fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/nfs_idmap.h>
  55 #include <linux/xattr.h>
  56 #include <linux/utsname.h>
  57 #include <linux/freezer.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4session.h"
  67 #include "fscache.h"
  68
  69 #define NFSDBG_FACILITY         NFSDBG_PROC
  70
  71 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  72 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  73
  74 struct nfs4_opendata;
  75 static int _nfs4_proc_open(struct nfs4_opendata *data);
  76 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  77 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  78 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  79 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  80 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *);
  81 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  82 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  83                             struct nfs_fattr *fattr, struct iattr *sattr,
  84                             struct nfs4_state *state);
  85 #ifdef CONFIG_NFS_V4_1
  86 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
  87 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
  88 #endif
  89 /* Prevent leaks of NFSv4 errors into userland */
  90 static int nfs4_map_errors(int err)
  91 {
  92         if (err >= -1000)
  93                 return err;
  94         switch (err) {
  95         case -NFS4ERR_RESOURCE:
  96         case -NFS4ERR_LAYOUTTRYLATER:
  97         case -NFS4ERR_RECALLCONFLICT:
  98                 return -EREMOTEIO;
  99         case -NFS4ERR_WRONGSEC:
 100                 return -EPERM;
 101         case -NFS4ERR_BADOWNER:
 102         case -NFS4ERR_BADNAME:
 103                 return -EINVAL;
 104         case -NFS4ERR_SHARE_DENIED:
 105                 return -EACCES;
 106         case -NFS4ERR_MINOR_VERS_MISMATCH:
 107                 return -EPROTONOSUPPORT;
 108         case -NFS4ERR_ACCESS:
 109                 return -EACCES;
 110         case -NFS4ERR_FILE_OPEN:
 111                 return -EBUSY;
 112         default:
 113                 dprintk("%s could not handle NFSv4 error %d\n",
 114                                 __func__, -err);
 115                 break;
 116         }
 117         return -EIO;
 118 }
 119
 120 /*
 121  * This is our standard bitmap for GETATTR requests.
 122  */
 123 const u32 nfs4_fattr_bitmap[3] = {
 124         FATTR4_WORD0_TYPE
 125         | FATTR4_WORD0_CHANGE
 126         | FATTR4_WORD0_SIZE
 127         | FATTR4_WORD0_FSID
 128         | FATTR4_WORD0_FILEID,
 129         FATTR4_WORD1_MODE
 130         | FATTR4_WORD1_NUMLINKS
 131         | FATTR4_WORD1_OWNER
 132         | FATTR4_WORD1_OWNER_GROUP
 133         | FATTR4_WORD1_RAWDEV
 134         | FATTR4_WORD1_SPACE_USED
 135         | FATTR4_WORD1_TIME_ACCESS
 136         | FATTR4_WORD1_TIME_METADATA
 137         | FATTR4_WORD1_TIME_MODIFY
 138 };
 139
 140 static const u32 nfs4_pnfs_open_bitmap[3] = {
 141         FATTR4_WORD0_TYPE
 142         | FATTR4_WORD0_CHANGE
 143         | FATTR4_WORD0_SIZE
 144         | FATTR4_WORD0_FSID
 145         | FATTR4_WORD0_FILEID,
 146         FATTR4_WORD1_MODE
 147         | FATTR4_WORD1_NUMLINKS
 148         | FATTR4_WORD1_OWNER
 149         | FATTR4_WORD1_OWNER_GROUP
 150         | FATTR4_WORD1_RAWDEV
 151         | FATTR4_WORD1_SPACE_USED
 152         | FATTR4_WORD1_TIME_ACCESS
 153         | FATTR4_WORD1_TIME_METADATA
 154         | FATTR4_WORD1_TIME_MODIFY,
 155         FATTR4_WORD2_MDSTHRESHOLD
 156 };
 157
 158 static const u32 nfs4_open_noattr_bitmap[3] = {
 159         FATTR4_WORD0_TYPE
 160         | FATTR4_WORD0_CHANGE
 161         | FATTR4_WORD0_FILEID,
 162 };
 163
 164 const u32 nfs4_statfs_bitmap[2] = {
 165         FATTR4_WORD0_FILES_AVAIL
 166         | FATTR4_WORD0_FILES_FREE
 167         | FATTR4_WORD0_FILES_TOTAL,
 168         FATTR4_WORD1_SPACE_AVAIL
 169         | FATTR4_WORD1_SPACE_FREE
 170         | FATTR4_WORD1_SPACE_TOTAL
 171 };
 172
 173 const u32 nfs4_pathconf_bitmap[2] = {
 174         FATTR4_WORD0_MAXLINK
 175         | FATTR4_WORD0_MAXNAME,
 176         0
 177 };
 178
 179 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 180                         | FATTR4_WORD0_MAXREAD
 181                         | FATTR4_WORD0_MAXWRITE
 182                         | FATTR4_WORD0_LEASE_TIME,
 183                         FATTR4_WORD1_TIME_DELTA
 184                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 185                         FATTR4_WORD2_LAYOUT_BLKSIZE
 186 };
 187
 188 const u32 nfs4_fs_locations_bitmap[2] = {
 189         FATTR4_WORD0_TYPE
 190         | FATTR4_WORD0_CHANGE
 191         | FATTR4_WORD0_SIZE
 192         | FATTR4_WORD0_FSID
 193         | FATTR4_WORD0_FILEID
 194         | FATTR4_WORD0_FS_LOCATIONS,
 195         FATTR4_WORD1_MODE
 196         | FATTR4_WORD1_NUMLINKS
 197         | FATTR4_WORD1_OWNER
 198         | FATTR4_WORD1_OWNER_GROUP
 199         | FATTR4_WORD1_RAWDEV
 200         | FATTR4_WORD1_SPACE_USED
 201         | FATTR4_WORD1_TIME_ACCESS
 202         | FATTR4_WORD1_TIME_METADATA
 203         | FATTR4_WORD1_TIME_MODIFY
 204         | FATTR4_WORD1_MOUNTED_ON_FILEID
 205 };
 206
 207 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 208                 struct nfs4_readdir_arg *readdir)
 209 {
 210         __be32 *start, *p;
 211
 212         if (cookie > 2) {
 213                 readdir->cookie = cookie;
 214                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 215                 return;
 216         }
 217
 218         readdir->cookie = 0;
 219         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 220         if (cookie == 2)
 221                 return;
 222
 223         /*
 224          * NFSv4 servers do not return entries for '.' and '..'
 225          * Therefore, we fake these entries here.  We let '.'
 226          * have cookie 0 and '..' have cookie 1.  Note that
 227          * when talking to the server, we always send cookie 0
 228          * instead of 1 or 2.
 229          */
 230         start = p = kmap_atomic(*readdir->pages);
 231
 232         if (cookie == 0) {
 233                 *p++ = xdr_one;                                  /* next */
 234                 *p++ = xdr_zero;                   /* cookie, first word */
 235                 *p++ = xdr_one;                   /* cookie, second word */
 236                 *p++ = xdr_one;                             /* entry len */
 237                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 238                 p++;
 239                 *p++ = xdr_one;                         /* bitmap length */
 240                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 241                 *p++ = htonl(8);              /* attribute buffer length */
 242                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 243         }
 244
 245         *p++ = xdr_one;                                  /* next */
 246         *p++ = xdr_zero;                   /* cookie, first word */
 247         *p++ = xdr_two;                   /* cookie, second word */
 248         *p++ = xdr_two;                             /* entry len */
 249         memcpy(p, "..\0\0", 4);                         /* entry */
 250         p++;
 251         *p++ = xdr_one;                         /* bitmap length */
 252         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 253         *p++ = htonl(8);              /* attribute buffer length */
 254         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 255
 256         readdir->pgbase = (char *)p - (char *)start;
 257         readdir->count -= readdir->pgbase;
 258         kunmap_atomic(start);
 259 }
 260
 261 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 262 {
 263         int res = 0;
 264
 265         might_sleep();
 266
 267         if (*timeout <= 0)
 268                 *timeout = NFS4_POLL_RETRY_MIN;
 269         if (*timeout > NFS4_POLL_RETRY_MAX)
 270                 *timeout = NFS4_POLL_RETRY_MAX;
 271         freezable_schedule_timeout_killable(*timeout);
 272         if (fatal_signal_pending(current))
 273                 res = -ERESTARTSYS;
 274         *timeout <<= 1;
 275         return res;
 276 }
 277
 278 /* This is the error handling routine for processes that are allowed
 279  * to sleep.
 280  */
 281 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 282 {
 283         struct nfs_client *clp = server->nfs_client;
 284         struct nfs4_state *state = exception->state;
 285         struct inode *inode = exception->inode;
 286         int ret = errorcode;
 287
 288         exception->retry = 0;
 289         switch(errorcode) {
 290                 case 0:
 291                         return 0;
 292                 case -NFS4ERR_OPENMODE:
 293                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 294                                 nfs4_inode_return_delegation(inode);
 295                                 exception->retry = 1;
 296                                 return 0;
 297                         }
 298                         if (state == NULL)
 299                                 break;
 300                         ret = nfs4_schedule_stateid_recovery(server, state);
 301                         if (ret < 0)
 302                                 break;
 303                         goto wait_on_recovery;
 304                 case -NFS4ERR_DELEG_REVOKED:
 305                 case -NFS4ERR_ADMIN_REVOKED:
 306                 case -NFS4ERR_BAD_STATEID:
 307                         if (inode != NULL && nfs4_have_delegation(inode, FMODE_READ)) {
 308                                 nfs_remove_bad_delegation(inode);
 309                                 exception->retry = 1;
 310                                 break;
 311                         }
 312                         if (state == NULL)
 313                                 break;
 314                         ret = nfs4_schedule_stateid_recovery(server, state);
 315                         if (ret < 0)
 316                                 break;
 317                         goto wait_on_recovery;
 318                 case -NFS4ERR_EXPIRED:
 319                         if (state != NULL) {
 320                                 ret = nfs4_schedule_stateid_recovery(server, state);
 321                                 if (ret < 0)
 322                                         break;
 323                         }
 324                 case -NFS4ERR_STALE_STATEID:
 325                 case -NFS4ERR_STALE_CLIENTID:
 326                         nfs4_schedule_lease_recovery(clp);
 327                         goto wait_on_recovery;
 328 #if defined(CONFIG_NFS_V4_1)
 329                 case -NFS4ERR_BADSESSION:
 330                 case -NFS4ERR_BADSLOT:
 331                 case -NFS4ERR_BAD_HIGH_SLOT:
 332                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 333                 case -NFS4ERR_DEADSESSION:
 334                 case -NFS4ERR_SEQ_FALSE_RETRY:
 335                 case -NFS4ERR_SEQ_MISORDERED:
 336                         dprintk("%s ERROR: %d Reset session\n", __func__,
 337                                 errorcode);
 338                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 339                         goto wait_on_recovery;
 340 #endif /* defined(CONFIG_NFS_V4_1) */
 341                 case -NFS4ERR_FILE_OPEN:
 342                         if (exception->timeout > HZ) {
 343                                 /* We have retried a decent amount, time to
 344                                  * fail
 345                                  */
 346                                 ret = -EBUSY;
 347                                 break;
 348                         }
 349                 case -NFS4ERR_GRACE:
 350                 case -NFS4ERR_DELAY:
 351                         ret = nfs4_delay(server->client, &exception->timeout);
 352                         if (ret != 0)
 353                                 break;
 354                 case -NFS4ERR_RETRY_UNCACHED_REP:
 355                 case -NFS4ERR_OLD_STATEID:
 356                         exception->retry = 1;
 357                         break;
 358                 case -NFS4ERR_BADOWNER:
 359                         /* The following works around a Linux server bug! */
 360                 case -NFS4ERR_BADNAME:
 361                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 362                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 363                                 exception->retry = 1;
 364                                 printk(KERN_WARNING "NFS: v4 server %s "
 365                                                 "does not accept raw "
 366                                                 "uid/gids. "
 367                                                 "Reenabling the idmapper.\n",
 368                                                 server->nfs_client->cl_hostname);
 369                         }
 370         }
 371         /* We failed to handle the error */
 372         return nfs4_map_errors(ret);
 373 wait_on_recovery:
 374         ret = nfs4_wait_clnt_recover(clp);
 375         if (ret == 0)
 376                 exception->retry = 1;
 377         return ret;
 378 }
 379
 380
 381 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 382 {
 383         spin_lock(&clp->cl_lock);
 384         if (time_before(clp->cl_last_renewal,timestamp))
 385                 clp->cl_last_renewal = timestamp;
 386         spin_unlock(&clp->cl_lock);
 387 }
 388
 389 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 390 {
 391         do_renew_lease(server->nfs_client, timestamp);
 392 }
 393
 394 #if defined(CONFIG_NFS_V4_1)
 395
 396 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 397 {
 398         struct nfs4_session *session;
 399         struct nfs4_slot_table *tbl;
 400         bool send_new_highest_used_slotid = false;
 401
 402         if (!res->sr_slot) {
 403                 /* just wake up the next guy waiting since
 404                  * we may have not consumed a slot after all */
 405                 dprintk("%s: No slot\n", __func__);
 406                 return;
 407         }
 408         tbl = res->sr_slot->table;
 409         session = tbl->session;
 410
 411         spin_lock(&tbl->slot_tbl_lock);
 412         /* Be nice to the server: try to ensure that the last transmitted
 413          * value for highest_user_slotid <= target_highest_slotid
 414          */
 415         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 416                 send_new_highest_used_slotid = true;
 417
 418         if (nfs41_wake_and_assign_slot(tbl, res->sr_slot)) {
 419                 send_new_highest_used_slotid = false;
 420                 goto out_unlock;
 421         }
 422         nfs4_free_slot(tbl, res->sr_slot);
 423
 424         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 425                 send_new_highest_used_slotid = false;
 426 out_unlock:
 427         spin_unlock(&tbl->slot_tbl_lock);
 428         res->sr_slot = NULL;
 429         if (send_new_highest_used_slotid)
 430                 nfs41_server_notify_highest_slotid_update(session->clp);
 431 }
 432
 433 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 434 {
 435         struct nfs4_session *session;
 436         struct nfs4_slot *slot;
 437         struct nfs_client *clp;
 438         bool interrupted = false;
 439         int ret = 1;
 440
 441         /* don't increment the sequence number if the task wasn't sent */
 442         if (!RPC_WAS_SENT(task))
 443                 goto out;
 444
 445         slot = res->sr_slot;
 446         session = slot->table->session;
 447
 448         if (slot->interrupted) {
 449                 slot->interrupted = 0;
 450                 interrupted = true;
 451         }
 452
 453         /* Check the SEQUENCE operation status */
 454         switch (res->sr_status) {
 455         case 0:
 456                 /* Update the slot's sequence and clientid lease timer */
 457                 ++slot->seq_nr;
 458                 clp = session->clp;
 459                 do_renew_lease(clp, res->sr_timestamp);
 460                 /* Check sequence flags */
 461                 if (res->sr_status_flags != 0)
 462                         nfs4_schedule_lease_recovery(clp);
 463                 nfs41_update_target_slotid(slot->table, slot, res);
 464                 break;
 465         case 1:
 466                 /*
 467                  * sr_status remains 1 if an RPC level error occurred.
 468                  * The server may or may not have processed the sequence
 469                  * operation..
 470                  * Mark the slot as having hosted an interrupted RPC call.
 471                  */
 472                 slot->interrupted = 1;
 473                 goto out;
 474         case -NFS4ERR_DELAY:
 475                 /* The server detected a resend of the RPC call and
 476                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 477                  * of RFC5661.
 478                  */
 479                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 480                         __func__,
 481                         slot->slot_nr,
 482                         slot->seq_nr);
 483                 goto out_retry;
 484         case -NFS4ERR_BADSLOT:
 485                 /*
 486                  * The slot id we used was probably retired. Try again
 487                  * using a different slot id.
 488                  */
 489                 goto retry_nowait;
 490         case -NFS4ERR_SEQ_MISORDERED:
 491                 /*
 492                  * Was the last operation on this sequence interrupted?
 493                  * If so, retry after bumping the sequence number.
 494                  */
 495                 if (interrupted) {
 496                         ++slot->seq_nr;
 497                         goto retry_nowait;
 498                 }
 499                 /*
 500                  * Could this slot have been previously retired?
 501                  * If so, then the server may be expecting seq_nr = 1!
 502                  */
 503                 if (slot->seq_nr != 1) {
 504                         slot->seq_nr = 1;
 505                         goto retry_nowait;
 506                 }
 507                 break;
 508         case -NFS4ERR_SEQ_FALSE_RETRY:
 509                 ++slot->seq_nr;
 510                 goto retry_nowait;
 511         default:
 512                 /* Just update the slot sequence no. */
 513                 ++slot->seq_nr;
 514         }
 515 out:
 516         /* The session may be reset by one of the error handlers. */
 517         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 518         nfs41_sequence_free_slot(res);
 519         return ret;
 520 retry_nowait:
 521         if (rpc_restart_call_prepare(task)) {
 522                 task->tk_status = 0;
 523                 ret = 0;
 524         }
 525         goto out;
 526 out_retry:
 527         if (!rpc_restart_call(task))
 528                 goto out;
 529         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 530         return 0;
 531 }
 532
 533 static int nfs4_sequence_done(struct rpc_task *task,
 534                                struct nfs4_sequence_res *res)
 535 {
 536         if (res->sr_slot == NULL)
 537                 return 1;
 538         return nfs41_sequence_done(task, res);
 539 }
 540
 541 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 542                 struct nfs4_sequence_res *res, int cache_reply)
 543 {
 544         args->sa_slot = NULL;
 545         args->sa_cache_this = 0;
 546         args->sa_privileged = 0;
 547         if (cache_reply)
 548                 args->sa_cache_this = 1;
 549         res->sr_slot = NULL;
 550 }
 551
 552 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 553 {
 554         args->sa_privileged = 1;
 555 }
 556
 557 int nfs41_setup_sequence(struct nfs4_session *session,
 558                                 struct nfs4_sequence_args *args,
 559                                 struct nfs4_sequence_res *res,
 560                                 struct rpc_task *task)
 561 {
 562         struct nfs4_slot *slot;
 563         struct nfs4_slot_table *tbl;
 564
 565         dprintk("--> %s\n", __func__);
 566         /* slot already allocated? */
 567         if (res->sr_slot != NULL)
 568                 goto out_success;
 569
 570         tbl = &session->fc_slot_table;
 571
 572         task->tk_timeout = 0;
 573
 574         spin_lock(&tbl->slot_tbl_lock);
 575         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 576             !args->sa_privileged) {
 577                 /* The state manager will wait until the slot table is empty */
 578                 dprintk("%s session is draining\n", __func__);
 579                 goto out_sleep;
 580         }
 581
 582         slot = nfs4_alloc_slot(tbl);
 583         if (IS_ERR(slot)) {
 584                 /* If out of memory, try again in 1/4 second */
 585                 if (slot == ERR_PTR(-ENOMEM))
 586                         task->tk_timeout = HZ >> 2;
 587                 dprintk("<-- %s: no free slots\n", __func__);
 588                 goto out_sleep;
 589         }
 590         spin_unlock(&tbl->slot_tbl_lock);
 591
 592         args->sa_slot = slot;
 593
 594         dprintk("<-- %s slotid=%d seqid=%d\n", __func__,
 595                         slot->slot_nr, slot->seq_nr);
 596
 597         res->sr_slot = slot;
 598         res->sr_timestamp = jiffies;
 599         res->sr_status_flags = 0;
 600         /*
 601          * sr_status is only set in decode_sequence, and so will remain
 602          * set to 1 if an rpc level failure occurs.
 603          */
 604         res->sr_status = 1;
 605 out_success:
 606         rpc_call_start(task);
 607         return 0;
 608 out_sleep:
 609         /* Privileged tasks are queued with top priority */
 610         if (args->sa_privileged)
 611                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 612                                 NULL, RPC_PRIORITY_PRIVILEGED);
 613         else
 614                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 615         spin_unlock(&tbl->slot_tbl_lock);
 616         return -EAGAIN;
 617 }
 618 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 619
 620 int nfs4_setup_sequence(const struct nfs_server *server,
 621                         struct nfs4_sequence_args *args,
 622                         struct nfs4_sequence_res *res,
 623                         struct rpc_task *task)
 624 {
 625         struct nfs4_session *session = nfs4_get_session(server);
 626         int ret = 0;
 627
 628         if (session == NULL) {
 629                 rpc_call_start(task);
 630                 goto out;
 631         }
 632
 633         dprintk("--> %s clp %p session %p sr_slot %d\n",
 634                 __func__, session->clp, session, res->sr_slot ?
 635                         res->sr_slot->slot_nr : -1);
 636
 637         ret = nfs41_setup_sequence(session, args, res, task);
 638 out:
 639         dprintk("<-- %s status=%d\n", __func__, ret);
 640         return ret;
 641 }
 642
 643 struct nfs41_call_sync_data {
 644         const struct nfs_server *seq_server;
 645         struct nfs4_sequence_args *seq_args;
 646         struct nfs4_sequence_res *seq_res;
 647 };
 648
 649 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 650 {
 651         struct nfs41_call_sync_data *data = calldata;
 652         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 653
 654         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 655
 656         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 657 }
 658
 659 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 660 {
 661         struct nfs41_call_sync_data *data = calldata;
 662
 663         nfs41_sequence_done(task, data->seq_res);
 664 }
 665
 666 static const struct rpc_call_ops nfs41_call_sync_ops = {
 667         .rpc_call_prepare = nfs41_call_sync_prepare,
 668         .rpc_call_done = nfs41_call_sync_done,
 669 };
 670
 671 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 672                                    struct nfs_server *server,
 673                                    struct rpc_message *msg,
 674                                    struct nfs4_sequence_args *args,
 675                                    struct nfs4_sequence_res *res)
 676 {
 677         int ret;
 678         struct rpc_task *task;
 679         struct nfs41_call_sync_data data = {
 680                 .seq_server = server,
 681                 .seq_args = args,
 682                 .seq_res = res,
 683         };
 684         struct rpc_task_setup task_setup = {
 685                 .rpc_client = clnt,
 686                 .rpc_message = msg,
 687                 .callback_ops = &nfs41_call_sync_ops,
 688                 .callback_data = &data
 689         };
 690
 691         task = rpc_run_task(&task_setup);
 692         if (IS_ERR(task))
 693                 ret = PTR_ERR(task);
 694         else {
 695                 ret = task->tk_status;
 696                 rpc_put_task(task);
 697         }
 698         return ret;
 699 }
 700
 701 #else
 702 static
 703 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 704                 struct nfs4_sequence_res *res, int cache_reply)
 705 {
 706 }
 707
 708 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 709 {
 710 }
 711
 712
 713 static int nfs4_sequence_done(struct rpc_task *task,
 714                                struct nfs4_sequence_res *res)
 715 {
 716         return 1;
 717 }
 718 #endif /* CONFIG_NFS_V4_1 */
 719
 720 static
 721 int _nfs4_call_sync(struct rpc_clnt *clnt,
 722                     struct nfs_server *server,
 723                     struct rpc_message *msg,
 724                     struct nfs4_sequence_args *args,
 725                     struct nfs4_sequence_res *res)
 726 {
 727         return rpc_call_sync(clnt, msg, 0);
 728 }
 729
 730 static
 731 int nfs4_call_sync(struct rpc_clnt *clnt,
 732                    struct nfs_server *server,
 733                    struct rpc_message *msg,
 734                    struct nfs4_sequence_args *args,
 735                    struct nfs4_sequence_res *res,
 736                    int cache_reply)
 737 {
 738         nfs41_init_sequence(args, res, cache_reply);
 739         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 740                                                 args, res);
 741 }
 742
 743 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 744 {
 745         struct nfs_inode *nfsi = NFS_I(dir);
 746
 747         spin_lock(&dir->i_lock);
 748         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 749         if (!cinfo->atomic || cinfo->before != dir->i_version)
 750                 nfs_force_lookup_revalidate(dir);
 751         dir->i_version = cinfo->after;
 752         nfs_fscache_invalidate(dir);
 753         spin_unlock(&dir->i_lock);
 754 }
 755
 756 struct nfs4_opendata {
 757         struct kref kref;
 758         struct nfs_openargs o_arg;
 759         struct nfs_openres o_res;
 760         struct nfs_open_confirmargs c_arg;
 761         struct nfs_open_confirmres c_res;
 762         struct nfs4_string owner_name;
 763         struct nfs4_string group_name;
 764         struct nfs_fattr f_attr;
 765         struct dentry *dir;
 766         struct dentry *dentry;
 767         struct nfs4_state_owner *owner;
 768         struct nfs4_state *state;
 769         struct iattr attrs;
 770         unsigned long timestamp;
 771         unsigned int rpc_done : 1;
 772         unsigned int is_recover : 1;
 773         int rpc_status;
 774         int cancelled;
 775 };
 776
 777 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 778                 int err, struct nfs4_exception *exception)
 779 {
 780         if (err != -EINVAL)
 781                 return false;
 782         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 783                 return false;
 784         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 785         exception->retry = 1;
 786         return true;
 787 }
 788
 789 static enum open_claim_type4
 790 nfs4_map_atomic_open_claim(struct nfs_server *server,
 791                 enum open_claim_type4 claim)
 792 {
 793         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 794                 return claim;
 795         switch (claim) {
 796         default:
 797                 return claim;
 798         case NFS4_OPEN_CLAIM_FH:
 799                 return NFS4_OPEN_CLAIM_NULL;
 800         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 801                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 802         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 803                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 804         }
 805 }
 806
 807 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 808 {
 809         p->o_res.f_attr = &p->f_attr;
 810         p->o_res.seqid = p->o_arg.seqid;
 811         p->c_res.seqid = p->c_arg.seqid;
 812         p->o_res.server = p->o_arg.server;
 813         p->o_res.access_request = p->o_arg.access;
 814         nfs_fattr_init(&p->f_attr);
 815         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 816 }
 817
 818 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 819                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 820                 const struct iattr *attrs,
 821                 enum open_claim_type4 claim,
 822                 gfp_t gfp_mask)
 823 {
 824         struct dentry *parent = dget_parent(dentry);
 825         struct inode *dir = parent->d_inode;
 826         struct nfs_server *server = NFS_SERVER(dir);
 827         struct nfs4_opendata *p;
 828
 829         p = kzalloc(sizeof(*p), gfp_mask);
 830         if (p == NULL)
 831                 goto err;
 832         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 833         if (p->o_arg.seqid == NULL)
 834                 goto err_free;
 835         nfs_sb_active(dentry->d_sb);
 836         p->dentry = dget(dentry);
 837         p->dir = parent;
 838         p->owner = sp;
 839         atomic_inc(&sp->so_count);
 840         p->o_arg.open_flags = flags;
 841         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 842         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
 843          * will return permission denied for all bits until close */
 844         if (!(flags & O_EXCL)) {
 845                 /* ask server to check for all possible rights as results
 846                  * are cached */
 847                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
 848                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
 849         }
 850         p->o_arg.clientid = server->nfs_client->cl_clientid;
 851         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
 852         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 853         p->o_arg.name = &dentry->d_name;
 854         p->o_arg.server = server;
 855         p->o_arg.bitmask = server->attr_bitmask;
 856         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
 857         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
 858         switch (p->o_arg.claim) {
 859         case NFS4_OPEN_CLAIM_NULL:
 860         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
 861         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
 862                 p->o_arg.fh = NFS_FH(dir);
 863                 break;
 864         case NFS4_OPEN_CLAIM_PREVIOUS:
 865         case NFS4_OPEN_CLAIM_FH:
 866         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 867         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 868                 p->o_arg.fh = NFS_FH(dentry->d_inode);
 869         }
 870         if (attrs != NULL && attrs->ia_valid != 0) {
 871                 __be32 verf[2];
 872
 873                 p->o_arg.u.attrs = &p->attrs;
 874                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 875
 876                 verf[0] = jiffies;
 877                 verf[1] = current->pid;
 878                 memcpy(p->o_arg.u.verifier.data, verf,
 879                                 sizeof(p->o_arg.u.verifier.data));
 880         }
 881         p->c_arg.fh = &p->o_res.fh;
 882         p->c_arg.stateid = &p->o_res.stateid;
 883         p->c_arg.seqid = p->o_arg.seqid;
 884         nfs4_init_opendata_res(p);
 885         kref_init(&p->kref);
 886         return p;
 887 err_free:
 888         kfree(p);
 889 err:
 890         dput(parent);
 891         return NULL;
 892 }
 893
 894 static void nfs4_opendata_free(struct kref *kref)
 895 {
 896         struct nfs4_opendata *p = container_of(kref,
 897                         struct nfs4_opendata, kref);
 898         struct super_block *sb = p->dentry->d_sb;
 899
 900         nfs_free_seqid(p->o_arg.seqid);
 901         if (p->state != NULL)
 902                 nfs4_put_open_state(p->state);
 903         nfs4_put_state_owner(p->owner);
 904         dput(p->dir);
 905         dput(p->dentry);
 906         nfs_sb_deactive(sb);
 907         nfs_fattr_free_names(&p->f_attr);
 908         kfree(p);
 909 }
 910
 911 static void nfs4_opendata_put(struct nfs4_opendata *p)
 912 {
 913         if (p != NULL)
 914                 kref_put(&p->kref, nfs4_opendata_free);
 915 }
 916
 917 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 918 {
 919         int ret;
 920
 921         ret = rpc_wait_for_completion_task(task);
 922         return ret;
 923 }
 924
 925 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 926 {
 927         int ret = 0;
 928
 929         if (open_mode & (O_EXCL|O_TRUNC))
 930                 goto out;
 931         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 932                 case FMODE_READ:
 933                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 934                                 && state->n_rdonly != 0;
 935                         break;
 936                 case FMODE_WRITE:
 937                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 938                                 && state->n_wronly != 0;
 939                         break;
 940                 case FMODE_READ|FMODE_WRITE:
 941                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 942                                 && state->n_rdwr != 0;
 943         }
 944 out:
 945         return ret;
 946 }
 947
 948 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 949 {
 950         if (delegation == NULL)
 951                 return 0;
 952         if ((delegation->type & fmode) != fmode)
 953                 return 0;
 954         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 955                 return 0;
 956         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
 957                 return 0;
 958         nfs_mark_delegation_referenced(delegation);
 959         return 1;
 960 }
 961
 962 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 963 {
 964         switch (fmode) {
 965                 case FMODE_WRITE:
 966                         state->n_wronly++;
 967                         break;
 968                 case FMODE_READ:
 969                         state->n_rdonly++;
 970                         break;
 971                 case FMODE_READ|FMODE_WRITE:
 972                         state->n_rdwr++;
 973         }
 974         nfs4_state_set_mode_locked(state, state->state | fmode);
 975 }
 976
 977 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 978 {
 979         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 980                 nfs4_stateid_copy(&state->stateid, stateid);
 981         nfs4_stateid_copy(&state->open_stateid, stateid);
 982         set_bit(NFS_OPEN_STATE, &state->flags);
 983         switch (fmode) {
 984                 case FMODE_READ:
 985                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 986                         break;
 987                 case FMODE_WRITE:
 988                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 989                         break;
 990                 case FMODE_READ|FMODE_WRITE:
 991                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 992         }
 993 }
 994
 995 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 996 {
 997         write_seqlock(&state->seqlock);
 998         nfs_set_open_stateid_locked(state, stateid, fmode);
 999         write_sequnlock(&state->seqlock);
1000 }
1001
1002 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1003 {
1004         /*
1005          * Protect the call to nfs4_state_set_mode_locked and
1006          * serialise the stateid update
1007          */
1008         write_seqlock(&state->seqlock);
1009         if (deleg_stateid != NULL) {
1010                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1011                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1012         }
1013         if (open_stateid != NULL)
1014                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1015         write_sequnlock(&state->seqlock);
1016         spin_lock(&state->owner->so_lock);
1017         update_open_stateflags(state, fmode);
1018         spin_unlock(&state->owner->so_lock);
1019 }
1020
1021 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1022 {
1023         struct nfs_inode *nfsi = NFS_I(state->inode);
1024         struct nfs_delegation *deleg_cur;
1025         int ret = 0;
1026
1027         fmode &= (FMODE_READ|FMODE_WRITE);
1028
1029         rcu_read_lock();
1030         deleg_cur = rcu_dereference(nfsi->delegation);
1031         if (deleg_cur == NULL)
1032                 goto no_delegation;
1033
1034         spin_lock(&deleg_cur->lock);
1035         if (nfsi->delegation != deleg_cur ||
1036            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1037             (deleg_cur->type & fmode) != fmode)
1038                 goto no_delegation_unlock;
1039
1040         if (delegation == NULL)
1041                 delegation = &deleg_cur->stateid;
1042         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1043                 goto no_delegation_unlock;
1044
1045         nfs_mark_delegation_referenced(deleg_cur);
1046         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1047         ret = 1;
1048 no_delegation_unlock:
1049         spin_unlock(&deleg_cur->lock);
1050 no_delegation:
1051         rcu_read_unlock();
1052
1053         if (!ret && open_stateid != NULL) {
1054                 __update_open_stateid(state, open_stateid, NULL, fmode);
1055                 ret = 1;
1056         }
1057
1058         return ret;
1059 }
1060
1061
1062 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1063 {
1064         struct nfs_delegation *delegation;
1065
1066         rcu_read_lock();
1067         delegation = rcu_dereference(NFS_I(inode)->delegation);
1068         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1069                 rcu_read_unlock();
1070                 return;
1071         }
1072         rcu_read_unlock();
1073         nfs4_inode_return_delegation(inode);
1074 }
1075
1076 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1077 {
1078         struct nfs4_state *state = opendata->state;
1079         struct nfs_inode *nfsi = NFS_I(state->inode);
1080         struct nfs_delegation *delegation;
1081         int open_mode = opendata->o_arg.open_flags;
1082         fmode_t fmode = opendata->o_arg.fmode;
1083         nfs4_stateid stateid;
1084         int ret = -EAGAIN;
1085
1086         for (;;) {
1087                 if (can_open_cached(state, fmode, open_mode)) {
1088                         spin_lock(&state->owner->so_lock);
1089                         if (can_open_cached(state, fmode, open_mode)) {
1090                                 update_open_stateflags(state, fmode);
1091                                 spin_unlock(&state->owner->so_lock);
1092                                 goto out_return_state;
1093                         }
1094                         spin_unlock(&state->owner->so_lock);
1095                 }
1096                 rcu_read_lock();
1097                 delegation = rcu_dereference(nfsi->delegation);
1098                 if (!can_open_delegated(delegation, fmode)) {
1099                         rcu_read_unlock();
1100                         break;
1101                 }
1102                 /* Save the delegation */
1103                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1104                 rcu_read_unlock();
1105                 nfs_release_seqid(opendata->o_arg.seqid);
1106                 if (!opendata->is_recover) {
1107                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1108                         if (ret != 0)
1109                                 goto out;
1110                 }
1111                 ret = -EAGAIN;
1112
1113                 /* Try to update the stateid using the delegation */
1114                 if (update_open_stateid(state, NULL, &stateid, fmode))
1115                         goto out_return_state;
1116         }
1117 out:
1118         return ERR_PTR(ret);
1119 out_return_state:
1120         atomic_inc(&state->count);
1121         return state;
1122 }
1123
1124 static void
1125 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1126 {
1127         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1128         struct nfs_delegation *delegation;
1129         int delegation_flags = 0;
1130
1131         rcu_read_lock();
1132         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1133         if (delegation)
1134                 delegation_flags = delegation->flags;
1135         rcu_read_unlock();
1136         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1137                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1138                                    "returning a delegation for "
1139                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1140                                    clp->cl_hostname);
1141         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1142                 nfs_inode_set_delegation(state->inode,
1143                                          data->owner->so_cred,
1144                                          &data->o_res);
1145         else
1146                 nfs_inode_reclaim_delegation(state->inode,
1147                                              data->owner->so_cred,
1148                                              &data->o_res);
1149 }
1150
1151 /*
1152  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1153  * and update the nfs4_state.
1154  */
1155 static struct nfs4_state *
1156 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1157 {
1158         struct inode *inode = data->state->inode;
1159         struct nfs4_state *state = data->state;
1160         int ret;
1161
1162         if (!data->rpc_done) {
1163                 if (data->rpc_status) {
1164                         ret = data->rpc_status;
1165                         goto err;
1166                 }
1167                 /* cached opens have already been processed */
1168                 goto update;
1169         }
1170
1171         ret = nfs_refresh_inode(inode, &data->f_attr);
1172         if (ret)
1173                 goto err;
1174
1175         if (data->o_res.delegation_type != 0)
1176                 nfs4_opendata_check_deleg(data, state);
1177 update:
1178         update_open_stateid(state, &data->o_res.stateid, NULL,
1179                             data->o_arg.fmode);
1180         atomic_inc(&state->count);
1181
1182         return state;
1183 err:
1184         return ERR_PTR(ret);
1185
1186 }
1187
1188 static struct nfs4_state *
1189 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1190 {
1191         struct inode *inode;
1192         struct nfs4_state *state = NULL;
1193         int ret;
1194
1195         if (!data->rpc_done) {
1196                 state = nfs4_try_open_cached(data);
1197                 goto out;
1198         }
1199
1200         ret = -EAGAIN;
1201         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1202                 goto err;
1203         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1204         ret = PTR_ERR(inode);
1205         if (IS_ERR(inode))
1206                 goto err;
1207         ret = -ENOMEM;
1208         state = nfs4_get_open_state(inode, data->owner);
1209         if (state == NULL)
1210                 goto err_put_inode;
1211         if (data->o_res.delegation_type != 0)
1212                 nfs4_opendata_check_deleg(data, state);
1213         update_open_stateid(state, &data->o_res.stateid, NULL,
1214                         data->o_arg.fmode);
1215         iput(inode);
1216 out:
1217         nfs_release_seqid(data->o_arg.seqid);
1218         return state;
1219 err_put_inode:
1220         iput(inode);
1221 err:
1222         return ERR_PTR(ret);
1223 }
1224
1225 static struct nfs4_state *
1226 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1227 {
1228         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1229                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1230         return _nfs4_opendata_to_nfs4_state(data);
1231 }
1232
1233 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1234 {
1235         struct nfs_inode *nfsi = NFS_I(state->inode);
1236         struct nfs_open_context *ctx;
1237
1238         spin_lock(&state->inode->i_lock);
1239         list_for_each_entry(ctx, &nfsi->open_files, list) {
1240                 if (ctx->state != state)
1241                         continue;
1242                 get_nfs_open_context(ctx);
1243                 spin_unlock(&state->inode->i_lock);
1244                 return ctx;
1245         }
1246         spin_unlock(&state->inode->i_lock);
1247         return ERR_PTR(-ENOENT);
1248 }
1249
1250 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1251                 struct nfs4_state *state, enum open_claim_type4 claim)
1252 {
1253         struct nfs4_opendata *opendata;
1254
1255         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1256                         NULL, claim, GFP_NOFS);
1257         if (opendata == NULL)
1258                 return ERR_PTR(-ENOMEM);
1259         opendata->state = state;
1260         atomic_inc(&state->count);
1261         return opendata;
1262 }
1263
1264 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1265 {
1266         struct nfs4_state *newstate;
1267         int ret;
1268
1269         opendata->o_arg.open_flags = 0;
1270         opendata->o_arg.fmode = fmode;
1271         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1272         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1273         nfs4_init_opendata_res(opendata);
1274         ret = _nfs4_recover_proc_open(opendata);
1275         if (ret != 0)
1276                 return ret;
1277         newstate = nfs4_opendata_to_nfs4_state(opendata);
1278         if (IS_ERR(newstate))
1279                 return PTR_ERR(newstate);
1280         nfs4_close_state(newstate, fmode);
1281         *res = newstate;
1282         return 0;
1283 }
1284
1285 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1286 {
1287         struct nfs4_state *newstate;
1288         int ret;
1289
1290         /* memory barrier prior to reading state->n_* */
1291         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1292         clear_bit(NFS_OPEN_STATE, &state->flags);
1293         smp_rmb();
1294         if (state->n_rdwr != 0) {
1295                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1296                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1297                 if (ret != 0)
1298                         return ret;
1299                 if (newstate != state)
1300                         return -ESTALE;
1301         }
1302         if (state->n_wronly != 0) {
1303                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1304                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1305                 if (ret != 0)
1306                         return ret;
1307                 if (newstate != state)
1308                         return -ESTALE;
1309         }
1310         if (state->n_rdonly != 0) {
1311                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1312                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1313                 if (ret != 0)
1314                         return ret;
1315                 if (newstate != state)
1316                         return -ESTALE;
1317         }
1318         /*
1319          * We may have performed cached opens for all three recoveries.
1320          * Check if we need to update the current stateid.
1321          */
1322         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1323             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1324                 write_seqlock(&state->seqlock);
1325                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1326                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1327                 write_sequnlock(&state->seqlock);
1328         }
1329         return 0;
1330 }
1331
1332 /*
1333  * OPEN_RECLAIM:
1334  *      reclaim state on the server after a reboot.
1335  */
1336 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1337 {
1338         struct nfs_delegation *delegation;
1339         struct nfs4_opendata *opendata;
1340         fmode_t delegation_type = 0;
1341         int status;
1342
1343         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1344                         NFS4_OPEN_CLAIM_PREVIOUS);
1345         if (IS_ERR(opendata))
1346                 return PTR_ERR(opendata);
1347         rcu_read_lock();
1348         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1349         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1350                 delegation_type = delegation->type;
1351         rcu_read_unlock();
1352         opendata->o_arg.u.delegation_type = delegation_type;
1353         status = nfs4_open_recover(opendata, state);
1354         nfs4_opendata_put(opendata);
1355         return status;
1356 }
1357
1358 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1359 {
1360         struct nfs_server *server = NFS_SERVER(state->inode);
1361         struct nfs4_exception exception = { };
1362         int err;
1363         do {
1364                 err = _nfs4_do_open_reclaim(ctx, state);
1365                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1366                         continue;
1367                 if (err != -NFS4ERR_DELAY)
1368                         break;
1369                 nfs4_handle_exception(server, err, &exception);
1370         } while (exception.retry);
1371         return err;
1372 }
1373
1374 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1375 {
1376         struct nfs_open_context *ctx;
1377         int ret;
1378
1379         ctx = nfs4_state_find_open_context(state);
1380         if (IS_ERR(ctx))
1381                 return -EAGAIN;
1382         ret = nfs4_do_open_reclaim(ctx, state);
1383         put_nfs_open_context(ctx);
1384         return ret;
1385 }
1386
1387 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1388 {
1389         switch (err) {
1390                 default:
1391                         printk(KERN_ERR "NFS: %s: unhandled error "
1392                                         "%d.\n", __func__, err);
1393                 case 0:
1394                 case -ENOENT:
1395                 case -ESTALE:
1396                         break;
1397                 case -NFS4ERR_BADSESSION:
1398                 case -NFS4ERR_BADSLOT:
1399                 case -NFS4ERR_BAD_HIGH_SLOT:
1400                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1401                 case -NFS4ERR_DEADSESSION:
1402                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1403                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1404                         return -EAGAIN;
1405                 case -NFS4ERR_STALE_CLIENTID:
1406                 case -NFS4ERR_STALE_STATEID:
1407                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1408                 case -NFS4ERR_EXPIRED:
1409                         /* Don't recall a delegation if it was lost */
1410                         nfs4_schedule_lease_recovery(server->nfs_client);
1411                         return -EAGAIN;
1412                 case -NFS4ERR_DELEG_REVOKED:
1413                 case -NFS4ERR_ADMIN_REVOKED:
1414                 case -NFS4ERR_BAD_STATEID:
1415                 case -NFS4ERR_OPENMODE:
1416                         nfs_inode_find_state_and_recover(state->inode,
1417                                         stateid);
1418                         nfs4_schedule_stateid_recovery(server, state);
1419                         return 0;
1420                 case -NFS4ERR_DELAY:
1421                 case -NFS4ERR_GRACE:
1422                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1423                         ssleep(1);
1424                         return -EAGAIN;
1425                 case -ENOMEM:
1426                 case -NFS4ERR_DENIED:
1427                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1428                         return 0;
1429         }
1430         return err;
1431 }
1432
1433 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1434 {
1435         struct nfs_server *server = NFS_SERVER(state->inode);
1436         struct nfs4_opendata *opendata;
1437         int err;
1438
1439         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1440                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1441         if (IS_ERR(opendata))
1442                 return PTR_ERR(opendata);
1443         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1444         err = nfs4_open_recover(opendata, state);
1445         nfs4_opendata_put(opendata);
1446         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1447 }
1448
1449 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1450 {
1451         struct nfs4_opendata *data = calldata;
1452
1453         data->rpc_status = task->tk_status;
1454         if (data->rpc_status == 0) {
1455                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1456                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1457                 renew_lease(data->o_res.server, data->timestamp);
1458                 data->rpc_done = 1;
1459         }
1460 }
1461
1462 static void nfs4_open_confirm_release(void *calldata)
1463 {
1464         struct nfs4_opendata *data = calldata;
1465         struct nfs4_state *state = NULL;
1466
1467         /* If this request hasn't been cancelled, do nothing */
1468         if (data->cancelled == 0)
1469                 goto out_free;
1470         /* In case of error, no cleanup! */
1471         if (!data->rpc_done)
1472                 goto out_free;
1473         state = nfs4_opendata_to_nfs4_state(data);
1474         if (!IS_ERR(state))
1475                 nfs4_close_state(state, data->o_arg.fmode);
1476 out_free:
1477         nfs4_opendata_put(data);
1478 }
1479
1480 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1481         .rpc_call_done = nfs4_open_confirm_done,
1482         .rpc_release = nfs4_open_confirm_release,
1483 };
1484
1485 /*
1486  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1487  */
1488 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1489 {
1490         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1491         struct rpc_task *task;
1492         struct  rpc_message msg = {
1493                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1494                 .rpc_argp = &data->c_arg,
1495                 .rpc_resp = &data->c_res,
1496                 .rpc_cred = data->owner->so_cred,
1497         };
1498         struct rpc_task_setup task_setup_data = {
1499                 .rpc_client = server->client,
1500                 .rpc_message = &msg,
1501                 .callback_ops = &nfs4_open_confirm_ops,
1502                 .callback_data = data,
1503                 .workqueue = nfsiod_workqueue,
1504                 .flags = RPC_TASK_ASYNC,
1505         };
1506         int status;
1507
1508         kref_get(&data->kref);
1509         data->rpc_done = 0;
1510         data->rpc_status = 0;
1511         data->timestamp = jiffies;
1512         task = rpc_run_task(&task_setup_data);
1513         if (IS_ERR(task))
1514                 return PTR_ERR(task);
1515         status = nfs4_wait_for_completion_rpc_task(task);
1516         if (status != 0) {
1517                 data->cancelled = 1;
1518                 smp_wmb();
1519         } else
1520                 status = data->rpc_status;
1521         rpc_put_task(task);
1522         return status;
1523 }
1524
1525 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1526 {
1527         struct nfs4_opendata *data = calldata;
1528         struct nfs4_state_owner *sp = data->owner;
1529         struct nfs_client *clp = sp->so_server->nfs_client;
1530
1531         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1532                 goto out_wait;
1533         /*
1534          * Check if we still need to send an OPEN call, or if we can use
1535          * a delegation instead.
1536          */
1537         if (data->state != NULL) {
1538                 struct nfs_delegation *delegation;
1539
1540                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1541                         goto out_no_action;
1542                 rcu_read_lock();
1543                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1544                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1545                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1546                     can_open_delegated(delegation, data->o_arg.fmode))
1547                         goto unlock_no_action;
1548                 rcu_read_unlock();
1549         }
1550         /* Update client id. */
1551         data->o_arg.clientid = clp->cl_clientid;
1552         switch (data->o_arg.claim) {
1553         case NFS4_OPEN_CLAIM_PREVIOUS:
1554         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1555         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1556                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1557         case NFS4_OPEN_CLAIM_FH:
1558                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1559                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1560         }
1561         data->timestamp = jiffies;
1562         if (nfs4_setup_sequence(data->o_arg.server,
1563                                 &data->o_arg.seq_args,
1564                                 &data->o_res.seq_res,
1565                                 task) != 0)
1566                 nfs_release_seqid(data->o_arg.seqid);
1567
1568         /* Set the create mode (note dependency on the session type) */
1569         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1570         if (data->o_arg.open_flags & O_EXCL) {
1571                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1572                 if (nfs4_has_persistent_session(clp))
1573                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1574                 else if (clp->cl_mvops->minor_version > 0)
1575                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1576         }
1577         return;
1578 unlock_no_action:
1579         rcu_read_unlock();
1580 out_no_action:
1581         task->tk_action = NULL;
1582 out_wait:
1583         nfs4_sequence_done(task, &data->o_res.seq_res);
1584 }
1585
1586 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1587 {
1588         struct nfs4_opendata *data = calldata;
1589
1590         data->rpc_status = task->tk_status;
1591
1592         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1593                 return;
1594
1595         if (task->tk_status == 0) {
1596                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1597                         switch (data->o_res.f_attr->mode & S_IFMT) {
1598                         case S_IFREG:
1599                                 break;
1600                         case S_IFLNK:
1601                                 data->rpc_status = -ELOOP;
1602                                 break;
1603                         case S_IFDIR:
1604                                 data->rpc_status = -EISDIR;
1605                                 break;
1606                         default:
1607                                 data->rpc_status = -ENOTDIR;
1608                         }
1609                 }
1610                 renew_lease(data->o_res.server, data->timestamp);
1611                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1612                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1613         }
1614         data->rpc_done = 1;
1615 }
1616
1617 static void nfs4_open_release(void *calldata)
1618 {
1619         struct nfs4_opendata *data = calldata;
1620         struct nfs4_state *state = NULL;
1621
1622         /* If this request hasn't been cancelled, do nothing */
1623         if (data->cancelled == 0)
1624                 goto out_free;
1625         /* In case of error, no cleanup! */
1626         if (data->rpc_status != 0 || !data->rpc_done)
1627                 goto out_free;
1628         /* In case we need an open_confirm, no cleanup! */
1629         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1630                 goto out_free;
1631         state = nfs4_opendata_to_nfs4_state(data);
1632         if (!IS_ERR(state))
1633                 nfs4_close_state(state, data->o_arg.fmode);
1634 out_free:
1635         nfs4_opendata_put(data);
1636 }
1637
1638 static const struct rpc_call_ops nfs4_open_ops = {
1639         .rpc_call_prepare = nfs4_open_prepare,
1640         .rpc_call_done = nfs4_open_done,
1641         .rpc_release = nfs4_open_release,
1642 };
1643
1644 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1645 {
1646         struct inode *dir = data->dir->d_inode;
1647         struct nfs_server *server = NFS_SERVER(dir);
1648         struct nfs_openargs *o_arg = &data->o_arg;
1649         struct nfs_openres *o_res = &data->o_res;
1650         struct rpc_task *task;
1651         struct rpc_message msg = {
1652                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1653                 .rpc_argp = o_arg,
1654                 .rpc_resp = o_res,
1655                 .rpc_cred = data->owner->so_cred,
1656         };
1657         struct rpc_task_setup task_setup_data = {
1658                 .rpc_client = server->client,
1659                 .rpc_message = &msg,
1660                 .callback_ops = &nfs4_open_ops,
1661                 .callback_data = data,
1662                 .workqueue = nfsiod_workqueue,
1663                 .flags = RPC_TASK_ASYNC,
1664         };
1665         int status;
1666
1667         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1668         kref_get(&data->kref);
1669         data->rpc_done = 0;
1670         data->rpc_status = 0;
1671         data->cancelled = 0;
1672         data->is_recover = 0;
1673         if (isrecover) {
1674                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1675                 data->is_recover = 1;
1676         }
1677         task = rpc_run_task(&task_setup_data);
1678         if (IS_ERR(task))
1679                 return PTR_ERR(task);
1680         status = nfs4_wait_for_completion_rpc_task(task);
1681         if (status != 0) {
1682                 data->cancelled = 1;
1683                 smp_wmb();
1684         } else
1685                 status = data->rpc_status;
1686         rpc_put_task(task);
1687
1688         return status;
1689 }
1690
1691 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1692 {
1693         struct inode *dir = data->dir->d_inode;
1694         struct nfs_openres *o_res = &data->o_res;
1695         int status;
1696
1697         status = nfs4_run_open_task(data, 1);
1698         if (status != 0 || !data->rpc_done)
1699                 return status;
1700
1701         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1702
1703         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1704                 status = _nfs4_proc_open_confirm(data);
1705                 if (status != 0)
1706                         return status;
1707         }
1708
1709         return status;
1710 }
1711
1712 static int nfs4_opendata_access(struct rpc_cred *cred,
1713                                 struct nfs4_opendata *opendata,
1714                                 struct nfs4_state *state, fmode_t fmode,
1715                                 int openflags)
1716 {
1717         struct nfs_access_entry cache;
1718         u32 mask;
1719
1720         /* access call failed or for some reason the server doesn't
1721          * support any access modes -- defer access call until later */
1722         if (opendata->o_res.access_supported == 0)
1723                 return 0;
1724
1725         mask = 0;
1726         /* don't check MAY_WRITE - a newly created file may not have
1727          * write mode bits, but POSIX allows the creating process to write.
1728          * use openflags to check for exec, because fmode won't
1729          * always have FMODE_EXEC set when file open for exec. */
1730         if (openflags & __FMODE_EXEC) {
1731                 /* ONLY check for exec rights */
1732                 mask = MAY_EXEC;
1733         } else if (fmode & FMODE_READ)
1734                 mask = MAY_READ;
1735
1736         cache.cred = cred;
1737         cache.jiffies = jiffies;
1738         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1739         nfs_access_add_cache(state->inode, &cache);
1740
1741         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1742                 return 0;
1743
1744         /* even though OPEN succeeded, access is denied. Close the file */
1745         nfs4_close_state(state, fmode);
1746         return -EACCES;
1747 }
1748
1749 /*
1750  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1751  */
1752 static int _nfs4_proc_open(struct nfs4_opendata *data)
1753 {
1754         struct inode *dir = data->dir->d_inode;
1755         struct nfs_server *server = NFS_SERVER(dir);
1756         struct nfs_openargs *o_arg = &data->o_arg;
1757         struct nfs_openres *o_res = &data->o_res;
1758         int status;
1759
1760         status = nfs4_run_open_task(data, 0);
1761         if (!data->rpc_done)
1762                 return status;
1763         if (status != 0) {
1764                 if (status == -NFS4ERR_BADNAME &&
1765                                 !(o_arg->open_flags & O_CREAT))
1766                         return -ENOENT;
1767                 return status;
1768         }
1769
1770         nfs_fattr_map_and_free_names(server, &data->f_attr);
1771
1772         if (o_arg->open_flags & O_CREAT)
1773                 update_changeattr(dir, &o_res->cinfo);
1774         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1775                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1776         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1777                 status = _nfs4_proc_open_confirm(data);
1778                 if (status != 0)
1779                         return status;
1780         }
1781         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1782                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1783         return 0;
1784 }
1785
1786 static int nfs4_recover_expired_lease(struct nfs_server *server)
1787 {
1788         return nfs4_client_recover_expired_lease(server->nfs_client);
1789 }
1790
1791 /*
1792  * OPEN_EXPIRED:
1793  *      reclaim state on the server after a network partition.
1794  *      Assumes caller holds the appropriate lock
1795  */
1796 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1797 {
1798         struct nfs4_opendata *opendata;
1799         int ret;
1800
1801         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1802                         NFS4_OPEN_CLAIM_FH);
1803         if (IS_ERR(opendata))
1804                 return PTR_ERR(opendata);
1805         ret = nfs4_open_recover(opendata, state);
1806         if (ret == -ESTALE)
1807                 d_drop(ctx->dentry);
1808         nfs4_opendata_put(opendata);
1809         return ret;
1810 }
1811
1812 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1813 {
1814         struct nfs_server *server = NFS_SERVER(state->inode);
1815         struct nfs4_exception exception = { };
1816         int err;
1817
1818         do {
1819                 err = _nfs4_open_expired(ctx, state);
1820                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1821                         continue;
1822                 switch (err) {
1823                 default:
1824                         goto out;
1825                 case -NFS4ERR_GRACE:
1826                 case -NFS4ERR_DELAY:
1827                         nfs4_handle_exception(server, err, &exception);
1828                         err = 0;
1829                 }
1830         } while (exception.retry);
1831 out:
1832         return err;
1833 }
1834
1835 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1836 {
1837         struct nfs_open_context *ctx;
1838         int ret;
1839
1840         ctx = nfs4_state_find_open_context(state);
1841         if (IS_ERR(ctx))
1842                 return -EAGAIN;
1843         ret = nfs4_do_open_expired(ctx, state);
1844         put_nfs_open_context(ctx);
1845         return ret;
1846 }
1847
1848 #if defined(CONFIG_NFS_V4_1)
1849 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1850 {
1851         struct nfs_server *server = NFS_SERVER(state->inode);
1852         nfs4_stateid *stateid = &state->stateid;
1853         int status;
1854
1855         /* If a state reset has been done, test_stateid is unneeded */
1856         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1857                 return;
1858
1859         status = nfs41_test_stateid(server, stateid);
1860         if (status != NFS_OK) {
1861                 /* Free the stateid unless the server explicitly
1862                  * informs us the stateid is unrecognized. */
1863                 if (status != -NFS4ERR_BAD_STATEID)
1864                         nfs41_free_stateid(server, stateid);
1865                 nfs_remove_bad_delegation(state->inode);
1866
1867                 write_seqlock(&state->seqlock);
1868                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1869                 write_sequnlock(&state->seqlock);
1870                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1871         }
1872 }
1873
1874 /**
1875  * nfs41_check_open_stateid - possibly free an open stateid
1876  *
1877  * @state: NFSv4 state for an inode
1878  *
1879  * Returns NFS_OK if recovery for this stateid is now finished.
1880  * Otherwise a negative NFS4ERR value is returned.
1881  */
1882 static int nfs41_check_open_stateid(struct nfs4_state *state)
1883 {
1884         struct nfs_server *server = NFS_SERVER(state->inode);
1885         nfs4_stateid *stateid = &state->open_stateid;
1886         int status;
1887
1888         /* If a state reset has been done, test_stateid is unneeded */
1889         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1890             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1891             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1892                 return -NFS4ERR_BAD_STATEID;
1893
1894         status = nfs41_test_stateid(server, stateid);
1895         if (status != NFS_OK) {
1896                 /* Free the stateid unless the server explicitly
1897                  * informs us the stateid is unrecognized. */
1898                 if (status != -NFS4ERR_BAD_STATEID)
1899                         nfs41_free_stateid(server, stateid);
1900
1901                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1902                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1903                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1904                 clear_bit(NFS_OPEN_STATE, &state->flags);
1905         }
1906         return status;
1907 }
1908
1909 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1910 {
1911         int status;
1912
1913         nfs41_clear_delegation_stateid(state);
1914         status = nfs41_check_open_stateid(state);
1915         if (status != NFS_OK)
1916                 status = nfs4_open_expired(sp, state);
1917         return status;
1918 }
1919 #endif
1920
1921 /*
1922  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1923  * fields corresponding to attributes that were used to store the verifier.
1924  * Make sure we clobber those fields in the later setattr call
1925  */
1926 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1927 {
1928         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1929             !(sattr->ia_valid & ATTR_ATIME_SET))
1930                 sattr->ia_valid |= ATTR_ATIME;
1931
1932         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1933             !(sattr->ia_valid & ATTR_MTIME_SET))
1934                 sattr->ia_valid |= ATTR_MTIME;
1935 }
1936
1937 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
1938                 fmode_t fmode,
1939                 int flags,
1940                 struct nfs4_state **res)
1941 {
1942         struct nfs4_state_owner *sp = opendata->owner;
1943         struct nfs_server *server = sp->so_server;
1944         struct nfs4_state *state;
1945         unsigned int seq;
1946         int ret;
1947
1948         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
1949
1950         ret = _nfs4_proc_open(opendata);
1951         if (ret != 0)
1952                 goto out;
1953
1954         state = nfs4_opendata_to_nfs4_state(opendata);
1955         ret = PTR_ERR(state);
1956         if (IS_ERR(state))
1957                 goto out;
1958         if (server->caps & NFS_CAP_POSIX_LOCK)
1959                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1960
1961         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
1962         if (ret != 0)
1963                 goto out;
1964
1965         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
1966                 nfs4_schedule_stateid_recovery(server, state);
1967         *res = state;
1968 out:
1969         return ret;
1970 }
1971
1972 /*
1973  * Returns a referenced nfs4_state
1974  */
1975 static int _nfs4_do_open(struct inode *dir,
1976                         struct dentry *dentry,
1977                         fmode_t fmode,
1978                         int flags,
1979                         struct iattr *sattr,
1980                         struct rpc_cred *cred,
1981                         struct nfs4_state **res,
1982                         struct nfs4_threshold **ctx_th)
1983 {
1984         struct nfs4_state_owner  *sp;
1985         struct nfs4_state     *state = NULL;
1986         struct nfs_server       *server = NFS_SERVER(dir);
1987         struct nfs4_opendata *opendata;
1988         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
1989         int status;
1990
1991         /* Protect against reboot recovery conflicts */
1992         status = -ENOMEM;
1993         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1994         if (sp == NULL) {
1995                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1996                 goto out_err;
1997         }
1998         status = nfs4_recover_expired_lease(server);
1999         if (status != 0)
2000                 goto err_put_state_owner;
2001         if (dentry->d_inode != NULL)
2002                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2003         status = -ENOMEM;
2004         if (dentry->d_inode)
2005                 claim = NFS4_OPEN_CLAIM_FH;
2006         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2007                         claim, GFP_KERNEL);
2008         if (opendata == NULL)
2009                 goto err_put_state_owner;
2010
2011         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2012                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2013                 if (!opendata->f_attr.mdsthreshold)
2014                         goto err_opendata_put;
2015                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2016         }
2017         if (dentry->d_inode != NULL)
2018                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2019
2020         status = _nfs4_open_and_get_state(opendata, fmode, flags, &state);
2021         if (status != 0)
2022                 goto err_opendata_put;
2023
2024         if ((opendata->o_arg.open_flags & O_EXCL) &&
2025             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2026                 nfs4_exclusive_attrset(opendata, sattr);
2027
2028                 nfs_fattr_init(opendata->o_res.f_attr);
2029                 status = nfs4_do_setattr(state->inode, cred,
2030                                 opendata->o_res.f_attr, sattr,
2031                                 state);
2032                 if (status == 0)
2033                         nfs_setattr_update_inode(state->inode, sattr);
2034                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2035         }
2036
2037         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2038                 *ctx_th = opendata->f_attr.mdsthreshold;
2039         else
2040                 kfree(opendata->f_attr.mdsthreshold);
2041         opendata->f_attr.mdsthreshold = NULL;
2042
2043         nfs4_opendata_put(opendata);
2044         nfs4_put_state_owner(sp);
2045         *res = state;
2046         return 0;
2047 err_opendata_put:
2048         kfree(opendata->f_attr.mdsthreshold);
2049         nfs4_opendata_put(opendata);
2050 err_put_state_owner:
2051         nfs4_put_state_owner(sp);
2052 out_err:
2053         *res = NULL;
2054         return status;
2055 }
2056
2057
2058 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2059                                         struct dentry *dentry,
2060                                         fmode_t fmode,
2061                                         int flags,
2062                                         struct iattr *sattr,
2063                                         struct rpc_cred *cred,
2064                                         struct nfs4_threshold **ctx_th)
2065 {
2066         struct nfs_server *server = NFS_SERVER(dir);
2067         struct nfs4_exception exception = { };
2068         struct nfs4_state *res;
2069         int status;
2070
2071         fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
2072         do {
2073                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
2074                                        &res, ctx_th);
2075                 if (status == 0)
2076                         break;
2077                 /* NOTE: BAD_SEQID means the server and client disagree about the
2078                  * book-keeping w.r.t. state-changing operations
2079                  * (OPEN/CLOSE/LOCK/LOCKU...)
2080                  * It is actually a sign of a bug on the client or on the server.
2081                  *
2082                  * If we receive a BAD_SEQID error in the particular case of
2083                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2084                  * have unhashed the old state_owner for us, and that we can
2085                  * therefore safely retry using a new one. We should still warn
2086                  * the user though...
2087                  */
2088                 if (status == -NFS4ERR_BAD_SEQID) {
2089                         pr_warn_ratelimited("NFS: v4 server %s "
2090                                         " returned a bad sequence-id error!\n",
2091                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2092                         exception.retry = 1;
2093                         continue;
2094                 }
2095                 /*
2096                  * BAD_STATEID on OPEN means that the server cancelled our
2097                  * state before it received the OPEN_CONFIRM.
2098                  * Recover by retrying the request as per the discussion
2099                  * on Page 181 of RFC3530.
2100                  */
2101                 if (status == -NFS4ERR_BAD_STATEID) {
2102                         exception.retry = 1;
2103                         continue;
2104                 }
2105                 if (status == -EAGAIN) {
2106                         /* We must have found a delegation */
2107                         exception.retry = 1;
2108                         continue;
2109                 }
2110                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2111                         continue;
2112                 res = ERR_PTR(nfs4_handle_exception(server,
2113                                         status, &exception));
2114         } while (exception.retry);
2115         return res;
2116 }
2117
2118 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2119                             struct nfs_fattr *fattr, struct iattr *sattr,
2120                             struct nfs4_state *state)
2121 {
2122         struct nfs_server *server = NFS_SERVER(inode);
2123         struct nfs_setattrargs  arg = {
2124                 .fh             = NFS_FH(inode),
2125                 .iap            = sattr,
2126                 .server         = server,
2127                 .bitmask = server->attr_bitmask,
2128         };
2129         struct nfs_setattrres  res = {
2130                 .fattr          = fattr,
2131                 .server         = server,
2132         };
2133         struct rpc_message msg = {
2134                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2135                 .rpc_argp       = &arg,
2136                 .rpc_resp       = &res,
2137                 .rpc_cred       = cred,
2138         };
2139         unsigned long timestamp = jiffies;
2140         fmode_t fmode;
2141         bool truncate;
2142         int status;
2143
2144         nfs_fattr_init(fattr);
2145
2146         /* Servers should only apply open mode checks for file size changes */
2147         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2148         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2149
2150         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2151                 /* Use that stateid */
2152         } else if (truncate && state != NULL && nfs4_valid_open_stateid(state)) {
2153                 struct nfs_lockowner lockowner = {
2154                         .l_owner = current->files,
2155                         .l_pid = current->tgid,
2156                 };
2157                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2158                                 &lockowner);
2159         } else
2160                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2161
2162         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2163         if (status == 0 && state != NULL)
2164                 renew_lease(server, timestamp);
2165         return status;
2166 }
2167
2168 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2169                            struct nfs_fattr *fattr, struct iattr *sattr,
2170                            struct nfs4_state *state)
2171 {
2172         struct nfs_server *server = NFS_SERVER(inode);
2173         struct nfs4_exception exception = {
2174                 .state = state,
2175                 .inode = inode,
2176         };
2177         int err;
2178         do {
2179                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2180                 switch (err) {
2181                 case -NFS4ERR_OPENMODE:
2182                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2183                                 pr_warn_once("NFSv4: server %s is incorrectly "
2184                                                 "applying open mode checks to "
2185                                                 "a SETATTR that is not "
2186                                                 "changing file size.\n",
2187                                                 server->nfs_client->cl_hostname);
2188                         }
2189                         if (state && !(state->state & FMODE_WRITE)) {
2190                                 err = -EBADF;
2191                                 if (sattr->ia_valid & ATTR_OPEN)
2192                                         err = -EACCES;
2193                                 goto out;
2194                         }
2195                 }
2196                 err = nfs4_handle_exception(server, err, &exception);
2197         } while (exception.retry);
2198 out:
2199         return err;
2200 }
2201
2202 struct nfs4_closedata {
2203         struct inode *inode;
2204         struct nfs4_state *state;
2205         struct nfs_closeargs arg;
2206         struct nfs_closeres res;
2207         struct nfs_fattr fattr;
2208         unsigned long timestamp;
2209         bool roc;
2210         u32 roc_barrier;
2211 };
2212
2213 static void nfs4_free_closedata(void *data)
2214 {
2215         struct nfs4_closedata *calldata = data;
2216         struct nfs4_state_owner *sp = calldata->state->owner;
2217         struct super_block *sb = calldata->state->inode->i_sb;
2218
2219         if (calldata->roc)
2220                 pnfs_roc_release(calldata->state->inode);
2221         nfs4_put_open_state(calldata->state);
2222         nfs_free_seqid(calldata->arg.seqid);
2223         nfs4_put_state_owner(sp);
2224         nfs_sb_deactive(sb);
2225         kfree(calldata);
2226 }
2227
2228 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2229                 fmode_t fmode)
2230 {
2231         spin_lock(&state->owner->so_lock);
2232         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2233         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
2234         case FMODE_WRITE:
2235                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2236                 break;
2237         case FMODE_READ:
2238                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2239                 break;
2240         case 0:
2241                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2242                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2243                 clear_bit(NFS_OPEN_STATE, &state->flags);
2244         }
2245         spin_unlock(&state->owner->so_lock);
2246 }
2247
2248 static void nfs4_close_done(struct rpc_task *task, void *data)
2249 {
2250         struct nfs4_closedata *calldata = data;
2251         struct nfs4_state *state = calldata->state;
2252         struct nfs_server *server = NFS_SERVER(calldata->inode);
2253
2254         dprintk("%s: begin!\n", __func__);
2255         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2256                 return;
2257         /* hmm. we are done with the inode, and in the process of freeing
2258          * the state_owner. we keep this around to process errors
2259          */
2260         switch (task->tk_status) {
2261                 case 0:
2262                         if (calldata->roc)
2263                                 pnfs_roc_set_barrier(state->inode,
2264                                                      calldata->roc_barrier);
2265                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2266                         renew_lease(server, calldata->timestamp);
2267                         nfs4_close_clear_stateid_flags(state,
2268                                         calldata->arg.fmode);
2269                         break;
2270                 case -NFS4ERR_STALE_STATEID:
2271                 case -NFS4ERR_OLD_STATEID:
2272                 case -NFS4ERR_BAD_STATEID:
2273                 case -NFS4ERR_EXPIRED:
2274                         if (calldata->arg.fmode == 0)
2275                                 break;
2276                 default:
2277                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2278                                 rpc_restart_call_prepare(task);
2279         }
2280         nfs_release_seqid(calldata->arg.seqid);
2281         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2282         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2283 }
2284
2285 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2286 {
2287         struct nfs4_closedata *calldata = data;
2288         struct nfs4_state *state = calldata->state;
2289         struct inode *inode = calldata->inode;
2290         int call_close = 0;
2291
2292         dprintk("%s: begin!\n", __func__);
2293         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2294                 goto out_wait;
2295
2296         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2297         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2298         spin_lock(&state->owner->so_lock);
2299         /* Calculate the change in open mode */
2300         if (state->n_rdwr == 0) {
2301                 if (state->n_rdonly == 0) {
2302                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2303                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2304                         calldata->arg.fmode &= ~FMODE_READ;
2305                 }
2306                 if (state->n_wronly == 0) {
2307                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2308                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2309                         calldata->arg.fmode &= ~FMODE_WRITE;
2310                 }
2311         }
2312         if (!nfs4_valid_open_stateid(state))
2313                 call_close = 0;
2314         spin_unlock(&state->owner->so_lock);
2315
2316         if (!call_close) {
2317                 /* Note: exit _without_ calling nfs4_close_done */
2318                 goto out_no_action;
2319         }
2320
2321         if (calldata->arg.fmode == 0) {
2322                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2323                 if (calldata->roc &&
2324                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2325                         nfs_release_seqid(calldata->arg.seqid);
2326                         goto out_wait;
2327                     }
2328         }
2329
2330         nfs_fattr_init(calldata->res.fattr);
2331         calldata->timestamp = jiffies;
2332         if (nfs4_setup_sequence(NFS_SERVER(inode),
2333                                 &calldata->arg.seq_args,
2334                                 &calldata->res.seq_res,
2335                                 task) != 0)
2336                 nfs_release_seqid(calldata->arg.seqid);
2337         dprintk("%s: done!\n", __func__);
2338         return;
2339 out_no_action:
2340         task->tk_action = NULL;
2341 out_wait:
2342         nfs4_sequence_done(task, &calldata->res.seq_res);
2343 }
2344
2345 static const struct rpc_call_ops nfs4_close_ops = {
2346         .rpc_call_prepare = nfs4_close_prepare,
2347         .rpc_call_done = nfs4_close_done,
2348         .rpc_release = nfs4_free_closedata,
2349 };
2350
2351 /*
2352  * It is possible for data to be read/written from a mem-mapped file
2353  * after the sys_close call (which hits the vfs layer as a flush).
2354  * This means that we can't safely call nfsv4 close on a file until
2355  * the inode is cleared. This in turn means that we are not good
2356  * NFSv4 citizens - we do not indicate to the server to update the file's
2357  * share state even when we are done with one of the three share
2358  * stateid's in the inode.
2359  *
2360  * NOTE: Caller must be holding the sp->so_owner semaphore!
2361  */
2362 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2363 {
2364         struct nfs_server *server = NFS_SERVER(state->inode);
2365         struct nfs4_closedata *calldata;
2366         struct nfs4_state_owner *sp = state->owner;
2367         struct rpc_task *task;
2368         struct rpc_message msg = {
2369                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2370                 .rpc_cred = state->owner->so_cred,
2371         };
2372         struct rpc_task_setup task_setup_data = {
2373                 .rpc_client = server->client,
2374                 .rpc_message = &msg,
2375                 .callback_ops = &nfs4_close_ops,
2376                 .workqueue = nfsiod_workqueue,
2377                 .flags = RPC_TASK_ASYNC,
2378         };
2379         int status = -ENOMEM;
2380
2381         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2382         if (calldata == NULL)
2383                 goto out;
2384         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2385         calldata->inode = state->inode;
2386         calldata->state = state;
2387         calldata->arg.fh = NFS_FH(state->inode);
2388         calldata->arg.stateid = &state->open_stateid;
2389         /* Serialization for the sequence id */
2390         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2391         if (calldata->arg.seqid == NULL)
2392                 goto out_free_calldata;
2393         calldata->arg.fmode = 0;
2394         calldata->arg.bitmask = server->cache_consistency_bitmask;
2395         calldata->res.fattr = &calldata->fattr;
2396         calldata->res.seqid = calldata->arg.seqid;
2397         calldata->res.server = server;
2398         calldata->roc = pnfs_roc(state->inode);
2399         nfs_sb_active(calldata->inode->i_sb);
2400
2401         msg.rpc_argp = &calldata->arg;
2402         msg.rpc_resp = &calldata->res;
2403         task_setup_data.callback_data = calldata;
2404         task = rpc_run_task(&task_setup_data);
2405         if (IS_ERR(task))
2406                 return PTR_ERR(task);
2407         status = 0;
2408         if (wait)
2409                 status = rpc_wait_for_completion_task(task);
2410         rpc_put_task(task);
2411         return status;
2412 out_free_calldata:
2413         kfree(calldata);
2414 out:
2415         nfs4_put_open_state(state);
2416         nfs4_put_state_owner(sp);
2417         return status;
2418 }
2419
2420 static struct inode *
2421 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2422 {
2423         struct nfs4_state *state;
2424
2425         /* Protect against concurrent sillydeletes */
2426         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2427                              ctx->cred, &ctx->mdsthreshold);
2428         if (IS_ERR(state))
2429                 return ERR_CAST(state);
2430         ctx->state = state;
2431         return igrab(state->inode);
2432 }
2433
2434 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2435 {
2436         if (ctx->state == NULL)
2437                 return;
2438         if (is_sync)
2439                 nfs4_close_sync(ctx->state, ctx->mode);
2440         else
2441                 nfs4_close_state(ctx->state, ctx->mode);
2442 }
2443
2444 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2445 {
2446         struct nfs4_server_caps_arg args = {
2447                 .fhandle = fhandle,
2448         };
2449         struct nfs4_server_caps_res res = {};
2450         struct rpc_message msg = {
2451                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2452                 .rpc_argp = &args,
2453                 .rpc_resp = &res,
2454         };
2455         int status;
2456
2457         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2458         if (status == 0) {
2459                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2460                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2461                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2462                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2463                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2464                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2465                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2466                         server->caps |= NFS_CAP_ACLS;
2467                 if (res.has_links != 0)
2468                         server->caps |= NFS_CAP_HARDLINKS;
2469                 if (res.has_symlinks != 0)
2470                         server->caps |= NFS_CAP_SYMLINKS;
2471                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2472                         server->caps |= NFS_CAP_FILEID;
2473                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2474                         server->caps |= NFS_CAP_MODE;
2475                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2476                         server->caps |= NFS_CAP_NLINK;
2477                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2478                         server->caps |= NFS_CAP_OWNER;
2479                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2480                         server->caps |= NFS_CAP_OWNER_GROUP;
2481                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2482                         server->caps |= NFS_CAP_ATIME;
2483                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2484                         server->caps |= NFS_CAP_CTIME;
2485                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2486                         server->caps |= NFS_CAP_MTIME;
2487
2488                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2489                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2490                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2491                 server->acl_bitmask = res.acl_bitmask;
2492                 server->fh_expire_type = res.fh_expire_type;
2493         }
2494
2495         return status;
2496 }
2497
2498 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2499 {
2500         struct nfs4_exception exception = { };
2501         int err;
2502         do {
2503                 err = nfs4_handle_exception(server,
2504                                 _nfs4_server_capabilities(server, fhandle),
2505                                 &exception);
2506         } while (exception.retry);
2507         return err;
2508 }
2509
2510 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2511                 struct nfs_fsinfo *info)
2512 {
2513         struct nfs4_lookup_root_arg args = {
2514                 .bitmask = nfs4_fattr_bitmap,
2515         };
2516         struct nfs4_lookup_res res = {
2517                 .server = server,
2518                 .fattr = info->fattr,
2519                 .fh = fhandle,
2520         };
2521         struct rpc_message msg = {
2522                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2523                 .rpc_argp = &args,
2524                 .rpc_resp = &res,
2525         };
2526
2527         nfs_fattr_init(info->fattr);
2528         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2529 }
2530
2531 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2532                 struct nfs_fsinfo *info)
2533 {
2534         struct nfs4_exception exception = { };
2535         int err;
2536         do {
2537                 err = _nfs4_lookup_root(server, fhandle, info);
2538                 switch (err) {
2539                 case 0:
2540                 case -NFS4ERR_WRONGSEC:
2541                         goto out;
2542                 default:
2543                         err = nfs4_handle_exception(server, err, &exception);
2544                 }
2545         } while (exception.retry);
2546 out:
2547         return err;
2548 }
2549
2550 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2551                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2552 {
2553         struct rpc_auth *auth;
2554         int ret;
2555
2556         auth = rpcauth_create(flavor, server->client);
2557         if (IS_ERR(auth)) {
2558                 ret = -EACCES;
2559                 goto out;
2560         }
2561         ret = nfs4_lookup_root(server, fhandle, info);
2562 out:
2563         return ret;
2564 }
2565
2566 /*
2567  * Retry pseudoroot lookup with various security flavors.  We do this when:
2568  *
2569  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2570  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2571  *
2572  * Returns zero on success, or a negative NFS4ERR value, or a
2573  * negative errno value.
2574  */
2575 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2576                               struct nfs_fsinfo *info)
2577 {
2578         /* Per 3530bis 15.33.5 */
2579         static const rpc_authflavor_t flav_array[] = {
2580                 RPC_AUTH_GSS_KRB5P,
2581                 RPC_AUTH_GSS_KRB5I,
2582                 RPC_AUTH_GSS_KRB5,
2583                 RPC_AUTH_UNIX,                  /* courtesy */
2584                 RPC_AUTH_NULL,
2585         };
2586         int status = -EPERM;
2587         size_t i;
2588
2589         for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
2590                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2591                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2592                         continue;
2593                 break;
2594         }
2595
2596         /*
2597          * -EACCESS could mean that the user doesn't have correct permissions
2598          * to access the mount.  It could also mean that we tried to mount
2599          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2600          * existing mount programs don't handle -EACCES very well so it should
2601          * be mapped to -EPERM instead.
2602          */
2603         if (status == -EACCES)
2604                 status = -EPERM;
2605         return status;
2606 }
2607
2608 static int nfs4_do_find_root_sec(struct nfs_server *server,
2609                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
2610 {
2611         int mv = server->nfs_client->cl_minorversion;
2612         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
2613 }
2614
2615 /**
2616  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
2617  * @server: initialized nfs_server handle
2618  * @fhandle: we fill in the pseudo-fs root file handle
2619  * @info: we fill in an FSINFO struct
2620  *
2621  * Returns zero on success, or a negative errno.
2622  */
2623 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2624                          struct nfs_fsinfo *info)
2625 {
2626         int status;
2627
2628         status = nfs4_lookup_root(server, fhandle, info);
2629         if ((status == -NFS4ERR_WRONGSEC) &&
2630             !(server->flags & NFS_MOUNT_SECFLAVOUR))
2631                 status = nfs4_do_find_root_sec(server, fhandle, info);
2632
2633         if (status == 0)
2634                 status = nfs4_server_capabilities(server, fhandle);
2635         if (status == 0)
2636                 status = nfs4_do_fsinfo(server, fhandle, info);
2637
2638         return nfs4_map_errors(status);
2639 }
2640
2641 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2642                               struct nfs_fsinfo *info)
2643 {
2644         int error;
2645         struct nfs_fattr *fattr = info->fattr;
2646
2647         error = nfs4_server_capabilities(server, mntfh);
2648         if (error < 0) {
2649                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2650                 return error;
2651         }
2652
2653         error = nfs4_proc_getattr(server, mntfh, fattr);
2654         if (error < 0) {
2655                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2656                 return error;
2657         }
2658
2659         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2660             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2661                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2662
2663         return error;
2664 }
2665
2666 /*
2667  * Get locations and (maybe) other attributes of a referral.
2668  * Note that we'll actually follow the referral later when
2669  * we detect fsid mismatch in inode revalidation
2670  */
2671 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2672                              const struct qstr *name, struct nfs_fattr *fattr,
2673                              struct nfs_fh *fhandle)
2674 {
2675         int status = -ENOMEM;
2676         struct page *page = NULL;
2677         struct nfs4_fs_locations *locations = NULL;
2678
2679         page = alloc_page(GFP_KERNEL);
2680         if (page == NULL)
2681                 goto out;
2682         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2683         if (locations == NULL)
2684                 goto out;
2685
2686         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2687         if (status != 0)
2688                 goto out;
2689         /* Make sure server returned a different fsid for the referral */
2690         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2691                 dprintk("%s: server did not return a different fsid for"
2692                         " a referral at %s\n", __func__, name->name);
2693                 status = -EIO;
2694                 goto out;
2695         }
2696         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2697         nfs_fixup_referral_attributes(&locations->fattr);
2698
2699         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2700         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2701         memset(fhandle, 0, sizeof(struct nfs_fh));
2702 out:
2703         if (page)
2704                 __free_page(page);
2705         kfree(locations);
2706         return status;
2707 }
2708
2709 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2710 {
2711         struct nfs4_getattr_arg args = {
2712                 .fh = fhandle,
2713                 .bitmask = server->attr_bitmask,
2714         };
2715         struct nfs4_getattr_res res = {
2716                 .fattr = fattr,
2717                 .server = server,
2718         };
2719         struct rpc_message msg = {
2720                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2721                 .rpc_argp = &args,
2722                 .rpc_resp = &res,
2723         };
2724
2725         nfs_fattr_init(fattr);
2726         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2727 }
2728
2729 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2730 {
2731         struct nfs4_exception exception = { };
2732         int err;
2733         do {
2734                 err = nfs4_handle_exception(server,
2735                                 _nfs4_proc_getattr(server, fhandle, fattr),
2736                                 &exception);
2737         } while (exception.retry);
2738         return err;
2739 }
2740
2741 /*
2742  * The file is not closed if it is opened due to the a request to change
2743  * the size of the file. The open call will not be needed once the
2744  * VFS layer lookup-intents are implemented.
2745  *
2746  * Close is called when the inode is destroyed.
2747  * If we haven't opened the file for O_WRONLY, we
2748  * need to in the size_change case to obtain a stateid.
2749  *
2750  * Got race?
2751  * Because OPEN is always done by name in nfsv4, it is
2752  * possible that we opened a different file by the same
2753  * name.  We can recognize this race condition, but we
2754  * can't do anything about it besides returning an error.
2755  *
2756  * This will be fixed with VFS changes (lookup-intent).
2757  */
2758 static int
2759 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2760                   struct iattr *sattr)
2761 {
2762         struct inode *inode = dentry->d_inode;
2763         struct rpc_cred *cred = NULL;
2764         struct nfs4_state *state = NULL;
2765         int status;
2766
2767         if (pnfs_ld_layoutret_on_setattr(inode))
2768                 pnfs_commit_and_return_layout(inode);
2769
2770         nfs_fattr_init(fattr);
2771
2772         /* Deal with open(O_TRUNC) */
2773         if (sattr->ia_valid & ATTR_OPEN)
2774                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2775
2776         /* Optimization: if the end result is no change, don't RPC */
2777         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2778                 return 0;
2779
2780         /* Search for an existing open(O_WRITE) file */
2781         if (sattr->ia_valid & ATTR_FILE) {
2782                 struct nfs_open_context *ctx;
2783
2784                 ctx = nfs_file_open_context(sattr->ia_file);
2785                 if (ctx) {
2786                         cred = ctx->cred;
2787                         state = ctx->state;
2788                 }
2789         }
2790
2791         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2792         if (status == 0)
2793                 nfs_setattr_update_inode(inode, sattr);
2794         return status;
2795 }
2796
2797 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2798                 const struct qstr *name, struct nfs_fh *fhandle,
2799                 struct nfs_fattr *fattr)
2800 {
2801         struct nfs_server *server = NFS_SERVER(dir);
2802         int                    status;
2803         struct nfs4_lookup_arg args = {
2804                 .bitmask = server->attr_bitmask,
2805                 .dir_fh = NFS_FH(dir),
2806                 .name = name,
2807         };
2808         struct nfs4_lookup_res res = {
2809                 .server = server,
2810                 .fattr = fattr,
2811                 .fh = fhandle,
2812         };
2813         struct rpc_message msg = {
2814                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2815                 .rpc_argp = &args,
2816                 .rpc_resp = &res,
2817         };
2818
2819         nfs_fattr_init(fattr);
2820
2821         dprintk("NFS call  lookup %s\n", name->name);
2822         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2823         dprintk("NFS reply lookup: %d\n", status);
2824         return status;
2825 }
2826
2827 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2828 {
2829         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2830                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2831         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2832         fattr->nlink = 2;
2833 }
2834
2835 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2836                                    struct qstr *name, struct nfs_fh *fhandle,
2837                                    struct nfs_fattr *fattr)
2838 {
2839         struct nfs4_exception exception = { };
2840         struct rpc_clnt *client = *clnt;
2841         int err;
2842         do {
2843                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2844                 switch (err) {
2845                 case -NFS4ERR_BADNAME:
2846                         err = -ENOENT;
2847                         goto out;
2848                 case -NFS4ERR_MOVED:
2849                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2850                         goto out;
2851                 case -NFS4ERR_WRONGSEC:
2852                         err = -EPERM;
2853                         if (client != *clnt)
2854                                 goto out;
2855
2856                         client = nfs4_create_sec_client(client, dir, name);
2857                         if (IS_ERR(client))
2858                                 return PTR_ERR(client);
2859
2860                         exception.retry = 1;
2861                         break;
2862                 default:
2863                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2864                 }
2865         } while (exception.retry);
2866
2867 out:
2868         if (err == 0)
2869                 *clnt = client;
2870         else if (client != *clnt)
2871                 rpc_shutdown_client(client);
2872
2873         return err;
2874 }
2875
2876 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2877                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2878 {
2879         int status;
2880         struct rpc_clnt *client = NFS_CLIENT(dir);
2881
2882         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2883         if (client != NFS_CLIENT(dir)) {
2884                 rpc_shutdown_client(client);
2885                 nfs_fixup_secinfo_attributes(fattr);
2886         }
2887         return status;
2888 }
2889
2890 struct rpc_clnt *
2891 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2892                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2893 {
2894         int status;
2895         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2896
2897         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2898         if (status < 0) {
2899                 rpc_shutdown_client(client);
2900                 return ERR_PTR(status);
2901         }
2902         return client;
2903 }
2904
2905 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2906 {
2907         struct nfs_server *server = NFS_SERVER(inode);
2908         struct nfs4_accessargs args = {
2909                 .fh = NFS_FH(inode),
2910                 .bitmask = server->cache_consistency_bitmask,
2911         };
2912         struct nfs4_accessres res = {
2913                 .server = server,
2914         };
2915         struct rpc_message msg = {
2916                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2917                 .rpc_argp = &args,
2918                 .rpc_resp = &res,
2919                 .rpc_cred = entry->cred,
2920         };
2921         int mode = entry->mask;
2922         int status;
2923
2924         /*
2925          * Determine which access bits we want to ask for...
2926          */
2927         if (mode & MAY_READ)
2928                 args.access |= NFS4_ACCESS_READ;
2929         if (S_ISDIR(inode->i_mode)) {
2930                 if (mode & MAY_WRITE)
2931                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2932                 if (mode & MAY_EXEC)
2933                         args.access |= NFS4_ACCESS_LOOKUP;
2934         } else {
2935                 if (mode & MAY_WRITE)
2936                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2937                 if (mode & MAY_EXEC)
2938                         args.access |= NFS4_ACCESS_EXECUTE;
2939         }
2940
2941         res.fattr = nfs_alloc_fattr();
2942         if (res.fattr == NULL)
2943                 return -ENOMEM;
2944
2945         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2946         if (!status) {
2947                 nfs_access_set_mask(entry, res.access);
2948                 nfs_refresh_inode(inode, res.fattr);
2949         }
2950         nfs_free_fattr(res.fattr);
2951         return status;
2952 }
2953
2954 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2955 {
2956         struct nfs4_exception exception = { };
2957         int err;
2958         do {
2959                 err = nfs4_handle_exception(NFS_SERVER(inode),
2960                                 _nfs4_proc_access(inode, entry),
2961                                 &exception);
2962         } while (exception.retry);
2963         return err;
2964 }
2965
2966 /*
2967  * TODO: For the time being, we don't try to get any attributes
2968  * along with any of the zero-copy operations READ, READDIR,
2969  * READLINK, WRITE.
2970  *
2971  * In the case of the first three, we want to put the GETATTR
2972  * after the read-type operation -- this is because it is hard
2973  * to predict the length of a GETATTR response in v4, and thus
2974  * align the READ data correctly.  This means that the GETATTR
2975  * may end up partially falling into the page cache, and we should
2976  * shift it into the 'tail' of the xdr_buf before processing.
2977  * To do this efficiently, we need to know the total length
2978  * of data received, which doesn't seem to be available outside
2979  * of the RPC layer.
2980  *
2981  * In the case of WRITE, we also want to put the GETATTR after
2982  * the operation -- in this case because we want to make sure
2983  * we get the post-operation mtime and size.
2984  *
2985  * Both of these changes to the XDR layer would in fact be quite
2986  * minor, but I decided to leave them for a subsequent patch.
2987  */
2988 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2989                 unsigned int pgbase, unsigned int pglen)
2990 {
2991         struct nfs4_readlink args = {
2992                 .fh       = NFS_FH(inode),
2993                 .pgbase   = pgbase,
2994                 .pglen    = pglen,
2995                 .pages    = &page,
2996         };
2997         struct nfs4_readlink_res res;
2998         struct rpc_message msg = {
2999                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3000                 .rpc_argp = &args,
3001                 .rpc_resp = &res,
3002         };
3003
3004         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3005 }
3006
3007 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3008                 unsigned int pgbase, unsigned int pglen)
3009 {
3010         struct nfs4_exception exception = { };
3011         int err;
3012         do {
3013                 err = nfs4_handle_exception(NFS_SERVER(inode),
3014                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
3015                                 &exception);
3016         } while (exception.retry);
3017         return err;
3018 }
3019
3020 /*
3021  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3022  */
3023 static int
3024 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3025                  int flags)
3026 {
3027         struct nfs_open_context *ctx;
3028         struct nfs4_state *state;
3029         int status = 0;
3030
3031         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3032         if (IS_ERR(ctx))
3033                 return PTR_ERR(ctx);
3034
3035         sattr->ia_mode &= ~current_umask();
3036         state = nfs4_do_open(dir, dentry, ctx->mode,
3037                         flags, sattr, ctx->cred,
3038                         &ctx->mdsthreshold);
3039         d_drop(dentry);
3040         if (IS_ERR(state)) {
3041                 status = PTR_ERR(state);
3042                 goto out;
3043         }
3044         d_add(dentry, igrab(state->inode));
3045         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
3046         ctx->state = state;
3047 out:
3048         put_nfs_open_context(ctx);
3049         return status;
3050 }
3051
3052 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3053 {
3054         struct nfs_server *server = NFS_SERVER(dir);
3055         struct nfs_removeargs args = {
3056                 .fh = NFS_FH(dir),
3057                 .name = *name,
3058         };
3059         struct nfs_removeres res = {
3060                 .server = server,
3061         };
3062         struct rpc_message msg = {
3063                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3064                 .rpc_argp = &args,
3065                 .rpc_resp = &res,
3066         };
3067         int status;
3068
3069         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3070         if (status == 0)
3071                 update_changeattr(dir, &res.cinfo);
3072         return status;
3073 }
3074
3075 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3076 {
3077         struct nfs4_exception exception = { };
3078         int err;
3079         do {
3080                 err = nfs4_handle_exception(NFS_SERVER(dir),
3081                                 _nfs4_proc_remove(dir, name),
3082                                 &exception);
3083         } while (exception.retry);
3084         return err;
3085 }
3086
3087 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3088 {
3089         struct nfs_server *server = NFS_SERVER(dir);
3090         struct nfs_removeargs *args = msg->rpc_argp;
3091         struct nfs_removeres *res = msg->rpc_resp;
3092
3093         res->server = server;
3094         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3095         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3096 }
3097
3098 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3099 {
3100         nfs4_setup_sequence(NFS_SERVER(data->dir),
3101                         &data->args.seq_args,
3102                         &data->res.seq_res,
3103                         task);
3104 }
3105
3106 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3107 {
3108         struct nfs_removeres *res = task->tk_msg.rpc_resp;
3109
3110         if (!nfs4_sequence_done(task, &res->seq_res))
3111                 return 0;
3112         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3113                 return 0;
3114         update_changeattr(dir, &res->cinfo);
3115         return 1;
3116 }
3117
3118 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3119 {
3120         struct nfs_server *server = NFS_SERVER(dir);
3121         struct nfs_renameargs *arg = msg->rpc_argp;
3122         struct nfs_renameres *res = msg->rpc_resp;
3123
3124         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3125         res->server = server;
3126         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3127 }
3128
3129 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3130 {
3131         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3132                         &data->args.seq_args,
3133                         &data->res.seq_res,
3134                         task);
3135 }
3136
3137 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3138                                  struct inode *new_dir)
3139 {
3140         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3141
3142         if (!nfs4_sequence_done(task, &res->seq_res))
3143                 return 0;
3144         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3145                 return 0;
3146
3147         update_changeattr(old_dir, &res->old_cinfo);
3148         update_changeattr(new_dir, &res->new_cinfo);
3149         return 1;
3150 }
3151
3152 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3153                 struct inode *new_dir, struct qstr *new_name)
3154 {
3155         struct nfs_server *server = NFS_SERVER(old_dir);
3156         struct nfs_renameargs arg = {
3157                 .old_dir = NFS_FH(old_dir),
3158                 .new_dir = NFS_FH(new_dir),
3159                 .old_name = old_name,
3160                 .new_name = new_name,
3161         };
3162         struct nfs_renameres res = {
3163                 .server = server,
3164         };
3165         struct rpc_message msg = {
3166                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3167                 .rpc_argp = &arg,
3168                 .rpc_resp = &res,
3169         };
3170         int status = -ENOMEM;
3171
3172         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3173         if (!status) {
3174                 update_changeattr(old_dir, &res.old_cinfo);
3175                 update_changeattr(new_dir, &res.new_cinfo);
3176         }
3177         return status;
3178 }
3179
3180 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3181                 struct inode *new_dir, struct qstr *new_name)
3182 {
3183         struct nfs4_exception exception = { };
3184         int err;
3185         do {
3186                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3187                                 _nfs4_proc_rename(old_dir, old_name,
3188                                         new_dir, new_name),
3189                                 &exception);
3190         } while (exception.retry);
3191         return err;
3192 }
3193
3194 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3195 {
3196         struct nfs_server *server = NFS_SERVER(inode);
3197         struct nfs4_link_arg arg = {
3198                 .fh     = NFS_FH(inode),
3199                 .dir_fh = NFS_FH(dir),
3200                 .name   = name,
3201                 .bitmask = server->attr_bitmask,
3202         };
3203         struct nfs4_link_res res = {
3204                 .server = server,
3205         };
3206         struct rpc_message msg = {
3207                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3208                 .rpc_argp = &arg,
3209                 .rpc_resp = &res,
3210         };
3211         int status = -ENOMEM;
3212
3213         res.fattr = nfs_alloc_fattr();
3214         if (res.fattr == NULL)
3215                 goto out;
3216
3217         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3218         if (!status) {
3219                 update_changeattr(dir, &res.cinfo);
3220                 nfs_post_op_update_inode(inode, res.fattr);
3221         }
3222 out:
3223         nfs_free_fattr(res.fattr);
3224         return status;
3225 }
3226
3227 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3228 {
3229         struct nfs4_exception exception = { };
3230         int err;
3231         do {
3232                 err = nfs4_handle_exception(NFS_SERVER(inode),
3233                                 _nfs4_proc_link(inode, dir, name),
3234                                 &exception);
3235         } while (exception.retry);
3236         return err;
3237 }
3238
3239 struct nfs4_createdata {
3240         struct rpc_message msg;
3241         struct nfs4_create_arg arg;
3242         struct nfs4_create_res res;
3243         struct nfs_fh fh;
3244         struct nfs_fattr fattr;
3245 };
3246
3247 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3248                 struct qstr *name, struct iattr *sattr, u32 ftype)
3249 {
3250         struct nfs4_createdata *data;
3251
3252         data = kzalloc(sizeof(*data), GFP_KERNEL);
3253         if (data != NULL) {
3254                 struct nfs_server *server = NFS_SERVER(dir);
3255
3256                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3257                 data->msg.rpc_argp = &data->arg;
3258                 data->msg.rpc_resp = &data->res;
3259                 data->arg.dir_fh = NFS_FH(dir);
3260                 data->arg.server = server;
3261                 data->arg.name = name;
3262                 data->arg.attrs = sattr;
3263                 data->arg.ftype = ftype;
3264                 data->arg.bitmask = server->attr_bitmask;
3265                 data->res.server = server;
3266                 data->res.fh = &data->fh;
3267                 data->res.fattr = &data->fattr;
3268                 nfs_fattr_init(data->res.fattr);
3269         }
3270         return data;
3271 }
3272
3273 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3274 {
3275         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3276                                     &data->arg.seq_args, &data->res.seq_res, 1);
3277         if (status == 0) {
3278                 update_changeattr(dir, &data->res.dir_cinfo);
3279                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3280         }
3281         return status;
3282 }
3283
3284 static void nfs4_free_createdata(struct nfs4_createdata *data)
3285 {
3286         kfree(data);
3287 }
3288
3289 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3290                 struct page *page, unsigned int len, struct iattr *sattr)
3291 {
3292         struct nfs4_createdata *data;
3293         int status = -ENAMETOOLONG;
3294
3295         if (len > NFS4_MAXPATHLEN)
3296                 goto out;
3297
3298         status = -ENOMEM;
3299         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3300         if (data == NULL)
3301                 goto out;
3302
3303         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3304         data->arg.u.symlink.pages = &page;
3305         data->arg.u.symlink.len = len;
3306
3307         status = nfs4_do_create(dir, dentry, data);
3308
3309         nfs4_free_createdata(data);
3310 out:
3311         return status;
3312 }
3313
3314 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3315                 struct page *page, unsigned int len, struct iattr *sattr)
3316 {
3317         struct nfs4_exception exception = { };
3318         int err;
3319         do {
3320                 err = nfs4_handle_exception(NFS_SERVER(dir),
3321                                 _nfs4_proc_symlink(dir, dentry, page,
3322                                                         len, sattr),
3323                                 &exception);
3324         } while (exception.retry);
3325         return err;
3326 }
3327
3328 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3329                 struct iattr *sattr)
3330 {
3331         struct nfs4_createdata *data;
3332         int status = -ENOMEM;
3333
3334         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3335         if (data == NULL)
3336                 goto out;
3337
3338         status = nfs4_do_create(dir, dentry, data);
3339
3340         nfs4_free_createdata(data);
3341 out:
3342         return status;
3343 }
3344
3345 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3346                 struct iattr *sattr)
3347 {
3348         struct nfs4_exception exception = { };
3349         int err;
3350
3351         sattr->ia_mode &= ~current_umask();
3352         do {
3353                 err = nfs4_handle_exception(NFS_SERVER(dir),
3354                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3355                                 &exception);
3356         } while (exception.retry);
3357         return err;
3358 }
3359
3360 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3361                 u64 cookie, struct page **pages, unsigned int count, int plus)
3362 {
3363         struct inode            *dir = dentry->d_inode;
3364         struct nfs4_readdir_arg args = {
3365                 .fh = NFS_FH(dir),
3366                 .pages = pages,
3367                 .pgbase = 0,
3368                 .count = count,
3369                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3370                 .plus = plus,
3371         };
3372         struct nfs4_readdir_res res;
3373         struct rpc_message msg = {
3374                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3375                 .rpc_argp = &args,
3376                 .rpc_resp = &res,
3377                 .rpc_cred = cred,
3378         };
3379         int                     status;
3380
3381         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3382                         dentry->d_parent->d_name.name,
3383                         dentry->d_name.name,
3384                         (unsigned long long)cookie);
3385         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3386         res.pgbase = args.pgbase;
3387         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3388         if (status >= 0) {
3389                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3390                 status += args.pgbase;
3391         }
3392
3393         nfs_invalidate_atime(dir);
3394
3395         dprintk("%s: returns %d\n", __func__, status);
3396         return status;
3397 }
3398
3399 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3400                 u64 cookie, struct page **pages, unsigned int count, int plus)
3401 {
3402         struct nfs4_exception exception = { };
3403         int err;
3404         do {
3405                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3406                                 _nfs4_proc_readdir(dentry, cred, cookie,
3407                                         pages, count, plus),
3408                                 &exception);
3409         } while (exception.retry);
3410         return err;
3411 }
3412
3413 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3414                 struct iattr *sattr, dev_t rdev)
3415 {
3416         struct nfs4_createdata *data;
3417         int mode = sattr->ia_mode;
3418         int status = -ENOMEM;
3419
3420         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3421         if (data == NULL)
3422                 goto out;
3423
3424         if (S_ISFIFO(mode))
3425                 data->arg.ftype = NF4FIFO;
3426         else if (S_ISBLK(mode)) {
3427                 data->arg.ftype = NF4BLK;
3428                 data->arg.u.device.specdata1 = MAJOR(rdev);
3429                 data->arg.u.device.specdata2 = MINOR(rdev);
3430         }
3431         else if (S_ISCHR(mode)) {
3432                 data->arg.ftype = NF4CHR;
3433                 data->arg.u.device.specdata1 = MAJOR(rdev);
3434                 data->arg.u.device.specdata2 = MINOR(rdev);
3435         } else if (!S_ISSOCK(mode)) {
3436                 status = -EINVAL;
3437                 goto out_free;
3438         }
3439
3440         status = nfs4_do_create(dir, dentry, data);
3441 out_free:
3442         nfs4_free_createdata(data);
3443 out:
3444         return status;
3445 }
3446
3447 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3448                 struct iattr *sattr, dev_t rdev)
3449 {
3450         struct nfs4_exception exception = { };
3451         int err;
3452
3453         sattr->ia_mode &= ~current_umask();
3454         do {
3455                 err = nfs4_handle_exception(NFS_SERVER(dir),
3456                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3457                                 &exception);
3458         } while (exception.retry);
3459         return err;
3460 }
3461
3462 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3463                  struct nfs_fsstat *fsstat)
3464 {
3465         struct nfs4_statfs_arg args = {
3466                 .fh = fhandle,
3467                 .bitmask = server->attr_bitmask,
3468         };
3469         struct nfs4_statfs_res res = {
3470                 .fsstat = fsstat,
3471         };
3472         struct rpc_message msg = {
3473                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3474                 .rpc_argp = &args,
3475                 .rpc_resp = &res,
3476         };
3477
3478         nfs_fattr_init(fsstat->fattr);
3479         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3480 }
3481
3482 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3483 {
3484         struct nfs4_exception exception = { };
3485         int err;
3486         do {
3487                 err = nfs4_handle_exception(server,
3488                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3489                                 &exception);
3490         } while (exception.retry);
3491         return err;
3492 }
3493
3494 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3495                 struct nfs_fsinfo *fsinfo)
3496 {
3497         struct nfs4_fsinfo_arg args = {
3498                 .fh = fhandle,
3499                 .bitmask = server->attr_bitmask,
3500         };
3501         struct nfs4_fsinfo_res res = {
3502                 .fsinfo = fsinfo,
3503         };
3504         struct rpc_message msg = {
3505                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3506                 .rpc_argp = &args,
3507                 .rpc_resp = &res,
3508         };
3509
3510         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3511 }
3512
3513 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3514 {
3515         struct nfs4_exception exception = { };
3516         unsigned long now = jiffies;
3517         int err;
3518
3519         do {
3520                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
3521                 if (err == 0) {
3522                         struct nfs_client *clp = server->nfs_client;
3523
3524                         spin_lock(&clp->cl_lock);
3525                         clp->cl_lease_time = fsinfo->lease_time * HZ;
3526                         clp->cl_last_renewal = now;
3527                         spin_unlock(&clp->cl_lock);
3528                         break;
3529                 }
3530                 err = nfs4_handle_exception(server, err, &exception);
3531         } while (exception.retry);
3532         return err;
3533 }
3534
3535 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3536 {
3537         int error;
3538
3539         nfs_fattr_init(fsinfo->fattr);
3540         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3541         if (error == 0) {
3542                 /* block layout checks this! */
3543                 server->pnfs_blksize = fsinfo->blksize;
3544                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3545         }
3546
3547         return error;
3548 }
3549
3550 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3551                 struct nfs_pathconf *pathconf)
3552 {
3553         struct nfs4_pathconf_arg args = {
3554                 .fh = fhandle,
3555                 .bitmask = server->attr_bitmask,
3556         };
3557         struct nfs4_pathconf_res res = {
3558                 .pathconf = pathconf,
3559         };
3560         struct rpc_message msg = {
3561                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3562                 .rpc_argp = &args,
3563                 .rpc_resp = &res,
3564         };
3565
3566         /* None of the pathconf attributes are mandatory to implement */
3567         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3568                 memset(pathconf, 0, sizeof(*pathconf));
3569                 return 0;
3570         }
3571
3572         nfs_fattr_init(pathconf->fattr);
3573         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3574 }
3575
3576 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3577                 struct nfs_pathconf *pathconf)
3578 {
3579         struct nfs4_exception exception = { };
3580         int err;
3581
3582         do {
3583                 err = nfs4_handle_exception(server,
3584                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3585                                 &exception);
3586         } while (exception.retry);
3587         return err;
3588 }
3589
3590 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
3591                 const struct nfs_open_context *ctx,
3592                 const struct nfs_lock_context *l_ctx,
3593                 fmode_t fmode)
3594 {
3595         const struct nfs_lockowner *lockowner = NULL;
3596
3597         if (l_ctx != NULL)
3598                 lockowner = &l_ctx->lockowner;
3599         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
3600 }
3601 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
3602
3603 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
3604                 const struct nfs_open_context *ctx,
3605                 const struct nfs_lock_context *l_ctx,
3606                 fmode_t fmode)
3607 {
3608         nfs4_stateid current_stateid;
3609
3610         /* If the current stateid represents a lost lock, then exit */
3611         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
3612                 return true;
3613         return nfs4_stateid_match(stateid, &current_stateid);
3614 }
3615
3616 static bool nfs4_error_stateid_expired(int err)
3617 {
3618         switch (err) {
3619         case -NFS4ERR_DELEG_REVOKED:
3620         case -NFS4ERR_ADMIN_REVOKED:
3621         case -NFS4ERR_BAD_STATEID:
3622         case -NFS4ERR_STALE_STATEID:
3623         case -NFS4ERR_OLD_STATEID:
3624         case -NFS4ERR_OPENMODE:
3625         case -NFS4ERR_EXPIRED:
3626                 return true;
3627         }
3628         return false;
3629 }
3630
3631 void __nfs4_read_done_cb(struct nfs_read_data *data)
3632 {
3633         nfs_invalidate_atime(data->header->inode);
3634 }
3635
3636 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3637 {
3638         struct nfs_server *server = NFS_SERVER(data->header->inode);
3639
3640         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3641                 rpc_restart_call_prepare(task);
3642                 return -EAGAIN;
3643         }
3644
3645         __nfs4_read_done_cb(data);
3646         if (task->tk_status > 0)
3647                 renew_lease(server, data->timestamp);
3648         return 0;
3649 }
3650
3651 static bool nfs4_read_stateid_changed(struct rpc_task *task,
3652                 struct nfs_readargs *args)
3653 {
3654
3655         if (!nfs4_error_stateid_expired(task->tk_status) ||
3656                 nfs4_stateid_is_current(&args->stateid,
3657                                 args->context,
3658                                 args->lock_context,
3659                                 FMODE_READ))
3660                 return false;
3661         rpc_restart_call_prepare(task);
3662         return true;
3663 }
3664
3665 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3666 {
3667
3668         dprintk("--> %s\n", __func__);
3669
3670         if (!nfs4_sequence_done(task, &data->res.seq_res))
3671                 return -EAGAIN;
3672         if (nfs4_read_stateid_changed(task, &data->args))
3673                 return -EAGAIN;
3674         return data->read_done_cb ? data->read_done_cb(task, data) :
3675                                     nfs4_read_done_cb(task, data);
3676 }
3677
3678 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3679 {
3680         data->timestamp   = jiffies;
3681         data->read_done_cb = nfs4_read_done_cb;
3682         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3683         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3684 }
3685
3686 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3687 {
3688         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3689                         &data->args.seq_args,
3690                         &data->res.seq_res,
3691                         task))
3692                 return;
3693         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3694                         data->args.lock_context, FMODE_READ);
3695 }
3696
3697 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3698 {
3699         struct inode *inode = data->header->inode;
3700
3701         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3702                 rpc_restart_call_prepare(task);
3703                 return -EAGAIN;
3704         }
3705         if (task->tk_status >= 0) {
3706                 renew_lease(NFS_SERVER(inode), data->timestamp);
3707                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3708         }
3709         return 0;
3710 }
3711
3712 static bool nfs4_write_stateid_changed(struct rpc_task *task,
3713                 struct nfs_writeargs *args)
3714 {
3715
3716         if (!nfs4_error_stateid_expired(task->tk_status) ||
3717                 nfs4_stateid_is_current(&args->stateid,
3718                                 args->context,
3719                                 args->lock_context,
3720                                 FMODE_WRITE))
3721                 return false;
3722         rpc_restart_call_prepare(task);
3723         return true;
3724 }
3725
3726 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3727 {
3728         if (!nfs4_sequence_done(task, &data->res.seq_res))
3729                 return -EAGAIN;
3730         if (nfs4_write_stateid_changed(task, &data->args))
3731                 return -EAGAIN;
3732         return data->write_done_cb ? data->write_done_cb(task, data) :
3733                 nfs4_write_done_cb(task, data);
3734 }
3735
3736 static
3737 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3738 {
3739         const struct nfs_pgio_header *hdr = data->header;
3740
3741         /* Don't request attributes for pNFS or O_DIRECT writes */
3742         if (data->ds_clp != NULL || hdr->dreq != NULL)
3743                 return false;
3744         /* Otherwise, request attributes if and only if we don't hold
3745          * a delegation
3746          */
3747         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3748 }
3749
3750 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3751 {
3752         struct nfs_server *server = NFS_SERVER(data->header->inode);
3753
3754         if (!nfs4_write_need_cache_consistency_data(data)) {
3755                 data->args.bitmask = NULL;
3756                 data->res.fattr = NULL;
3757         } else
3758                 data->args.bitmask = server->cache_consistency_bitmask;
3759
3760         if (!data->write_done_cb)
3761                 data->write_done_cb = nfs4_write_done_cb;
3762         data->res.server = server;
3763         data->timestamp   = jiffies;
3764
3765         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3766         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3767 }
3768
3769 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3770 {
3771         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3772                         &data->args.seq_args,
3773                         &data->res.seq_res,
3774                         task))
3775                 return;
3776         nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3777                         data->args.lock_context, FMODE_WRITE);
3778 }
3779
3780 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3781 {
3782         nfs4_setup_sequence(NFS_SERVER(data->inode),
3783                         &data->args.seq_args,
3784                         &data->res.seq_res,
3785                         task);
3786 }
3787
3788 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3789 {
3790         struct inode *inode = data->inode;
3791
3792         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3793                 rpc_restart_call_prepare(task);
3794                 return -EAGAIN;
3795         }
3796         return 0;
3797 }
3798
3799 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3800 {
3801         if (!nfs4_sequence_done(task, &data->res.seq_res))
3802                 return -EAGAIN;
3803         return data->commit_done_cb(task, data);
3804 }
3805
3806 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3807 {
3808         struct nfs_server *server = NFS_SERVER(data->inode);
3809
3810         if (data->commit_done_cb == NULL)
3811                 data->commit_done_cb = nfs4_commit_done_cb;
3812         data->res.server = server;
3813         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3814         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3815 }
3816
3817 struct nfs4_renewdata {
3818         struct nfs_client       *client;
3819         unsigned long           timestamp;
3820 };
3821
3822 /*
3823  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3824  * standalone procedure for queueing an asynchronous RENEW.
3825  */
3826 static void nfs4_renew_release(void *calldata)
3827 {
3828         struct nfs4_renewdata *data = calldata;
3829         struct nfs_client *clp = data->client;
3830
3831         if (atomic_read(&clp->cl_count) > 1)
3832                 nfs4_schedule_state_renewal(clp);
3833         nfs_put_client(clp);
3834         kfree(data);
3835 }
3836
3837 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3838 {
3839         struct nfs4_renewdata *data = calldata;
3840         struct nfs_client *clp = data->client;
3841         unsigned long timestamp = data->timestamp;
3842
3843         if (task->tk_status < 0) {
3844                 /* Unless we're shutting down, schedule state recovery! */
3845                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3846                         return;
3847                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3848                         nfs4_schedule_lease_recovery(clp);
3849                         return;
3850                 }
3851                 nfs4_schedule_path_down_recovery(clp);
3852         }
3853         do_renew_lease(clp, timestamp);
3854 }
3855
3856 static const struct rpc_call_ops nfs4_renew_ops = {
3857         .rpc_call_done = nfs4_renew_done,
3858         .rpc_release = nfs4_renew_release,
3859 };
3860
3861 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3862 {
3863         struct rpc_message msg = {
3864                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3865                 .rpc_argp       = clp,
3866                 .rpc_cred       = cred,
3867         };
3868         struct nfs4_renewdata *data;
3869
3870         if (renew_flags == 0)
3871                 return 0;
3872         if (!atomic_inc_not_zero(&clp->cl_count))
3873                 return -EIO;
3874         data = kmalloc(sizeof(*data), GFP_NOFS);
3875         if (data == NULL)
3876                 return -ENOMEM;
3877         data->client = clp;
3878         data->timestamp = jiffies;
3879         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
3880                         &nfs4_renew_ops, data);
3881 }
3882
3883 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3884 {
3885         struct rpc_message msg = {
3886                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3887                 .rpc_argp       = clp,
3888                 .rpc_cred       = cred,
3889         };
3890         unsigned long now = jiffies;
3891         int status;
3892
3893         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3894         if (status < 0)
3895                 return status;
3896         do_renew_lease(clp, now);
3897         return 0;
3898 }
3899
3900 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3901 {
3902         return (server->caps & NFS_CAP_ACLS)
3903                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3904                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3905 }
3906
3907 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3908  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3909  * the stack.
3910  */
3911 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3912
3913 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3914                 struct page **pages, unsigned int *pgbase)
3915 {
3916         struct page *newpage, **spages;
3917         int rc = 0;
3918         size_t len;
3919         spages = pages;
3920
3921         do {
3922                 len = min_t(size_t, PAGE_SIZE, buflen);
3923                 newpage = alloc_page(GFP_KERNEL);
3924
3925                 if (newpage == NULL)
3926                         goto unwind;
3927                 memcpy(page_address(newpage), buf, len);
3928                 buf += len;
3929                 buflen -= len;
3930                 *pages++ = newpage;
3931                 rc++;
3932         } while (buflen != 0);
3933
3934         return rc;
3935
3936 unwind:
3937         for(; rc > 0; rc--)
3938                 __free_page(spages[rc-1]);
3939         return -ENOMEM;
3940 }
3941
3942 struct nfs4_cached_acl {
3943         int cached;
3944         size_t len;
3945         char data[0];
3946 };
3947
3948 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3949 {
3950         struct nfs_inode *nfsi = NFS_I(inode);
3951
3952         spin_lock(&inode->i_lock);
3953         kfree(nfsi->nfs4_acl);
3954         nfsi->nfs4_acl = acl;
3955         spin_unlock(&inode->i_lock);
3956 }
3957
3958 static void nfs4_zap_acl_attr(struct inode *inode)
3959 {
3960         nfs4_set_cached_acl(inode, NULL);
3961 }
3962
3963 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3964 {
3965         struct nfs_inode *nfsi = NFS_I(inode);
3966         struct nfs4_cached_acl *acl;
3967         int ret = -ENOENT;
3968
3969         spin_lock(&inode->i_lock);
3970         acl = nfsi->nfs4_acl;
3971         if (acl == NULL)
3972                 goto out;
3973         if (buf == NULL) /* user is just asking for length */
3974                 goto out_len;
3975         if (acl->cached == 0)
3976                 goto out;
3977         ret = -ERANGE; /* see getxattr(2) man page */
3978         if (acl->len > buflen)
3979                 goto out;
3980         memcpy(buf, acl->data, acl->len);
3981 out_len:
3982         ret = acl->len;
3983 out:
3984         spin_unlock(&inode->i_lock);
3985         return ret;
3986 }
3987
3988 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3989 {
3990         struct nfs4_cached_acl *acl;
3991         size_t buflen = sizeof(*acl) + acl_len;
3992
3993         if (buflen <= PAGE_SIZE) {
3994                 acl = kmalloc(buflen, GFP_KERNEL);
3995                 if (acl == NULL)
3996                         goto out;
3997                 acl->cached = 1;
3998                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3999         } else {
4000                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4001                 if (acl == NULL)
4002                         goto out;
4003                 acl->cached = 0;
4004         }
4005         acl->len = acl_len;
4006 out:
4007         nfs4_set_cached_acl(inode, acl);
4008 }
4009
4010 /*
4011  * The getxattr API returns the required buffer length when called with a
4012  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4013  * the required buf.  On a NULL buf, we send a page of data to the server
4014  * guessing that the ACL request can be serviced by a page. If so, we cache
4015  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4016  * the cache. If not so, we throw away the page, and cache the required
4017  * length. The next getxattr call will then produce another round trip to
4018  * the server, this time with the input buf of the required size.
4019  */
4020 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4021 {
4022         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4023         struct nfs_getaclargs args = {
4024                 .fh = NFS_FH(inode),
4025                 .acl_pages = pages,
4026                 .acl_len = buflen,
4027         };
4028         struct nfs_getaclres res = {
4029                 .acl_len = buflen,
4030         };
4031         struct rpc_message msg = {
4032                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4033                 .rpc_argp = &args,
4034                 .rpc_resp = &res,
4035         };
4036         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4037         int ret = -ENOMEM, i;
4038
4039         /* As long as we're doing a round trip to the server anyway,
4040          * let's be prepared for a page of acl data. */
4041         if (npages == 0)
4042                 npages = 1;
4043         if (npages > ARRAY_SIZE(pages))
4044                 return -ERANGE;
4045
4046         for (i = 0; i < npages; i++) {
4047                 pages[i] = alloc_page(GFP_KERNEL);
4048                 if (!pages[i])
4049                         goto out_free;
4050         }
4051
4052         /* for decoding across pages */
4053         res.acl_scratch = alloc_page(GFP_KERNEL);
4054         if (!res.acl_scratch)
4055                 goto out_free;
4056
4057         args.acl_len = npages * PAGE_SIZE;
4058         args.acl_pgbase = 0;
4059
4060         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4061                 __func__, buf, buflen, npages, args.acl_len);
4062         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4063                              &msg, &args.seq_args, &res.seq_res, 0);
4064         if (ret)
4065                 goto out_free;
4066
4067         /* Handle the case where the passed-in buffer is too short */
4068         if (res.acl_flags & NFS4_ACL_TRUNC) {
4069                 /* Did the user only issue a request for the acl length? */
4070                 if (buf == NULL)
4071                         goto out_ok;
4072                 ret = -ERANGE;
4073                 goto out_free;
4074         }
4075         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4076         if (buf) {
4077                 if (res.acl_len > buflen) {
4078                         ret = -ERANGE;
4079                         goto out_free;
4080                 }
4081                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4082         }
4083 out_ok:
4084         ret = res.acl_len;
4085 out_free:
4086         for (i = 0; i < npages; i++)
4087                 if (pages[i])
4088                         __free_page(pages[i]);
4089         if (res.acl_scratch)
4090                 __free_page(res.acl_scratch);
4091         return ret;
4092 }
4093
4094 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4095 {
4096         struct nfs4_exception exception = { };
4097         ssize_t ret;
4098         do {
4099                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4100                 if (ret >= 0)
4101                         break;
4102                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4103         } while (exception.retry);
4104         return ret;
4105 }
4106
4107 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4108 {
4109         struct nfs_server *server = NFS_SERVER(inode);
4110         int ret;
4111
4112         if (!nfs4_server_supports_acls(server))
4113                 return -EOPNOTSUPP;
4114         ret = nfs_revalidate_inode(server, inode);
4115         if (ret < 0)
4116                 return ret;
4117         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4118                 nfs_zap_acl_cache(inode);
4119         ret = nfs4_read_cached_acl(inode, buf, buflen);
4120         if (ret != -ENOENT)
4121                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4122                  * but no cached acl data, just the acl length */
4123                 return ret;
4124         return nfs4_get_acl_uncached(inode, buf, buflen);
4125 }
4126
4127 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4128 {
4129         struct nfs_server *server = NFS_SERVER(inode);
4130         struct page *pages[NFS4ACL_MAXPAGES];
4131         struct nfs_setaclargs arg = {
4132                 .fh             = NFS_FH(inode),
4133                 .acl_pages      = pages,
4134                 .acl_len        = buflen,
4135         };
4136         struct nfs_setaclres res;
4137         struct rpc_message msg = {
4138                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4139                 .rpc_argp       = &arg,
4140                 .rpc_resp       = &res,
4141         };
4142         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4143         int ret, i;
4144
4145         if (!nfs4_server_supports_acls(server))
4146                 return -EOPNOTSUPP;
4147         if (npages > ARRAY_SIZE(pages))
4148                 return -ERANGE;
4149         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4150         if (i < 0)
4151                 return i;
4152         nfs4_inode_return_delegation(inode);
4153         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4154
4155         /*
4156          * Free each page after tx, so the only ref left is
4157          * held by the network stack
4158          */
4159         for (; i > 0; i--)
4160                 put_page(pages[i-1]);
4161
4162         /*
4163          * Acl update can result in inode attribute update.
4164          * so mark the attribute cache invalid.
4165          */
4166         spin_lock(&inode->i_lock);
4167         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4168         spin_unlock(&inode->i_lock);
4169         nfs_access_zap_cache(inode);
4170         nfs_zap_acl_cache(inode);
4171         return ret;
4172 }
4173
4174 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4175 {
4176         struct nfs4_exception exception = { };
4177         int err;
4178         do {
4179                 err = nfs4_handle_exception(NFS_SERVER(inode),
4180                                 __nfs4_proc_set_acl(inode, buf, buflen),
4181                                 &exception);
4182         } while (exception.retry);
4183         return err;
4184 }
4185
4186 static int
4187 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4188 {
4189         struct nfs_client *clp = server->nfs_client;
4190
4191         if (task->tk_status >= 0)
4192                 return 0;
4193         switch(task->tk_status) {
4194                 case -NFS4ERR_DELEG_REVOKED:
4195                 case -NFS4ERR_ADMIN_REVOKED:
4196                 case -NFS4ERR_BAD_STATEID:
4197                         if (state == NULL)
4198                                 break;
4199                         nfs_remove_bad_delegation(state->inode);
4200                 case -NFS4ERR_OPENMODE:
4201                         if (state == NULL)
4202                                 break;
4203                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4204                                 goto stateid_invalid;
4205                         goto wait_on_recovery;
4206                 case -NFS4ERR_EXPIRED:
4207                         if (state != NULL) {
4208                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4209                                         goto stateid_invalid;
4210                         }
4211                 case -NFS4ERR_STALE_STATEID:
4212                 case -NFS4ERR_STALE_CLIENTID:
4213                         nfs4_schedule_lease_recovery(clp);
4214                         goto wait_on_recovery;
4215 #if defined(CONFIG_NFS_V4_1)
4216                 case -NFS4ERR_BADSESSION:
4217                 case -NFS4ERR_BADSLOT:
4218                 case -NFS4ERR_BAD_HIGH_SLOT:
4219                 case -NFS4ERR_DEADSESSION:
4220                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4221                 case -NFS4ERR_SEQ_FALSE_RETRY:
4222                 case -NFS4ERR_SEQ_MISORDERED:
4223                         dprintk("%s ERROR %d, Reset session\n", __func__,
4224                                 task->tk_status);
4225                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4226                         goto wait_on_recovery;
4227 #endif /* CONFIG_NFS_V4_1 */
4228                 case -NFS4ERR_DELAY:
4229                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4230                 case -NFS4ERR_GRACE:
4231                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4232                         task->tk_status = 0;
4233                         return -EAGAIN;
4234                 case -NFS4ERR_RETRY_UNCACHED_REP:
4235                 case -NFS4ERR_OLD_STATEID:
4236                         task->tk_status = 0;
4237                         return -EAGAIN;
4238         }
4239         task->tk_status = nfs4_map_errors(task->tk_status);
4240         return 0;
4241 stateid_invalid:
4242         task->tk_status = -EIO;
4243         return 0;
4244 wait_on_recovery:
4245         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4246         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4247                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4248         task->tk_status = 0;
4249         return -EAGAIN;
4250 }
4251
4252 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4253                                     nfs4_verifier *bootverf)
4254 {
4255         __be32 verf[2];
4256
4257         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4258                 /* An impossible timestamp guarantees this value
4259                  * will never match a generated boot time. */
4260                 verf[0] = 0;
4261                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4262         } else {
4263                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4264                 verf[0] = (__be32)nn->boot_time.tv_sec;
4265                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4266         }
4267         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4268 }
4269
4270 static unsigned int
4271 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4272                                    char *buf, size_t len)
4273 {
4274         unsigned int result;
4275
4276         rcu_read_lock();
4277         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4278                                 clp->cl_ipaddr,
4279                                 rpc_peeraddr2str(clp->cl_rpcclient,
4280                                                         RPC_DISPLAY_ADDR),
4281                                 rpc_peeraddr2str(clp->cl_rpcclient,
4282                                                         RPC_DISPLAY_PROTO));
4283         rcu_read_unlock();
4284         return result;
4285 }
4286
4287 static unsigned int
4288 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4289                                 char *buf, size_t len)
4290 {
4291         char *nodename = clp->cl_rpcclient->cl_nodename;
4292
4293         if (nfs4_client_id_uniquifier[0] != '\0')
4294                 nodename = nfs4_client_id_uniquifier;
4295         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4296                                 clp->rpc_ops->version, clp->cl_minorversion,
4297                                 nodename);
4298 }
4299
4300 /**
4301  * nfs4_proc_setclientid - Negotiate client ID
4302  * @clp: state data structure
4303  * @program: RPC program for NFSv4 callback service
4304  * @port: IP port number for NFS4 callback service
4305  * @cred: RPC credential to use for this call
4306  * @res: where to place the result
4307  *
4308  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4309  */
4310 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4311                 unsigned short port, struct rpc_cred *cred,
4312                 struct nfs4_setclientid_res *res)
4313 {
4314         nfs4_verifier sc_verifier;
4315         struct nfs4_setclientid setclientid = {
4316                 .sc_verifier = &sc_verifier,
4317                 .sc_prog = program,
4318                 .sc_cb_ident = clp->cl_cb_ident,
4319         };
4320         struct rpc_message msg = {
4321                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4322                 .rpc_argp = &setclientid,
4323                 .rpc_resp = res,
4324                 .rpc_cred = cred,
4325         };
4326         int status;
4327
4328         /* nfs_client_id4 */
4329         nfs4_init_boot_verifier(clp, &sc_verifier);
4330         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4331                 setclientid.sc_name_len =
4332                                 nfs4_init_uniform_client_string(clp,
4333                                                 setclientid.sc_name,
4334                                                 sizeof(setclientid.sc_name));
4335         else
4336                 setclientid.sc_name_len =
4337                                 nfs4_init_nonuniform_client_string(clp,
4338                                                 setclientid.sc_name,
4339                                                 sizeof(setclientid.sc_name));
4340         /* cb_client4 */
4341         rcu_read_lock();
4342         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4343                                 sizeof(setclientid.sc_netid),
4344                                 rpc_peeraddr2str(clp->cl_rpcclient,
4345                                                         RPC_DISPLAY_NETID));
4346         rcu_read_unlock();
4347         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4348                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4349                                 clp->cl_ipaddr, port >> 8, port & 255);
4350
4351         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4352                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4353                 setclientid.sc_name_len, setclientid.sc_name);
4354         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4355         dprintk("NFS reply setclientid: %d\n", status);
4356         return status;
4357 }
4358
4359 /**
4360  * nfs4_proc_setclientid_confirm - Confirm client ID
4361  * @clp: state data structure
4362  * @res: result of a previous SETCLIENTID
4363  * @cred: RPC credential to use for this call
4364  *
4365  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4366  */
4367 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4368                 struct nfs4_setclientid_res *arg,
4369                 struct rpc_cred *cred)
4370 {
4371         struct rpc_message msg = {
4372                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4373                 .rpc_argp = arg,
4374                 .rpc_cred = cred,
4375         };
4376         int status;
4377
4378         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4379                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4380                 clp->cl_clientid);
4381         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4382         dprintk("NFS reply setclientid_confirm: %d\n", status);
4383         return status;
4384 }
4385
4386 struct nfs4_delegreturndata {
4387         struct nfs4_delegreturnargs args;
4388         struct nfs4_delegreturnres res;
4389         struct nfs_fh fh;
4390         nfs4_stateid stateid;
4391         unsigned long timestamp;
4392         struct nfs_fattr fattr;
4393         int rpc_status;
4394 };
4395
4396 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4397 {
4398         struct nfs4_delegreturndata *data = calldata;
4399
4400         if (!nfs4_sequence_done(task, &data->res.seq_res))
4401                 return;
4402
4403         switch (task->tk_status) {
4404         case 0:
4405                 renew_lease(data->res.server, data->timestamp);
4406                 break;
4407         case -NFS4ERR_ADMIN_REVOKED:
4408         case -NFS4ERR_DELEG_REVOKED:
4409         case -NFS4ERR_BAD_STATEID:
4410         case -NFS4ERR_OLD_STATEID:
4411         case -NFS4ERR_STALE_STATEID:
4412         case -NFS4ERR_EXPIRED:
4413                 task->tk_status = 0;
4414                 break;
4415         default:
4416                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4417                                 -EAGAIN) {
4418                         rpc_restart_call_prepare(task);
4419                         return;
4420                 }
4421         }
4422         data->rpc_status = task->tk_status;
4423 }
4424
4425 static void nfs4_delegreturn_release(void *calldata)
4426 {
4427         kfree(calldata);
4428 }
4429
4430 #if defined(CONFIG_NFS_V4_1)
4431 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4432 {
4433         struct nfs4_delegreturndata *d_data;
4434
4435         d_data = (struct nfs4_delegreturndata *)data;
4436
4437         nfs4_setup_sequence(d_data->res.server,
4438                         &d_data->args.seq_args,
4439                         &d_data->res.seq_res,
4440                         task);
4441 }
4442 #endif /* CONFIG_NFS_V4_1 */
4443
4444 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4445 #if defined(CONFIG_NFS_V4_1)
4446         .rpc_call_prepare = nfs4_delegreturn_prepare,
4447 #endif /* CONFIG_NFS_V4_1 */
4448         .rpc_call_done = nfs4_delegreturn_done,
4449         .rpc_release = nfs4_delegreturn_release,
4450 };
4451
4452 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4453 {
4454         struct nfs4_delegreturndata *data;
4455         struct nfs_server *server = NFS_SERVER(inode);
4456         struct rpc_task *task;
4457         struct rpc_message msg = {
4458                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4459                 .rpc_cred = cred,
4460         };
4461         struct rpc_task_setup task_setup_data = {
4462                 .rpc_client = server->client,
4463                 .rpc_message = &msg,
4464                 .callback_ops = &nfs4_delegreturn_ops,
4465                 .flags = RPC_TASK_ASYNC,
4466         };
4467         int status = 0;
4468
4469         data = kzalloc(sizeof(*data), GFP_NOFS);
4470         if (data == NULL)
4471                 return -ENOMEM;
4472         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4473         data->args.fhandle = &data->fh;
4474         data->args.stateid = &data->stateid;
4475         data->args.bitmask = server->cache_consistency_bitmask;
4476         nfs_copy_fh(&data->fh, NFS_FH(inode));
4477         nfs4_stateid_copy(&data->stateid, stateid);
4478         data->res.fattr = &data->fattr;
4479         data->res.server = server;
4480         nfs_fattr_init(data->res.fattr);
4481         data->timestamp = jiffies;
4482         data->rpc_status = 0;
4483
4484         task_setup_data.callback_data = data;
4485         msg.rpc_argp = &data->args;
4486         msg.rpc_resp = &data->res;
4487         task = rpc_run_task(&task_setup_data);
4488         if (IS_ERR(task))
4489                 return PTR_ERR(task);
4490         if (!issync)
4491                 goto out;
4492         status = nfs4_wait_for_completion_rpc_task(task);
4493         if (status != 0)
4494                 goto out;
4495         status = data->rpc_status;
4496         if (status == 0)
4497                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4498         else
4499                 nfs_refresh_inode(inode, &data->fattr);
4500 out:
4501         rpc_put_task(task);
4502         return status;
4503 }
4504
4505 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4506 {
4507         struct nfs_server *server = NFS_SERVER(inode);
4508         struct nfs4_exception exception = { };
4509         int err;
4510         do {
4511                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4512                 switch (err) {
4513                         case -NFS4ERR_STALE_STATEID:
4514                         case -NFS4ERR_EXPIRED:
4515                         case 0:
4516                                 return 0;
4517                 }
4518                 err = nfs4_handle_exception(server, err, &exception);
4519         } while (exception.retry);
4520         return err;
4521 }
4522
4523 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4524 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4525
4526 /*
4527  * sleep, with exponential backoff, and retry the LOCK operation.
4528  */
4529 static unsigned long
4530 nfs4_set_lock_task_retry(unsigned long timeout)
4531 {
4532         freezable_schedule_timeout_killable(timeout);
4533         timeout <<= 1;
4534         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4535                 return NFS4_LOCK_MAXTIMEOUT;
4536         return timeout;
4537 }
4538
4539 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4540 {
4541         struct inode *inode = state->inode;
4542         struct nfs_server *server = NFS_SERVER(inode);
4543         struct nfs_client *clp = server->nfs_client;
4544         struct nfs_lockt_args arg = {
4545                 .fh = NFS_FH(inode),
4546                 .fl = request,
4547         };
4548         struct nfs_lockt_res res = {
4549                 .denied = request,
4550         };
4551         struct rpc_message msg = {
4552                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4553                 .rpc_argp       = &arg,
4554                 .rpc_resp       = &res,
4555                 .rpc_cred       = state->owner->so_cred,
4556         };
4557         struct nfs4_lock_state *lsp;
4558         int status;
4559
4560         arg.lock_owner.clientid = clp->cl_clientid;
4561         status = nfs4_set_lock_state(state, request);
4562         if (status != 0)
4563                 goto out;
4564         lsp = request->fl_u.nfs4_fl.owner;
4565         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4566         arg.lock_owner.s_dev = server->s_dev;
4567         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4568         switch (status) {
4569                 case 0:
4570                         request->fl_type = F_UNLCK;
4571                         break;
4572                 case -NFS4ERR_DENIED:
4573                         status = 0;
4574         }
4575         request->fl_ops->fl_release_private(request);
4576         request->fl_ops = NULL;
4577 out:
4578         return status;
4579 }
4580
4581 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4582 {
4583         struct nfs4_exception exception = { };
4584         int err;
4585
4586         do {
4587                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4588                                 _nfs4_proc_getlk(state, cmd, request),
4589                                 &exception);
4590         } while (exception.retry);
4591         return err;
4592 }
4593
4594 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4595 {
4596         int res = 0;
4597         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4598                 case FL_POSIX:
4599                         res = posix_lock_file_wait(file, fl);
4600                         break;
4601                 case FL_FLOCK:
4602                         res = flock_lock_file_wait(file, fl);
4603                         break;
4604                 default:
4605                         BUG();
4606         }
4607         return res;
4608 }
4609
4610 struct nfs4_unlockdata {
4611         struct nfs_locku_args arg;
4612         struct nfs_locku_res res;
4613         struct nfs4_lock_state *lsp;
4614         struct nfs_open_context *ctx;
4615         struct file_lock fl;
4616         const struct nfs_server *server;
4617         unsigned long timestamp;
4618 };
4619
4620 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4621                 struct nfs_open_context *ctx,
4622                 struct nfs4_lock_state *lsp,
4623                 struct nfs_seqid *seqid)
4624 {
4625         struct nfs4_unlockdata *p;
4626         struct inode *inode = lsp->ls_state->inode;
4627
4628         p = kzalloc(sizeof(*p), GFP_NOFS);
4629         if (p == NULL)
4630                 return NULL;
4631         p->arg.fh = NFS_FH(inode);
4632         p->arg.fl = &p->fl;
4633         p->arg.seqid = seqid;
4634         p->res.seqid = seqid;
4635         p->arg.stateid = &lsp->ls_stateid;
4636         p->lsp = lsp;
4637         atomic_inc(&lsp->ls_count);
4638         /* Ensure we don't close file until we're done freeing locks! */
4639         p->ctx = get_nfs_open_context(ctx);
4640         memcpy(&p->fl, fl, sizeof(p->fl));
4641         p->server = NFS_SERVER(inode);
4642         return p;
4643 }
4644
4645 static void nfs4_locku_release_calldata(void *data)
4646 {
4647         struct nfs4_unlockdata *calldata = data;
4648         nfs_free_seqid(calldata->arg.seqid);
4649         nfs4_put_lock_state(calldata->lsp);
4650         put_nfs_open_context(calldata->ctx);
4651         kfree(calldata);
4652 }
4653
4654 static void nfs4_locku_done(struct rpc_task *task, void *data)
4655 {
4656         struct nfs4_unlockdata *calldata = data;
4657
4658         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4659                 return;
4660         switch (task->tk_status) {
4661                 case 0:
4662                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4663                                         &calldata->res.stateid);
4664                         renew_lease(calldata->server, calldata->timestamp);
4665                         break;
4666                 case -NFS4ERR_BAD_STATEID:
4667                 case -NFS4ERR_OLD_STATEID:
4668                 case -NFS4ERR_STALE_STATEID:
4669                 case -NFS4ERR_EXPIRED:
4670                         break;
4671                 default:
4672                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4673                                 rpc_restart_call_prepare(task);
4674         }
4675         nfs_release_seqid(calldata->arg.seqid);
4676 }
4677
4678 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4679 {
4680         struct nfs4_unlockdata *calldata = data;
4681
4682         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4683                 goto out_wait;
4684         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4685                 /* Note: exit _without_ running nfs4_locku_done */
4686                 goto out_no_action;
4687         }
4688         calldata->timestamp = jiffies;
4689         if (nfs4_setup_sequence(calldata->server,
4690                                 &calldata->arg.seq_args,
4691                                 &calldata->res.seq_res,
4692                                 task) != 0)
4693                 nfs_release_seqid(calldata->arg.seqid);
4694         return;
4695 out_no_action:
4696         task->tk_action = NULL;
4697 out_wait:
4698         nfs4_sequence_done(task, &calldata->res.seq_res);
4699 }
4700
4701 static const struct rpc_call_ops nfs4_locku_ops = {
4702         .rpc_call_prepare = nfs4_locku_prepare,
4703         .rpc_call_done = nfs4_locku_done,
4704         .rpc_release = nfs4_locku_release_calldata,
4705 };
4706
4707 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4708                 struct nfs_open_context *ctx,
4709                 struct nfs4_lock_state *lsp,
4710                 struct nfs_seqid *seqid)
4711 {
4712         struct nfs4_unlockdata *data;
4713         struct rpc_message msg = {
4714                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4715                 .rpc_cred = ctx->cred,
4716         };
4717         struct rpc_task_setup task_setup_data = {
4718                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4719                 .rpc_message = &msg,
4720                 .callback_ops = &nfs4_locku_ops,
4721                 .workqueue = nfsiod_workqueue,
4722                 .flags = RPC_TASK_ASYNC,
4723         };
4724
4725         /* Ensure this is an unlock - when canceling a lock, the
4726          * canceled lock is passed in, and it won't be an unlock.
4727          */
4728         fl->fl_type = F_UNLCK;
4729
4730         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4731         if (data == NULL) {
4732                 nfs_free_seqid(seqid);
4733                 return ERR_PTR(-ENOMEM);
4734         }
4735
4736         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4737         msg.rpc_argp = &data->arg;
4738         msg.rpc_resp = &data->res;
4739         task_setup_data.callback_data = data;
4740         return rpc_run_task(&task_setup_data);
4741 }
4742
4743 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4744 {
4745         struct inode *inode = state->inode;
4746         struct nfs4_state_owner *sp = state->owner;
4747         struct nfs_inode *nfsi = NFS_I(inode);
4748         struct nfs_seqid *seqid;
4749         struct nfs4_lock_state *lsp;
4750         struct rpc_task *task;
4751         int status = 0;
4752         unsigned char fl_flags = request->fl_flags;
4753
4754         status = nfs4_set_lock_state(state, request);
4755         /* Unlock _before_ we do the RPC call */
4756         request->fl_flags |= FL_EXISTS;
4757         /* Exclude nfs_delegation_claim_locks() */
4758         mutex_lock(&sp->so_delegreturn_mutex);
4759         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
4760         down_read(&nfsi->rwsem);
4761         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4762                 up_read(&nfsi->rwsem);
4763                 mutex_unlock(&sp->so_delegreturn_mutex);
4764                 goto out;
4765         }
4766         up_read(&nfsi->rwsem);
4767         mutex_unlock(&sp->so_delegreturn_mutex);
4768         if (status != 0)
4769                 goto out;
4770         /* Is this a delegated lock? */
4771         lsp = request->fl_u.nfs4_fl.owner;
4772         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
4773                 goto out;
4774         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4775         status = -ENOMEM;
4776         if (seqid == NULL)
4777                 goto out;
4778         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4779         status = PTR_ERR(task);
4780         if (IS_ERR(task))
4781                 goto out;
4782         status = nfs4_wait_for_completion_rpc_task(task);
4783         rpc_put_task(task);
4784 out:
4785         request->fl_flags = fl_flags;
4786         return status;
4787 }
4788
4789 struct nfs4_lockdata {
4790         struct nfs_lock_args arg;
4791         struct nfs_lock_res res;
4792         struct nfs4_lock_state *lsp;
4793         struct nfs_open_context *ctx;
4794         struct file_lock fl;
4795         unsigned long timestamp;
4796         int rpc_status;
4797         int cancelled;
4798         struct nfs_server *server;
4799 };
4800
4801 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4802                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4803                 gfp_t gfp_mask)
4804 {
4805         struct nfs4_lockdata *p;
4806         struct inode *inode = lsp->ls_state->inode;
4807         struct nfs_server *server = NFS_SERVER(inode);
4808
4809         p = kzalloc(sizeof(*p), gfp_mask);
4810         if (p == NULL)
4811                 return NULL;
4812
4813         p->arg.fh = NFS_FH(inode);
4814         p->arg.fl = &p->fl;
4815         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4816         if (p->arg.open_seqid == NULL)
4817                 goto out_free;
4818         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4819         if (p->arg.lock_seqid == NULL)
4820                 goto out_free_seqid;
4821         p->arg.lock_stateid = &lsp->ls_stateid;
4822         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4823         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4824         p->arg.lock_owner.s_dev = server->s_dev;
4825         p->res.lock_seqid = p->arg.lock_seqid;
4826         p->lsp = lsp;
4827         p->server = server;
4828         atomic_inc(&lsp->ls_count);
4829         p->ctx = get_nfs_open_context(ctx);
4830         memcpy(&p->fl, fl, sizeof(p->fl));
4831         return p;
4832 out_free_seqid:
4833         nfs_free_seqid(p->arg.open_seqid);
4834 out_free:
4835         kfree(p);
4836         return NULL;
4837 }
4838
4839 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4840 {
4841         struct nfs4_lockdata *data = calldata;
4842         struct nfs4_state *state = data->lsp->ls_state;
4843
4844         dprintk("%s: begin!\n", __func__);
4845         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4846                 goto out_wait;
4847         /* Do we need to do an open_to_lock_owner? */
4848         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4849                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
4850                         goto out_release_lock_seqid;
4851                 }
4852                 data->arg.open_stateid = &state->open_stateid;
4853                 data->arg.new_lock_owner = 1;
4854                 data->res.open_seqid = data->arg.open_seqid;
4855         } else
4856                 data->arg.new_lock_owner = 0;
4857         if (!nfs4_valid_open_stateid(state)) {
4858                 data->rpc_status = -EBADF;
4859                 task->tk_action = NULL;
4860                 goto out_release_open_seqid;
4861         }
4862         data->timestamp = jiffies;
4863         if (nfs4_setup_sequence(data->server,
4864                                 &data->arg.seq_args,
4865                                 &data->res.seq_res,
4866                                 task) == 0)
4867                 return;
4868 out_release_open_seqid:
4869         nfs_release_seqid(data->arg.open_seqid);
4870 out_release_lock_seqid:
4871         nfs_release_seqid(data->arg.lock_seqid);
4872 out_wait:
4873         nfs4_sequence_done(task, &data->res.seq_res);
4874         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4875 }
4876
4877 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4878 {
4879         struct nfs4_lockdata *data = calldata;
4880
4881         dprintk("%s: begin!\n", __func__);
4882
4883         if (!nfs4_sequence_done(task, &data->res.seq_res))
4884                 return;
4885
4886         data->rpc_status = task->tk_status;
4887         if (data->arg.new_lock_owner != 0) {
4888                 if (data->rpc_status == 0)
4889                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4890                 else
4891                         goto out;
4892         }
4893         if (data->rpc_status == 0) {
4894                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4895                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4896                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4897         }
4898 out:
4899         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4900 }
4901
4902 static void nfs4_lock_release(void *calldata)
4903 {
4904         struct nfs4_lockdata *data = calldata;
4905
4906         dprintk("%s: begin!\n", __func__);
4907         nfs_free_seqid(data->arg.open_seqid);
4908         if (data->cancelled != 0) {
4909                 struct rpc_task *task;
4910                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4911                                 data->arg.lock_seqid);
4912                 if (!IS_ERR(task))
4913                         rpc_put_task_async(task);
4914                 dprintk("%s: cancelling lock!\n", __func__);
4915         } else
4916                 nfs_free_seqid(data->arg.lock_seqid);
4917         nfs4_put_lock_state(data->lsp);
4918         put_nfs_open_context(data->ctx);
4919         kfree(data);
4920         dprintk("%s: done!\n", __func__);
4921 }
4922
4923 static const struct rpc_call_ops nfs4_lock_ops = {
4924         .rpc_call_prepare = nfs4_lock_prepare,
4925         .rpc_call_done = nfs4_lock_done,
4926         .rpc_release = nfs4_lock_release,
4927 };
4928
4929 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4930 {
4931         switch (error) {
4932         case -NFS4ERR_ADMIN_REVOKED:
4933         case -NFS4ERR_BAD_STATEID:
4934                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4935                 if (new_lock_owner != 0 ||
4936                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4937                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4938                 break;
4939         case -NFS4ERR_STALE_STATEID:
4940                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4941         case -NFS4ERR_EXPIRED:
4942                 nfs4_schedule_lease_recovery(server->nfs_client);
4943         };
4944 }
4945
4946 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4947 {
4948         struct nfs4_lockdata *data;
4949         struct rpc_task *task;
4950         struct rpc_message msg = {
4951                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4952                 .rpc_cred = state->owner->so_cred,
4953         };
4954         struct rpc_task_setup task_setup_data = {
4955                 .rpc_client = NFS_CLIENT(state->inode),
4956                 .rpc_message = &msg,
4957                 .callback_ops = &nfs4_lock_ops,
4958                 .workqueue = nfsiod_workqueue,
4959                 .flags = RPC_TASK_ASYNC,
4960         };
4961         int ret;
4962
4963         dprintk("%s: begin!\n", __func__);
4964         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4965                         fl->fl_u.nfs4_fl.owner,
4966                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4967         if (data == NULL)
4968                 return -ENOMEM;
4969         if (IS_SETLKW(cmd))
4970                 data->arg.block = 1;
4971         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4972         msg.rpc_argp = &data->arg;
4973         msg.rpc_resp = &data->res;
4974         task_setup_data.callback_data = data;
4975         if (recovery_type > NFS_LOCK_NEW) {
4976                 if (recovery_type == NFS_LOCK_RECLAIM)
4977                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4978                 nfs4_set_sequence_privileged(&data->arg.seq_args);
4979         }
4980         task = rpc_run_task(&task_setup_data);
4981         if (IS_ERR(task))
4982                 return PTR_ERR(task);
4983         ret = nfs4_wait_for_completion_rpc_task(task);
4984         if (ret == 0) {
4985                 ret = data->rpc_status;
4986                 if (ret)
4987                         nfs4_handle_setlk_error(data->server, data->lsp,
4988                                         data->arg.new_lock_owner, ret);
4989         } else
4990                 data->cancelled = 1;
4991         rpc_put_task(task);
4992         dprintk("%s: done, ret = %d!\n", __func__, ret);
4993         return ret;
4994 }
4995
4996 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4997 {
4998         struct nfs_server *server = NFS_SERVER(state->inode);
4999         struct nfs4_exception exception = {
5000                 .inode = state->inode,
5001         };
5002         int err;
5003
5004         do {
5005                 /* Cache the lock if possible... */
5006                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5007                         return 0;
5008                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5009                 if (err != -NFS4ERR_DELAY)
5010                         break;
5011                 nfs4_handle_exception(server, err, &exception);
5012         } while (exception.retry);
5013         return err;
5014 }
5015
5016 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5017 {
5018         struct nfs_server *server = NFS_SERVER(state->inode);
5019         struct nfs4_exception exception = {
5020                 .inode = state->inode,
5021         };
5022         int err;
5023
5024         err = nfs4_set_lock_state(state, request);
5025         if (err != 0)
5026                 return err;
5027         do {
5028                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5029                         return 0;
5030                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5031                 switch (err) {
5032                 default:
5033                         goto out;
5034                 case -NFS4ERR_GRACE:
5035                 case -NFS4ERR_DELAY:
5036                         nfs4_handle_exception(server, err, &exception);
5037                         err = 0;
5038                 }
5039         } while (exception.retry);
5040 out:
5041         return err;
5042 }
5043
5044 #if defined(CONFIG_NFS_V4_1)
5045 /**
5046  * nfs41_check_expired_locks - possibly free a lock stateid
5047  *
5048  * @state: NFSv4 state for an inode
5049  *
5050  * Returns NFS_OK if recovery for this stateid is now finished.
5051  * Otherwise a negative NFS4ERR value is returned.
5052  */
5053 static int nfs41_check_expired_locks(struct nfs4_state *state)
5054 {
5055         int status, ret = -NFS4ERR_BAD_STATEID;
5056         struct nfs4_lock_state *lsp;
5057         struct nfs_server *server = NFS_SERVER(state->inode);
5058
5059         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5060                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5061                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
5062                         if (status != NFS_OK) {
5063                                 /* Free the stateid unless the server
5064                                  * informs us the stateid is unrecognized. */
5065                                 if (status != -NFS4ERR_BAD_STATEID)
5066                                         nfs41_free_stateid(server,
5067                                                         &lsp->ls_stateid);
5068                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5069                                 ret = status;
5070                         }
5071                 }
5072         };
5073
5074         return ret;
5075 }
5076
5077 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5078 {
5079         int status = NFS_OK;
5080
5081         if (test_bit(LK_STATE_IN_USE, &state->flags))
5082                 status = nfs41_check_expired_locks(state);
5083         if (status != NFS_OK)
5084                 status = nfs4_lock_expired(state, request);
5085         return status;
5086 }
5087 #endif
5088
5089 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5090 {
5091         struct nfs4_state_owner *sp = state->owner;
5092         struct nfs_inode *nfsi = NFS_I(state->inode);
5093         unsigned char fl_flags = request->fl_flags;
5094         unsigned int seq;
5095         int status = -ENOLCK;
5096
5097         if ((fl_flags & FL_POSIX) &&
5098                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5099                 goto out;
5100         /* Is this a delegated open? */
5101         status = nfs4_set_lock_state(state, request);
5102         if (status != 0)
5103                 goto out;
5104         request->fl_flags |= FL_ACCESS;
5105         status = do_vfs_lock(request->fl_file, request);
5106         if (status < 0)
5107                 goto out;
5108         down_read(&nfsi->rwsem);
5109         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5110                 /* Yes: cache locks! */
5111                 /* ...but avoid races with delegation recall... */
5112                 request->fl_flags = fl_flags & ~FL_SLEEP;
5113                 status = do_vfs_lock(request->fl_file, request);
5114                 goto out_unlock;
5115         }
5116         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5117         up_read(&nfsi->rwsem);
5118         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5119         if (status != 0)
5120                 goto out;
5121         down_read(&nfsi->rwsem);
5122         if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5123                 status = -NFS4ERR_DELAY;
5124                 goto out_unlock;
5125         }
5126         /* Note: we always want to sleep here! */
5127         request->fl_flags = fl_flags | FL_SLEEP;
5128         if (do_vfs_lock(request->fl_file, request) < 0)
5129                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5130                         "manager!\n", __func__);
5131 out_unlock:
5132         up_read(&nfsi->rwsem);
5133 out:
5134         request->fl_flags = fl_flags;
5135         return status;
5136 }
5137
5138 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5139 {
5140         struct nfs4_exception exception = {
5141                 .state = state,
5142                 .inode = state->inode,
5143         };
5144         int err;
5145
5146         do {
5147                 err = _nfs4_proc_setlk(state, cmd, request);
5148                 if (err == -NFS4ERR_DENIED)
5149                         err = -EAGAIN;
5150                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
5151                                 err, &exception);
5152         } while (exception.retry);
5153         return err;
5154 }
5155
5156 static int
5157 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5158 {
5159         struct nfs_open_context *ctx;
5160         struct nfs4_state *state;
5161         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5162         int status;
5163
5164         /* verify open state */
5165         ctx = nfs_file_open_context(filp);
5166         state = ctx->state;
5167
5168         if (request->fl_start < 0 || request->fl_end < 0)
5169                 return -EINVAL;
5170
5171         if (IS_GETLK(cmd)) {
5172                 if (state != NULL)
5173                         return nfs4_proc_getlk(state, F_GETLK, request);
5174                 return 0;
5175         }
5176
5177         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5178                 return -EINVAL;
5179
5180         if (request->fl_type == F_UNLCK) {
5181                 if (state != NULL)
5182                         return nfs4_proc_unlck(state, cmd, request);
5183                 return 0;
5184         }
5185
5186         if (state == NULL)
5187                 return -ENOLCK;
5188         /*
5189          * Don't rely on the VFS having checked the file open mode,
5190          * since it won't do this for flock() locks.
5191          */
5192         switch (request->fl_type) {
5193         case F_RDLCK:
5194                 if (!(filp->f_mode & FMODE_READ))
5195                         return -EBADF;
5196                 break;
5197         case F_WRLCK:
5198                 if (!(filp->f_mode & FMODE_WRITE))
5199                         return -EBADF;
5200         }
5201
5202         do {
5203                 status = nfs4_proc_setlk(state, cmd, request);
5204                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5205                         break;
5206                 timeout = nfs4_set_lock_task_retry(timeout);
5207                 status = -ERESTARTSYS;
5208                 if (signalled())
5209                         break;
5210         } while(status < 0);
5211         return status;
5212 }
5213
5214 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5215 {
5216         struct nfs_server *server = NFS_SERVER(state->inode);
5217         int err;
5218
5219         err = nfs4_set_lock_state(state, fl);
5220         if (err != 0)
5221                 return err;
5222         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5223         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5224 }
5225
5226 struct nfs_release_lockowner_data {
5227         struct nfs4_lock_state *lsp;
5228         struct nfs_server *server;
5229         struct nfs_release_lockowner_args args;
5230 };
5231
5232 static void nfs4_release_lockowner_release(void *calldata)
5233 {
5234         struct nfs_release_lockowner_data *data = calldata;
5235         nfs4_free_lock_state(data->server, data->lsp);
5236         kfree(calldata);
5237 }
5238
5239 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5240         .rpc_release = nfs4_release_lockowner_release,
5241 };
5242
5243 static int nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
5244 {
5245         struct nfs_release_lockowner_data *data;
5246         struct rpc_message msg = {
5247                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5248         };
5249
5250         if (server->nfs_client->cl_mvops->minor_version != 0)
5251                 return -EINVAL;
5252         data = kmalloc(sizeof(*data), GFP_NOFS);
5253         if (!data)
5254                 return -ENOMEM;
5255         data->lsp = lsp;
5256         data->server = server;
5257         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5258         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5259         data->args.lock_owner.s_dev = server->s_dev;
5260         msg.rpc_argp = &data->args;
5261         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5262         return 0;
5263 }
5264
5265 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5266
5267 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5268                                    const void *buf, size_t buflen,
5269                                    int flags, int type)
5270 {
5271         if (strcmp(key, "") != 0)
5272                 return -EINVAL;
5273
5274         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5275 }
5276
5277 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5278                                    void *buf, size_t buflen, int type)
5279 {
5280         if (strcmp(key, "") != 0)
5281                 return -EINVAL;
5282
5283         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5284 }
5285
5286 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5287                                        size_t list_len, const char *name,
5288                                        size_t name_len, int type)
5289 {
5290         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5291
5292         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5293                 return 0;
5294
5295         if (list && len <= list_len)
5296                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5297         return len;
5298 }
5299
5300 /*
5301  * nfs_fhget will use either the mounted_on_fileid or the fileid
5302  */
5303 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5304 {
5305         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5306                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5307               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5308               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5309                 return;
5310
5311         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5312                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5313         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5314         fattr->nlink = 2;
5315 }
5316
5317 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5318                                    const struct qstr *name,
5319                                    struct nfs4_fs_locations *fs_locations,
5320                                    struct page *page)
5321 {
5322         struct nfs_server *server = NFS_SERVER(dir);
5323         u32 bitmask[2] = {
5324                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5325         };
5326         struct nfs4_fs_locations_arg args = {
5327                 .dir_fh = NFS_FH(dir),
5328                 .name = name,
5329                 .page = page,
5330                 .bitmask = bitmask,
5331         };
5332         struct nfs4_fs_locations_res res = {
5333                 .fs_locations = fs_locations,
5334         };
5335         struct rpc_message msg = {
5336                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5337                 .rpc_argp = &args,
5338                 .rpc_resp = &res,
5339         };
5340         int status;
5341
5342         dprintk("%s: start\n", __func__);
5343
5344         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5345          * is not supported */
5346         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5347                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5348         else
5349                 bitmask[0] |= FATTR4_WORD0_FILEID;
5350
5351         nfs_fattr_init(&fs_locations->fattr);
5352         fs_locations->server = server;
5353         fs_locations->nlocations = 0;
5354         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5355         dprintk("%s: returned status = %d\n", __func__, status);
5356         return status;
5357 }
5358
5359 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5360                            const struct qstr *name,
5361                            struct nfs4_fs_locations *fs_locations,
5362                            struct page *page)
5363 {
5364         struct nfs4_exception exception = { };
5365         int err;
5366         do {
5367                 err = nfs4_handle_exception(NFS_SERVER(dir),
5368                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5369                                 &exception);
5370         } while (exception.retry);
5371         return err;
5372 }
5373
5374 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5375 {
5376         int status;
5377         struct nfs4_secinfo_arg args = {
5378                 .dir_fh = NFS_FH(dir),
5379                 .name   = name,
5380         };
5381         struct nfs4_secinfo_res res = {
5382                 .flavors     = flavors,
5383         };
5384         struct rpc_message msg = {
5385                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5386                 .rpc_argp = &args,
5387                 .rpc_resp = &res,
5388         };
5389
5390         dprintk("NFS call  secinfo %s\n", name->name);
5391         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5392         dprintk("NFS reply  secinfo: %d\n", status);
5393         return status;
5394 }
5395
5396 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5397                       struct nfs4_secinfo_flavors *flavors)
5398 {
5399         struct nfs4_exception exception = { };
5400         int err;
5401         do {
5402                 err = nfs4_handle_exception(NFS_SERVER(dir),
5403                                 _nfs4_proc_secinfo(dir, name, flavors),
5404                                 &exception);
5405         } while (exception.retry);
5406         return err;
5407 }
5408
5409 #ifdef CONFIG_NFS_V4_1
5410 /*
5411  * Check the exchange flags returned by the server for invalid flags, having
5412  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5413  * DS flags set.
5414  */
5415 static int nfs4_check_cl_exchange_flags(u32 flags)
5416 {
5417         if (flags & ~EXCHGID4_FLAG_MASK_R)
5418                 goto out_inval;
5419         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5420             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5421                 goto out_inval;
5422         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5423                 goto out_inval;
5424         return NFS_OK;
5425 out_inval:
5426         return -NFS4ERR_INVAL;
5427 }
5428
5429 static bool
5430 nfs41_same_server_scope(struct nfs41_server_scope *a,
5431                         struct nfs41_server_scope *b)
5432 {
5433         if (a->server_scope_sz == b->server_scope_sz &&
5434             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5435                 return true;
5436
5437         return false;
5438 }
5439
5440 /*
5441  * nfs4_proc_bind_conn_to_session()
5442  *
5443  * The 4.1 client currently uses the same TCP connection for the
5444  * fore and backchannel.
5445  */
5446 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5447 {
5448         int status;
5449         struct nfs41_bind_conn_to_session_res res;
5450         struct rpc_message msg = {
5451                 .rpc_proc =
5452                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5453                 .rpc_argp = clp,
5454                 .rpc_resp = &res,
5455                 .rpc_cred = cred,
5456         };
5457
5458         dprintk("--> %s\n", __func__);
5459
5460         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5461         if (unlikely(res.session == NULL)) {
5462                 status = -ENOMEM;
5463                 goto out;
5464         }
5465
5466         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5467         if (status == 0) {
5468                 if (memcmp(res.session->sess_id.data,
5469                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5470                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5471                         status = -EIO;
5472                         goto out_session;
5473                 }
5474                 if (res.dir != NFS4_CDFS4_BOTH) {
5475                         dprintk("NFS: %s: Unexpected direction from server\n",
5476                                 __func__);
5477                         status = -EIO;
5478                         goto out_session;
5479                 }
5480                 if (res.use_conn_in_rdma_mode) {
5481                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5482                                 __func__);
5483                         status = -EIO;
5484                         goto out_session;
5485                 }
5486         }
5487 out_session:
5488         kfree(res.session);
5489 out:
5490         dprintk("<-- %s status= %d\n", __func__, status);
5491         return status;
5492 }
5493
5494 /*
5495  * nfs4_proc_exchange_id()
5496  *
5497  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5498  *
5499  * Since the clientid has expired, all compounds using sessions
5500  * associated with the stale clientid will be returning
5501  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5502  * be in some phase of session reset.
5503  */
5504 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5505 {
5506         nfs4_verifier verifier;
5507         struct nfs41_exchange_id_args args = {
5508                 .verifier = &verifier,
5509                 .client = clp,
5510                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5511         };
5512         struct nfs41_exchange_id_res res = {
5513                 0
5514         };
5515         int status;
5516         struct rpc_message msg = {
5517                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5518                 .rpc_argp = &args,
5519                 .rpc_resp = &res,
5520                 .rpc_cred = cred,
5521         };
5522
5523         nfs4_init_boot_verifier(clp, &verifier);
5524         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5525                                                         sizeof(args.id));
5526         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5527                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5528                 args.id_len, args.id);
5529
5530         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5531                                         GFP_NOFS);
5532         if (unlikely(res.server_owner == NULL)) {
5533                 status = -ENOMEM;
5534                 goto out;
5535         }
5536
5537         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5538                                         GFP_NOFS);
5539         if (unlikely(res.server_scope == NULL)) {
5540                 status = -ENOMEM;
5541                 goto out_server_owner;
5542         }
5543
5544         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5545         if (unlikely(res.impl_id == NULL)) {
5546                 status = -ENOMEM;
5547                 goto out_server_scope;
5548         }
5549
5550         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5551         if (status == 0)
5552                 status = nfs4_check_cl_exchange_flags(res.flags);
5553
5554         if (status == 0) {
5555                 clp->cl_clientid = res.clientid;
5556                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5557                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5558                         clp->cl_seqid = res.seqid;
5559
5560                 kfree(clp->cl_serverowner);
5561                 clp->cl_serverowner = res.server_owner;
5562                 res.server_owner = NULL;
5563
5564                 /* use the most recent implementation id */
5565                 kfree(clp->cl_implid);
5566                 clp->cl_implid = res.impl_id;
5567
5568                 if (clp->cl_serverscope != NULL &&
5569                     !nfs41_same_server_scope(clp->cl_serverscope,
5570                                              res.server_scope)) {
5571                         dprintk("%s: server_scope mismatch detected\n",
5572                                 __func__);
5573                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5574                         kfree(clp->cl_serverscope);
5575                         clp->cl_serverscope = NULL;
5576                 }
5577
5578                 if (clp->cl_serverscope == NULL) {
5579                         clp->cl_serverscope = res.server_scope;
5580                         goto out;
5581                 }
5582         } else
5583                 kfree(res.impl_id);
5584
5585 out_server_owner:
5586         kfree(res.server_owner);
5587 out_server_scope:
5588         kfree(res.server_scope);
5589 out:
5590         if (clp->cl_implid != NULL)
5591                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5592                         "domain: %s, name: %s, date: %llu,%u\n",
5593                         clp->cl_implid->domain, clp->cl_implid->name,
5594                         clp->cl_implid->date.seconds,
5595                         clp->cl_implid->date.nseconds);
5596         dprintk("NFS reply exchange_id: %d\n", status);
5597         return status;
5598 }
5599
5600 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5601                 struct rpc_cred *cred)
5602 {
5603         struct rpc_message msg = {
5604                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5605                 .rpc_argp = clp,
5606                 .rpc_cred = cred,
5607         };
5608         int status;
5609
5610         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5611         if (status)
5612                 dprintk("NFS: Got error %d from the server %s on "
5613                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5614         return status;
5615 }
5616
5617 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5618                 struct rpc_cred *cred)
5619 {
5620         unsigned int loop;
5621         int ret;
5622
5623         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5624                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5625                 switch (ret) {
5626                 case -NFS4ERR_DELAY:
5627                 case -NFS4ERR_CLIENTID_BUSY:
5628                         ssleep(1);
5629                         break;
5630                 default:
5631                         return ret;
5632                 }
5633         }
5634         return 0;
5635 }
5636
5637 int nfs4_destroy_clientid(struct nfs_client *clp)
5638 {
5639         struct rpc_cred *cred;
5640         int ret = 0;
5641
5642         if (clp->cl_mvops->minor_version < 1)
5643                 goto out;
5644         if (clp->cl_exchange_flags == 0)
5645                 goto out;
5646         if (clp->cl_preserve_clid)
5647                 goto out;
5648         cred = nfs4_get_exchange_id_cred(clp);
5649         ret = nfs4_proc_destroy_clientid(clp, cred);
5650         if (cred)
5651                 put_rpccred(cred);
5652         switch (ret) {
5653         case 0:
5654         case -NFS4ERR_STALE_CLIENTID:
5655                 clp->cl_exchange_flags = 0;
5656         }
5657 out:
5658         return ret;
5659 }
5660
5661 struct nfs4_get_lease_time_data {
5662         struct nfs4_get_lease_time_args *args;
5663         struct nfs4_get_lease_time_res *res;
5664         struct nfs_client *clp;
5665 };
5666
5667 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5668                                         void *calldata)
5669 {
5670         struct nfs4_get_lease_time_data *data =
5671                         (struct nfs4_get_lease_time_data *)calldata;
5672
5673         dprintk("--> %s\n", __func__);
5674         /* just setup sequence, do not trigger session recovery
5675            since we're invoked within one */
5676         nfs41_setup_sequence(data->clp->cl_session,
5677                         &data->args->la_seq_args,
5678                         &data->res->lr_seq_res,
5679                         task);
5680         dprintk("<-- %s\n", __func__);
5681 }
5682
5683 /*
5684  * Called from nfs4_state_manager thread for session setup, so don't recover
5685  * from sequence operation or clientid errors.
5686  */
5687 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5688 {
5689         struct nfs4_get_lease_time_data *data =
5690                         (struct nfs4_get_lease_time_data *)calldata;
5691
5692         dprintk("--> %s\n", __func__);
5693         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5694                 return;
5695         switch (task->tk_status) {
5696         case -NFS4ERR_DELAY:
5697         case -NFS4ERR_GRACE:
5698                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5699                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5700                 task->tk_status = 0;
5701                 /* fall through */
5702         case -NFS4ERR_RETRY_UNCACHED_REP:
5703                 rpc_restart_call_prepare(task);
5704                 return;
5705         }
5706         dprintk("<-- %s\n", __func__);
5707 }
5708
5709 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5710         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5711         .rpc_call_done = nfs4_get_lease_time_done,
5712 };
5713
5714 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5715 {
5716         struct rpc_task *task;
5717         struct nfs4_get_lease_time_args args;
5718         struct nfs4_get_lease_time_res res = {
5719                 .lr_fsinfo = fsinfo,
5720         };
5721         struct nfs4_get_lease_time_data data = {
5722                 .args = &args,
5723                 .res = &res,
5724                 .clp = clp,
5725         };
5726         struct rpc_message msg = {
5727                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5728                 .rpc_argp = &args,
5729                 .rpc_resp = &res,
5730         };
5731         struct rpc_task_setup task_setup = {
5732                 .rpc_client = clp->cl_rpcclient,
5733                 .rpc_message = &msg,
5734                 .callback_ops = &nfs4_get_lease_time_ops,
5735                 .callback_data = &data,
5736                 .flags = RPC_TASK_TIMEOUT,
5737         };
5738         int status;
5739
5740         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5741         nfs4_set_sequence_privileged(&args.la_seq_args);
5742         dprintk("--> %s\n", __func__);
5743         task = rpc_run_task(&task_setup);
5744
5745         if (IS_ERR(task))
5746                 status = PTR_ERR(task);
5747         else {
5748                 status = task->tk_status;
5749                 rpc_put_task(task);
5750         }
5751         dprintk("<-- %s return %d\n", __func__, status);
5752
5753         return status;
5754 }
5755
5756 /*
5757  * Initialize the values to be used by the client in CREATE_SESSION
5758  * If nfs4_init_session set the fore channel request and response sizes,
5759  * use them.
5760  *
5761  * Set the back channel max_resp_sz_cached to zero to force the client to
5762  * always set csa_cachethis to FALSE because the current implementation
5763  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5764  */
5765 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5766 {
5767         struct nfs4_session *session = args->client->cl_session;
5768         unsigned int mxrqst_sz = session->fc_target_max_rqst_sz,
5769                      mxresp_sz = session->fc_target_max_resp_sz;
5770
5771         if (mxrqst_sz == 0)
5772                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5773         if (mxresp_sz == 0)
5774                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5775         /* Fore channel attributes */
5776         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5777         args->fc_attrs.max_resp_sz = mxresp_sz;
5778         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5779         args->fc_attrs.max_reqs = max_session_slots;
5780
5781         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5782                 "max_ops=%u max_reqs=%u\n",
5783                 __func__,
5784                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5785                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5786
5787         /* Back channel attributes */
5788         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5789         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5790         args->bc_attrs.max_resp_sz_cached = 0;
5791         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5792         args->bc_attrs.max_reqs = 1;
5793
5794         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5795                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5796                 __func__,
5797                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5798                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5799                 args->bc_attrs.max_reqs);
5800 }
5801
5802 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5803 {
5804         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5805         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5806
5807         if (rcvd->max_resp_sz > sent->max_resp_sz)
5808                 return -EINVAL;
5809         /*
5810          * Our requested max_ops is the minimum we need; we're not
5811          * prepared to break up compounds into smaller pieces than that.
5812          * So, no point even trying to continue if the server won't
5813          * cooperate:
5814          */
5815         if (rcvd->max_ops < sent->max_ops)
5816                 return -EINVAL;
5817         if (rcvd->max_reqs == 0)
5818                 return -EINVAL;
5819         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5820                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5821         return 0;
5822 }
5823
5824 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5825 {
5826         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5827         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5828
5829         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5830                 return -EINVAL;
5831         if (rcvd->max_resp_sz < sent->max_resp_sz)
5832                 return -EINVAL;
5833         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5834                 return -EINVAL;
5835         /* These would render the backchannel useless: */
5836         if (rcvd->max_ops != sent->max_ops)
5837                 return -EINVAL;
5838         if (rcvd->max_reqs != sent->max_reqs)
5839                 return -EINVAL;
5840         return 0;
5841 }
5842
5843 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5844                                      struct nfs4_session *session)
5845 {
5846         int ret;
5847
5848         ret = nfs4_verify_fore_channel_attrs(args, session);
5849         if (ret)
5850                 return ret;
5851         return nfs4_verify_back_channel_attrs(args, session);
5852 }
5853
5854 static int _nfs4_proc_create_session(struct nfs_client *clp,
5855                 struct rpc_cred *cred)
5856 {
5857         struct nfs4_session *session = clp->cl_session;
5858         struct nfs41_create_session_args args = {
5859                 .client = clp,
5860                 .cb_program = NFS4_CALLBACK,
5861         };
5862         struct nfs41_create_session_res res = {
5863                 .client = clp,
5864         };
5865         struct rpc_message msg = {
5866                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5867                 .rpc_argp = &args,
5868                 .rpc_resp = &res,
5869                 .rpc_cred = cred,
5870         };
5871         int status;
5872
5873         nfs4_init_channel_attrs(&args);
5874         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5875
5876         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5877
5878         if (!status) {
5879                 /* Verify the session's negotiated channel_attrs values */
5880                 status = nfs4_verify_channel_attrs(&args, session);
5881                 /* Increment the clientid slot sequence id */
5882                 clp->cl_seqid++;
5883         }
5884
5885         return status;
5886 }
5887
5888 /*
5889  * Issues a CREATE_SESSION operation to the server.
5890  * It is the responsibility of the caller to verify the session is
5891  * expired before calling this routine.
5892  */
5893 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5894 {
5895         int status;
5896         unsigned *ptr;
5897         struct nfs4_session *session = clp->cl_session;
5898
5899         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5900
5901         status = _nfs4_proc_create_session(clp, cred);
5902         if (status)
5903                 goto out;
5904
5905         /* Init or reset the session slot tables */
5906         status = nfs4_setup_session_slot_tables(session);
5907         dprintk("slot table setup returned %d\n", status);
5908         if (status)
5909                 goto out;
5910
5911         ptr = (unsigned *)&session->sess_id.data[0];
5912         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5913                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5914 out:
5915         dprintk("<-- %s\n", __func__);
5916         return status;
5917 }
5918
5919 /*
5920  * Issue the over-the-wire RPC DESTROY_SESSION.
5921  * The caller must serialize access to this routine.
5922  */
5923 int nfs4_proc_destroy_session(struct nfs4_session *session,
5924                 struct rpc_cred *cred)
5925 {
5926         struct rpc_message msg = {
5927                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5928                 .rpc_argp = session,
5929                 .rpc_cred = cred,
5930         };
5931         int status = 0;
5932
5933         dprintk("--> nfs4_proc_destroy_session\n");
5934
5935         /* session is still being setup */
5936         if (session->clp->cl_cons_state != NFS_CS_READY)
5937                 return status;
5938
5939         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5940
5941         if (status)
5942                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5943                         "Session has been destroyed regardless...\n", status);
5944
5945         dprintk("<-- nfs4_proc_destroy_session\n");
5946         return status;
5947 }
5948
5949 /*
5950  * Renew the cl_session lease.
5951  */
5952 struct nfs4_sequence_data {
5953         struct nfs_client *clp;
5954         struct nfs4_sequence_args args;
5955         struct nfs4_sequence_res res;
5956 };
5957
5958 static void nfs41_sequence_release(void *data)
5959 {
5960         struct nfs4_sequence_data *calldata = data;
5961         struct nfs_client *clp = calldata->clp;
5962
5963         if (atomic_read(&clp->cl_count) > 1)
5964                 nfs4_schedule_state_renewal(clp);
5965         nfs_put_client(clp);
5966         kfree(calldata);
5967 }
5968
5969 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5970 {
5971         switch(task->tk_status) {
5972         case -NFS4ERR_DELAY:
5973                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5974                 return -EAGAIN;
5975         default:
5976                 nfs4_schedule_lease_recovery(clp);
5977         }
5978         return 0;
5979 }
5980
5981 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5982 {
5983         struct nfs4_sequence_data *calldata = data;
5984         struct nfs_client *clp = calldata->clp;
5985
5986         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5987                 return;
5988
5989         if (task->tk_status < 0) {
5990                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5991                 if (atomic_read(&clp->cl_count) == 1)
5992                         goto out;
5993
5994                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5995                         rpc_restart_call_prepare(task);
5996                         return;
5997                 }
5998         }
5999         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
6000 out:
6001         dprintk("<-- %s\n", __func__);
6002 }
6003
6004 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
6005 {
6006         struct nfs4_sequence_data *calldata = data;
6007         struct nfs_client *clp = calldata->clp;
6008         struct nfs4_sequence_args *args;
6009         struct nfs4_sequence_res *res;
6010
6011         args = task->tk_msg.rpc_argp;
6012         res = task->tk_msg.rpc_resp;
6013
6014         nfs41_setup_sequence(clp->cl_session, args, res, task);
6015 }
6016
6017 static const struct rpc_call_ops nfs41_sequence_ops = {
6018         .rpc_call_done = nfs41_sequence_call_done,
6019         .rpc_call_prepare = nfs41_sequence_prepare,
6020         .rpc_release = nfs41_sequence_release,
6021 };
6022
6023 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
6024                 struct rpc_cred *cred,
6025                 bool is_privileged)
6026 {
6027         struct nfs4_sequence_data *calldata;
6028         struct rpc_message msg = {
6029                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
6030                 .rpc_cred = cred,
6031         };
6032         struct rpc_task_setup task_setup_data = {
6033                 .rpc_client = clp->cl_rpcclient,
6034                 .rpc_message = &msg,
6035                 .callback_ops = &nfs41_sequence_ops,
6036                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6037         };
6038
6039         if (!atomic_inc_not_zero(&clp->cl_count))
6040                 return ERR_PTR(-EIO);
6041         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6042         if (calldata == NULL) {
6043                 nfs_put_client(clp);
6044                 return ERR_PTR(-ENOMEM);
6045         }
6046         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6047         if (is_privileged)
6048                 nfs4_set_sequence_privileged(&calldata->args);
6049         msg.rpc_argp = &calldata->args;
6050         msg.rpc_resp = &calldata->res;
6051         calldata->clp = clp;
6052         task_setup_data.callback_data = calldata;
6053
6054         return rpc_run_task(&task_setup_data);
6055 }
6056
6057 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6058 {
6059         struct rpc_task *task;
6060         int ret = 0;
6061
6062         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6063                 return 0;
6064         task = _nfs41_proc_sequence(clp, cred, false);
6065         if (IS_ERR(task))
6066                 ret = PTR_ERR(task);
6067         else
6068                 rpc_put_task_async(task);
6069         dprintk("<-- %s status=%d\n", __func__, ret);
6070         return ret;
6071 }
6072
6073 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6074 {
6075         struct rpc_task *task;
6076         int ret;
6077
6078         task = _nfs41_proc_sequence(clp, cred, true);
6079         if (IS_ERR(task)) {
6080                 ret = PTR_ERR(task);
6081                 goto out;
6082         }
6083         ret = rpc_wait_for_completion_task(task);
6084         if (!ret) {
6085                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6086
6087                 if (task->tk_status == 0)
6088                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6089                 ret = task->tk_status;
6090         }
6091         rpc_put_task(task);
6092 out:
6093         dprintk("<-- %s status=%d\n", __func__, ret);
6094         return ret;
6095 }
6096
6097 struct nfs4_reclaim_complete_data {
6098         struct nfs_client *clp;
6099         struct nfs41_reclaim_complete_args arg;
6100         struct nfs41_reclaim_complete_res res;
6101 };
6102
6103 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6104 {
6105         struct nfs4_reclaim_complete_data *calldata = data;
6106
6107         nfs41_setup_sequence(calldata->clp->cl_session,
6108                         &calldata->arg.seq_args,
6109                         &calldata->res.seq_res,
6110                         task);
6111 }
6112
6113 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6114 {
6115         switch(task->tk_status) {
6116         case 0:
6117         case -NFS4ERR_COMPLETE_ALREADY:
6118         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6119                 break;
6120         case -NFS4ERR_DELAY:
6121                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6122                 /* fall through */
6123         case -NFS4ERR_RETRY_UNCACHED_REP:
6124                 return -EAGAIN;
6125         default:
6126                 nfs4_schedule_lease_recovery(clp);
6127         }
6128         return 0;
6129 }
6130
6131 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6132 {
6133         struct nfs4_reclaim_complete_data *calldata = data;
6134         struct nfs_client *clp = calldata->clp;
6135         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6136
6137         dprintk("--> %s\n", __func__);
6138         if (!nfs41_sequence_done(task, res))
6139                 return;
6140
6141         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6142                 rpc_restart_call_prepare(task);
6143                 return;
6144         }
6145         dprintk("<-- %s\n", __func__);
6146 }
6147
6148 static void nfs4_free_reclaim_complete_data(void *data)
6149 {
6150         struct nfs4_reclaim_complete_data *calldata = data;
6151
6152         kfree(calldata);
6153 }
6154
6155 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6156         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6157         .rpc_call_done = nfs4_reclaim_complete_done,
6158         .rpc_release = nfs4_free_reclaim_complete_data,
6159 };
6160
6161 /*
6162  * Issue a global reclaim complete.
6163  */
6164 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6165 {
6166         struct nfs4_reclaim_complete_data *calldata;
6167         struct rpc_task *task;
6168         struct rpc_message msg = {
6169                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6170         };
6171         struct rpc_task_setup task_setup_data = {
6172                 .rpc_client = clp->cl_rpcclient,
6173                 .rpc_message = &msg,
6174                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6175                 .flags = RPC_TASK_ASYNC,
6176         };
6177         int status = -ENOMEM;
6178
6179         dprintk("--> %s\n", __func__);
6180         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6181         if (calldata == NULL)
6182                 goto out;
6183         calldata->clp = clp;
6184         calldata->arg.one_fs = 0;
6185
6186         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6187         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
6188         msg.rpc_argp = &calldata->arg;
6189         msg.rpc_resp = &calldata->res;
6190         task_setup_data.callback_data = calldata;
6191         task = rpc_run_task(&task_setup_data);
6192         if (IS_ERR(task)) {
6193                 status = PTR_ERR(task);
6194                 goto out;
6195         }
6196         status = nfs4_wait_for_completion_rpc_task(task);
6197         if (status == 0)
6198                 status = task->tk_status;
6199         rpc_put_task(task);
6200         return 0;
6201 out:
6202         dprintk("<-- %s status=%d\n", __func__, status);
6203         return status;
6204 }
6205
6206 static void
6207 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6208 {
6209         struct nfs4_layoutget *lgp = calldata;
6210         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6211         struct nfs4_session *session = nfs4_get_session(server);
6212
6213         dprintk("--> %s\n", __func__);
6214         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6215          * right now covering the LAYOUTGET we are about to send.
6216          * However, that is not so catastrophic, and there seems
6217          * to be no way to prevent it completely.
6218          */
6219         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
6220                                 &lgp->res.seq_res, task))
6221                 return;
6222         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6223                                           NFS_I(lgp->args.inode)->layout,
6224                                           lgp->args.ctx->state)) {
6225                 rpc_exit(task, NFS4_OK);
6226         }
6227 }
6228
6229 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6230 {
6231         struct nfs4_layoutget *lgp = calldata;
6232         struct inode *inode = lgp->args.inode;
6233         struct nfs_server *server = NFS_SERVER(inode);
6234         struct pnfs_layout_hdr *lo;
6235         struct nfs4_state *state = NULL;
6236         unsigned long timeo, now, giveup;
6237
6238         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
6239
6240         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
6241                 goto out;
6242
6243         switch (task->tk_status) {
6244         case 0:
6245                 goto out;
6246         /*
6247          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
6248          * (or clients) writing to the same RAID stripe
6249          */
6250         case -NFS4ERR_LAYOUTTRYLATER:
6251         /*
6252          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
6253          * existing layout before getting a new one).
6254          */
6255         case -NFS4ERR_RECALLCONFLICT:
6256                 timeo = rpc_get_timeout(task->tk_client);
6257                 giveup = lgp->args.timestamp + timeo;
6258                 now = jiffies;
6259                 if (time_after(giveup, now)) {
6260                         unsigned long delay;
6261
6262                         /* Delay for:
6263                          * - Not less then NFS4_POLL_RETRY_MIN.
6264                          * - One last time a jiffie before we give up
6265                          * - exponential backoff (time_now minus start_attempt)
6266                          */
6267                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
6268                                     min((giveup - now - 1),
6269                                         now - lgp->args.timestamp));
6270
6271                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
6272                                 __func__, delay);
6273                         rpc_delay(task, delay);
6274                         task->tk_status = 0;
6275                         rpc_restart_call_prepare(task);
6276                         goto out; /* Do not call nfs4_async_handle_error() */
6277                 }
6278                 break;
6279         case -NFS4ERR_EXPIRED:
6280         case -NFS4ERR_BAD_STATEID:
6281                 spin_lock(&inode->i_lock);
6282                 lo = NFS_I(inode)->layout;
6283                 if (!lo || list_empty(&lo->plh_segs)) {
6284                         spin_unlock(&inode->i_lock);
6285                         /* If the open stateid was bad, then recover it. */
6286                         state = lgp->args.ctx->state;
6287                 } else {
6288                         LIST_HEAD(head);
6289
6290                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6291                         spin_unlock(&inode->i_lock);
6292                         /* Mark the bad layout state as invalid, then
6293                          * retry using the open stateid. */
6294                         pnfs_free_lseg_list(&head);
6295                 }
6296         }
6297         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6298                 rpc_restart_call_prepare(task);
6299 out:
6300         dprintk("<-- %s\n", __func__);
6301 }
6302
6303 static size_t max_response_pages(struct nfs_server *server)
6304 {
6305         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6306         return nfs_page_array_len(0, max_resp_sz);
6307 }
6308
6309 static void nfs4_free_pages(struct page **pages, size_t size)
6310 {
6311         int i;
6312
6313         if (!pages)
6314                 return;
6315
6316         for (i = 0; i < size; i++) {
6317                 if (!pages[i])
6318                         break;
6319                 __free_page(pages[i]);
6320         }
6321         kfree(pages);
6322 }
6323
6324 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6325 {
6326         struct page **pages;
6327         int i;
6328
6329         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6330         if (!pages) {
6331                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6332                 return NULL;
6333         }
6334
6335         for (i = 0; i < size; i++) {
6336                 pages[i] = alloc_page(gfp_flags);
6337                 if (!pages[i]) {
6338                         dprintk("%s: failed to allocate page\n", __func__);
6339                         nfs4_free_pages(pages, size);
6340                         return NULL;
6341                 }
6342         }
6343
6344         return pages;
6345 }
6346
6347 static void nfs4_layoutget_release(void *calldata)
6348 {
6349         struct nfs4_layoutget *lgp = calldata;
6350         struct inode *inode = lgp->args.inode;
6351         struct nfs_server *server = NFS_SERVER(inode);
6352         size_t max_pages = max_response_pages(server);
6353
6354         dprintk("--> %s\n", __func__);
6355         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6356         pnfs_put_layout_hdr(NFS_I(inode)->layout);
6357         put_nfs_open_context(lgp->args.ctx);
6358         kfree(calldata);
6359         dprintk("<-- %s\n", __func__);
6360 }
6361
6362 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6363         .rpc_call_prepare = nfs4_layoutget_prepare,
6364         .rpc_call_done = nfs4_layoutget_done,
6365         .rpc_release = nfs4_layoutget_release,
6366 };
6367
6368 struct pnfs_layout_segment *
6369 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6370 {
6371         struct inode *inode = lgp->args.inode;
6372         struct nfs_server *server = NFS_SERVER(inode);
6373         size_t max_pages = max_response_pages(server);
6374         struct rpc_task *task;
6375         struct rpc_message msg = {
6376                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6377                 .rpc_argp = &lgp->args,
6378                 .rpc_resp = &lgp->res,
6379         };
6380         struct rpc_task_setup task_setup_data = {
6381                 .rpc_client = server->client,
6382                 .rpc_message = &msg,
6383                 .callback_ops = &nfs4_layoutget_call_ops,
6384                 .callback_data = lgp,
6385                 .flags = RPC_TASK_ASYNC,
6386         };
6387         struct pnfs_layout_segment *lseg = NULL;
6388         int status = 0;
6389
6390         dprintk("--> %s\n", __func__);
6391
6392         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6393         if (!lgp->args.layout.pages) {
6394                 nfs4_layoutget_release(lgp);
6395                 return ERR_PTR(-ENOMEM);
6396         }
6397         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6398         lgp->args.timestamp = jiffies;
6399
6400         lgp->res.layoutp = &lgp->args.layout;
6401         lgp->res.seq_res.sr_slot = NULL;
6402         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6403
6404         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
6405         pnfs_get_layout_hdr(NFS_I(inode)->layout);
6406
6407         task = rpc_run_task(&task_setup_data);
6408         if (IS_ERR(task))
6409                 return ERR_CAST(task);
6410         status = nfs4_wait_for_completion_rpc_task(task);
6411         if (status == 0)
6412                 status = task->tk_status;
6413         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
6414         if (status == 0 && lgp->res.layoutp->len)
6415                 lseg = pnfs_layout_process(lgp);
6416         rpc_put_task(task);
6417         dprintk("<-- %s status=%d\n", __func__, status);
6418         if (status)
6419                 return ERR_PTR(status);
6420         return lseg;
6421 }
6422
6423 static void
6424 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6425 {
6426         struct nfs4_layoutreturn *lrp = calldata;
6427
6428         dprintk("--> %s\n", __func__);
6429         nfs41_setup_sequence(lrp->clp->cl_session,
6430                         &lrp->args.seq_args,
6431                         &lrp->res.seq_res,
6432                         task);
6433 }
6434
6435 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6436 {
6437         struct nfs4_layoutreturn *lrp = calldata;
6438         struct nfs_server *server;
6439
6440         dprintk("--> %s\n", __func__);
6441
6442         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
6443                 return;
6444
6445         server = NFS_SERVER(lrp->args.inode);
6446         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6447                 rpc_restart_call_prepare(task);
6448                 return;
6449         }
6450         dprintk("<-- %s\n", __func__);
6451 }
6452
6453 static void nfs4_layoutreturn_release(void *calldata)
6454 {
6455         struct nfs4_layoutreturn *lrp = calldata;
6456         struct pnfs_layout_hdr *lo = lrp->args.layout;
6457
6458         dprintk("--> %s\n", __func__);
6459         spin_lock(&lo->plh_inode->i_lock);
6460         if (lrp->res.lrs_present)
6461                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6462         lo->plh_block_lgets--;
6463         spin_unlock(&lo->plh_inode->i_lock);
6464         pnfs_put_layout_hdr(lrp->args.layout);
6465         kfree(calldata);
6466         dprintk("<-- %s\n", __func__);
6467 }
6468
6469 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6470         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6471         .rpc_call_done = nfs4_layoutreturn_done,
6472         .rpc_release = nfs4_layoutreturn_release,
6473 };
6474
6475 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6476 {
6477         struct rpc_task *task;
6478         struct rpc_message msg = {
6479                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6480                 .rpc_argp = &lrp->args,
6481                 .rpc_resp = &lrp->res,
6482         };
6483         struct rpc_task_setup task_setup_data = {
6484                 .rpc_client = lrp->clp->cl_rpcclient,
6485                 .rpc_message = &msg,
6486                 .callback_ops = &nfs4_layoutreturn_call_ops,
6487                 .callback_data = lrp,
6488         };
6489         int status;
6490
6491         dprintk("--> %s\n", __func__);
6492         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6493         task = rpc_run_task(&task_setup_data);
6494         if (IS_ERR(task))
6495                 return PTR_ERR(task);
6496         status = task->tk_status;
6497         dprintk("<-- %s status=%d\n", __func__, status);
6498         rpc_put_task(task);
6499         return status;
6500 }
6501
6502 /*
6503  * Retrieve the list of Data Server devices from the MDS.
6504  */
6505 static int _nfs4_getdevicelist(struct nfs_server *server,
6506                                     const struct nfs_fh *fh,
6507                                     struct pnfs_devicelist *devlist)
6508 {
6509         struct nfs4_getdevicelist_args args = {
6510                 .fh = fh,
6511                 .layoutclass = server->pnfs_curr_ld->id,
6512         };
6513         struct nfs4_getdevicelist_res res = {
6514                 .devlist = devlist,
6515         };
6516         struct rpc_message msg = {
6517                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6518                 .rpc_argp = &args,
6519                 .rpc_resp = &res,
6520         };
6521         int status;
6522
6523         dprintk("--> %s\n", __func__);
6524         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6525                                 &res.seq_res, 0);
6526         dprintk("<-- %s status=%d\n", __func__, status);
6527         return status;
6528 }
6529
6530 int nfs4_proc_getdevicelist(struct nfs_server *server,
6531                             const struct nfs_fh *fh,
6532                             struct pnfs_devicelist *devlist)
6533 {
6534         struct nfs4_exception exception = { };
6535         int err;
6536
6537         do {
6538                 err = nfs4_handle_exception(server,
6539                                 _nfs4_getdevicelist(server, fh, devlist),
6540                                 &exception);
6541         } while (exception.retry);
6542
6543         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6544                 err, devlist->num_devs);
6545
6546         return err;
6547 }
6548 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6549
6550 static int
6551 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6552 {
6553         struct nfs4_getdeviceinfo_args args = {
6554                 .pdev = pdev,
6555         };
6556         struct nfs4_getdeviceinfo_res res = {
6557                 .pdev = pdev,
6558         };
6559         struct rpc_message msg = {
6560                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6561                 .rpc_argp = &args,
6562                 .rpc_resp = &res,
6563         };
6564         int status;
6565
6566         dprintk("--> %s\n", __func__);
6567         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6568         dprintk("<-- %s status=%d\n", __func__, status);
6569
6570         return status;
6571 }
6572
6573 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6574 {
6575         struct nfs4_exception exception = { };
6576         int err;
6577
6578         do {
6579                 err = nfs4_handle_exception(server,
6580                                         _nfs4_proc_getdeviceinfo(server, pdev),
6581                                         &exception);
6582         } while (exception.retry);
6583         return err;
6584 }
6585 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6586
6587 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6588 {
6589         struct nfs4_layoutcommit_data *data = calldata;
6590         struct nfs_server *server = NFS_SERVER(data->args.inode);
6591         struct nfs4_session *session = nfs4_get_session(server);
6592
6593         nfs41_setup_sequence(session,
6594                         &data->args.seq_args,
6595                         &data->res.seq_res,
6596                         task);
6597 }
6598
6599 static void
6600 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6601 {
6602         struct nfs4_layoutcommit_data *data = calldata;
6603         struct nfs_server *server = NFS_SERVER(data->args.inode);
6604
6605         if (!nfs41_sequence_done(task, &data->res.seq_res))
6606                 return;
6607
6608         switch (task->tk_status) { /* Just ignore these failures */
6609         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6610         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6611         case -NFS4ERR_BADLAYOUT:     /* no layout */
6612         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6613                 task->tk_status = 0;
6614                 break;
6615         case 0:
6616                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6617                                                    data->res.fattr);
6618                 break;
6619         default:
6620                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6621                         rpc_restart_call_prepare(task);
6622                         return;
6623                 }
6624         }
6625 }
6626
6627 static void nfs4_layoutcommit_release(void *calldata)
6628 {
6629         struct nfs4_layoutcommit_data *data = calldata;
6630
6631         pnfs_cleanup_layoutcommit(data);
6632         put_rpccred(data->cred);
6633         kfree(data);
6634 }
6635
6636 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6637         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6638         .rpc_call_done = nfs4_layoutcommit_done,
6639         .rpc_release = nfs4_layoutcommit_release,
6640 };
6641
6642 int
6643 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6644 {
6645         struct rpc_message msg = {
6646                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6647                 .rpc_argp = &data->args,
6648                 .rpc_resp = &data->res,
6649                 .rpc_cred = data->cred,
6650         };
6651         struct rpc_task_setup task_setup_data = {
6652                 .task = &data->task,
6653                 .rpc_client = NFS_CLIENT(data->args.inode),
6654                 .rpc_message = &msg,
6655                 .callback_ops = &nfs4_layoutcommit_ops,
6656                 .callback_data = data,
6657                 .flags = RPC_TASK_ASYNC,
6658         };
6659         struct rpc_task *task;
6660         int status = 0;
6661
6662         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6663                 "lbw: %llu inode %lu\n",
6664                 data->task.tk_pid, sync,
6665                 data->args.lastbytewritten,
6666                 data->args.inode->i_ino);
6667
6668         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6669         task = rpc_run_task(&task_setup_data);
6670         if (IS_ERR(task))
6671                 return PTR_ERR(task);
6672         if (sync == false)
6673                 goto out;
6674         status = nfs4_wait_for_completion_rpc_task(task);
6675         if (status != 0)
6676                 goto out;
6677         status = task->tk_status;
6678 out:
6679         dprintk("%s: status %d\n", __func__, status);
6680         rpc_put_task(task);
6681         return status;
6682 }
6683
6684 static int
6685 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6686                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6687 {
6688         struct nfs41_secinfo_no_name_args args = {
6689                 .style = SECINFO_STYLE_CURRENT_FH,
6690         };
6691         struct nfs4_secinfo_res res = {
6692                 .flavors = flavors,
6693         };
6694         struct rpc_message msg = {
6695                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6696                 .rpc_argp = &args,
6697                 .rpc_resp = &res,
6698         };
6699         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6700 }
6701
6702 static int
6703 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6704                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6705 {
6706         struct nfs4_exception exception = { };
6707         int err;
6708         do {
6709                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6710                 switch (err) {
6711                 case 0:
6712                 case -NFS4ERR_WRONGSEC:
6713                 case -ENOTSUPP:
6714                         goto out;
6715                 default:
6716                         err = nfs4_handle_exception(server, err, &exception);
6717                 }
6718         } while (exception.retry);
6719 out:
6720         return err;
6721 }
6722
6723 static int
6724 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6725                     struct nfs_fsinfo *info)
6726 {
6727         int err;
6728         struct page *page;
6729         rpc_authflavor_t flavor;
6730         struct nfs4_secinfo_flavors *flavors;
6731
6732         page = alloc_page(GFP_KERNEL);
6733         if (!page) {
6734                 err = -ENOMEM;
6735                 goto out;
6736         }
6737
6738         flavors = page_address(page);
6739         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6740
6741         /*
6742          * Fall back on "guess and check" method if
6743          * the server doesn't support SECINFO_NO_NAME
6744          */
6745         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
6746                 err = nfs4_find_root_sec(server, fhandle, info);
6747                 goto out_freepage;
6748         }
6749         if (err)
6750                 goto out_freepage;
6751
6752         flavor = nfs_find_best_sec(flavors);
6753         if (err == 0)
6754                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6755
6756 out_freepage:
6757         put_page(page);
6758         if (err == -EACCES)
6759                 return -EPERM;
6760 out:
6761         return err;
6762 }
6763
6764 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6765 {
6766         int status;
6767         struct nfs41_test_stateid_args args = {
6768                 .stateid = stateid,
6769         };
6770         struct nfs41_test_stateid_res res;
6771         struct rpc_message msg = {
6772                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6773                 .rpc_argp = &args,
6774                 .rpc_resp = &res,
6775         };
6776
6777         dprintk("NFS call  test_stateid %p\n", stateid);
6778         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6779         nfs4_set_sequence_privileged(&args.seq_args);
6780         status = nfs4_call_sync_sequence(server->client, server, &msg,
6781                         &args.seq_args, &res.seq_res);
6782         if (status != NFS_OK) {
6783                 dprintk("NFS reply test_stateid: failed, %d\n", status);
6784                 return status;
6785         }
6786         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
6787         return -res.status;
6788 }
6789
6790 /**
6791  * nfs41_test_stateid - perform a TEST_STATEID operation
6792  *
6793  * @server: server / transport on which to perform the operation
6794  * @stateid: state ID to test
6795  *
6796  * Returns NFS_OK if the server recognizes that "stateid" is valid.
6797  * Otherwise a negative NFS4ERR value is returned if the operation
6798  * failed or the state ID is not currently valid.
6799  */
6800 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6801 {
6802         struct nfs4_exception exception = { };
6803         int err;
6804         do {
6805                 err = _nfs41_test_stateid(server, stateid);
6806                 if (err != -NFS4ERR_DELAY)
6807                         break;
6808                 nfs4_handle_exception(server, err, &exception);
6809         } while (exception.retry);
6810         return err;
6811 }
6812
6813 struct nfs_free_stateid_data {
6814         struct nfs_server *server;
6815         struct nfs41_free_stateid_args args;
6816         struct nfs41_free_stateid_res res;
6817 };
6818
6819 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
6820 {
6821         struct nfs_free_stateid_data *data = calldata;
6822         nfs41_setup_sequence(nfs4_get_session(data->server),
6823                         &data->args.seq_args,
6824                         &data->res.seq_res,
6825                         task);
6826 }
6827
6828 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
6829 {
6830         struct nfs_free_stateid_data *data = calldata;
6831
6832         nfs41_sequence_done(task, &data->res.seq_res);
6833
6834         switch (task->tk_status) {
6835         case -NFS4ERR_DELAY:
6836                 if (nfs4_async_handle_error(task, data->server, NULL) == -EAGAIN)
6837                         rpc_restart_call_prepare(task);
6838         }
6839 }
6840
6841 static void nfs41_free_stateid_release(void *calldata)
6842 {
6843         kfree(calldata);
6844 }
6845
6846 const struct rpc_call_ops nfs41_free_stateid_ops = {
6847         .rpc_call_prepare = nfs41_free_stateid_prepare,
6848         .rpc_call_done = nfs41_free_stateid_done,
6849         .rpc_release = nfs41_free_stateid_release,
6850 };
6851
6852 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
6853                 nfs4_stateid *stateid,
6854                 bool privileged)
6855 {
6856         struct rpc_message msg = {
6857                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6858         };
6859         struct rpc_task_setup task_setup = {
6860                 .rpc_client = server->client,
6861                 .rpc_message = &msg,
6862                 .callback_ops = &nfs41_free_stateid_ops,
6863                 .flags = RPC_TASK_ASYNC,
6864         };
6865         struct nfs_free_stateid_data *data;
6866
6867         dprintk("NFS call  free_stateid %p\n", stateid);
6868         data = kmalloc(sizeof(*data), GFP_NOFS);
6869         if (!data)
6870                 return ERR_PTR(-ENOMEM);
6871         data->server = server;
6872         nfs4_stateid_copy(&data->args.stateid, stateid);
6873
6874         task_setup.callback_data = data;
6875
6876         msg.rpc_argp = &data->args;
6877         msg.rpc_resp = &data->res;
6878         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6879         if (privileged)
6880                 nfs4_set_sequence_privileged(&data->args.seq_args);
6881
6882         return rpc_run_task(&task_setup);
6883 }
6884
6885 /**
6886  * nfs41_free_stateid - perform a FREE_STATEID operation
6887  *
6888  * @server: server / transport on which to perform the operation
6889  * @stateid: state ID to release
6890  *
6891  * Returns NFS_OK if the server freed "stateid".  Otherwise a
6892  * negative NFS4ERR value is returned.
6893  */
6894 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6895 {
6896         struct rpc_task *task;
6897         int ret;
6898
6899         task = _nfs41_free_stateid(server, stateid, true);
6900         if (IS_ERR(task))
6901                 return PTR_ERR(task);
6902         ret = rpc_wait_for_completion_task(task);
6903         if (!ret)
6904                 ret = task->tk_status;
6905         rpc_put_task(task);
6906         return ret;
6907 }
6908
6909 static int nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
6910 {
6911         struct rpc_task *task;
6912
6913         task = _nfs41_free_stateid(server, &lsp->ls_stateid, false);
6914         nfs4_free_lock_state(server, lsp);
6915         if (IS_ERR(task))
6916                 return PTR_ERR(task);
6917         rpc_put_task(task);
6918         return 0;
6919 }
6920
6921 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6922                 const nfs4_stateid *s2)
6923 {
6924         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6925                 return false;
6926
6927         if (s1->seqid == s2->seqid)
6928                 return true;
6929         if (s1->seqid == 0 || s2->seqid == 0)
6930                 return true;
6931
6932         return false;
6933 }
6934
6935 #endif /* CONFIG_NFS_V4_1 */
6936
6937 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6938                 const nfs4_stateid *s2)
6939 {
6940         return nfs4_stateid_match(s1, s2);
6941 }
6942
6943
6944 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6945         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6946         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6947         .recover_open   = nfs4_open_reclaim,
6948         .recover_lock   = nfs4_lock_reclaim,
6949         .establish_clid = nfs4_init_clientid,
6950         .get_clid_cred  = nfs4_get_setclientid_cred,
6951         .detect_trunking = nfs40_discover_server_trunking,
6952 };
6953
6954 #if defined(CONFIG_NFS_V4_1)
6955 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6956         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6957         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6958         .recover_open   = nfs4_open_reclaim,
6959         .recover_lock   = nfs4_lock_reclaim,
6960         .establish_clid = nfs41_init_clientid,
6961         .get_clid_cred  = nfs4_get_exchange_id_cred,
6962         .reclaim_complete = nfs41_proc_reclaim_complete,
6963         .detect_trunking = nfs41_discover_server_trunking,
6964 };
6965 #endif /* CONFIG_NFS_V4_1 */
6966
6967 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6968         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6969         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6970         .recover_open   = nfs4_open_expired,
6971         .recover_lock   = nfs4_lock_expired,
6972         .establish_clid = nfs4_init_clientid,
6973         .get_clid_cred  = nfs4_get_setclientid_cred,
6974 };
6975
6976 #if defined(CONFIG_NFS_V4_1)
6977 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6978         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6979         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6980         .recover_open   = nfs41_open_expired,
6981         .recover_lock   = nfs41_lock_expired,
6982         .establish_clid = nfs41_init_clientid,
6983         .get_clid_cred  = nfs4_get_exchange_id_cred,
6984 };
6985 #endif /* CONFIG_NFS_V4_1 */
6986
6987 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6988         .sched_state_renewal = nfs4_proc_async_renew,
6989         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6990         .renew_lease = nfs4_proc_renew,
6991 };
6992
6993 #if defined(CONFIG_NFS_V4_1)
6994 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6995         .sched_state_renewal = nfs41_proc_async_sequence,
6996         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6997         .renew_lease = nfs4_proc_sequence,
6998 };
6999 #endif
7000
7001 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
7002         .minor_version = 0,
7003         .init_caps = NFS_CAP_READDIRPLUS
7004                 | NFS_CAP_ATOMIC_OPEN
7005                 | NFS_CAP_CHANGE_ATTR
7006                 | NFS_CAP_POSIX_LOCK,
7007         .call_sync = _nfs4_call_sync,
7008         .match_stateid = nfs4_match_stateid,
7009         .find_root_sec = nfs4_find_root_sec,
7010         .free_lock_state = nfs4_release_lockowner,
7011         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
7012         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
7013         .state_renewal_ops = &nfs40_state_renewal_ops,
7014 };
7015
7016 #if defined(CONFIG_NFS_V4_1)
7017 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
7018         .minor_version = 1,
7019         .init_caps = NFS_CAP_READDIRPLUS
7020                 | NFS_CAP_ATOMIC_OPEN
7021                 | NFS_CAP_CHANGE_ATTR
7022                 | NFS_CAP_POSIX_LOCK
7023                 | NFS_CAP_STATEID_NFSV41
7024                 | NFS_CAP_ATOMIC_OPEN_V1,
7025         .call_sync = nfs4_call_sync_sequence,
7026         .match_stateid = nfs41_match_stateid,
7027         .find_root_sec = nfs41_find_root_sec,
7028         .free_lock_state = nfs41_free_lock_state,
7029         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
7030         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
7031         .state_renewal_ops = &nfs41_state_renewal_ops,
7032 };
7033 #endif
7034
7035 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
7036         [0] = &nfs_v4_0_minor_ops,
7037 #if defined(CONFIG_NFS_V4_1)
7038         [1] = &nfs_v4_1_minor_ops,
7039 #endif
7040 };
7041
7042 const struct inode_operations nfs4_dir_inode_operations = {
7043         .create         = nfs_create,
7044         .lookup         = nfs_lookup,
7045         .atomic_open    = nfs_atomic_open,
7046         .link           = nfs_link,
7047         .unlink         = nfs_unlink,
7048         .symlink        = nfs_symlink,
7049         .mkdir          = nfs_mkdir,
7050         .rmdir          = nfs_rmdir,
7051         .mknod          = nfs_mknod,
7052         .rename         = nfs_rename,
7053         .permission     = nfs_permission,
7054         .getattr        = nfs_getattr,
7055         .setattr        = nfs_setattr,
7056         .getxattr       = generic_getxattr,
7057         .setxattr       = generic_setxattr,
7058         .listxattr      = generic_listxattr,
7059         .removexattr    = generic_removexattr,
7060 };
7061
7062 static const struct inode_operations nfs4_file_inode_operations = {
7063         .permission     = nfs_permission,
7064         .getattr        = nfs_getattr,
7065         .setattr        = nfs_setattr,
7066         .getxattr       = generic_getxattr,
7067         .setxattr       = generic_setxattr,
7068         .listxattr      = generic_listxattr,
7069         .removexattr    = generic_removexattr,
7070 };
7071
7072 const struct nfs_rpc_ops nfs_v4_clientops = {
7073         .version        = 4,                    /* protocol version */
7074         .dentry_ops     = &nfs4_dentry_operations,
7075         .dir_inode_ops  = &nfs4_dir_inode_operations,
7076         .file_inode_ops = &nfs4_file_inode_operations,
7077         .file_ops       = &nfs4_file_operations,
7078         .getroot        = nfs4_proc_get_root,
7079         .submount       = nfs4_submount,
7080         .try_mount      = nfs4_try_mount,
7081         .getattr        = nfs4_proc_getattr,
7082         .setattr        = nfs4_proc_setattr,
7083         .lookup         = nfs4_proc_lookup,
7084         .access         = nfs4_proc_access,
7085         .readlink       = nfs4_proc_readlink,
7086         .create         = nfs4_proc_create,
7087         .remove         = nfs4_proc_remove,
7088         .unlink_setup   = nfs4_proc_unlink_setup,
7089         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
7090         .unlink_done    = nfs4_proc_unlink_done,
7091         .rename         = nfs4_proc_rename,
7092         .rename_setup   = nfs4_proc_rename_setup,
7093         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
7094         .rename_done    = nfs4_proc_rename_done,
7095         .link           = nfs4_proc_link,
7096         .symlink        = nfs4_proc_symlink,
7097         .mkdir          = nfs4_proc_mkdir,
7098         .rmdir          = nfs4_proc_remove,
7099         .readdir        = nfs4_proc_readdir,
7100         .mknod          = nfs4_proc_mknod,
7101         .statfs         = nfs4_proc_statfs,
7102         .fsinfo         = nfs4_proc_fsinfo,
7103         .pathconf       = nfs4_proc_pathconf,
7104         .set_capabilities = nfs4_server_capabilities,
7105         .decode_dirent  = nfs4_decode_dirent,
7106         .read_setup     = nfs4_proc_read_setup,
7107         .read_pageio_init = pnfs_pageio_init_read,
7108         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
7109         .read_done      = nfs4_read_done,
7110         .write_setup    = nfs4_proc_write_setup,
7111         .write_pageio_init = pnfs_pageio_init_write,
7112         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
7113         .write_done     = nfs4_write_done,
7114         .commit_setup   = nfs4_proc_commit_setup,
7115         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
7116         .commit_done    = nfs4_commit_done,
7117         .lock           = nfs4_proc_lock,
7118         .clear_acl_cache = nfs4_zap_acl_attr,
7119         .close_context  = nfs4_close_context,
7120         .open_context   = nfs4_atomic_open,
7121         .have_delegation = nfs4_have_delegation,
7122         .return_delegation = nfs4_inode_return_delegation,
7123         .alloc_client   = nfs4_alloc_client,
7124         .init_client    = nfs4_init_client,
7125         .free_client    = nfs4_free_client,
7126         .create_server  = nfs4_create_server,
7127         .clone_server   = nfs_clone_server,
7128 };
7129
7130 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7131         .prefix = XATTR_NAME_NFSV4_ACL,
7132         .list   = nfs4_xattr_list_nfs4_acl,
7133         .get    = nfs4_xattr_get_nfs4_acl,
7134         .set    = nfs4_xattr_set_nfs4_acl,
7135 };
7136
7137 const struct xattr_handler *nfs4_xattr_handlers[] = {
7138         &nfs4_xattr_nfs4_acl_handler,
7139         NULL
7140 };
7141
7142 /*
7143  * Local variables:
7144  *  c-basic-offset: 8
7145  * End:
7146  */