         return err;
6537 }
6538 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6539
6540 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6541 {
6542         struct nfs4_layoutcommit_data *data = calldata;
6543         struct nfs_server *server = NFS_SERVER(data->args.inode);
6544         struct nfs4_session *session = nfs4_get_session(server);
6545
6546         nfs41_setup_sequence(session,
6547                         &data->args.seq_args,
6548                         &data->res.seq_res,
6549                         task);
6550 }
6551
6552 static void
6553 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6554 {
6555         struct nfs4_layoutcommit_data *data = calldata;
6556         struct nfs_server *server = NFS_SERVER(data->args.inode);
6557
6558         if (!nfs41_sequence_done(task, &data->res.seq_res))
6559                 return;
6560
6561         switch (task->tk_status) { /* Just ignore these failures */
6562         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6563         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6564         case -NFS4ERR_BADLAYOUT:     /* no layout */
6565         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6566                 task->tk_status = 0;
6567                 break;
6568         case 0:
6569                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6570                                                    data->res.fattr);
6571                 break;
6572         default:
6573                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6574                         rpc_restart_call_prepare(task);
6575                         return;
6576                 }
6577         }
6578 }
6579
6580 static void nfs4_layoutcommit_release(void *calldata)
6581 {
6582         struct nfs4_layoutcommit_data *data = calldata;
6583
6584         pnfs_cleanup_layoutcommit(data);
6585         put_rpccred(data->cred);
6586         kfree(data);
6587 }
6588
6589 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6590         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6591         .rpc_call_done = nfs4_layoutcommit_done,
6592         .rpc_release = nfs4_layoutcommit_release,
6593 };
6594
6595 int
6596 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6597 {
6598         struct rpc_message msg = {
6599                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6600                 .rpc_argp = &data->args,
6601                 .rpc_resp = &data->res,
6602                 .rpc_cred = data->cred,
6603         };
6604         struct rpc_task_setup task_setup_data = {
6605                 .task = &data->task,
6606                 .rpc_client = NFS_CLIENT(data->args.inode),
6607                 .rpc_message = &msg,
6608                 .callback_ops = &nfs4_layoutcommit_ops,
6609                 .callback_data = data,
6610                 .flags = RPC_TASK_ASYNC,
6611         };
6612         struct rpc_task *task;
6613         int status = 0;
6614
6615         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6616                 "lbw: %llu inode %lu\n",
6617                 data->task.tk_pid, sync,
6618                 data->args.lastbytewritten,
6619                 data->args.inode->i_ino);
6620
6621         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6622         task = rpc_run_task(&task_setup_data);
6623         if (IS_ERR(task))
6624                 return PTR_ERR(task);
6625         if (sync == false)
6626                 goto out;
6627         status = nfs4_wait_for_completion_rpc_task(task);
6628         if (status != 0)
6629                 goto out;
6630         status = task->tk_status;
6631 out:
6632         dprintk("%s: status %d\n", __func__, status);
6633         rpc_put_task(task);
6634         return status;
6635 }
6636
6637 static int
6638 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6639                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6640 {
6641         struct nfs41_secinfo_no_name_args args = {
6642                 .style = SECINFO_STYLE_CURRENT_FH,
6643         };
6644         struct nfs4_secinfo_res res = {
6645                 .flavors = flavors,
6646         };
6647         struct rpc_message msg = {
6648                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6649                 .rpc_argp = &args,
6650                 .rpc_resp = &res,
6651         };
6652         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6653 }
6654
6655 static int
6656 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6657                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6658 {
6659         struct nfs4_exception exception = { };
6660         int err;
6661         do {
6662                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6663                 switch (err) {
6664                 case 0:
6665                 case -NFS4ERR_WRONGSEC:
6666                 case -NFS4ERR_NOTSUPP:
6667                         goto out;
6668                 default:
6669                         err = nfs4_handle_exception(server, err, &exception);
6670                 }
6671         } while (exception.retry);
6672 out:
6673         return err;
6674 }
6675
6676 static int
6677 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6678                     struct nfs_fsinfo *info)
6679 {
6680         int err;
6681         struct page *page;
6682         rpc_authflavor_t flavor;
6683         struct nfs4_secinfo_flavors *flavors;
6684
6685         page = alloc_page(GFP_KERNEL);
6686         if (!page) {
6687                 err = -ENOMEM;
6688                 goto out;
6689         }
6690
6691         flavors = page_address(page);
6692         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6693
6694         /*
6695          * Fall back on "guess and check" method if
6696          * the server doesn't support SECINFO_NO_NAME
6697          */
6698         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6699                 err = nfs4_find_root_sec(server, fhandle, info);
6700                 goto out_freepage;
6701         }
6702         if (err)
6703                 goto out_freepage;
6704
6705         flavor = nfs_find_best_sec(flavors);
6706         if (err == 0)
6707                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6708
6709 out_freepage:
6710         put_page(page);
6711         if (err == -EACCES)
6712                 return -EPERM;
6713 out:
6714         return err;
6715 }
6716
6717 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6718 {
6719         int status;
6720         struct nfs41_test_stateid_args args = {
6721                 .stateid = stateid,
6722         };
6723         struct nfs41_test_stateid_res res;
6724         struct rpc_message msg = {
6725                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6726                 .rpc_argp = &args,
6727                 .rpc_resp = &res,
6728         };
6729
6730         dprintk("NFS call  test_stateid %p\n", stateid);
6731         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6732         nfs4_set_sequence_privileged(&args.seq_args);
6733         status = nfs4_call_sync_sequence(server->client, server, &msg,
6734                         &args.seq_args, &res.seq_res);
6735         if (status != NFS_OK) {
6736                 dprintk("NFS reply test_stateid: failed, %d\n", status);
6737                 return status;
6738         }
6739         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
6740         return -res.status;
6741 }
6742
6743 /**
6744  * nfs41_test_stateid - perform a TEST_STATEID operation
6745  *
6746  * @server: server / transport on which to perform the operation
6747  * @stateid: state ID to test
6748  *
6749  * Returns NFS_OK if the server recognizes that "stateid" is valid.
6750  * Otherwise a negative NFS4ERR value is returned if the operation
6751  * failed or the state ID is not currently valid.
6752  */
6753 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6754 {
6755         struct nfs4_exception exception = { };
6756         int err;
6757         do {
6758                 err = _nfs41_test_stateid(server, stateid);
6759                 if (err != -NFS4ERR_DELAY)
6760                         break;
6761                 nfs4_handle_exception(server, err, &exception);
6762         } while (exception.retry);
6763         return err;
6764 }
6765
6766 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6767 {
6768         struct nfs41_free_stateid_args args = {
6769                 .stateid = stateid,
6770         };
6771         struct nfs41_free_stateid_res res;
6772         struct rpc_message msg = {
6773                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6774                 .rpc_argp = &args,
6775                 .rpc_resp = &res,
6776         };
6777         int status;
6778
6779         dprintk("NFS call  free_stateid %p\n", stateid);
6780         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6781         nfs4_set_sequence_privileged(&args.seq_args);
6782         status = nfs4_call_sync_sequence(server->client, server, &msg,
6783                         &args.seq_args, &res.seq_res);
6784         dprintk("NFS reply free_stateid: %d\n", status);
6785         return status;
6786 }
6787
6788 /**
6789  * nfs41_free_stateid - perform a FREE_STATEID operation
6790  *
6791  * @server: server / transport on which to perform the operation
6792  * @stateid: state ID to release
6793  *
6794  * Returns NFS_OK if the server freed "stateid".  Otherwise a
6795  * negative NFS4ERR value is returned.
6796  */
6797 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6798 {
6799         struct nfs4_exception exception = { };
6800         int err;
6801         do {
6802                 err = _nfs4_free_stateid(server, stateid);
6803                 if (err != -NFS4ERR_DELAY)
6804                         break;
6805                 nfs4_handle_exception(server, err, &exception);
6806         } while (exception.retry);
6807         return err;
6808 }
6809
6810 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6811                 const nfs4_stateid *s2)
6812 {
6813         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6814                 return false;
6815
6816         if (s1->seqid == s2->seqid)
6817                 return true;
6818         if (s1->seqid == 0 || s2->seqid == 0)
6819                 return true;
6820
6821         return false;
6822 }
6823
6824 #endif /* CONFIG_NFS_V4_1 */
6825
6826 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6827                 const nfs4_stateid *s2)
6828 {
6829         return nfs4_stateid_match(s1, s2);
6830 }
6831
6832
6833 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6834         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6835         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6836         .recover_open   = nfs4_open_reclaim,
6837         .recover_lock   = nfs4_lock_reclaim,
6838         .establish_clid = nfs4_init_clientid,
6839         .get_clid_cred  = nfs4_get_setclientid_cred,
6840         .detect_trunking = nfs40_discover_server_trunking,
6841 };
6842
6843 #if defined(CONFIG_NFS_V4_1)
6844 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6845         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6846         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6847         .recover_open   = nfs4_open_reclaim,
6848         .recover_lock   = nfs4_lock_reclaim,
6849         .establish_clid = nfs41_init_clientid,
6850         .get_clid_cred  = nfs4_get_exchange_id_cred,
6851         .reclaim_complete = nfs41_proc_reclaim_complete,
6852         .detect_trunking = nfs41_discover_server_trunking,
6853 };
6854 #endif /* CONFIG_NFS_V4_1 */
6855
6856 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6857         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6858         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6859         .recover_open   = nfs4_open_expired,
6860         .recover_lock   = nfs4_lock_expired,
6861         .establish_clid = nfs4_init_clientid,
6862         .get_clid_cred  = nfs4_get_setclientid_cred,
6863 };
6864
6865 #if defined(CONFIG_NFS_V4_1)
6866 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6867         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6868         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6869         .recover_open   = nfs41_open_expired,
6870         .recover_lock   = nfs41_lock_expired,
6871         .establish_clid = nfs41_init_clientid,
6872         .get_clid_cred  = nfs4_get_exchange_id_cred,
6873 };
6874 #endif /* CONFIG_NFS_V4_1 */
6875
6876 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6877         .sched_state_renewal = nfs4_proc_async_renew,
6878         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6879         .renew_lease = nfs4_proc_renew,
6880 };
6881
6882 #if defined(CONFIG_NFS_V4_1)
6883 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6884         .sched_state_renewal = nfs41_proc_async_sequence,
6885         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6886         .renew_lease = nfs4_proc_sequence,
6887 };
6888 #endif
6889
6890 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6891         .minor_version = 0,
6892         .init_caps = NFS_CAP_READDIRPLUS
6893                 | NFS_CAP_ATOMIC_OPEN
6894                 | NFS_CAP_CHANGE_ATTR
6895                 | NFS_CAP_POSIX_LOCK,
6896         .call_sync = _nfs4_call_sync,
6897         .match_stateid = nfs4_match_stateid,
6898         .find_root_sec = nfs4_find_root_sec,
6899         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6900         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6901         .state_renewal_ops = &nfs40_state_renewal_ops,
6902 };
6903
6904 #if defined(CONFIG_NFS_V4_1)
6905 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6906         .minor_version = 1,
6907         .init_caps = NFS_CAP_READDIRPLUS
6908                 | NFS_CAP_ATOMIC_OPEN
6909                 | NFS_CAP_CHANGE_ATTR
6910                 | NFS_CAP_POSIX_LOCK
6911                 | NFS_CAP_STATEID_NFSV41
6912                 | NFS_CAP_ATOMIC_OPEN_V1,
6913         .call_sync = nfs4_call_sync_sequence,
6914         .match_stateid = nfs41_match_stateid,
6915         .find_root_sec = nfs41_find_root_sec,
6916         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6917         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6918         .state_renewal_ops = &nfs41_state_renewal_ops,
6919 };
6920 #endif
6921
6922 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6923         [0] = &nfs_v4_0_minor_ops,
6924 #if defined(CONFIG_NFS_V4_1)
6925         [1] = &nfs_v4_1_minor_ops,
6926 #endif
6927 };
6928
6929 const struct inode_operations nfs4_dir_inode_operations = {
6930         .create         = nfs_create,
6931         .lookup         = nfs_lookup,
6932         .atomic_open    = nfs_atomic_open,
6933         .link           = nfs_link,
6934         .unlink         = nfs_unlink,
6935         .symlink        = nfs_symlink,
6936         .mkdir          = nfs_mkdir,
6937         .rmdir          = nfs_rmdir,
6938         .mknod          = nfs_mknod,
6939         .rename         = nfs_rename,
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 static const struct inode_operations nfs4_file_inode_operations = {
6950         .permission     = nfs_permission,
6951         .getattr        = nfs_getattr,
6952         .setattr        = nfs_setattr,
6953         .getxattr       = generic_getxattr,
6954         .setxattr       = generic_setxattr,
6955         .listxattr      = generic_listxattr,
6956         .removexattr    = generic_removexattr,
6957 };
6958
6959 const struct nfs_rpc_ops nfs_v4_clientops = {
6960         .version        = 4,                    /* protocol version */
6961         .dentry_ops     = &nfs4_dentry_operations,
6962         .dir_inode_ops  = &nfs4_dir_inode_operations,
6963         .file_inode_ops = &nfs4_file_inode_operations,
6964         .file_ops       = &nfs4_file_operations,
6965         .getroot        = nfs4_proc_get_root,
6966         .submount       = nfs4_submount,
6967         .try_mount      = nfs4_try_mount,
6968         .getattr        = nfs4_proc_getattr,
6969         .setattr        = nfs4_proc_setattr,
6970         .lookup         = nfs4_proc_lookup,
6971         .access         = nfs4_proc_access,
6972         .readlink       = nfs4_proc_readlink,
6973         .create         = nfs4_proc_create,
6974         .remove         = nfs4_proc_remove,
6975         .unlink_setup   = nfs4_proc_unlink_setup,
6976         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
6977         .unlink_done    = nfs4_proc_unlink_done,
6978         .rename         = nfs4_proc_rename,
6979         .rename_setup   = nfs4_proc_rename_setup,
6980         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
6981         .rename_done    = nfs4_proc_rename_done,
6982         .link           = nfs4_proc_link,
6983         .symlink        = nfs4_proc_symlink,
6984         .mkdir          = nfs4_proc_mkdir,
6985         .rmdir          = nfs4_proc_remove,
6986         .readdir        = nfs4_proc_readdir,
6987         .mknod          = nfs4_proc_mknod,
6988         .statfs         = nfs4_proc_statfs,
6989         .fsinfo         = nfs4_proc_fsinfo,
6990         .pathconf       = nfs4_proc_pathconf,
6991         .set_capabilities = nfs4_server_capabilities,
6992         .decode_dirent  = nfs4_decode_dirent,
6993         .read_setup     = nfs4_proc_read_setup,
6994         .read_pageio_init = pnfs_pageio_init_read,
6995         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
6996         .read_done      = nfs4_read_done,
6997         .write_setup    = nfs4_proc_write_setup,
6998         .write_pageio_init = pnfs_pageio_init_write,
6999         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
7000         .write_done     = nfs4_write_done,
7001         .commit_setup   = nfs4_proc_commit_setup,
7002         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
7003         .commit_done    = nfs4_commit_done,
7004         .lock           = nfs4_proc_lock,
7005         .clear_acl_cache = nfs4_zap_acl_attr,
7006         .close_context  = nfs4_close_context,
7007         .open_context   = nfs4_atomic_open,
7008         .have_delegation = nfs4_have_delegation,
7009         .return_delegation = nfs4_inode_return_delegation,
7010         .alloc_client   = nfs4_alloc_client,
7011         .init_client    = nfs4_init_client,
7012         .free_client    = nfs4_free_client,
7013         .create_server  = nfs4_create_server,
7014         .clone_server   = nfs_clone_server,
7015 };
7016
7017 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7018         .prefix = XATTR_NAME_NFSV4_ACL,
7019         .list   = nfs4_xattr_list_nfs4_acl,
7020         .get    = nfs4_xattr_get_nfs4_acl,
7021         .set    = nfs4_xattr_set_nfs4_acl,
7022 };
7023
7024 const struct xattr_handler *nfs4_xattr_handlers[] = {
7025         &nfs4_xattr_nfs4_acl_handler,
7026         NULL
7027 };
7028
7029 /*
7030  * Local variables:
7031  *  c-basic-offset: 8
7032  * End:
7033  */