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/sunrpc/bc_xprt.h>
  56 #include <linux/xattr.h>
  57 #include <linux/utsname.h>
  58 #include <linux/freezer.h>
  59
  60 #include "nfs4_fs.h"
  61 #include "delegation.h"
  62 #include "internal.h"
  63 #include "iostat.h"
  64 #include "callback.h"
  65 #include "pnfs.h"
  66 #include "netns.h"
  67
  68 #define NFSDBG_FACILITY         NFSDBG_PROC
  69
  70 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  71 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  72
  73 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  74
  75 struct nfs4_opendata;
  76 static int _nfs4_proc_open(struct nfs4_opendata *data);
  77 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  78 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  79 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  80 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  81 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *);
  82 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  83 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  84                             struct nfs_fattr *fattr, struct iattr *sattr,
  85                             struct nfs4_state *state);
  86 #ifdef CONFIG_NFS_V4_1
  87 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
  88 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
  89 #endif
  90 /* Prevent leaks of NFSv4 errors into userland */
  91 static int nfs4_map_errors(int err)
  92 {
  93         if (err >= -1000)
  94                 return err;
  95         switch (err) {
  96         case -NFS4ERR_RESOURCE:
  97                 return -EREMOTEIO;
  98         case -NFS4ERR_WRONGSEC:
  99                 return -EPERM;
 100         case -NFS4ERR_BADOWNER:
 101         case -NFS4ERR_BADNAME:
 102                 return -EINVAL;
 103         case -NFS4ERR_SHARE_DENIED:
 104                 return -EACCES;
 105         case -NFS4ERR_MINOR_VERS_MISMATCH:
 106                 return -EPROTONOSUPPORT;
 107         case -NFS4ERR_ACCESS:
 108                 return -EACCES;
 109         default:
 110                 dprintk("%s could not handle NFSv4 error %d\n",
 111                                 __func__, -err);
 112                 break;
 113         }
 114         return -EIO;
 115 }
 116
 117 /*
 118  * This is our standard bitmap for GETATTR requests.
 119  */
 120 const u32 nfs4_fattr_bitmap[3] = {
 121         FATTR4_WORD0_TYPE
 122         | FATTR4_WORD0_CHANGE
 123         | FATTR4_WORD0_SIZE
 124         | FATTR4_WORD0_FSID
 125         | FATTR4_WORD0_FILEID,
 126         FATTR4_WORD1_MODE
 127         | FATTR4_WORD1_NUMLINKS
 128         | FATTR4_WORD1_OWNER
 129         | FATTR4_WORD1_OWNER_GROUP
 130         | FATTR4_WORD1_RAWDEV
 131         | FATTR4_WORD1_SPACE_USED
 132         | FATTR4_WORD1_TIME_ACCESS
 133         | FATTR4_WORD1_TIME_METADATA
 134         | FATTR4_WORD1_TIME_MODIFY
 135 };
 136
 137 static const u32 nfs4_pnfs_open_bitmap[3] = {
 138         FATTR4_WORD0_TYPE
 139         | FATTR4_WORD0_CHANGE
 140         | FATTR4_WORD0_SIZE
 141         | FATTR4_WORD0_FSID
 142         | FATTR4_WORD0_FILEID,
 143         FATTR4_WORD1_MODE
 144         | FATTR4_WORD1_NUMLINKS
 145         | FATTR4_WORD1_OWNER
 146         | FATTR4_WORD1_OWNER_GROUP
 147         | FATTR4_WORD1_RAWDEV
 148         | FATTR4_WORD1_SPACE_USED
 149         | FATTR4_WORD1_TIME_ACCESS
 150         | FATTR4_WORD1_TIME_METADATA
 151         | FATTR4_WORD1_TIME_MODIFY,
 152         FATTR4_WORD2_MDSTHRESHOLD
 153 };
 154
 155 static const u32 nfs4_open_noattr_bitmap[3] = {
 156         FATTR4_WORD0_TYPE
 157         | FATTR4_WORD0_CHANGE
 158         | FATTR4_WORD0_FILEID,
 159 };
 160
 161 const u32 nfs4_statfs_bitmap[2] = {
 162         FATTR4_WORD0_FILES_AVAIL
 163         | FATTR4_WORD0_FILES_FREE
 164         | FATTR4_WORD0_FILES_TOTAL,
 165         FATTR4_WORD1_SPACE_AVAIL
 166         | FATTR4_WORD1_SPACE_FREE
 167         | FATTR4_WORD1_SPACE_TOTAL
 168 };
 169
 170 const u32 nfs4_pathconf_bitmap[2] = {
 171         FATTR4_WORD0_MAXLINK
 172         | FATTR4_WORD0_MAXNAME,
 173         0
 174 };
 175
 176 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 177                         | FATTR4_WORD0_MAXREAD
 178                         | FATTR4_WORD0_MAXWRITE
 179                         | FATTR4_WORD0_LEASE_TIME,
 180                         FATTR4_WORD1_TIME_DELTA
 181                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 182                         FATTR4_WORD2_LAYOUT_BLKSIZE
 183 };
 184
 185 const u32 nfs4_fs_locations_bitmap[2] = {
 186         FATTR4_WORD0_TYPE
 187         | FATTR4_WORD0_CHANGE
 188         | FATTR4_WORD0_SIZE
 189         | FATTR4_WORD0_FSID
 190         | FATTR4_WORD0_FILEID
 191         | FATTR4_WORD0_FS_LOCATIONS,
 192         FATTR4_WORD1_MODE
 193         | FATTR4_WORD1_NUMLINKS
 194         | FATTR4_WORD1_OWNER
 195         | FATTR4_WORD1_OWNER_GROUP
 196         | FATTR4_WORD1_RAWDEV
 197         | FATTR4_WORD1_SPACE_USED
 198         | FATTR4_WORD1_TIME_ACCESS
 199         | FATTR4_WORD1_TIME_METADATA
 200         | FATTR4_WORD1_TIME_MODIFY
 201         | FATTR4_WORD1_MOUNTED_ON_FILEID
 202 };
 203
 204 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 205                 struct nfs4_readdir_arg *readdir)
 206 {
 207         __be32 *start, *p;
 208
 209         if (cookie > 2) {
 210                 readdir->cookie = cookie;
 211                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 212                 return;
 213         }
 214
 215         readdir->cookie = 0;
 216         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 217         if (cookie == 2)
 218                 return;
 219
 220         /*
 221          * NFSv4 servers do not return entries for '.' and '..'
 222          * Therefore, we fake these entries here.  We let '.'
 223          * have cookie 0 and '..' have cookie 1.  Note that
 224          * when talking to the server, we always send cookie 0
 225          * instead of 1 or 2.
 226          */
 227         start = p = kmap_atomic(*readdir->pages);
 228
 229         if (cookie == 0) {
 230                 *p++ = xdr_one;                                  /* next */
 231                 *p++ = xdr_zero;                   /* cookie, first word */
 232                 *p++ = xdr_one;                   /* cookie, second word */
 233                 *p++ = xdr_one;                             /* entry len */
 234                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 235                 p++;
 236                 *p++ = xdr_one;                         /* bitmap length */
 237                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 238                 *p++ = htonl(8);              /* attribute buffer length */
 239                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 240         }
 241
 242         *p++ = xdr_one;                                  /* next */
 243         *p++ = xdr_zero;                   /* cookie, first word */
 244         *p++ = xdr_two;                   /* cookie, second word */
 245         *p++ = xdr_two;                             /* entry len */
 246         memcpy(p, "..\0\0", 4);                         /* entry */
 247         p++;
 248         *p++ = xdr_one;                         /* bitmap length */
 249         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 250         *p++ = htonl(8);              /* attribute buffer length */
 251         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 252
 253         readdir->pgbase = (char *)p - (char *)start;
 254         readdir->count -= readdir->pgbase;
 255         kunmap_atomic(start);
 256 }
 257
 258 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 259 {
 260         int res;
 261
 262         might_sleep();
 263
 264         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 265                         nfs_wait_bit_killable, TASK_KILLABLE);
 266         if (res)
 267                 return res;
 268
 269         if (clp->cl_cons_state < 0)
 270                 return clp->cl_cons_state;
 271         return 0;
 272 }
 273
 274 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 275 {
 276         int res = 0;
 277
 278         might_sleep();
 279
 280         if (*timeout <= 0)
 281                 *timeout = NFS4_POLL_RETRY_MIN;
 282         if (*timeout > NFS4_POLL_RETRY_MAX)
 283                 *timeout = NFS4_POLL_RETRY_MAX;
 284         freezable_schedule_timeout_killable(*timeout);
 285         if (fatal_signal_pending(current))
 286                 res = -ERESTARTSYS;
 287         *timeout <<= 1;
 288         return res;
 289 }
 290
 291 /* This is the error handling routine for processes that are allowed
 292  * to sleep.
 293  */
 294 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 295 {
 296         struct nfs_client *clp = server->nfs_client;
 297         struct nfs4_state *state = exception->state;
 298         struct inode *inode = exception->inode;
 299         int ret = errorcode;
 300
 301         exception->retry = 0;
 302         switch(errorcode) {
 303                 case 0:
 304                         return 0;
 305                 case -NFS4ERR_OPENMODE:
 306                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 307                                 nfs4_inode_return_delegation(inode);
 308                                 exception->retry = 1;
 309                                 return 0;
 310                         }
 311                         if (state == NULL)
 312                                 break;
 313                         nfs4_schedule_stateid_recovery(server, state);
 314                         goto wait_on_recovery;
 315                 case -NFS4ERR_DELEG_REVOKED:
 316                 case -NFS4ERR_ADMIN_REVOKED:
 317                 case -NFS4ERR_BAD_STATEID:
 318                         if (state == NULL)
 319                                 break;
 320                         nfs_remove_bad_delegation(state->inode);
 321                         nfs4_schedule_stateid_recovery(server, state);
 322                         goto wait_on_recovery;
 323                 case -NFS4ERR_EXPIRED:
 324                         if (state != NULL)
 325                                 nfs4_schedule_stateid_recovery(server, state);
 326                 case -NFS4ERR_STALE_STATEID:
 327                 case -NFS4ERR_STALE_CLIENTID:
 328                         nfs4_schedule_lease_recovery(clp);
 329                         goto wait_on_recovery;
 330 #if defined(CONFIG_NFS_V4_1)
 331                 case -NFS4ERR_BADSESSION:
 332                 case -NFS4ERR_BADSLOT:
 333                 case -NFS4ERR_BAD_HIGH_SLOT:
 334                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 335                 case -NFS4ERR_DEADSESSION:
 336                 case -NFS4ERR_SEQ_FALSE_RETRY:
 337                 case -NFS4ERR_SEQ_MISORDERED:
 338                         dprintk("%s ERROR: %d Reset session\n", __func__,
 339                                 errorcode);
 340                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 341                         goto wait_on_recovery;
 342 #endif /* defined(CONFIG_NFS_V4_1) */
 343                 case -NFS4ERR_FILE_OPEN:
 344                         if (exception->timeout > HZ) {
 345                                 /* We have retried a decent amount, time to
 346                                  * fail
 347                                  */
 348                                 ret = -EBUSY;
 349                                 break;
 350                         }
 351                 case -NFS4ERR_GRACE:
 352                 case -NFS4ERR_DELAY:
 353                 case -EKEYEXPIRED:
 354                         ret = nfs4_delay(server->client, &exception->timeout);
 355                         if (ret != 0)
 356                                 break;
 357                 case -NFS4ERR_RETRY_UNCACHED_REP:
 358                 case -NFS4ERR_OLD_STATEID:
 359                         exception->retry = 1;
 360                         break;
 361                 case -NFS4ERR_BADOWNER:
 362                         /* The following works around a Linux server bug! */
 363                 case -NFS4ERR_BADNAME:
 364                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 365                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 366                                 exception->retry = 1;
 367                                 printk(KERN_WARNING "NFS: v4 server %s "
 368                                                 "does not accept raw "
 369                                                 "uid/gids. "
 370                                                 "Reenabling the idmapper.\n",
 371                                                 server->nfs_client->cl_hostname);
 372                         }
 373         }
 374         /* We failed to handle the error */
 375         return nfs4_map_errors(ret);
 376 wait_on_recovery:
 377         ret = nfs4_wait_clnt_recover(clp);
 378         if (ret == 0)
 379                 exception->retry = 1;
 380         return ret;
 381 }
 382
 383
 384 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 385 {
 386         spin_lock(&clp->cl_lock);
 387         if (time_before(clp->cl_last_renewal,timestamp))
 388                 clp->cl_last_renewal = timestamp;
 389         spin_unlock(&clp->cl_lock);
 390 }
 391
 392 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 393 {
 394         do_renew_lease(server->nfs_client, timestamp);
 395 }
 396
 397 #if defined(CONFIG_NFS_V4_1)
 398
 399 /*
 400  * nfs4_free_slot - free a slot and efficiently update slot table.
 401  *
 402  * freeing a slot is trivially done by clearing its respective bit
 403  * in the bitmap.
 404  * If the freed slotid equals highest_used_slotid we want to update it
 405  * so that the server would be able to size down the slot table if needed,
 406  * otherwise we know that the highest_used_slotid is still in use.
 407  * When updating highest_used_slotid there may be "holes" in the bitmap
 408  * so we need to scan down from highest_used_slotid to 0 looking for the now
 409  * highest slotid in use.
 410  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
 411  *
 412  * Must be called while holding tbl->slot_tbl_lock
 413  */
 414 static void
 415 nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
 416 {
 417         /* clear used bit in bitmap */
 418         __clear_bit(slotid, tbl->used_slots);
 419
 420         /* update highest_used_slotid when it is freed */
 421         if (slotid == tbl->highest_used_slotid) {
 422                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 423                 if (slotid < tbl->max_slots)
 424                         tbl->highest_used_slotid = slotid;
 425                 else
 426                         tbl->highest_used_slotid = NFS4_NO_SLOT;
 427         }
 428         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
 429                 slotid, tbl->highest_used_slotid);
 430 }
 431
 432 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
 433 {
 434         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 435         return true;
 436 }
 437
 438 /*
 439  * Signal state manager thread if session fore channel is drained
 440  */
 441 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 442 {
 443         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 444                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
 445                                 nfs4_set_task_privileged, NULL);
 446                 return;
 447         }
 448
 449         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 450                 return;
 451
 452         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 453         complete(&ses->fc_slot_table.complete);
 454 }
 455
 456 /*
 457  * Signal state manager thread if session back channel is drained
 458  */
 459 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 460 {
 461         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 462             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 463                 return;
 464         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 465         complete(&ses->bc_slot_table.complete);
 466 }
 467
 468 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 469 {
 470         struct nfs4_slot_table *tbl;
 471
 472         tbl = &res->sr_session->fc_slot_table;
 473         if (!res->sr_slot) {
 474                 /* just wake up the next guy waiting since
 475                  * we may have not consumed a slot after all */
 476                 dprintk("%s: No slot\n", __func__);
 477                 return;
 478         }
 479
 480         spin_lock(&tbl->slot_tbl_lock);
 481         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 482         nfs4_check_drain_fc_complete(res->sr_session);
 483         spin_unlock(&tbl->slot_tbl_lock);
 484         res->sr_slot = NULL;
 485 }
 486
 487 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 488 {
 489         unsigned long timestamp;
 490         struct nfs_client *clp;
 491
 492         /*
 493          * sr_status remains 1 if an RPC level error occurred. The server
 494          * may or may not have processed the sequence operation..
 495          * Proceed as if the server received and processed the sequence
 496          * operation.
 497          */
 498         if (res->sr_status == 1)
 499                 res->sr_status = NFS_OK;
 500
 501         /* don't increment the sequence number if the task wasn't sent */
 502         if (!RPC_WAS_SENT(task))
 503                 goto out;
 504
 505         /* Check the SEQUENCE operation status */
 506         switch (res->sr_status) {
 507         case 0:
 508                 /* Update the slot's sequence and clientid lease timer */
 509                 ++res->sr_slot->seq_nr;
 510                 timestamp = res->sr_renewal_time;
 511                 clp = res->sr_session->clp;
 512                 do_renew_lease(clp, timestamp);
 513                 /* Check sequence flags */
 514                 if (res->sr_status_flags != 0)
 515                         nfs4_schedule_lease_recovery(clp);
 516                 break;
 517         case -NFS4ERR_DELAY:
 518                 /* The server detected a resend of the RPC call and
 519                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 520                  * of RFC5661.
 521                  */
 522                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 523                         __func__,
 524                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 525                         res->sr_slot->seq_nr);
 526                 goto out_retry;
 527         default:
 528                 /* Just update the slot sequence no. */
 529                 ++res->sr_slot->seq_nr;
 530         }
 531 out:
 532         /* The session may be reset by one of the error handlers. */
 533         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 534         nfs41_sequence_free_slot(res);
 535         return 1;
 536 out_retry:
 537         if (!rpc_restart_call(task))
 538                 goto out;
 539         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 540         return 0;
 541 }
 542
 543 static int nfs4_sequence_done(struct rpc_task *task,
 544                                struct nfs4_sequence_res *res)
 545 {
 546         if (res->sr_session == NULL)
 547                 return 1;
 548         return nfs41_sequence_done(task, res);
 549 }
 550
 551 /*
 552  * nfs4_find_slot - efficiently look for a free slot
 553  *
 554  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 555  * If found, we mark the slot as used, update the highest_used_slotid,
 556  * and respectively set up the sequence operation args.
 557  * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
 558  *
 559  * Note: must be called with under the slot_tbl_lock.
 560  */
 561 static u32
 562 nfs4_find_slot(struct nfs4_slot_table *tbl)
 563 {
 564         u32 slotid;
 565         u32 ret_id = NFS4_NO_SLOT;
 566
 567         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
 568                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 569                 tbl->max_slots);
 570         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 571         if (slotid >= tbl->max_slots)
 572                 goto out;
 573         __set_bit(slotid, tbl->used_slots);
 574         if (slotid > tbl->highest_used_slotid ||
 575                         tbl->highest_used_slotid == NFS4_NO_SLOT)
 576                 tbl->highest_used_slotid = slotid;
 577         ret_id = slotid;
 578 out:
 579         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 580                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 581         return ret_id;
 582 }
 583
 584 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 585                 struct nfs4_sequence_res *res, int cache_reply)
 586 {
 587         args->sa_session = NULL;
 588         args->sa_cache_this = 0;
 589         if (cache_reply)
 590                 args->sa_cache_this = 1;
 591         res->sr_session = NULL;
 592         res->sr_slot = NULL;
 593 }
 594
 595 int nfs41_setup_sequence(struct nfs4_session *session,
 596                                 struct nfs4_sequence_args *args,
 597                                 struct nfs4_sequence_res *res,
 598                                 struct rpc_task *task)
 599 {
 600         struct nfs4_slot *slot;
 601         struct nfs4_slot_table *tbl;
 602         u32 slotid;
 603
 604         dprintk("--> %s\n", __func__);
 605         /* slot already allocated? */
 606         if (res->sr_slot != NULL)
 607                 return 0;
 608
 609         tbl = &session->fc_slot_table;
 610
 611         spin_lock(&tbl->slot_tbl_lock);
 612         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 613             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 614                 /* The state manager will wait until the slot table is empty */
 615                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 616                 spin_unlock(&tbl->slot_tbl_lock);
 617                 dprintk("%s session is draining\n", __func__);
 618                 return -EAGAIN;
 619         }
 620
 621         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 622             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 623                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 624                 spin_unlock(&tbl->slot_tbl_lock);
 625                 dprintk("%s enforce FIFO order\n", __func__);
 626                 return -EAGAIN;
 627         }
 628
 629         slotid = nfs4_find_slot(tbl);
 630         if (slotid == NFS4_NO_SLOT) {
 631                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 632                 spin_unlock(&tbl->slot_tbl_lock);
 633                 dprintk("<-- %s: no free slots\n", __func__);
 634                 return -EAGAIN;
 635         }
 636         spin_unlock(&tbl->slot_tbl_lock);
 637
 638         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 639         slot = tbl->slots + slotid;
 640         args->sa_session = session;
 641         args->sa_slotid = slotid;
 642
 643         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 644
 645         res->sr_session = session;
 646         res->sr_slot = slot;
 647         res->sr_renewal_time = jiffies;
 648         res->sr_status_flags = 0;
 649         /*
 650          * sr_status is only set in decode_sequence, and so will remain
 651          * set to 1 if an rpc level failure occurs.
 652          */
 653         res->sr_status = 1;
 654         return 0;
 655 }
 656 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 657
 658 int nfs4_setup_sequence(const struct nfs_server *server,
 659                         struct nfs4_sequence_args *args,
 660                         struct nfs4_sequence_res *res,
 661                         struct rpc_task *task)
 662 {
 663         struct nfs4_session *session = nfs4_get_session(server);
 664         int ret = 0;
 665
 666         if (session == NULL)
 667                 goto out;
 668
 669         dprintk("--> %s clp %p session %p sr_slot %td\n",
 670                 __func__, session->clp, session, res->sr_slot ?
 671                         res->sr_slot - session->fc_slot_table.slots : -1);
 672
 673         ret = nfs41_setup_sequence(session, args, res, task);
 674 out:
 675         dprintk("<-- %s status=%d\n", __func__, ret);
 676         return ret;
 677 }
 678
 679 struct nfs41_call_sync_data {
 680         const struct nfs_server *seq_server;
 681         struct nfs4_sequence_args *seq_args;
 682         struct nfs4_sequence_res *seq_res;
 683 };
 684
 685 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 686 {
 687         struct nfs41_call_sync_data *data = calldata;
 688
 689         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 690
 691         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 692                                 data->seq_res, task))
 693                 return;
 694         rpc_call_start(task);
 695 }
 696
 697 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 698 {
 699         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 700         nfs41_call_sync_prepare(task, calldata);
 701 }
 702
 703 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 704 {
 705         struct nfs41_call_sync_data *data = calldata;
 706
 707         nfs41_sequence_done(task, data->seq_res);
 708 }
 709
 710 static const struct rpc_call_ops nfs41_call_sync_ops = {
 711         .rpc_call_prepare = nfs41_call_sync_prepare,
 712         .rpc_call_done = nfs41_call_sync_done,
 713 };
 714
 715 static const struct rpc_call_ops nfs41_call_priv_sync_ops = {
 716         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 717         .rpc_call_done = nfs41_call_sync_done,
 718 };
 719
 720 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 721                                    struct nfs_server *server,
 722                                    struct rpc_message *msg,
 723                                    struct nfs4_sequence_args *args,
 724                                    struct nfs4_sequence_res *res,
 725                                    int privileged)
 726 {
 727         int ret;
 728         struct rpc_task *task;
 729         struct nfs41_call_sync_data data = {
 730                 .seq_server = server,
 731                 .seq_args = args,
 732                 .seq_res = res,
 733         };
 734         struct rpc_task_setup task_setup = {
 735                 .rpc_client = clnt,
 736                 .rpc_message = msg,
 737                 .callback_ops = &nfs41_call_sync_ops,
 738                 .callback_data = &data
 739         };
 740
 741         if (privileged)
 742                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 743         task = rpc_run_task(&task_setup);
 744         if (IS_ERR(task))
 745                 ret = PTR_ERR(task);
 746         else {
 747                 ret = task->tk_status;
 748                 rpc_put_task(task);
 749         }
 750         return ret;
 751 }
 752
 753 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 754                             struct nfs_server *server,
 755                             struct rpc_message *msg,
 756                             struct nfs4_sequence_args *args,
 757                             struct nfs4_sequence_res *res,
 758                             int cache_reply)
 759 {
 760         nfs41_init_sequence(args, res, cache_reply);
 761         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
 762 }
 763
 764 #else
 765 static inline
 766 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 767                 struct nfs4_sequence_res *res, int cache_reply)
 768 {
 769 }
 770
 771 static int nfs4_sequence_done(struct rpc_task *task,
 772                                struct nfs4_sequence_res *res)
 773 {
 774         return 1;
 775 }
 776 #endif /* CONFIG_NFS_V4_1 */
 777
 778 int _nfs4_call_sync(struct rpc_clnt *clnt,
 779                     struct nfs_server *server,
 780                     struct rpc_message *msg,
 781                     struct nfs4_sequence_args *args,
 782                     struct nfs4_sequence_res *res,
 783                     int cache_reply)
 784 {
 785         nfs41_init_sequence(args, res, cache_reply);
 786         return rpc_call_sync(clnt, msg, 0);
 787 }
 788
 789 static inline
 790 int nfs4_call_sync(struct rpc_clnt *clnt,
 791                    struct nfs_server *server,
 792                    struct rpc_message *msg,
 793                    struct nfs4_sequence_args *args,
 794                    struct nfs4_sequence_res *res,
 795                    int cache_reply)
 796 {
 797         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 798                                                 args, res, cache_reply);
 799 }
 800
 801 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 802 {
 803         struct nfs_inode *nfsi = NFS_I(dir);
 804
 805         spin_lock(&dir->i_lock);
 806         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 807         if (!cinfo->atomic || cinfo->before != dir->i_version)
 808                 nfs_force_lookup_revalidate(dir);
 809         dir->i_version = cinfo->after;
 810         spin_unlock(&dir->i_lock);
 811 }
 812
 813 struct nfs4_opendata {
 814         struct kref kref;
 815         struct nfs_openargs o_arg;
 816         struct nfs_openres o_res;
 817         struct nfs_open_confirmargs c_arg;
 818         struct nfs_open_confirmres c_res;
 819         struct nfs4_string owner_name;
 820         struct nfs4_string group_name;
 821         struct nfs_fattr f_attr;
 822         struct dentry *dir;
 823         struct dentry *dentry;
 824         struct nfs4_state_owner *owner;
 825         struct nfs4_state *state;
 826         struct iattr attrs;
 827         unsigned long timestamp;
 828         unsigned int rpc_done : 1;
 829         int rpc_status;
 830         int cancelled;
 831 };
 832
 833
 834 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 835 {
 836         p->o_res.f_attr = &p->f_attr;
 837         p->o_res.seqid = p->o_arg.seqid;
 838         p->c_res.seqid = p->c_arg.seqid;
 839         p->o_res.server = p->o_arg.server;
 840         p->o_res.access_request = p->o_arg.access;
 841         nfs_fattr_init(&p->f_attr);
 842         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 843 }
 844
 845 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 846                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 847                 const struct iattr *attrs,
 848                 gfp_t gfp_mask)
 849 {
 850         struct dentry *parent = dget_parent(dentry);
 851         struct inode *dir = parent->d_inode;
 852         struct nfs_server *server = NFS_SERVER(dir);
 853         struct nfs4_opendata *p;
 854
 855         p = kzalloc(sizeof(*p), gfp_mask);
 856         if (p == NULL)
 857                 goto err;
 858         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 859         if (p->o_arg.seqid == NULL)
 860                 goto err_free;
 861         nfs_sb_active(dentry->d_sb);
 862         p->dentry = dget(dentry);
 863         p->dir = parent;
 864         p->owner = sp;
 865         atomic_inc(&sp->so_count);
 866         p->o_arg.fh = NFS_FH(dir);
 867         p->o_arg.open_flags = flags;
 868         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 869         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
 870          * will return permission denied for all bits until close */
 871         if (!(flags & O_EXCL)) {
 872                 /* ask server to check for all possible rights as results
 873                  * are cached */
 874                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
 875                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
 876         }
 877         p->o_arg.clientid = server->nfs_client->cl_clientid;
 878         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
 879         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 880         p->o_arg.name = &dentry->d_name;
 881         p->o_arg.server = server;
 882         p->o_arg.bitmask = server->attr_bitmask;
 883         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
 884         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 885         if (attrs != NULL && attrs->ia_valid != 0) {
 886                 __be32 verf[2];
 887
 888                 p->o_arg.u.attrs = &p->attrs;
 889                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 890
 891                 verf[0] = jiffies;
 892                 verf[1] = current->pid;
 893                 memcpy(p->o_arg.u.verifier.data, verf,
 894                                 sizeof(p->o_arg.u.verifier.data));
 895         }
 896         p->c_arg.fh = &p->o_res.fh;
 897         p->c_arg.stateid = &p->o_res.stateid;
 898         p->c_arg.seqid = p->o_arg.seqid;
 899         nfs4_init_opendata_res(p);
 900         kref_init(&p->kref);
 901         return p;
 902 err_free:
 903         kfree(p);
 904 err:
 905         dput(parent);
 906         return NULL;
 907 }
 908
 909 static void nfs4_opendata_free(struct kref *kref)
 910 {
 911         struct nfs4_opendata *p = container_of(kref,
 912                         struct nfs4_opendata, kref);
 913         struct super_block *sb = p->dentry->d_sb;
 914
 915         nfs_free_seqid(p->o_arg.seqid);
 916         if (p->state != NULL)
 917                 nfs4_put_open_state(p->state);
 918         nfs4_put_state_owner(p->owner);
 919         dput(p->dir);
 920         dput(p->dentry);
 921         nfs_sb_deactive(sb);
 922         nfs_fattr_free_names(&p->f_attr);
 923         kfree(p);
 924 }
 925
 926 static void nfs4_opendata_put(struct nfs4_opendata *p)
 927 {
 928         if (p != NULL)
 929                 kref_put(&p->kref, nfs4_opendata_free);
 930 }
 931
 932 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 933 {
 934         int ret;
 935
 936         ret = rpc_wait_for_completion_task(task);
 937         return ret;
 938 }
 939
 940 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 941 {
 942         int ret = 0;
 943
 944         if (open_mode & (O_EXCL|O_TRUNC))
 945                 goto out;
 946         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 947                 case FMODE_READ:
 948                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 949                                 && state->n_rdonly != 0;
 950                         break;
 951                 case FMODE_WRITE:
 952                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 953                                 && state->n_wronly != 0;
 954                         break;
 955                 case FMODE_READ|FMODE_WRITE:
 956                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 957                                 && state->n_rdwr != 0;
 958         }
 959 out:
 960         return ret;
 961 }
 962
 963 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 964 {
 965         if (delegation == NULL)
 966                 return 0;
 967         if ((delegation->type & fmode) != fmode)
 968                 return 0;
 969         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 970                 return 0;
 971         nfs_mark_delegation_referenced(delegation);
 972         return 1;
 973 }
 974
 975 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 976 {
 977         switch (fmode) {
 978                 case FMODE_WRITE:
 979                         state->n_wronly++;
 980                         break;
 981                 case FMODE_READ:
 982                         state->n_rdonly++;
 983                         break;
 984                 case FMODE_READ|FMODE_WRITE:
 985                         state->n_rdwr++;
 986         }
 987         nfs4_state_set_mode_locked(state, state->state | fmode);
 988 }
 989
 990 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 991 {
 992         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 993                 nfs4_stateid_copy(&state->stateid, stateid);
 994         nfs4_stateid_copy(&state->open_stateid, stateid);
 995         switch (fmode) {
 996                 case FMODE_READ:
 997                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 998                         break;
 999                 case FMODE_WRITE:
1000                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1001                         break;
1002                 case FMODE_READ|FMODE_WRITE:
1003                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1004         }
1005 }
1006
1007 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1008 {
1009         write_seqlock(&state->seqlock);
1010         nfs_set_open_stateid_locked(state, stateid, fmode);
1011         write_sequnlock(&state->seqlock);
1012 }
1013
1014 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1015 {
1016         /*
1017          * Protect the call to nfs4_state_set_mode_locked and
1018          * serialise the stateid update
1019          */
1020         write_seqlock(&state->seqlock);
1021         if (deleg_stateid != NULL) {
1022                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1023                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1024         }
1025         if (open_stateid != NULL)
1026                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1027         write_sequnlock(&state->seqlock);
1028         spin_lock(&state->owner->so_lock);
1029         update_open_stateflags(state, fmode);
1030         spin_unlock(&state->owner->so_lock);
1031 }
1032
1033 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1034 {
1035         struct nfs_inode *nfsi = NFS_I(state->inode);
1036         struct nfs_delegation *deleg_cur;
1037         int ret = 0;
1038
1039         fmode &= (FMODE_READ|FMODE_WRITE);
1040
1041         rcu_read_lock();
1042         deleg_cur = rcu_dereference(nfsi->delegation);
1043         if (deleg_cur == NULL)
1044                 goto no_delegation;
1045
1046         spin_lock(&deleg_cur->lock);
1047         if (nfsi->delegation != deleg_cur ||
1048             (deleg_cur->type & fmode) != fmode)
1049                 goto no_delegation_unlock;
1050
1051         if (delegation == NULL)
1052                 delegation = &deleg_cur->stateid;
1053         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1054                 goto no_delegation_unlock;
1055
1056         nfs_mark_delegation_referenced(deleg_cur);
1057         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1058         ret = 1;
1059 no_delegation_unlock:
1060         spin_unlock(&deleg_cur->lock);
1061 no_delegation:
1062         rcu_read_unlock();
1063
1064         if (!ret && open_stateid != NULL) {
1065                 __update_open_stateid(state, open_stateid, NULL, fmode);
1066                 ret = 1;
1067         }
1068
1069         return ret;
1070 }
1071
1072
1073 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1074 {
1075         struct nfs_delegation *delegation;
1076
1077         rcu_read_lock();
1078         delegation = rcu_dereference(NFS_I(inode)->delegation);
1079         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1080                 rcu_read_unlock();
1081                 return;
1082         }
1083         rcu_read_unlock();
1084         nfs4_inode_return_delegation(inode);
1085 }
1086
1087 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1088 {
1089         struct nfs4_state *state = opendata->state;
1090         struct nfs_inode *nfsi = NFS_I(state->inode);
1091         struct nfs_delegation *delegation;
1092         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1093         fmode_t fmode = opendata->o_arg.fmode;
1094         nfs4_stateid stateid;
1095         int ret = -EAGAIN;
1096
1097         for (;;) {
1098                 if (can_open_cached(state, fmode, open_mode)) {
1099                         spin_lock(&state->owner->so_lock);
1100                         if (can_open_cached(state, fmode, open_mode)) {
1101                                 update_open_stateflags(state, fmode);
1102                                 spin_unlock(&state->owner->so_lock);
1103                                 goto out_return_state;
1104                         }
1105                         spin_unlock(&state->owner->so_lock);
1106                 }
1107                 rcu_read_lock();
1108                 delegation = rcu_dereference(nfsi->delegation);
1109                 if (!can_open_delegated(delegation, fmode)) {
1110                         rcu_read_unlock();
1111                         break;
1112                 }
1113                 /* Save the delegation */
1114                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1115                 rcu_read_unlock();
1116                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1117                 if (ret != 0)
1118                         goto out;
1119                 ret = -EAGAIN;
1120
1121                 /* Try to update the stateid using the delegation */
1122                 if (update_open_stateid(state, NULL, &stateid, fmode))
1123                         goto out_return_state;
1124         }
1125 out:
1126         return ERR_PTR(ret);
1127 out_return_state:
1128         atomic_inc(&state->count);
1129         return state;
1130 }
1131
1132 static void
1133 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1134 {
1135         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1136         struct nfs_delegation *delegation;
1137         int delegation_flags = 0;
1138
1139         rcu_read_lock();
1140         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1141         if (delegation)
1142                 delegation_flags = delegation->flags;
1143         rcu_read_unlock();
1144         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1145                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1146                                    "returning a delegation for "
1147                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1148                                    clp->cl_hostname);
1149         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1150                 nfs_inode_set_delegation(state->inode,
1151                                          data->owner->so_cred,
1152                                          &data->o_res);
1153         else
1154                 nfs_inode_reclaim_delegation(state->inode,
1155                                              data->owner->so_cred,
1156                                              &data->o_res);
1157 }
1158
1159 /*
1160  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1161  * and update the nfs4_state.
1162  */
1163 static struct nfs4_state *
1164 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1165 {
1166         struct inode *inode = data->state->inode;
1167         struct nfs4_state *state = data->state;
1168         int ret;
1169
1170         if (!data->rpc_done) {
1171                 ret = data->rpc_status;
1172                 goto err;
1173         }
1174
1175         ret = -ESTALE;
1176         if (!(data->f_attr.valid & NFS_ATTR_FATTR_TYPE) ||
1177             !(data->f_attr.valid & NFS_ATTR_FATTR_FILEID) ||
1178             !(data->f_attr.valid & NFS_ATTR_FATTR_CHANGE))
1179                 goto err;
1180
1181         ret = -ENOMEM;
1182         state = nfs4_get_open_state(inode, data->owner);
1183         if (state == NULL)
1184                 goto err;
1185
1186         ret = nfs_refresh_inode(inode, &data->f_attr);
1187         if (ret)
1188                 goto err;
1189
1190         if (data->o_res.delegation_type != 0)
1191                 nfs4_opendata_check_deleg(data, state);
1192         update_open_stateid(state, &data->o_res.stateid, NULL,
1193                             data->o_arg.fmode);
1194
1195         return state;
1196 err:
1197         return ERR_PTR(ret);
1198
1199 }
1200
1201 static struct nfs4_state *
1202 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1203 {
1204         struct inode *inode;
1205         struct nfs4_state *state = NULL;
1206         int ret;
1207
1208         if (!data->rpc_done) {
1209                 state = nfs4_try_open_cached(data);
1210                 goto out;
1211         }
1212
1213         ret = -EAGAIN;
1214         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1215                 goto err;
1216         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1217         ret = PTR_ERR(inode);
1218         if (IS_ERR(inode))
1219                 goto err;
1220         ret = -ENOMEM;
1221         state = nfs4_get_open_state(inode, data->owner);
1222         if (state == NULL)
1223                 goto err_put_inode;
1224         if (data->o_res.delegation_type != 0)
1225                 nfs4_opendata_check_deleg(data, state);
1226         update_open_stateid(state, &data->o_res.stateid, NULL,
1227                         data->o_arg.fmode);
1228         iput(inode);
1229 out:
1230         return state;
1231 err_put_inode:
1232         iput(inode);
1233 err:
1234         return ERR_PTR(ret);
1235 }
1236
1237 static struct nfs4_state *
1238 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1239 {
1240         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1241                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1242         return _nfs4_opendata_to_nfs4_state(data);
1243 }
1244
1245 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1246 {
1247         struct nfs_inode *nfsi = NFS_I(state->inode);
1248         struct nfs_open_context *ctx;
1249
1250         spin_lock(&state->inode->i_lock);
1251         list_for_each_entry(ctx, &nfsi->open_files, list) {
1252                 if (ctx->state != state)
1253                         continue;
1254                 get_nfs_open_context(ctx);
1255                 spin_unlock(&state->inode->i_lock);
1256                 return ctx;
1257         }
1258         spin_unlock(&state->inode->i_lock);
1259         return ERR_PTR(-ENOENT);
1260 }
1261
1262 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1263 {
1264         struct nfs4_opendata *opendata;
1265
1266         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1267         if (opendata == NULL)
1268                 return ERR_PTR(-ENOMEM);
1269         opendata->state = state;
1270         atomic_inc(&state->count);
1271         return opendata;
1272 }
1273
1274 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1275 {
1276         struct nfs4_state *newstate;
1277         int ret;
1278
1279         opendata->o_arg.open_flags = 0;
1280         opendata->o_arg.fmode = fmode;
1281         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1282         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1283         nfs4_init_opendata_res(opendata);
1284         ret = _nfs4_recover_proc_open(opendata);
1285         if (ret != 0)
1286                 return ret;
1287         newstate = nfs4_opendata_to_nfs4_state(opendata);
1288         if (IS_ERR(newstate))
1289                 return PTR_ERR(newstate);
1290         nfs4_close_state(newstate, fmode);
1291         *res = newstate;
1292         return 0;
1293 }
1294
1295 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1296 {
1297         struct nfs4_state *newstate;
1298         int ret;
1299
1300         /* memory barrier prior to reading state->n_* */
1301         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1302         smp_rmb();
1303         if (state->n_rdwr != 0) {
1304                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1305                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1306                 if (ret != 0)
1307                         return ret;
1308                 if (newstate != state)
1309                         return -ESTALE;
1310         }
1311         if (state->n_wronly != 0) {
1312                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1313                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1314                 if (ret != 0)
1315                         return ret;
1316                 if (newstate != state)
1317                         return -ESTALE;
1318         }
1319         if (state->n_rdonly != 0) {
1320                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1321                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1322                 if (ret != 0)
1323                         return ret;
1324                 if (newstate != state)
1325                         return -ESTALE;
1326         }
1327         /*
1328          * We may have performed cached opens for all three recoveries.
1329          * Check if we need to update the current stateid.
1330          */
1331         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1332             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1333                 write_seqlock(&state->seqlock);
1334                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1335                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1336                 write_sequnlock(&state->seqlock);
1337         }
1338         return 0;
1339 }
1340
1341 /*
1342  * OPEN_RECLAIM:
1343  *      reclaim state on the server after a reboot.
1344  */
1345 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1346 {
1347         struct nfs_delegation *delegation;
1348         struct nfs4_opendata *opendata;
1349         fmode_t delegation_type = 0;
1350         int status;
1351
1352         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1353         if (IS_ERR(opendata))
1354                 return PTR_ERR(opendata);
1355         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1356         opendata->o_arg.fh = NFS_FH(state->inode);
1357         rcu_read_lock();
1358         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1359         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1360                 delegation_type = delegation->type;
1361         rcu_read_unlock();
1362         opendata->o_arg.u.delegation_type = delegation_type;
1363         status = nfs4_open_recover(opendata, state);
1364         nfs4_opendata_put(opendata);
1365         return status;
1366 }
1367
1368 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1369 {
1370         struct nfs_server *server = NFS_SERVER(state->inode);
1371         struct nfs4_exception exception = { };
1372         int err;
1373         do {
1374                 err = _nfs4_do_open_reclaim(ctx, state);
1375                 if (err != -NFS4ERR_DELAY)
1376                         break;
1377                 nfs4_handle_exception(server, err, &exception);
1378         } while (exception.retry);
1379         return err;
1380 }
1381
1382 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1383 {
1384         struct nfs_open_context *ctx;
1385         int ret;
1386
1387         ctx = nfs4_state_find_open_context(state);
1388         if (IS_ERR(ctx))
1389                 return PTR_ERR(ctx);
1390         ret = nfs4_do_open_reclaim(ctx, state);
1391         put_nfs_open_context(ctx);
1392         return ret;
1393 }
1394
1395 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1396 {
1397         struct nfs4_opendata *opendata;
1398         int ret;
1399
1400         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1401         if (IS_ERR(opendata))
1402                 return PTR_ERR(opendata);
1403         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1404         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1405         ret = nfs4_open_recover(opendata, state);
1406         nfs4_opendata_put(opendata);
1407         return ret;
1408 }
1409
1410 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1411 {
1412         struct nfs4_exception exception = { };
1413         struct nfs_server *server = NFS_SERVER(state->inode);
1414         int err;
1415         do {
1416                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1417                 switch (err) {
1418                         case 0:
1419                         case -ENOENT:
1420                         case -ESTALE:
1421                                 goto out;
1422                         case -NFS4ERR_BADSESSION:
1423                         case -NFS4ERR_BADSLOT:
1424                         case -NFS4ERR_BAD_HIGH_SLOT:
1425                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1426                         case -NFS4ERR_DEADSESSION:
1427                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1428                                 goto out;
1429                         case -NFS4ERR_STALE_CLIENTID:
1430                         case -NFS4ERR_STALE_STATEID:
1431                         case -NFS4ERR_EXPIRED:
1432                                 /* Don't recall a delegation if it was lost */
1433                                 nfs4_schedule_lease_recovery(server->nfs_client);
1434                                 goto out;
1435                         case -ERESTARTSYS:
1436                                 /*
1437                                  * The show must go on: exit, but mark the
1438                                  * stateid as needing recovery.
1439                                  */
1440                         case -NFS4ERR_DELEG_REVOKED:
1441                         case -NFS4ERR_ADMIN_REVOKED:
1442                         case -NFS4ERR_BAD_STATEID:
1443                                 nfs_inode_find_state_and_recover(state->inode,
1444                                                 stateid);
1445                                 nfs4_schedule_stateid_recovery(server, state);
1446                         case -EKEYEXPIRED:
1447                                 /*
1448                                  * User RPCSEC_GSS context has expired.
1449                                  * We cannot recover this stateid now, so
1450                                  * skip it and allow recovery thread to
1451                                  * proceed.
1452                                  */
1453                         case -ENOMEM:
1454                                 err = 0;
1455                                 goto out;
1456                 }
1457                 err = nfs4_handle_exception(server, err, &exception);
1458         } while (exception.retry);
1459 out:
1460         return err;
1461 }
1462
1463 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1464 {
1465         struct nfs4_opendata *data = calldata;
1466
1467         data->rpc_status = task->tk_status;
1468         if (data->rpc_status == 0) {
1469                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1470                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1471                 renew_lease(data->o_res.server, data->timestamp);
1472                 data->rpc_done = 1;
1473         }
1474 }
1475
1476 static void nfs4_open_confirm_release(void *calldata)
1477 {
1478         struct nfs4_opendata *data = calldata;
1479         struct nfs4_state *state = NULL;
1480
1481         /* If this request hasn't been cancelled, do nothing */
1482         if (data->cancelled == 0)
1483                 goto out_free;
1484         /* In case of error, no cleanup! */
1485         if (!data->rpc_done)
1486                 goto out_free;
1487         state = nfs4_opendata_to_nfs4_state(data);
1488         if (!IS_ERR(state))
1489                 nfs4_close_state(state, data->o_arg.fmode);
1490 out_free:
1491         nfs4_opendata_put(data);
1492 }
1493
1494 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1495         .rpc_call_done = nfs4_open_confirm_done,
1496         .rpc_release = nfs4_open_confirm_release,
1497 };
1498
1499 /*
1500  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1501  */
1502 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1503 {
1504         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1505         struct rpc_task *task;
1506         struct  rpc_message msg = {
1507                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1508                 .rpc_argp = &data->c_arg,
1509                 .rpc_resp = &data->c_res,
1510                 .rpc_cred = data->owner->so_cred,
1511         };
1512         struct rpc_task_setup task_setup_data = {
1513                 .rpc_client = server->client,
1514                 .rpc_message = &msg,
1515                 .callback_ops = &nfs4_open_confirm_ops,
1516                 .callback_data = data,
1517                 .workqueue = nfsiod_workqueue,
1518                 .flags = RPC_TASK_ASYNC,
1519         };
1520         int status;
1521
1522         kref_get(&data->kref);
1523         data->rpc_done = 0;
1524         data->rpc_status = 0;
1525         data->timestamp = jiffies;
1526         task = rpc_run_task(&task_setup_data);
1527         if (IS_ERR(task))
1528                 return PTR_ERR(task);
1529         status = nfs4_wait_for_completion_rpc_task(task);
1530         if (status != 0) {
1531                 data->cancelled = 1;
1532                 smp_wmb();
1533         } else
1534                 status = data->rpc_status;
1535         rpc_put_task(task);
1536         return status;
1537 }
1538
1539 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1540 {
1541         struct nfs4_opendata *data = calldata;
1542         struct nfs4_state_owner *sp = data->owner;
1543
1544         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1545                 return;
1546         /*
1547          * Check if we still need to send an OPEN call, or if we can use
1548          * a delegation instead.
1549          */
1550         if (data->state != NULL) {
1551                 struct nfs_delegation *delegation;
1552
1553                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1554                         goto out_no_action;
1555                 rcu_read_lock();
1556                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1557                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1558                     can_open_delegated(delegation, data->o_arg.fmode))
1559                         goto unlock_no_action;
1560                 rcu_read_unlock();
1561         }
1562         /* Update client id. */
1563         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1564         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1565                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1566                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1567                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1568         }
1569         data->timestamp = jiffies;
1570         if (nfs4_setup_sequence(data->o_arg.server,
1571                                 &data->o_arg.seq_args,
1572                                 &data->o_res.seq_res,
1573                                 task) != 0)
1574                 nfs_release_seqid(data->o_arg.seqid);
1575         else
1576                 rpc_call_start(task);
1577         return;
1578 unlock_no_action:
1579         rcu_read_unlock();
1580 out_no_action:
1581         task->tk_action = NULL;
1582
1583 }
1584
1585 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1586 {
1587         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1588         nfs4_open_prepare(task, calldata);
1589 }
1590
1591 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1592 {
1593         struct nfs4_opendata *data = calldata;
1594
1595         data->rpc_status = task->tk_status;
1596
1597         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1598                 return;
1599
1600         if (task->tk_status == 0) {
1601                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1602                         switch (data->o_res.f_attr->mode & S_IFMT) {
1603                         case S_IFREG:
1604                                 break;
1605                         case S_IFLNK:
1606                                 data->rpc_status = -ELOOP;
1607                                 break;
1608                         case S_IFDIR:
1609                                 data->rpc_status = -EISDIR;
1610                                 break;
1611                         default:
1612                                 data->rpc_status = -ENOTDIR;
1613                         }
1614                 }
1615                 renew_lease(data->o_res.server, data->timestamp);
1616                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1617                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1618         }
1619         data->rpc_done = 1;
1620 }
1621
1622 static void nfs4_open_release(void *calldata)
1623 {
1624         struct nfs4_opendata *data = calldata;
1625         struct nfs4_state *state = NULL;
1626
1627         /* If this request hasn't been cancelled, do nothing */
1628         if (data->cancelled == 0)
1629                 goto out_free;
1630         /* In case of error, no cleanup! */
1631         if (data->rpc_status != 0 || !data->rpc_done)
1632                 goto out_free;
1633         /* In case we need an open_confirm, no cleanup! */
1634         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1635                 goto out_free;
1636         state = nfs4_opendata_to_nfs4_state(data);
1637         if (!IS_ERR(state))
1638                 nfs4_close_state(state, data->o_arg.fmode);
1639 out_free:
1640         nfs4_opendata_put(data);
1641 }
1642
1643 static const struct rpc_call_ops nfs4_open_ops = {
1644         .rpc_call_prepare = nfs4_open_prepare,
1645         .rpc_call_done = nfs4_open_done,
1646         .rpc_release = nfs4_open_release,
1647 };
1648
1649 static const struct rpc_call_ops nfs4_recover_open_ops = {
1650         .rpc_call_prepare = nfs4_recover_open_prepare,
1651         .rpc_call_done = nfs4_open_done,
1652         .rpc_release = nfs4_open_release,
1653 };
1654
1655 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1656 {
1657         struct inode *dir = data->dir->d_inode;
1658         struct nfs_server *server = NFS_SERVER(dir);
1659         struct nfs_openargs *o_arg = &data->o_arg;
1660         struct nfs_openres *o_res = &data->o_res;
1661         struct rpc_task *task;
1662         struct rpc_message msg = {
1663                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1664                 .rpc_argp = o_arg,
1665                 .rpc_resp = o_res,
1666                 .rpc_cred = data->owner->so_cred,
1667         };
1668         struct rpc_task_setup task_setup_data = {
1669                 .rpc_client = server->client,
1670                 .rpc_message = &msg,
1671                 .callback_ops = &nfs4_open_ops,
1672                 .callback_data = data,
1673                 .workqueue = nfsiod_workqueue,
1674                 .flags = RPC_TASK_ASYNC,
1675         };
1676         int status;
1677
1678         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1679         kref_get(&data->kref);
1680         data->rpc_done = 0;
1681         data->rpc_status = 0;
1682         data->cancelled = 0;
1683         if (isrecover)
1684                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1685         task = rpc_run_task(&task_setup_data);
1686         if (IS_ERR(task))
1687                 return PTR_ERR(task);
1688         status = nfs4_wait_for_completion_rpc_task(task);
1689         if (status != 0) {
1690                 data->cancelled = 1;
1691                 smp_wmb();
1692         } else
1693                 status = data->rpc_status;
1694         rpc_put_task(task);
1695
1696         return status;
1697 }
1698
1699 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1700 {
1701         struct inode *dir = data->dir->d_inode;
1702         struct nfs_openres *o_res = &data->o_res;
1703         int status;
1704
1705         status = nfs4_run_open_task(data, 1);
1706         if (status != 0 || !data->rpc_done)
1707                 return status;
1708
1709         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1710
1711         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1712                 status = _nfs4_proc_open_confirm(data);
1713                 if (status != 0)
1714                         return status;
1715         }
1716
1717         return status;
1718 }
1719
1720 static int nfs4_opendata_access(struct rpc_cred *cred,
1721                                 struct nfs4_opendata *opendata,
1722                                 struct nfs4_state *state, fmode_t fmode)
1723 {
1724         struct nfs_access_entry cache;
1725         u32 mask;
1726
1727         /* access call failed or for some reason the server doesn't
1728          * support any access modes -- defer access call until later */
1729         if (opendata->o_res.access_supported == 0)
1730                 return 0;
1731
1732         mask = 0;
1733         /* don't check MAY_WRITE - a newly created file may not have
1734          * write mode bits, but POSIX allows the creating process to write */
1735         if (fmode & FMODE_READ)
1736                 mask |= MAY_READ;
1737         if (fmode & FMODE_EXEC)
1738                 mask |= MAY_EXEC;
1739
1740         cache.cred = cred;
1741         cache.jiffies = jiffies;
1742         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1743         nfs_access_add_cache(state->inode, &cache);
1744
1745         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1746                 return 0;
1747
1748         /* even though OPEN succeeded, access is denied. Close the file */
1749         nfs4_close_state(state, fmode);
1750         return -EACCES;
1751 }
1752
1753 /*
1754  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1755  */
1756 static int _nfs4_proc_open(struct nfs4_opendata *data)
1757 {
1758         struct inode *dir = data->dir->d_inode;
1759         struct nfs_server *server = NFS_SERVER(dir);
1760         struct nfs_openargs *o_arg = &data->o_arg;
1761         struct nfs_openres *o_res = &data->o_res;
1762         int status;
1763
1764         status = nfs4_run_open_task(data, 0);
1765         if (!data->rpc_done)
1766                 return status;
1767         if (status != 0) {
1768                 if (status == -NFS4ERR_BADNAME &&
1769                                 !(o_arg->open_flags & O_CREAT))
1770                         return -ENOENT;
1771                 return status;
1772         }
1773
1774         nfs_fattr_map_and_free_names(server, &data->f_attr);
1775
1776         if (o_arg->open_flags & O_CREAT)
1777                 update_changeattr(dir, &o_res->cinfo);
1778         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1779                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1780         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1781                 status = _nfs4_proc_open_confirm(data);
1782                 if (status != 0)
1783                         return status;
1784         }
1785         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1786                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1787         return 0;
1788 }
1789
1790 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1791 {
1792         unsigned int loop;
1793         int ret;
1794
1795         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1796                 ret = nfs4_wait_clnt_recover(clp);
1797                 if (ret != 0)
1798                         break;
1799                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1800                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1801                         break;
1802                 nfs4_schedule_state_manager(clp);
1803                 ret = -EIO;
1804         }
1805         return ret;
1806 }
1807
1808 static int nfs4_recover_expired_lease(struct nfs_server *server)
1809 {
1810         return nfs4_client_recover_expired_lease(server->nfs_client);
1811 }
1812
1813 /*
1814  * OPEN_EXPIRED:
1815  *      reclaim state on the server after a network partition.
1816  *      Assumes caller holds the appropriate lock
1817  */
1818 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1819 {
1820         struct nfs4_opendata *opendata;
1821         int ret;
1822
1823         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1824         if (IS_ERR(opendata))
1825                 return PTR_ERR(opendata);
1826         ret = nfs4_open_recover(opendata, state);
1827         if (ret == -ESTALE)
1828                 d_drop(ctx->dentry);
1829         nfs4_opendata_put(opendata);
1830         return ret;
1831 }
1832
1833 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1834 {
1835         struct nfs_server *server = NFS_SERVER(state->inode);
1836         struct nfs4_exception exception = { };
1837         int err;
1838
1839         do {
1840                 err = _nfs4_open_expired(ctx, state);
1841                 switch (err) {
1842                 default:
1843                         goto out;
1844                 case -NFS4ERR_GRACE:
1845                 case -NFS4ERR_DELAY:
1846                         nfs4_handle_exception(server, err, &exception);
1847                         err = 0;
1848                 }
1849         } while (exception.retry);
1850 out:
1851         return err;
1852 }
1853
1854 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1855 {
1856         struct nfs_open_context *ctx;
1857         int ret;
1858
1859         ctx = nfs4_state_find_open_context(state);
1860         if (IS_ERR(ctx))
1861                 return PTR_ERR(ctx);
1862         ret = nfs4_do_open_expired(ctx, state);
1863         put_nfs_open_context(ctx);
1864         return ret;
1865 }
1866
1867 #if defined(CONFIG_NFS_V4_1)
1868 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1869 {
1870         struct nfs_server *server = NFS_SERVER(state->inode);
1871         nfs4_stateid *stateid = &state->stateid;
1872         int status;
1873
1874         /* If a state reset has been done, test_stateid is unneeded */
1875         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1876                 return;
1877
1878         status = nfs41_test_stateid(server, stateid);
1879         if (status != NFS_OK) {
1880                 /* Free the stateid unless the server explicitly
1881                  * informs us the stateid is unrecognized. */
1882                 if (status != -NFS4ERR_BAD_STATEID)
1883                         nfs41_free_stateid(server, stateid);
1884                 nfs_remove_bad_delegation(state->inode);
1885
1886                 write_seqlock(&state->seqlock);
1887                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1888                 write_sequnlock(&state->seqlock);
1889                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1890         }
1891 }
1892
1893 /**
1894  * nfs41_check_open_stateid - possibly free an open stateid
1895  *
1896  * @state: NFSv4 state for an inode
1897  *
1898  * Returns NFS_OK if recovery for this stateid is now finished.
1899  * Otherwise a negative NFS4ERR value is returned.
1900  */
1901 static int nfs41_check_open_stateid(struct nfs4_state *state)
1902 {
1903         struct nfs_server *server = NFS_SERVER(state->inode);
1904         nfs4_stateid *stateid = &state->open_stateid;
1905         int status;
1906
1907         /* If a state reset has been done, test_stateid is unneeded */
1908         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1909             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1910             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1911                 return -NFS4ERR_BAD_STATEID;
1912
1913         status = nfs41_test_stateid(server, stateid);
1914         if (status != NFS_OK) {
1915                 /* Free the stateid unless the server explicitly
1916                  * informs us the stateid is unrecognized. */
1917                 if (status != -NFS4ERR_BAD_STATEID)
1918                         nfs41_free_stateid(server, stateid);
1919
1920                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1921                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1922                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1923         }
1924         return status;
1925 }
1926
1927 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1928 {
1929         int status;
1930
1931         nfs41_clear_delegation_stateid(state);
1932         status = nfs41_check_open_stateid(state);
1933         if (status != NFS_OK)
1934                 status = nfs4_open_expired(sp, state);
1935         return status;
1936 }
1937 #endif
1938
1939 /*
1940  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1941  * fields corresponding to attributes that were used to store the verifier.
1942  * Make sure we clobber those fields in the later setattr call
1943  */
1944 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1945 {
1946         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1947             !(sattr->ia_valid & ATTR_ATIME_SET))
1948                 sattr->ia_valid |= ATTR_ATIME;
1949
1950         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1951             !(sattr->ia_valid & ATTR_MTIME_SET))
1952                 sattr->ia_valid |= ATTR_MTIME;
1953 }
1954
1955 /*
1956  * Returns a referenced nfs4_state
1957  */
1958 static int _nfs4_do_open(struct inode *dir,
1959                         struct dentry *dentry,
1960                         fmode_t fmode,
1961                         int flags,
1962                         struct iattr *sattr,
1963                         struct rpc_cred *cred,
1964                         struct nfs4_state **res,
1965                         struct nfs4_threshold **ctx_th)
1966 {
1967         struct nfs4_state_owner  *sp;
1968         struct nfs4_state     *state = NULL;
1969         struct nfs_server       *server = NFS_SERVER(dir);
1970         struct nfs4_opendata *opendata;
1971         int status;
1972
1973         /* Protect against reboot recovery conflicts */
1974         status = -ENOMEM;
1975         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1976         if (sp == NULL) {
1977                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1978                 goto out_err;
1979         }
1980         status = nfs4_recover_expired_lease(server);
1981         if (status != 0)
1982                 goto err_put_state_owner;
1983         if (dentry->d_inode != NULL)
1984                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1985         status = -ENOMEM;
1986         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1987         if (opendata == NULL)
1988                 goto err_put_state_owner;
1989
1990         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1991                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1992                 if (!opendata->f_attr.mdsthreshold)
1993                         goto err_opendata_put;
1994                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
1995         }
1996         if (dentry->d_inode != NULL)
1997                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1998
1999         status = _nfs4_proc_open(opendata);
2000         if (status != 0)
2001                 goto err_opendata_put;
2002
2003         state = nfs4_opendata_to_nfs4_state(opendata);
2004         status = PTR_ERR(state);
2005         if (IS_ERR(state))
2006                 goto err_opendata_put;
2007         if (server->caps & NFS_CAP_POSIX_LOCK)
2008                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2009
2010         status = nfs4_opendata_access(cred, opendata, state, fmode);
2011         if (status != 0)
2012                 goto err_opendata_put;
2013
2014         if (opendata->o_arg.open_flags & O_EXCL) {
2015                 nfs4_exclusive_attrset(opendata, sattr);
2016
2017                 nfs_fattr_init(opendata->o_res.f_attr);
2018                 status = nfs4_do_setattr(state->inode, cred,
2019                                 opendata->o_res.f_attr, sattr,
2020                                 state);
2021                 if (status == 0)
2022                         nfs_setattr_update_inode(state->inode, sattr);
2023                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2024         }
2025
2026         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2027                 *ctx_th = opendata->f_attr.mdsthreshold;
2028         else
2029                 kfree(opendata->f_attr.mdsthreshold);
2030         opendata->f_attr.mdsthreshold = NULL;
2031
2032         nfs4_opendata_put(opendata);
2033         nfs4_put_state_owner(sp);
2034         *res = state;
2035         return 0;
2036 err_opendata_put:
2037         kfree(opendata->f_attr.mdsthreshold);
2038         nfs4_opendata_put(opendata);
2039 err_put_state_owner:
2040         nfs4_put_state_owner(sp);
2041 out_err:
2042         *res = NULL;
2043         return status;
2044 }
2045
2046
2047 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2048                                         struct dentry *dentry,
2049                                         fmode_t fmode,
2050                                         int flags,
2051                                         struct iattr *sattr,
2052                                         struct rpc_cred *cred,
2053                                         struct nfs4_threshold **ctx_th)
2054 {
2055         struct nfs4_exception exception = { };
2056         struct nfs4_state *res;
2057         int status;
2058
2059         fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
2060         do {
2061                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
2062                                        &res, ctx_th);
2063                 if (status == 0)
2064                         break;
2065                 /* NOTE: BAD_SEQID means the server and client disagree about the
2066                  * book-keeping w.r.t. state-changing operations
2067                  * (OPEN/CLOSE/LOCK/LOCKU...)
2068                  * It is actually a sign of a bug on the client or on the server.
2069                  *
2070                  * If we receive a BAD_SEQID error in the particular case of
2071                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2072                  * have unhashed the old state_owner for us, and that we can
2073                  * therefore safely retry using a new one. We should still warn
2074                  * the user though...
2075                  */
2076                 if (status == -NFS4ERR_BAD_SEQID) {
2077                         pr_warn_ratelimited("NFS: v4 server %s "
2078                                         " returned a bad sequence-id error!\n",
2079                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2080                         exception.retry = 1;
2081                         continue;
2082                 }
2083                 /*
2084                  * BAD_STATEID on OPEN means that the server cancelled our
2085                  * state before it received the OPEN_CONFIRM.
2086                  * Recover by retrying the request as per the discussion
2087                  * on Page 181 of RFC3530.
2088                  */
2089                 if (status == -NFS4ERR_BAD_STATEID) {
2090                         exception.retry = 1;
2091                         continue;
2092                 }
2093                 if (status == -EAGAIN) {
2094                         /* We must have found a delegation */
2095                         exception.retry = 1;
2096                         continue;
2097                 }
2098                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
2099                                         status, &exception));
2100         } while (exception.retry);
2101         return res;
2102 }
2103
2104 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2105                             struct nfs_fattr *fattr, struct iattr *sattr,
2106                             struct nfs4_state *state)
2107 {
2108         struct nfs_server *server = NFS_SERVER(inode);
2109         struct nfs_setattrargs  arg = {
2110                 .fh             = NFS_FH(inode),
2111                 .iap            = sattr,
2112                 .server         = server,
2113                 .bitmask = server->attr_bitmask,
2114         };
2115         struct nfs_setattrres  res = {
2116                 .fattr          = fattr,
2117                 .server         = server,
2118         };
2119         struct rpc_message msg = {
2120                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2121                 .rpc_argp       = &arg,
2122                 .rpc_resp       = &res,
2123                 .rpc_cred       = cred,
2124         };
2125         unsigned long timestamp = jiffies;
2126         int status;
2127
2128         nfs_fattr_init(fattr);
2129
2130         if (state != NULL) {
2131                 struct nfs_lockowner lockowner = {
2132                         .l_owner = current->files,
2133                         .l_pid = current->tgid,
2134                 };
2135                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2136                                 &lockowner);
2137         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
2138                                 FMODE_WRITE)) {
2139                 /* Use that stateid */
2140         } else
2141                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2142
2143         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2144         if (status == 0 && state != NULL)
2145                 renew_lease(server, timestamp);
2146         return status;
2147 }
2148
2149 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2150                            struct nfs_fattr *fattr, struct iattr *sattr,
2151                            struct nfs4_state *state)
2152 {
2153         struct nfs_server *server = NFS_SERVER(inode);
2154         struct nfs4_exception exception = {
2155                 .state = state,
2156                 .inode = inode,
2157         };
2158         int err;
2159         do {
2160                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2161                 switch (err) {
2162                 case -NFS4ERR_OPENMODE:
2163                         if (state && !(state->state & FMODE_WRITE)) {
2164                                 err = -EBADF;
2165                                 if (sattr->ia_valid & ATTR_OPEN)
2166                                         err = -EACCES;
2167                                 goto out;
2168                         }
2169                 }
2170                 err = nfs4_handle_exception(server, err, &exception);
2171         } while (exception.retry);
2172 out:
2173         return err;
2174 }
2175
2176 struct nfs4_closedata {
2177         struct inode *inode;
2178         struct nfs4_state *state;
2179         struct nfs_closeargs arg;
2180         struct nfs_closeres res;
2181         struct nfs_fattr fattr;
2182         unsigned long timestamp;
2183         bool roc;
2184         u32 roc_barrier;
2185 };
2186
2187 static void nfs4_free_closedata(void *data)
2188 {
2189         struct nfs4_closedata *calldata = data;
2190         struct nfs4_state_owner *sp = calldata->state->owner;
2191         struct super_block *sb = calldata->state->inode->i_sb;
2192
2193         if (calldata->roc)
2194                 pnfs_roc_release(calldata->state->inode);
2195         nfs4_put_open_state(calldata->state);
2196         nfs_free_seqid(calldata->arg.seqid);
2197         nfs4_put_state_owner(sp);
2198         nfs_sb_deactive_async(sb);
2199         kfree(calldata);
2200 }
2201
2202 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2203                 fmode_t fmode)
2204 {
2205         spin_lock(&state->owner->so_lock);
2206         if (!(fmode & FMODE_READ))
2207                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2208         if (!(fmode & FMODE_WRITE))
2209                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2210         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2211         spin_unlock(&state->owner->so_lock);
2212 }
2213
2214 static void nfs4_close_done(struct rpc_task *task, void *data)
2215 {
2216         struct nfs4_closedata *calldata = data;
2217         struct nfs4_state *state = calldata->state;
2218         struct nfs_server *server = NFS_SERVER(calldata->inode);
2219
2220         dprintk("%s: begin!\n", __func__);
2221         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2222                 return;
2223         /* hmm. we are done with the inode, and in the process of freeing
2224          * the state_owner. we keep this around to process errors
2225          */
2226         switch (task->tk_status) {
2227                 case 0:
2228                         if (calldata->roc)
2229                                 pnfs_roc_set_barrier(state->inode,
2230                                                      calldata->roc_barrier);
2231                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2232                         renew_lease(server, calldata->timestamp);
2233                         nfs4_close_clear_stateid_flags(state,
2234                                         calldata->arg.fmode);
2235                         break;
2236                 case -NFS4ERR_STALE_STATEID:
2237                 case -NFS4ERR_OLD_STATEID:
2238                 case -NFS4ERR_BAD_STATEID:
2239                 case -NFS4ERR_EXPIRED:
2240                         if (calldata->arg.fmode == 0)
2241                                 break;
2242                 default:
2243                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2244                                 rpc_restart_call_prepare(task);
2245         }
2246         nfs_release_seqid(calldata->arg.seqid);
2247         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2248         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2249 }
2250
2251 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2252 {
2253         struct nfs4_closedata *calldata = data;
2254         struct nfs4_state *state = calldata->state;
2255         struct inode *inode = calldata->inode;
2256         int call_close = 0;
2257
2258         dprintk("%s: begin!\n", __func__);
2259         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2260                 return;
2261
2262         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2263         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2264         spin_lock(&state->owner->so_lock);
2265         /* Calculate the change in open mode */
2266         if (state->n_rdwr == 0) {
2267                 if (state->n_rdonly == 0) {
2268                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2269                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2270                         calldata->arg.fmode &= ~FMODE_READ;
2271                 }
2272                 if (state->n_wronly == 0) {
2273                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2274                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2275                         calldata->arg.fmode &= ~FMODE_WRITE;
2276                 }
2277         }
2278         spin_unlock(&state->owner->so_lock);
2279
2280         if (!call_close) {
2281                 /* Note: exit _without_ calling nfs4_close_done */
2282                 task->tk_action = NULL;
2283                 goto out;
2284         }
2285
2286         if (calldata->arg.fmode == 0) {
2287                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2288                 if (calldata->roc &&
2289                     pnfs_roc_drain(inode, &calldata->roc_barrier, task))
2290                         goto out;
2291         }
2292
2293         nfs_fattr_init(calldata->res.fattr);
2294         calldata->timestamp = jiffies;
2295         if (nfs4_setup_sequence(NFS_SERVER(inode),
2296                                 &calldata->arg.seq_args,
2297                                 &calldata->res.seq_res,
2298                                 task) != 0)
2299                 nfs_release_seqid(calldata->arg.seqid);
2300         else
2301                 rpc_call_start(task);
2302 out:
2303         dprintk("%s: done!\n", __func__);
2304 }
2305
2306 static const struct rpc_call_ops nfs4_close_ops = {
2307         .rpc_call_prepare = nfs4_close_prepare,
2308         .rpc_call_done = nfs4_close_done,
2309         .rpc_release = nfs4_free_closedata,
2310 };
2311
2312 /*
2313  * It is possible for data to be read/written from a mem-mapped file
2314  * after the sys_close call (which hits the vfs layer as a flush).
2315  * This means that we can't safely call nfsv4 close on a file until
2316  * the inode is cleared. This in turn means that we are not good
2317  * NFSv4 citizens - we do not indicate to the server to update the file's
2318  * share state even when we are done with one of the three share
2319  * stateid's in the inode.
2320  *
2321  * NOTE: Caller must be holding the sp->so_owner semaphore!
2322  */
2323 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2324 {
2325         struct nfs_server *server = NFS_SERVER(state->inode);
2326         struct nfs4_closedata *calldata;
2327         struct nfs4_state_owner *sp = state->owner;
2328         struct rpc_task *task;
2329         struct rpc_message msg = {
2330                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2331                 .rpc_cred = state->owner->so_cred,
2332         };
2333         struct rpc_task_setup task_setup_data = {
2334                 .rpc_client = server->client,
2335                 .rpc_message = &msg,
2336                 .callback_ops = &nfs4_close_ops,
2337                 .workqueue = nfsiod_workqueue,
2338                 .flags = RPC_TASK_ASYNC,
2339         };
2340         int status = -ENOMEM;
2341
2342         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2343         if (calldata == NULL)
2344                 goto out;
2345         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2346         calldata->inode = state->inode;
2347         calldata->state = state;
2348         calldata->arg.fh = NFS_FH(state->inode);
2349         calldata->arg.stateid = &state->open_stateid;
2350         /* Serialization for the sequence id */
2351         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2352         if (calldata->arg.seqid == NULL)
2353                 goto out_free_calldata;
2354         calldata->arg.fmode = 0;
2355         calldata->arg.bitmask = server->cache_consistency_bitmask;
2356         calldata->res.fattr = &calldata->fattr;
2357         calldata->res.seqid = calldata->arg.seqid;
2358         calldata->res.server = server;
2359         calldata->roc = pnfs_roc(state->inode);
2360         nfs_sb_active(calldata->inode->i_sb);
2361
2362         msg.rpc_argp = &calldata->arg;
2363         msg.rpc_resp = &calldata->res;
2364         task_setup_data.callback_data = calldata;
2365         task = rpc_run_task(&task_setup_data);
2366         if (IS_ERR(task))
2367                 return PTR_ERR(task);
2368         status = 0;
2369         if (wait)
2370                 status = rpc_wait_for_completion_task(task);
2371         rpc_put_task(task);
2372         return status;
2373 out_free_calldata:
2374         kfree(calldata);
2375 out:
2376         nfs4_put_open_state(state);
2377         nfs4_put_state_owner(sp);
2378         return status;
2379 }
2380
2381 static struct inode *
2382 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2383 {
2384         struct nfs4_state *state;
2385
2386         /* Protect against concurrent sillydeletes */
2387         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2388                              ctx->cred, &ctx->mdsthreshold);
2389         if (IS_ERR(state))
2390                 return ERR_CAST(state);
2391         ctx->state = state;
2392         return igrab(state->inode);
2393 }
2394
2395 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2396 {
2397         if (ctx->state == NULL)
2398                 return;
2399         if (is_sync)
2400                 nfs4_close_sync(ctx->state, ctx->mode);
2401         else
2402                 nfs4_close_state(ctx->state, ctx->mode);
2403 }
2404
2405 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2406 {
2407         struct nfs4_server_caps_arg args = {
2408                 .fhandle = fhandle,
2409         };
2410         struct nfs4_server_caps_res res = {};
2411         struct rpc_message msg = {
2412                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2413                 .rpc_argp = &args,
2414                 .rpc_resp = &res,
2415         };
2416         int status;
2417
2418         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2419         if (status == 0) {
2420                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2421                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2422                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2423                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2424                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2425                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2426                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2427                         server->caps |= NFS_CAP_ACLS;
2428                 if (res.has_links != 0)
2429                         server->caps |= NFS_CAP_HARDLINKS;
2430                 if (res.has_symlinks != 0)
2431                         server->caps |= NFS_CAP_SYMLINKS;
2432                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2433                         server->caps |= NFS_CAP_FILEID;
2434                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2435                         server->caps |= NFS_CAP_MODE;
2436                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2437                         server->caps |= NFS_CAP_NLINK;
2438                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2439                         server->caps |= NFS_CAP_OWNER;
2440                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2441                         server->caps |= NFS_CAP_OWNER_GROUP;
2442                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2443                         server->caps |= NFS_CAP_ATIME;
2444                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2445                         server->caps |= NFS_CAP_CTIME;
2446                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2447                         server->caps |= NFS_CAP_MTIME;
2448
2449                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2450                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2451                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2452                 server->acl_bitmask = res.acl_bitmask;
2453                 server->fh_expire_type = res.fh_expire_type;
2454         }
2455
2456         return status;
2457 }
2458
2459 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2460 {
2461         struct nfs4_exception exception = { };
2462         int err;
2463         do {
2464                 err = nfs4_handle_exception(server,
2465                                 _nfs4_server_capabilities(server, fhandle),
2466                                 &exception);
2467         } while (exception.retry);
2468         return err;
2469 }
2470
2471 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2472                 struct nfs_fsinfo *info)
2473 {
2474         struct nfs4_lookup_root_arg args = {
2475                 .bitmask = nfs4_fattr_bitmap,
2476         };
2477         struct nfs4_lookup_res res = {
2478                 .server = server,
2479                 .fattr = info->fattr,
2480                 .fh = fhandle,
2481         };
2482         struct rpc_message msg = {
2483                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2484                 .rpc_argp = &args,
2485                 .rpc_resp = &res,
2486         };
2487
2488         nfs_fattr_init(info->fattr);
2489         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2490 }
2491
2492 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2493                 struct nfs_fsinfo *info)
2494 {
2495         struct nfs4_exception exception = { };
2496         int err;
2497         do {
2498                 err = _nfs4_lookup_root(server, fhandle, info);
2499                 switch (err) {
2500                 case 0:
2501                 case -NFS4ERR_WRONGSEC:
2502                         goto out;
2503                 default:
2504                         err = nfs4_handle_exception(server, err, &exception);
2505                 }
2506         } while (exception.retry);
2507 out:
2508         return err;
2509 }
2510
2511 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2512                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2513 {
2514         struct rpc_auth *auth;
2515         int ret;
2516
2517         auth = rpcauth_create(flavor, server->client);
2518         if (IS_ERR(auth)) {
2519                 ret = -EIO;
2520                 goto out;
2521         }
2522         ret = nfs4_lookup_root(server, fhandle, info);
2523 out:
2524         return ret;
2525 }
2526
2527 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2528                               struct nfs_fsinfo *info)
2529 {
2530         int i, len, status = 0;
2531         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2532
2533         len = rpcauth_list_flavors(flav_array, ARRAY_SIZE(flav_array));
2534         if (len < 0)
2535                 return len;
2536
2537         for (i = 0; i < len; i++) {
2538                 /* AUTH_UNIX is the default flavor if none was specified,
2539                  * thus has already been tried. */
2540                 if (flav_array[i] == RPC_AUTH_UNIX)
2541                         continue;
2542
2543                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2544                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2545                         continue;
2546                 break;
2547         }
2548         /*
2549          * -EACCESS could mean that the user doesn't have correct permissions
2550          * to access the mount.  It could also mean that we tried to mount
2551          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2552          * existing mount programs don't handle -EACCES very well so it should
2553          * be mapped to -EPERM instead.
2554          */
2555         if (status == -EACCES)
2556                 status = -EPERM;
2557         return status;
2558 }
2559
2560 /*
2561  * get the file handle for the "/" directory on the server
2562  */
2563 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2564                          struct nfs_fsinfo *info)
2565 {
2566         int minor_version = server->nfs_client->cl_minorversion;
2567         int status = nfs4_lookup_root(server, fhandle, info);
2568         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2569                 /*
2570                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2571                  * by nfs4_map_errors() as this function exits.
2572                  */
2573                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2574         if (status == 0)
2575                 status = nfs4_server_capabilities(server, fhandle);
2576         if (status == 0)
2577                 status = nfs4_do_fsinfo(server, fhandle, info);
2578         return nfs4_map_errors(status);
2579 }
2580
2581 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2582                               struct nfs_fsinfo *info)
2583 {
2584         int error;
2585         struct nfs_fattr *fattr = info->fattr;
2586
2587         error = nfs4_server_capabilities(server, mntfh);
2588         if (error < 0) {
2589                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2590                 return error;
2591         }
2592
2593         error = nfs4_proc_getattr(server, mntfh, fattr);
2594         if (error < 0) {
2595                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2596                 return error;
2597         }
2598
2599         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2600             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2601                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2602
2603         return error;
2604 }
2605
2606 /*
2607  * Get locations and (maybe) other attributes of a referral.
2608  * Note that we'll actually follow the referral later when
2609  * we detect fsid mismatch in inode revalidation
2610  */
2611 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2612                              const struct qstr *name, struct nfs_fattr *fattr,
2613                              struct nfs_fh *fhandle)
2614 {
2615         int status = -ENOMEM;
2616         struct page *page = NULL;
2617         struct nfs4_fs_locations *locations = NULL;
2618
2619         page = alloc_page(GFP_KERNEL);
2620         if (page == NULL)
2621                 goto out;
2622         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2623         if (locations == NULL)
2624                 goto out;
2625
2626         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2627         if (status != 0)
2628                 goto out;
2629         /* Make sure server returned a different fsid for the referral */
2630         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2631                 dprintk("%s: server did not return a different fsid for"
2632                         " a referral at %s\n", __func__, name->name);
2633                 status = -EIO;
2634                 goto out;
2635         }
2636         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2637         nfs_fixup_referral_attributes(&locations->fattr);
2638
2639         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2640         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2641         memset(fhandle, 0, sizeof(struct nfs_fh));
2642 out:
2643         if (page)
2644                 __free_page(page);
2645         kfree(locations);
2646         return status;
2647 }
2648
2649 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2650 {
2651         struct nfs4_getattr_arg args = {
2652                 .fh = fhandle,
2653                 .bitmask = server->attr_bitmask,
2654         };
2655         struct nfs4_getattr_res res = {
2656                 .fattr = fattr,
2657                 .server = server,
2658         };
2659         struct rpc_message msg = {
2660                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2661                 .rpc_argp = &args,
2662                 .rpc_resp = &res,
2663         };
2664
2665         nfs_fattr_init(fattr);
2666         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2667 }
2668
2669 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2670 {
2671         struct nfs4_exception exception = { };
2672         int err;
2673         do {
2674                 err = nfs4_handle_exception(server,
2675                                 _nfs4_proc_getattr(server, fhandle, fattr),
2676                                 &exception);
2677         } while (exception.retry);
2678         return err;
2679 }
2680
2681 /*
2682  * The file is not closed if it is opened due to the a request to change
2683  * the size of the file. The open call will not be needed once the
2684  * VFS layer lookup-intents are implemented.
2685  *
2686  * Close is called when the inode is destroyed.
2687  * If we haven't opened the file for O_WRONLY, we
2688  * need to in the size_change case to obtain a stateid.
2689  *
2690  * Got race?
2691  * Because OPEN is always done by name in nfsv4, it is
2692  * possible that we opened a different file by the same
2693  * name.  We can recognize this race condition, but we
2694  * can't do anything about it besides returning an error.
2695  *
2696  * This will be fixed with VFS changes (lookup-intent).
2697  */
2698 static int
2699 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2700                   struct iattr *sattr)
2701 {
2702         struct inode *inode = dentry->d_inode;
2703         struct rpc_cred *cred = NULL;
2704         struct nfs4_state *state = NULL;
2705         int status;
2706
2707         if (pnfs_ld_layoutret_on_setattr(inode))
2708                 pnfs_return_layout(inode);
2709
2710         nfs_fattr_init(fattr);
2711
2712         /* Deal with open(O_TRUNC) */
2713         if (sattr->ia_valid & ATTR_OPEN)
2714                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2715
2716         /* Optimization: if the end result is no change, don't RPC */
2717         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2718                 return 0;
2719
2720         /* Search for an existing open(O_WRITE) file */
2721         if (sattr->ia_valid & ATTR_FILE) {
2722                 struct nfs_open_context *ctx;
2723
2724                 ctx = nfs_file_open_context(sattr->ia_file);
2725                 if (ctx) {
2726                         cred = ctx->cred;
2727                         state = ctx->state;
2728                 }
2729         }
2730
2731         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2732         if (status == 0)
2733                 nfs_setattr_update_inode(inode, sattr);
2734         return status;
2735 }
2736
2737 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2738                 const struct qstr *name, struct nfs_fh *fhandle,
2739                 struct nfs_fattr *fattr)
2740 {
2741         struct nfs_server *server = NFS_SERVER(dir);
2742         int                    status;
2743         struct nfs4_lookup_arg args = {
2744                 .bitmask = server->attr_bitmask,
2745                 .dir_fh = NFS_FH(dir),
2746                 .name = name,
2747         };
2748         struct nfs4_lookup_res res = {
2749                 .server = server,
2750                 .fattr = fattr,
2751                 .fh = fhandle,
2752         };
2753         struct rpc_message msg = {
2754                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2755                 .rpc_argp = &args,
2756                 .rpc_resp = &res,
2757         };
2758
2759         nfs_fattr_init(fattr);
2760
2761         dprintk("NFS call  lookup %s\n", name->name);
2762         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2763         dprintk("NFS reply lookup: %d\n", status);
2764         return status;
2765 }
2766
2767 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2768 {
2769         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2770                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2771         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2772         fattr->nlink = 2;
2773 }
2774
2775 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2776                                    struct qstr *name, struct nfs_fh *fhandle,
2777                                    struct nfs_fattr *fattr)
2778 {
2779         struct nfs4_exception exception = { };
2780         struct rpc_clnt *client = *clnt;
2781         int err;
2782         do {
2783                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2784                 switch (err) {
2785                 case -NFS4ERR_BADNAME:
2786                         err = -ENOENT;
2787                         goto out;
2788                 case -NFS4ERR_MOVED:
2789                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2790                         goto out;
2791                 case -NFS4ERR_WRONGSEC:
2792                         err = -EPERM;
2793                         if (client != *clnt)
2794                                 goto out;
2795
2796                         client = nfs4_create_sec_client(client, dir, name);
2797                         if (IS_ERR(client))
2798                                 return PTR_ERR(client);
2799
2800                         exception.retry = 1;
2801                         break;
2802                 default:
2803                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2804                 }
2805         } while (exception.retry);
2806
2807 out:
2808         if (err == 0)
2809                 *clnt = client;
2810         else if (client != *clnt)
2811                 rpc_shutdown_client(client);
2812
2813         return err;
2814 }
2815
2816 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2817                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2818 {
2819         int status;
2820         struct rpc_clnt *client = NFS_CLIENT(dir);
2821
2822         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2823         if (client != NFS_CLIENT(dir)) {
2824                 rpc_shutdown_client(client);
2825                 nfs_fixup_secinfo_attributes(fattr);
2826         }
2827         return status;
2828 }
2829
2830 struct rpc_clnt *
2831 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2832                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2833 {
2834         int status;
2835         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2836
2837         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2838         if (status < 0) {
2839                 rpc_shutdown_client(client);
2840                 return ERR_PTR(status);
2841         }
2842         return client;
2843 }
2844
2845 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2846 {
2847         struct nfs_server *server = NFS_SERVER(inode);
2848         struct nfs4_accessargs args = {
2849                 .fh = NFS_FH(inode),
2850                 .bitmask = server->cache_consistency_bitmask,
2851         };
2852         struct nfs4_accessres res = {
2853                 .server = server,
2854         };
2855         struct rpc_message msg = {
2856                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2857                 .rpc_argp = &args,
2858                 .rpc_resp = &res,
2859                 .rpc_cred = entry->cred,
2860         };
2861         int mode = entry->mask;
2862         int status;
2863
2864         /*
2865          * Determine which access bits we want to ask for...
2866          */
2867         if (mode & MAY_READ)
2868                 args.access |= NFS4_ACCESS_READ;
2869         if (S_ISDIR(inode->i_mode)) {
2870                 if (mode & MAY_WRITE)
2871                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2872                 if (mode & MAY_EXEC)
2873                         args.access |= NFS4_ACCESS_LOOKUP;
2874         } else {
2875                 if (mode & MAY_WRITE)
2876                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2877                 if (mode & MAY_EXEC)
2878                         args.access |= NFS4_ACCESS_EXECUTE;
2879         }
2880
2881         res.fattr = nfs_alloc_fattr();
2882         if (res.fattr == NULL)
2883                 return -ENOMEM;
2884
2885         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2886         if (!status) {
2887                 nfs_access_set_mask(entry, res.access);
2888                 nfs_refresh_inode(inode, res.fattr);
2889         }
2890         nfs_free_fattr(res.fattr);
2891         return status;
2892 }
2893
2894 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2895 {
2896         struct nfs4_exception exception = { };
2897         int err;
2898         do {
2899                 err = nfs4_handle_exception(NFS_SERVER(inode),
2900                                 _nfs4_proc_access(inode, entry),
2901                                 &exception);
2902         } while (exception.retry);
2903         return err;
2904 }
2905
2906 /*
2907  * TODO: For the time being, we don't try to get any attributes
2908  * along with any of the zero-copy operations READ, READDIR,
2909  * READLINK, WRITE.
2910  *
2911  * In the case of the first three, we want to put the GETATTR
2912  * after the read-type operation -- this is because it is hard
2913  * to predict the length of a GETATTR response in v4, and thus
2914  * align the READ data correctly.  This means that the GETATTR
2915  * may end up partially falling into the page cache, and we should
2916  * shift it into the 'tail' of the xdr_buf before processing.
2917  * To do this efficiently, we need to know the total length
2918  * of data received, which doesn't seem to be available outside
2919  * of the RPC layer.
2920  *
2921  * In the case of WRITE, we also want to put the GETATTR after
2922  * the operation -- in this case because we want to make sure
2923  * we get the post-operation mtime and size.
2924  *
2925  * Both of these changes to the XDR layer would in fact be quite
2926  * minor, but I decided to leave them for a subsequent patch.
2927  */
2928 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2929                 unsigned int pgbase, unsigned int pglen)
2930 {
2931         struct nfs4_readlink args = {
2932                 .fh       = NFS_FH(inode),
2933                 .pgbase   = pgbase,
2934                 .pglen    = pglen,
2935                 .pages    = &page,
2936         };
2937         struct nfs4_readlink_res res;
2938         struct rpc_message msg = {
2939                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2940                 .rpc_argp = &args,
2941                 .rpc_resp = &res,
2942         };
2943
2944         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2945 }
2946
2947 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2948                 unsigned int pgbase, unsigned int pglen)
2949 {
2950         struct nfs4_exception exception = { };
2951         int err;
2952         do {
2953                 err = nfs4_handle_exception(NFS_SERVER(inode),
2954                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2955                                 &exception);
2956         } while (exception.retry);
2957         return err;
2958 }
2959
2960 /*
2961  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
2962  */
2963 static int
2964 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2965                  int flags)
2966 {
2967         struct nfs_open_context *ctx;
2968         struct nfs4_state *state;
2969         int status = 0;
2970
2971         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
2972         if (IS_ERR(ctx))
2973                 return PTR_ERR(ctx);
2974
2975         sattr->ia_mode &= ~current_umask();
2976         state = nfs4_do_open(dir, dentry, ctx->mode,
2977                         flags, sattr, ctx->cred,
2978                         &ctx->mdsthreshold);
2979         d_drop(dentry);
2980         if (IS_ERR(state)) {
2981                 status = PTR_ERR(state);
2982                 goto out;
2983         }
2984         d_add(dentry, igrab(state->inode));
2985         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2986         ctx->state = state;
2987 out:
2988         put_nfs_open_context(ctx);
2989         return status;
2990 }
2991
2992 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2993 {
2994         struct nfs_server *server = NFS_SERVER(dir);
2995         struct nfs_removeargs args = {
2996                 .fh = NFS_FH(dir),
2997                 .name = *name,
2998         };
2999         struct nfs_removeres res = {
3000                 .server = server,
3001         };
3002         struct rpc_message msg = {
3003                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3004                 .rpc_argp = &args,
3005                 .rpc_resp = &res,
3006         };
3007         int status;
3008
3009         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3010         if (status == 0)
3011                 update_changeattr(dir, &res.cinfo);
3012         return status;
3013 }
3014
3015 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3016 {
3017         struct nfs4_exception exception = { };
3018         int err;
3019         do {
3020                 err = nfs4_handle_exception(NFS_SERVER(dir),
3021                                 _nfs4_proc_remove(dir, name),
3022                                 &exception);
3023         } while (exception.retry);
3024         return err;
3025 }
3026
3027 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3028 {
3029         struct nfs_server *server = NFS_SERVER(dir);
3030         struct nfs_removeargs *args = msg->rpc_argp;
3031         struct nfs_removeres *res = msg->rpc_resp;
3032
3033         res->server = server;
3034         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3035         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3036 }
3037
3038 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3039 {
3040         if (nfs4_setup_sequence(NFS_SERVER(data->dir),
3041                                 &data->args.seq_args,
3042                                 &data->res.seq_res,
3043                                 task))
3044                 return;
3045         rpc_call_start(task);
3046 }
3047
3048 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3049 {
3050         struct nfs_removeres *res = task->tk_msg.rpc_resp;
3051
3052         if (!nfs4_sequence_done(task, &res->seq_res))
3053                 return 0;
3054         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3055                 return 0;
3056         update_changeattr(dir, &res->cinfo);
3057         return 1;
3058 }
3059
3060 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3061 {
3062         struct nfs_server *server = NFS_SERVER(dir);
3063         struct nfs_renameargs *arg = msg->rpc_argp;
3064         struct nfs_renameres *res = msg->rpc_resp;
3065
3066         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3067         res->server = server;
3068         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3069 }
3070
3071 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3072 {
3073         if (nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3074                                 &data->args.seq_args,
3075                                 &data->res.seq_res,
3076                                 task))
3077                 return;
3078         rpc_call_start(task);
3079 }
3080
3081 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3082                                  struct inode *new_dir)
3083 {
3084         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3085
3086         if (!nfs4_sequence_done(task, &res->seq_res))
3087                 return 0;
3088         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3089                 return 0;
3090
3091         update_changeattr(old_dir, &res->old_cinfo);
3092         update_changeattr(new_dir, &res->new_cinfo);
3093         return 1;
3094 }
3095
3096 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3097                 struct inode *new_dir, struct qstr *new_name)
3098 {
3099         struct nfs_server *server = NFS_SERVER(old_dir);
3100         struct nfs_renameargs arg = {
3101                 .old_dir = NFS_FH(old_dir),
3102                 .new_dir = NFS_FH(new_dir),
3103                 .old_name = old_name,
3104                 .new_name = new_name,
3105         };
3106         struct nfs_renameres res = {
3107                 .server = server,
3108         };
3109         struct rpc_message msg = {
3110                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3111                 .rpc_argp = &arg,
3112                 .rpc_resp = &res,
3113         };
3114         int status = -ENOMEM;
3115
3116         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3117         if (!status) {
3118                 update_changeattr(old_dir, &res.old_cinfo);
3119                 update_changeattr(new_dir, &res.new_cinfo);
3120         }
3121         return status;
3122 }
3123
3124 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3125                 struct inode *new_dir, struct qstr *new_name)
3126 {
3127         struct nfs4_exception exception = { };
3128         int err;
3129         do {
3130                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3131                                 _nfs4_proc_rename(old_dir, old_name,
3132                                         new_dir, new_name),
3133                                 &exception);
3134         } while (exception.retry);
3135         return err;
3136 }
3137
3138 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3139 {
3140         struct nfs_server *server = NFS_SERVER(inode);
3141         struct nfs4_link_arg arg = {
3142                 .fh     = NFS_FH(inode),
3143                 .dir_fh = NFS_FH(dir),
3144                 .name   = name,
3145                 .bitmask = server->attr_bitmask,
3146         };
3147         struct nfs4_link_res res = {
3148                 .server = server,
3149         };
3150         struct rpc_message msg = {
3151                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3152                 .rpc_argp = &arg,
3153                 .rpc_resp = &res,
3154         };
3155         int status = -ENOMEM;
3156
3157         res.fattr = nfs_alloc_fattr();
3158         if (res.fattr == NULL)
3159                 goto out;
3160
3161         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3162         if (!status) {
3163                 update_changeattr(dir, &res.cinfo);
3164                 nfs_post_op_update_inode(inode, res.fattr);
3165         }
3166 out:
3167         nfs_free_fattr(res.fattr);
3168         return status;
3169 }
3170
3171 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3172 {
3173         struct nfs4_exception exception = { };
3174         int err;
3175         do {
3176                 err = nfs4_handle_exception(NFS_SERVER(inode),
3177                                 _nfs4_proc_link(inode, dir, name),
3178                                 &exception);
3179         } while (exception.retry);
3180         return err;
3181 }
3182
3183 struct nfs4_createdata {
3184         struct rpc_message msg;
3185         struct nfs4_create_arg arg;
3186         struct nfs4_create_res res;
3187         struct nfs_fh fh;
3188         struct nfs_fattr fattr;
3189 };
3190
3191 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3192                 struct qstr *name, struct iattr *sattr, u32 ftype)
3193 {
3194         struct nfs4_createdata *data;
3195
3196         data = kzalloc(sizeof(*data), GFP_KERNEL);
3197         if (data != NULL) {
3198                 struct nfs_server *server = NFS_SERVER(dir);
3199
3200                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3201                 data->msg.rpc_argp = &data->arg;
3202                 data->msg.rpc_resp = &data->res;
3203                 data->arg.dir_fh = NFS_FH(dir);
3204                 data->arg.server = server;
3205                 data->arg.name = name;
3206                 data->arg.attrs = sattr;
3207                 data->arg.ftype = ftype;
3208                 data->arg.bitmask = server->attr_bitmask;
3209                 data->res.server = server;
3210                 data->res.fh = &data->fh;
3211                 data->res.fattr = &data->fattr;
3212                 nfs_fattr_init(data->res.fattr);
3213         }
3214         return data;
3215 }
3216
3217 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3218 {
3219         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3220                                     &data->arg.seq_args, &data->res.seq_res, 1);
3221         if (status == 0) {
3222                 update_changeattr(dir, &data->res.dir_cinfo);
3223                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3224         }
3225         return status;
3226 }
3227
3228 static void nfs4_free_createdata(struct nfs4_createdata *data)
3229 {
3230         kfree(data);
3231 }
3232
3233 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3234                 struct page *page, unsigned int len, struct iattr *sattr)
3235 {
3236         struct nfs4_createdata *data;
3237         int status = -ENAMETOOLONG;
3238
3239         if (len > NFS4_MAXPATHLEN)
3240                 goto out;
3241
3242         status = -ENOMEM;
3243         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3244         if (data == NULL)
3245                 goto out;
3246
3247         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3248         data->arg.u.symlink.pages = &page;
3249         data->arg.u.symlink.len = len;
3250
3251         status = nfs4_do_create(dir, dentry, data);
3252
3253         nfs4_free_createdata(data);
3254 out:
3255         return status;
3256 }
3257
3258 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3259                 struct page *page, unsigned int len, struct iattr *sattr)
3260 {
3261         struct nfs4_exception exception = { };
3262         int err;
3263         do {
3264                 err = nfs4_handle_exception(NFS_SERVER(dir),
3265                                 _nfs4_proc_symlink(dir, dentry, page,
3266                                                         len, sattr),
3267                                 &exception);
3268         } while (exception.retry);
3269         return err;
3270 }
3271
3272 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3273                 struct iattr *sattr)
3274 {
3275         struct nfs4_createdata *data;
3276         int status = -ENOMEM;
3277
3278         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3279         if (data == NULL)
3280                 goto out;
3281
3282         status = nfs4_do_create(dir, dentry, data);
3283
3284         nfs4_free_createdata(data);
3285 out:
3286         return status;
3287 }
3288
3289 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3290                 struct iattr *sattr)
3291 {
3292         struct nfs4_exception exception = { };
3293         int err;
3294
3295         sattr->ia_mode &= ~current_umask();
3296         do {
3297                 err = nfs4_handle_exception(NFS_SERVER(dir),
3298                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3299                                 &exception);
3300         } while (exception.retry);
3301         return err;
3302 }
3303
3304 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3305                 u64 cookie, struct page **pages, unsigned int count, int plus)
3306 {
3307         struct inode            *dir = dentry->d_inode;
3308         struct nfs4_readdir_arg args = {
3309                 .fh = NFS_FH(dir),
3310                 .pages = pages,
3311                 .pgbase = 0,
3312                 .count = count,
3313                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3314                 .plus = plus,
3315         };
3316         struct nfs4_readdir_res res;
3317         struct rpc_message msg = {
3318                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3319                 .rpc_argp = &args,
3320                 .rpc_resp = &res,
3321                 .rpc_cred = cred,
3322         };
3323         int                     status;
3324
3325         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3326                         dentry->d_parent->d_name.name,
3327                         dentry->d_name.name,
3328                         (unsigned long long)cookie);
3329         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3330         res.pgbase = args.pgbase;
3331         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3332         if (status >= 0) {
3333                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3334                 status += args.pgbase;
3335         }
3336
3337         nfs_invalidate_atime(dir);
3338
3339         dprintk("%s: returns %d\n", __func__, status);
3340         return status;
3341 }
3342
3343 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3344                 u64 cookie, struct page **pages, unsigned int count, int plus)
3345 {
3346         struct nfs4_exception exception = { };
3347         int err;
3348         do {
3349                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3350                                 _nfs4_proc_readdir(dentry, cred, cookie,
3351                                         pages, count, plus),
3352                                 &exception);
3353         } while (exception.retry);
3354         return err;
3355 }
3356
3357 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3358                 struct iattr *sattr, dev_t rdev)
3359 {
3360         struct nfs4_createdata *data;
3361         int mode = sattr->ia_mode;
3362         int status = -ENOMEM;
3363
3364         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3365         if (data == NULL)
3366                 goto out;
3367
3368         if (S_ISFIFO(mode))
3369                 data->arg.ftype = NF4FIFO;
3370         else if (S_ISBLK(mode)) {
3371                 data->arg.ftype = NF4BLK;
3372                 data->arg.u.device.specdata1 = MAJOR(rdev);
3373                 data->arg.u.device.specdata2 = MINOR(rdev);
3374         }
3375         else if (S_ISCHR(mode)) {
3376                 data->arg.ftype = NF4CHR;
3377                 data->arg.u.device.specdata1 = MAJOR(rdev);
3378                 data->arg.u.device.specdata2 = MINOR(rdev);
3379         } else if (!S_ISSOCK(mode)) {
3380                 status = -EINVAL;
3381                 goto out_free;
3382         }
3383
3384         status = nfs4_do_create(dir, dentry, data);
3385 out_free:
3386         nfs4_free_createdata(data);
3387 out:
3388         return status;
3389 }
3390
3391 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3392                 struct iattr *sattr, dev_t rdev)
3393 {
3394         struct nfs4_exception exception = { };
3395         int err;
3396
3397         sattr->ia_mode &= ~current_umask();
3398         do {
3399                 err = nfs4_handle_exception(NFS_SERVER(dir),
3400                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3401                                 &exception);
3402         } while (exception.retry);
3403         return err;
3404 }
3405
3406 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3407                  struct nfs_fsstat *fsstat)
3408 {
3409         struct nfs4_statfs_arg args = {
3410                 .fh = fhandle,
3411                 .bitmask = server->attr_bitmask,
3412         };
3413         struct nfs4_statfs_res res = {
3414                 .fsstat = fsstat,
3415         };
3416         struct rpc_message msg = {
3417                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3418                 .rpc_argp = &args,
3419                 .rpc_resp = &res,
3420         };
3421
3422         nfs_fattr_init(fsstat->fattr);
3423         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3424 }
3425
3426 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3427 {
3428         struct nfs4_exception exception = { };
3429         int err;
3430         do {
3431                 err = nfs4_handle_exception(server,
3432                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3433                                 &exception);
3434         } while (exception.retry);
3435         return err;
3436 }
3437
3438 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3439                 struct nfs_fsinfo *fsinfo)
3440 {
3441         struct nfs4_fsinfo_arg args = {
3442                 .fh = fhandle,
3443                 .bitmask = server->attr_bitmask,
3444         };
3445         struct nfs4_fsinfo_res res = {
3446                 .fsinfo = fsinfo,
3447         };
3448         struct rpc_message msg = {
3449                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3450                 .rpc_argp = &args,
3451                 .rpc_resp = &res,
3452         };
3453
3454         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3455 }
3456
3457 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3458 {
3459         struct nfs4_exception exception = { };
3460         int err;
3461
3462         do {
3463                 err = nfs4_handle_exception(server,
3464                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3465                                 &exception);
3466         } while (exception.retry);
3467         return err;
3468 }
3469
3470 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3471 {
3472         int error;
3473
3474         nfs_fattr_init(fsinfo->fattr);
3475         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3476         if (error == 0) {
3477                 /* block layout checks this! */
3478                 server->pnfs_blksize = fsinfo->blksize;
3479                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3480         }
3481
3482         return error;
3483 }
3484
3485 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3486                 struct nfs_pathconf *pathconf)
3487 {
3488         struct nfs4_pathconf_arg args = {
3489                 .fh = fhandle,
3490                 .bitmask = server->attr_bitmask,
3491         };
3492         struct nfs4_pathconf_res res = {
3493                 .pathconf = pathconf,
3494         };
3495         struct rpc_message msg = {
3496                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3497                 .rpc_argp = &args,
3498                 .rpc_resp = &res,
3499         };
3500
3501         /* None of the pathconf attributes are mandatory to implement */
3502         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3503                 memset(pathconf, 0, sizeof(*pathconf));
3504                 return 0;
3505         }
3506
3507         nfs_fattr_init(pathconf->fattr);
3508         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3509 }
3510
3511 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3512                 struct nfs_pathconf *pathconf)
3513 {
3514         struct nfs4_exception exception = { };
3515         int err;
3516
3517         do {
3518                 err = nfs4_handle_exception(server,
3519                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3520                                 &exception);
3521         } while (exception.retry);
3522         return err;
3523 }
3524
3525 void __nfs4_read_done_cb(struct nfs_read_data *data)
3526 {
3527         nfs_invalidate_atime(data->header->inode);
3528 }
3529
3530 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3531 {
3532         struct nfs_server *server = NFS_SERVER(data->header->inode);
3533
3534         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3535                 rpc_restart_call_prepare(task);
3536                 return -EAGAIN;
3537         }
3538
3539         __nfs4_read_done_cb(data);
3540         if (task->tk_status > 0)
3541                 renew_lease(server, data->timestamp);
3542         return 0;
3543 }
3544
3545 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3546 {
3547
3548         dprintk("--> %s\n", __func__);
3549
3550         if (!nfs4_sequence_done(task, &data->res.seq_res))
3551                 return -EAGAIN;
3552
3553         return data->read_done_cb ? data->read_done_cb(task, data) :
3554                                     nfs4_read_done_cb(task, data);
3555 }
3556
3557 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3558 {
3559         data->timestamp   = jiffies;
3560         data->read_done_cb = nfs4_read_done_cb;
3561         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3562         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3563 }
3564
3565 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3566 {
3567         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3568                                 &data->args.seq_args,
3569                                 &data->res.seq_res,
3570                                 task))
3571                 return;
3572         rpc_call_start(task);
3573 }
3574
3575 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3576 {
3577         struct inode *inode = data->header->inode;
3578
3579         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3580                 rpc_restart_call_prepare(task);
3581                 return -EAGAIN;
3582         }
3583         if (task->tk_status >= 0) {
3584                 renew_lease(NFS_SERVER(inode), data->timestamp);
3585                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3586         }
3587         return 0;
3588 }
3589
3590 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3591 {
3592         if (!nfs4_sequence_done(task, &data->res.seq_res))
3593                 return -EAGAIN;
3594         return data->write_done_cb ? data->write_done_cb(task, data) :
3595                 nfs4_write_done_cb(task, data);
3596 }
3597
3598 static
3599 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3600 {
3601         const struct nfs_pgio_header *hdr = data->header;
3602
3603         /* Don't request attributes for pNFS or O_DIRECT writes */
3604         if (data->ds_clp != NULL || hdr->dreq != NULL)
3605                 return false;
3606         /* Otherwise, request attributes if and only if we don't hold
3607          * a delegation
3608          */
3609         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3610 }
3611
3612 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3613 {
3614         struct nfs_server *server = NFS_SERVER(data->header->inode);
3615
3616         if (!nfs4_write_need_cache_consistency_data(data)) {
3617                 data->args.bitmask = NULL;
3618                 data->res.fattr = NULL;
3619         } else
3620                 data->args.bitmask = server->cache_consistency_bitmask;
3621
3622         if (!data->write_done_cb)
3623                 data->write_done_cb = nfs4_write_done_cb;
3624         data->res.server = server;
3625         data->timestamp   = jiffies;
3626
3627         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3628         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3629 }
3630
3631 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3632 {
3633         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3634                                 &data->args.seq_args,
3635                                 &data->res.seq_res,
3636                                 task))
3637                 return;
3638         rpc_call_start(task);
3639 }
3640
3641 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3642 {
3643         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3644                                 &data->args.seq_args,
3645                                 &data->res.seq_res,
3646                                 task))
3647                 return;
3648         rpc_call_start(task);
3649 }
3650
3651 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3652 {
3653         struct inode *inode = data->inode;
3654
3655         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3656                 rpc_restart_call_prepare(task);
3657                 return -EAGAIN;
3658         }
3659         return 0;
3660 }
3661
3662 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3663 {
3664         if (!nfs4_sequence_done(task, &data->res.seq_res))
3665                 return -EAGAIN;
3666         return data->commit_done_cb(task, data);
3667 }
3668
3669 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3670 {
3671         struct nfs_server *server = NFS_SERVER(data->inode);
3672
3673         if (data->commit_done_cb == NULL)
3674                 data->commit_done_cb = nfs4_commit_done_cb;
3675         data->res.server = server;
3676         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3677         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3678 }
3679
3680 struct nfs4_renewdata {
3681         struct nfs_client       *client;
3682         unsigned long           timestamp;
3683 };
3684
3685 /*
3686  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3687  * standalone procedure for queueing an asynchronous RENEW.
3688  */
3689 static void nfs4_renew_release(void *calldata)
3690 {
3691         struct nfs4_renewdata *data = calldata;
3692         struct nfs_client *clp = data->client;
3693
3694         if (atomic_read(&clp->cl_count) > 1)
3695                 nfs4_schedule_state_renewal(clp);
3696         nfs_put_client(clp);
3697         kfree(data);
3698 }
3699
3700 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3701 {
3702         struct nfs4_renewdata *data = calldata;
3703         struct nfs_client *clp = data->client;
3704         unsigned long timestamp = data->timestamp;
3705
3706         if (task->tk_status < 0) {
3707                 /* Unless we're shutting down, schedule state recovery! */
3708                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3709                         return;
3710                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3711                         nfs4_schedule_lease_recovery(clp);
3712                         return;
3713                 }
3714                 nfs4_schedule_path_down_recovery(clp);
3715         }
3716         do_renew_lease(clp, timestamp);
3717 }
3718
3719 static const struct rpc_call_ops nfs4_renew_ops = {
3720         .rpc_call_done = nfs4_renew_done,
3721         .rpc_release = nfs4_renew_release,
3722 };
3723
3724 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3725 {
3726         struct rpc_message msg = {
3727                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3728                 .rpc_argp       = clp,
3729                 .rpc_cred       = cred,
3730         };
3731         struct nfs4_renewdata *data;
3732
3733         if (renew_flags == 0)
3734                 return 0;
3735         if (!atomic_inc_not_zero(&clp->cl_count))
3736                 return -EIO;
3737         data = kmalloc(sizeof(*data), GFP_NOFS);
3738         if (data == NULL)
3739                 return -ENOMEM;
3740         data->client = clp;
3741         data->timestamp = jiffies;
3742         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3743                         &nfs4_renew_ops, data);
3744 }
3745
3746 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3747 {
3748         struct rpc_message msg = {
3749                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3750                 .rpc_argp       = clp,
3751                 .rpc_cred       = cred,
3752         };
3753         unsigned long now = jiffies;
3754         int status;
3755
3756         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3757         if (status < 0)
3758                 return status;
3759         do_renew_lease(clp, now);
3760         return 0;
3761 }
3762
3763 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3764 {
3765         return (server->caps & NFS_CAP_ACLS)
3766                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3767                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3768 }
3769
3770 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3771  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3772  * the stack.
3773  */
3774 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3775
3776 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3777                 struct page **pages, unsigned int *pgbase)
3778 {
3779         struct page *newpage, **spages;
3780         int rc = 0;
3781         size_t len;
3782         spages = pages;
3783
3784         do {
3785                 len = min_t(size_t, PAGE_SIZE, buflen);
3786                 newpage = alloc_page(GFP_KERNEL);
3787
3788                 if (newpage == NULL)
3789                         goto unwind;
3790                 memcpy(page_address(newpage), buf, len);
3791                 buf += len;
3792                 buflen -= len;
3793                 *pages++ = newpage;
3794                 rc++;
3795         } while (buflen != 0);
3796
3797         return rc;
3798
3799 unwind:
3800         for(; rc > 0; rc--)
3801                 __free_page(spages[rc-1]);
3802         return -ENOMEM;
3803 }
3804
3805 struct nfs4_cached_acl {
3806         int cached;
3807         size_t len;
3808         char data[0];
3809 };
3810
3811 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3812 {
3813         struct nfs_inode *nfsi = NFS_I(inode);
3814
3815         spin_lock(&inode->i_lock);
3816         kfree(nfsi->nfs4_acl);
3817         nfsi->nfs4_acl = acl;
3818         spin_unlock(&inode->i_lock);
3819 }
3820
3821 static void nfs4_zap_acl_attr(struct inode *inode)
3822 {
3823         nfs4_set_cached_acl(inode, NULL);
3824 }
3825
3826 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3827 {
3828         struct nfs_inode *nfsi = NFS_I(inode);
3829         struct nfs4_cached_acl *acl;
3830         int ret = -ENOENT;
3831
3832         spin_lock(&inode->i_lock);
3833         acl = nfsi->nfs4_acl;
3834         if (acl == NULL)
3835                 goto out;
3836         if (buf == NULL) /* user is just asking for length */
3837                 goto out_len;
3838         if (acl->cached == 0)
3839                 goto out;
3840         ret = -ERANGE; /* see getxattr(2) man page */
3841         if (acl->len > buflen)
3842                 goto out;
3843         memcpy(buf, acl->data, acl->len);
3844 out_len:
3845         ret = acl->len;
3846 out:
3847         spin_unlock(&inode->i_lock);
3848         return ret;
3849 }
3850
3851 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3852 {
3853         struct nfs4_cached_acl *acl;
3854         size_t buflen = sizeof(*acl) + acl_len;
3855
3856         if (buflen <= PAGE_SIZE) {
3857                 acl = kmalloc(buflen, GFP_KERNEL);
3858                 if (acl == NULL)
3859                         goto out;
3860                 acl->cached = 1;
3861                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3862         } else {
3863                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3864                 if (acl == NULL)
3865                         goto out;
3866                 acl->cached = 0;
3867         }
3868         acl->len = acl_len;
3869 out:
3870         nfs4_set_cached_acl(inode, acl);
3871 }
3872
3873 /*
3874  * The getxattr API returns the required buffer length when called with a
3875  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3876  * the required buf.  On a NULL buf, we send a page of data to the server
3877  * guessing that the ACL request can be serviced by a page. If so, we cache
3878  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3879  * the cache. If not so, we throw away the page, and cache the required
3880  * length. The next getxattr call will then produce another round trip to
3881  * the server, this time with the input buf of the required size.
3882  */
3883 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3884 {
3885         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3886         struct nfs_getaclargs args = {
3887                 .fh = NFS_FH(inode),
3888                 .acl_pages = pages,
3889                 .acl_len = buflen,
3890         };
3891         struct nfs_getaclres res = {
3892                 .acl_len = buflen,
3893         };
3894         struct rpc_message msg = {
3895                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3896                 .rpc_argp = &args,
3897                 .rpc_resp = &res,
3898         };
3899         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3900         int ret = -ENOMEM, i;
3901
3902         /* As long as we're doing a round trip to the server anyway,
3903          * let's be prepared for a page of acl data. */
3904         if (npages == 0)
3905                 npages = 1;
3906         if (npages > ARRAY_SIZE(pages))
3907                 return -ERANGE;
3908
3909         for (i = 0; i < npages; i++) {
3910                 pages[i] = alloc_page(GFP_KERNEL);
3911                 if (!pages[i])
3912                         goto out_free;
3913         }
3914
3915         /* for decoding across pages */
3916         res.acl_scratch = alloc_page(GFP_KERNEL);
3917         if (!res.acl_scratch)
3918                 goto out_free;
3919
3920         args.acl_len = npages * PAGE_SIZE;
3921         args.acl_pgbase = 0;
3922
3923         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3924                 __func__, buf, buflen, npages, args.acl_len);
3925         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3926                              &msg, &args.seq_args, &res.seq_res, 0);
3927         if (ret)
3928                 goto out_free;
3929
3930         /* Handle the case where the passed-in buffer is too short */
3931         if (res.acl_flags & NFS4_ACL_TRUNC) {
3932                 /* Did the user only issue a request for the acl length? */
3933                 if (buf == NULL)
3934                         goto out_ok;
3935                 ret = -ERANGE;
3936                 goto out_free;
3937         }
3938         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
3939         if (buf)
3940                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
3941 out_ok:
3942         ret = res.acl_len;
3943 out_free:
3944         for (i = 0; i < npages; i++)
3945                 if (pages[i])
3946                         __free_page(pages[i]);
3947         if (res.acl_scratch)
3948                 __free_page(res.acl_scratch);
3949         return ret;
3950 }
3951
3952 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3953 {
3954         struct nfs4_exception exception = { };
3955         ssize_t ret;
3956         do {
3957                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3958                 if (ret >= 0)
3959                         break;
3960                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3961         } while (exception.retry);
3962         return ret;
3963 }
3964
3965 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3966 {
3967         struct nfs_server *server = NFS_SERVER(inode);
3968         int ret;
3969
3970         if (!nfs4_server_supports_acls(server))
3971                 return -EOPNOTSUPP;
3972         ret = nfs_revalidate_inode(server, inode);
3973         if (ret < 0)
3974                 return ret;
3975         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3976                 nfs_zap_acl_cache(inode);
3977         ret = nfs4_read_cached_acl(inode, buf, buflen);
3978         if (ret != -ENOENT)
3979                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3980                  * but no cached acl data, just the acl length */
3981                 return ret;
3982         return nfs4_get_acl_uncached(inode, buf, buflen);
3983 }
3984
3985 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3986 {
3987         struct nfs_server *server = NFS_SERVER(inode);
3988         struct page *pages[NFS4ACL_MAXPAGES];
3989         struct nfs_setaclargs arg = {
3990                 .fh             = NFS_FH(inode),
3991                 .acl_pages      = pages,
3992                 .acl_len        = buflen,
3993         };
3994         struct nfs_setaclres res;
3995         struct rpc_message msg = {
3996                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3997                 .rpc_argp       = &arg,
3998                 .rpc_resp       = &res,
3999         };
4000         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4001         int ret, i;
4002
4003         if (!nfs4_server_supports_acls(server))
4004                 return -EOPNOTSUPP;
4005         if (npages > ARRAY_SIZE(pages))
4006                 return -ERANGE;
4007         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4008         if (i < 0)
4009                 return i;
4010         nfs4_inode_return_delegation(inode);
4011         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4012
4013         /*
4014          * Free each page after tx, so the only ref left is
4015          * held by the network stack
4016          */
4017         for (; i > 0; i--)
4018                 put_page(pages[i-1]);
4019
4020         /*
4021          * Acl update can result in inode attribute update.
4022          * so mark the attribute cache invalid.
4023          */
4024         spin_lock(&inode->i_lock);
4025         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4026         spin_unlock(&inode->i_lock);
4027         nfs_access_zap_cache(inode);
4028         nfs_zap_acl_cache(inode);
4029         return ret;
4030 }
4031
4032 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4033 {
4034         struct nfs4_exception exception = { };
4035         int err;
4036         do {
4037                 err = nfs4_handle_exception(NFS_SERVER(inode),
4038                                 __nfs4_proc_set_acl(inode, buf, buflen),
4039                                 &exception);
4040         } while (exception.retry);
4041         return err;
4042 }
4043
4044 static int
4045 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4046 {
4047         struct nfs_client *clp = server->nfs_client;
4048
4049         if (task->tk_status >= 0)
4050                 return 0;
4051         switch(task->tk_status) {
4052                 case -NFS4ERR_DELEG_REVOKED:
4053                 case -NFS4ERR_ADMIN_REVOKED:
4054                 case -NFS4ERR_BAD_STATEID:
4055                         if (state == NULL)
4056                                 break;
4057                         nfs_remove_bad_delegation(state->inode);
4058                 case -NFS4ERR_OPENMODE:
4059                         if (state == NULL)
4060                                 break;
4061                         nfs4_schedule_stateid_recovery(server, state);
4062                         goto wait_on_recovery;
4063                 case -NFS4ERR_EXPIRED:
4064                         if (state != NULL)
4065                                 nfs4_schedule_stateid_recovery(server, state);
4066                 case -NFS4ERR_STALE_STATEID:
4067                 case -NFS4ERR_STALE_CLIENTID:
4068                         nfs4_schedule_lease_recovery(clp);
4069                         goto wait_on_recovery;
4070 #if defined(CONFIG_NFS_V4_1)
4071                 case -NFS4ERR_BADSESSION:
4072                 case -NFS4ERR_BADSLOT:
4073                 case -NFS4ERR_BAD_HIGH_SLOT:
4074                 case -NFS4ERR_DEADSESSION:
4075                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4076                 case -NFS4ERR_SEQ_FALSE_RETRY:
4077                 case -NFS4ERR_SEQ_MISORDERED:
4078                         dprintk("%s ERROR %d, Reset session\n", __func__,
4079                                 task->tk_status);
4080                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4081                         task->tk_status = 0;
4082                         return -EAGAIN;
4083 #endif /* CONFIG_NFS_V4_1 */
4084                 case -NFS4ERR_DELAY:
4085                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4086                 case -NFS4ERR_GRACE:
4087                 case -EKEYEXPIRED:
4088                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4089                         task->tk_status = 0;
4090                         return -EAGAIN;
4091                 case -NFS4ERR_RETRY_UNCACHED_REP:
4092                 case -NFS4ERR_OLD_STATEID:
4093                         task->tk_status = 0;
4094                         return -EAGAIN;
4095         }
4096         task->tk_status = nfs4_map_errors(task->tk_status);
4097         return 0;
4098 wait_on_recovery:
4099         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4100         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4101                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4102         task->tk_status = 0;
4103         return -EAGAIN;
4104 }
4105
4106 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4107                                     nfs4_verifier *bootverf)
4108 {
4109         __be32 verf[2];
4110
4111         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4112                 /* An impossible timestamp guarantees this value
4113                  * will never match a generated boot time. */
4114                 verf[0] = 0;
4115                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4116         } else {
4117                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4118                 verf[0] = (__be32)nn->boot_time.tv_sec;
4119                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4120         }
4121         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4122 }
4123
4124 static unsigned int
4125 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4126                                    char *buf, size_t len)
4127 {
4128         unsigned int result;
4129
4130         rcu_read_lock();
4131         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4132                                 clp->cl_ipaddr,
4133                                 rpc_peeraddr2str(clp->cl_rpcclient,
4134                                                         RPC_DISPLAY_ADDR),
4135                                 rpc_peeraddr2str(clp->cl_rpcclient,
4136                                                         RPC_DISPLAY_PROTO));
4137         rcu_read_unlock();
4138         return result;
4139 }
4140
4141 static unsigned int
4142 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4143                                 char *buf, size_t len)
4144 {
4145         char *nodename = clp->cl_rpcclient->cl_nodename;
4146
4147         if (nfs4_client_id_uniquifier[0] != '\0')
4148                 nodename = nfs4_client_id_uniquifier;
4149         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4150                                 clp->rpc_ops->version, clp->cl_minorversion,
4151                                 nodename);
4152 }
4153
4154 /**
4155  * nfs4_proc_setclientid - Negotiate client ID
4156  * @clp: state data structure
4157  * @program: RPC program for NFSv4 callback service
4158  * @port: IP port number for NFS4 callback service
4159  * @cred: RPC credential to use for this call
4160  * @res: where to place the result
4161  *
4162  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4163  */
4164 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4165                 unsigned short port, struct rpc_cred *cred,
4166                 struct nfs4_setclientid_res *res)
4167 {
4168         nfs4_verifier sc_verifier;
4169         struct nfs4_setclientid setclientid = {
4170                 .sc_verifier = &sc_verifier,
4171                 .sc_prog = program,
4172                 .sc_cb_ident = clp->cl_cb_ident,
4173         };
4174         struct rpc_message msg = {
4175                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4176                 .rpc_argp = &setclientid,
4177                 .rpc_resp = res,
4178                 .rpc_cred = cred,
4179         };
4180         int status;
4181
4182         /* nfs_client_id4 */
4183         nfs4_init_boot_verifier(clp, &sc_verifier);
4184         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4185                 setclientid.sc_name_len =
4186                                 nfs4_init_uniform_client_string(clp,
4187                                                 setclientid.sc_name,
4188                                                 sizeof(setclientid.sc_name));
4189         else
4190                 setclientid.sc_name_len =
4191                                 nfs4_init_nonuniform_client_string(clp,
4192                                                 setclientid.sc_name,
4193                                                 sizeof(setclientid.sc_name));
4194         /* cb_client4 */
4195         rcu_read_lock();
4196         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4197                                 sizeof(setclientid.sc_netid),
4198                                 rpc_peeraddr2str(clp->cl_rpcclient,
4199                                                         RPC_DISPLAY_NETID));
4200         rcu_read_unlock();
4201         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4202                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4203                                 clp->cl_ipaddr, port >> 8, port & 255);
4204
4205         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4206                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4207                 setclientid.sc_name_len, setclientid.sc_name);
4208         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4209         dprintk("NFS reply setclientid: %d\n", status);
4210         return status;
4211 }
4212
4213 /**
4214  * nfs4_proc_setclientid_confirm - Confirm client ID
4215  * @clp: state data structure
4216  * @res: result of a previous SETCLIENTID
4217  * @cred: RPC credential to use for this call
4218  *
4219  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4220  */
4221 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4222                 struct nfs4_setclientid_res *arg,
4223                 struct rpc_cred *cred)
4224 {
4225         struct nfs_fsinfo fsinfo;
4226         struct rpc_message msg = {
4227                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4228                 .rpc_argp = arg,
4229                 .rpc_resp = &fsinfo,
4230                 .rpc_cred = cred,
4231         };
4232         unsigned long now;
4233         int status;
4234
4235         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4236                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4237                 clp->cl_clientid);
4238         now = jiffies;
4239         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4240         if (status == 0) {
4241                 spin_lock(&clp->cl_lock);
4242                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4243                 clp->cl_last_renewal = now;
4244                 spin_unlock(&clp->cl_lock);
4245         }
4246         dprintk("NFS reply setclientid_confirm: %d\n", status);
4247         return status;
4248 }
4249
4250 struct nfs4_delegreturndata {
4251         struct nfs4_delegreturnargs args;
4252         struct nfs4_delegreturnres res;
4253         struct nfs_fh fh;
4254         nfs4_stateid stateid;
4255         unsigned long timestamp;
4256         struct nfs_fattr fattr;
4257         int rpc_status;
4258 };
4259
4260 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4261 {
4262         struct nfs4_delegreturndata *data = calldata;
4263
4264         if (!nfs4_sequence_done(task, &data->res.seq_res))
4265                 return;
4266
4267         switch (task->tk_status) {
4268         case -NFS4ERR_STALE_STATEID:
4269         case -NFS4ERR_EXPIRED:
4270         case 0:
4271                 renew_lease(data->res.server, data->timestamp);
4272                 break;
4273         default:
4274                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4275                                 -EAGAIN) {
4276                         rpc_restart_call_prepare(task);
4277                         return;
4278                 }
4279         }
4280         data->rpc_status = task->tk_status;
4281 }
4282
4283 static void nfs4_delegreturn_release(void *calldata)
4284 {
4285         kfree(calldata);
4286 }
4287
4288 #if defined(CONFIG_NFS_V4_1)
4289 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4290 {
4291         struct nfs4_delegreturndata *d_data;
4292
4293         d_data = (struct nfs4_delegreturndata *)data;
4294
4295         if (nfs4_setup_sequence(d_data->res.server,
4296                                 &d_data->args.seq_args,
4297                                 &d_data->res.seq_res, task))
4298                 return;
4299         rpc_call_start(task);
4300 }
4301 #endif /* CONFIG_NFS_V4_1 */
4302
4303 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4304 #if defined(CONFIG_NFS_V4_1)
4305         .rpc_call_prepare = nfs4_delegreturn_prepare,
4306 #endif /* CONFIG_NFS_V4_1 */
4307         .rpc_call_done = nfs4_delegreturn_done,
4308         .rpc_release = nfs4_delegreturn_release,
4309 };
4310
4311 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4312 {
4313         struct nfs4_delegreturndata *data;
4314         struct nfs_server *server = NFS_SERVER(inode);
4315         struct rpc_task *task;
4316         struct rpc_message msg = {
4317                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4318                 .rpc_cred = cred,
4319         };
4320         struct rpc_task_setup task_setup_data = {
4321                 .rpc_client = server->client,
4322                 .rpc_message = &msg,
4323                 .callback_ops = &nfs4_delegreturn_ops,
4324                 .flags = RPC_TASK_ASYNC,
4325         };
4326         int status = 0;
4327
4328         data = kzalloc(sizeof(*data), GFP_NOFS);
4329         if (data == NULL)
4330                 return -ENOMEM;
4331         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4332         data->args.fhandle = &data->fh;
4333         data->args.stateid = &data->stateid;
4334         data->args.bitmask = server->cache_consistency_bitmask;
4335         nfs_copy_fh(&data->fh, NFS_FH(inode));
4336         nfs4_stateid_copy(&data->stateid, stateid);
4337         data->res.fattr = &data->fattr;
4338         data->res.server = server;
4339         nfs_fattr_init(data->res.fattr);
4340         data->timestamp = jiffies;
4341         data->rpc_status = 0;
4342
4343         task_setup_data.callback_data = data;
4344         msg.rpc_argp = &data->args;
4345         msg.rpc_resp = &data->res;
4346         task = rpc_run_task(&task_setup_data);
4347         if (IS_ERR(task))
4348                 return PTR_ERR(task);
4349         if (!issync)
4350                 goto out;
4351         status = nfs4_wait_for_completion_rpc_task(task);
4352         if (status != 0)
4353                 goto out;
4354         status = data->rpc_status;
4355         if (status == 0)
4356                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4357         else
4358                 nfs_refresh_inode(inode, &data->fattr);
4359 out:
4360         rpc_put_task(task);
4361         return status;
4362 }
4363
4364 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4365 {
4366         struct nfs_server *server = NFS_SERVER(inode);
4367         struct nfs4_exception exception = { };
4368         int err;
4369         do {
4370                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4371                 switch (err) {
4372                         case -NFS4ERR_STALE_STATEID:
4373                         case -NFS4ERR_EXPIRED:
4374                         case 0:
4375                                 return 0;
4376                 }
4377                 err = nfs4_handle_exception(server, err, &exception);
4378         } while (exception.retry);
4379         return err;
4380 }
4381
4382 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4383 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4384
4385 /*
4386  * sleep, with exponential backoff, and retry the LOCK operation.
4387  */
4388 static unsigned long
4389 nfs4_set_lock_task_retry(unsigned long timeout)
4390 {
4391         freezable_schedule_timeout_killable(timeout);
4392         timeout <<= 1;
4393         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4394                 return NFS4_LOCK_MAXTIMEOUT;
4395         return timeout;
4396 }
4397
4398 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4399 {
4400         struct inode *inode = state->inode;
4401         struct nfs_server *server = NFS_SERVER(inode);
4402         struct nfs_client *clp = server->nfs_client;
4403         struct nfs_lockt_args arg = {
4404                 .fh = NFS_FH(inode),
4405                 .fl = request,
4406         };
4407         struct nfs_lockt_res res = {
4408                 .denied = request,
4409         };
4410         struct rpc_message msg = {
4411                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4412                 .rpc_argp       = &arg,
4413                 .rpc_resp       = &res,
4414                 .rpc_cred       = state->owner->so_cred,
4415         };
4416         struct nfs4_lock_state *lsp;
4417         int status;
4418
4419         arg.lock_owner.clientid = clp->cl_clientid;
4420         status = nfs4_set_lock_state(state, request);
4421         if (status != 0)
4422                 goto out;
4423         lsp = request->fl_u.nfs4_fl.owner;
4424         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4425         arg.lock_owner.s_dev = server->s_dev;
4426         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4427         switch (status) {
4428                 case 0:
4429                         request->fl_type = F_UNLCK;
4430                         break;
4431                 case -NFS4ERR_DENIED:
4432                         status = 0;
4433         }
4434         request->fl_ops->fl_release_private(request);
4435 out:
4436         return status;
4437 }
4438
4439 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4440 {
4441         struct nfs4_exception exception = { };
4442         int err;
4443
4444         do {
4445                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4446                                 _nfs4_proc_getlk(state, cmd, request),
4447                                 &exception);
4448         } while (exception.retry);
4449         return err;
4450 }
4451
4452 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4453 {
4454         int res = 0;
4455         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4456                 case FL_POSIX:
4457                         res = posix_lock_file_wait(file, fl);
4458                         break;
4459                 case FL_FLOCK:
4460                         res = flock_lock_file_wait(file, fl);
4461                         break;
4462                 default:
4463                         BUG();
4464         }
4465         return res;
4466 }
4467
4468 struct nfs4_unlockdata {
4469         struct nfs_locku_args arg;
4470         struct nfs_locku_res res;
4471         struct nfs4_lock_state *lsp;
4472         struct nfs_open_context *ctx;
4473         struct file_lock fl;
4474         const struct nfs_server *server;
4475         unsigned long timestamp;
4476 };
4477
4478 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4479                 struct nfs_open_context *ctx,
4480                 struct nfs4_lock_state *lsp,
4481                 struct nfs_seqid *seqid)
4482 {
4483         struct nfs4_unlockdata *p;
4484         struct inode *inode = lsp->ls_state->inode;
4485
4486         p = kzalloc(sizeof(*p), GFP_NOFS);
4487         if (p == NULL)
4488                 return NULL;
4489         p->arg.fh = NFS_FH(inode);
4490         p->arg.fl = &p->fl;
4491         p->arg.seqid = seqid;
4492         p->res.seqid = seqid;
4493         p->arg.stateid = &lsp->ls_stateid;
4494         p->lsp = lsp;
4495         atomic_inc(&lsp->ls_count);
4496         /* Ensure we don't close file until we're done freeing locks! */
4497         p->ctx = get_nfs_open_context(ctx);
4498         memcpy(&p->fl, fl, sizeof(p->fl));
4499         p->server = NFS_SERVER(inode);
4500         return p;
4501 }
4502
4503 static void nfs4_locku_release_calldata(void *data)
4504 {
4505         struct nfs4_unlockdata *calldata = data;
4506         nfs_free_seqid(calldata->arg.seqid);
4507         nfs4_put_lock_state(calldata->lsp);
4508         put_nfs_open_context(calldata->ctx);
4509         kfree(calldata);
4510 }
4511
4512 static void nfs4_locku_done(struct rpc_task *task, void *data)
4513 {
4514         struct nfs4_unlockdata *calldata = data;
4515
4516         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4517                 return;
4518         switch (task->tk_status) {
4519                 case 0:
4520                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4521                                         &calldata->res.stateid);
4522                         renew_lease(calldata->server, calldata->timestamp);
4523                         break;
4524                 case -NFS4ERR_BAD_STATEID:
4525                 case -NFS4ERR_OLD_STATEID:
4526                 case -NFS4ERR_STALE_STATEID:
4527                 case -NFS4ERR_EXPIRED:
4528                         break;
4529                 default:
4530                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4531                                 rpc_restart_call_prepare(task);
4532         }
4533         nfs_release_seqid(calldata->arg.seqid);
4534 }
4535
4536 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4537 {
4538         struct nfs4_unlockdata *calldata = data;
4539
4540         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4541                 return;
4542         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4543                 /* Note: exit _without_ running nfs4_locku_done */
4544                 task->tk_action = NULL;
4545                 return;
4546         }
4547         calldata->timestamp = jiffies;
4548         if (nfs4_setup_sequence(calldata->server,
4549                                 &calldata->arg.seq_args,
4550                                 &calldata->res.seq_res,
4551                                 task) != 0)
4552                 nfs_release_seqid(calldata->arg.seqid);
4553         else
4554                 rpc_call_start(task);
4555 }
4556
4557 static const struct rpc_call_ops nfs4_locku_ops = {
4558         .rpc_call_prepare = nfs4_locku_prepare,
4559         .rpc_call_done = nfs4_locku_done,
4560         .rpc_release = nfs4_locku_release_calldata,
4561 };
4562
4563 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4564                 struct nfs_open_context *ctx,
4565                 struct nfs4_lock_state *lsp,
4566                 struct nfs_seqid *seqid)
4567 {
4568         struct nfs4_unlockdata *data;
4569         struct rpc_message msg = {
4570                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4571                 .rpc_cred = ctx->cred,
4572         };
4573         struct rpc_task_setup task_setup_data = {
4574                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4575                 .rpc_message = &msg,
4576                 .callback_ops = &nfs4_locku_ops,
4577                 .workqueue = nfsiod_workqueue,
4578                 .flags = RPC_TASK_ASYNC,
4579         };
4580
4581         /* Ensure this is an unlock - when canceling a lock, the
4582          * canceled lock is passed in, and it won't be an unlock.
4583          */
4584         fl->fl_type = F_UNLCK;
4585
4586         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4587         if (data == NULL) {
4588                 nfs_free_seqid(seqid);
4589                 return ERR_PTR(-ENOMEM);
4590         }
4591
4592         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4593         msg.rpc_argp = &data->arg;
4594         msg.rpc_resp = &data->res;
4595         task_setup_data.callback_data = data;
4596         return rpc_run_task(&task_setup_data);
4597 }
4598
4599 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4600 {
4601         struct nfs_inode *nfsi = NFS_I(state->inode);
4602         struct nfs_seqid *seqid;
4603         struct nfs4_lock_state *lsp;
4604         struct rpc_task *task;
4605         int status = 0;
4606         unsigned char fl_flags = request->fl_flags;
4607
4608         status = nfs4_set_lock_state(state, request);
4609         /* Unlock _before_ we do the RPC call */
4610         request->fl_flags |= FL_EXISTS;
4611         down_read(&nfsi->rwsem);
4612         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4613                 up_read(&nfsi->rwsem);
4614                 goto out;
4615         }
4616         up_read(&nfsi->rwsem);
4617         if (status != 0)
4618                 goto out;
4619         /* Is this a delegated lock? */
4620         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4621                 goto out;
4622         lsp = request->fl_u.nfs4_fl.owner;
4623         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4624         status = -ENOMEM;
4625         if (seqid == NULL)
4626                 goto out;
4627         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4628         status = PTR_ERR(task);
4629         if (IS_ERR(task))
4630                 goto out;
4631         status = nfs4_wait_for_completion_rpc_task(task);
4632         rpc_put_task(task);
4633 out:
4634         request->fl_flags = fl_flags;
4635         return status;
4636 }
4637
4638 struct nfs4_lockdata {
4639         struct nfs_lock_args arg;
4640         struct nfs_lock_res res;
4641         struct nfs4_lock_state *lsp;
4642         struct nfs_open_context *ctx;
4643         struct file_lock fl;
4644         unsigned long timestamp;
4645         int rpc_status;
4646         int cancelled;
4647         struct nfs_server *server;
4648 };
4649
4650 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4651                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4652                 gfp_t gfp_mask)
4653 {
4654         struct nfs4_lockdata *p;
4655         struct inode *inode = lsp->ls_state->inode;
4656         struct nfs_server *server = NFS_SERVER(inode);
4657
4658         p = kzalloc(sizeof(*p), gfp_mask);
4659         if (p == NULL)
4660                 return NULL;
4661
4662         p->arg.fh = NFS_FH(inode);
4663         p->arg.fl = &p->fl;
4664         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4665         if (p->arg.open_seqid == NULL)
4666                 goto out_free;
4667         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4668         if (p->arg.lock_seqid == NULL)
4669                 goto out_free_seqid;
4670         p->arg.lock_stateid = &lsp->ls_stateid;
4671         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4672         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4673         p->arg.lock_owner.s_dev = server->s_dev;
4674         p->res.lock_seqid = p->arg.lock_seqid;
4675         p->lsp = lsp;
4676         p->server = server;
4677         atomic_inc(&lsp->ls_count);
4678         p->ctx = get_nfs_open_context(ctx);
4679         memcpy(&p->fl, fl, sizeof(p->fl));
4680         return p;
4681 out_free_seqid:
4682         nfs_free_seqid(p->arg.open_seqid);
4683 out_free:
4684         kfree(p);
4685         return NULL;
4686 }
4687
4688 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4689 {
4690         struct nfs4_lockdata *data = calldata;
4691         struct nfs4_state *state = data->lsp->ls_state;
4692
4693         dprintk("%s: begin!\n", __func__);
4694         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4695                 return;
4696         /* Do we need to do an open_to_lock_owner? */
4697         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4698                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4699                         goto out_release_lock_seqid;
4700                 data->arg.open_stateid = &state->stateid;
4701                 data->arg.new_lock_owner = 1;
4702                 data->res.open_seqid = data->arg.open_seqid;
4703         } else
4704                 data->arg.new_lock_owner = 0;
4705         data->timestamp = jiffies;
4706         if (nfs4_setup_sequence(data->server,
4707                                 &data->arg.seq_args,
4708                                 &data->res.seq_res,
4709                                 task) == 0) {
4710                 rpc_call_start(task);
4711                 return;
4712         }
4713         nfs_release_seqid(data->arg.open_seqid);
4714 out_release_lock_seqid:
4715         nfs_release_seqid(data->arg.lock_seqid);
4716         dprintk("%s: done!, ret = %d\n", __func__, task->tk_status);
4717 }
4718
4719 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4720 {
4721         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4722         nfs4_lock_prepare(task, calldata);
4723 }
4724
4725 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4726 {
4727         struct nfs4_lockdata *data = calldata;
4728
4729         dprintk("%s: begin!\n", __func__);
4730
4731         if (!nfs4_sequence_done(task, &data->res.seq_res))
4732                 return;
4733
4734         data->rpc_status = task->tk_status;
4735         if (data->arg.new_lock_owner != 0) {
4736                 if (data->rpc_status == 0)
4737                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4738                 else
4739                         goto out;
4740         }
4741         if (data->rpc_status == 0) {
4742                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4743                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4744                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4745         }
4746 out:
4747         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4748 }
4749
4750 static void nfs4_lock_release(void *calldata)
4751 {
4752         struct nfs4_lockdata *data = calldata;
4753
4754         dprintk("%s: begin!\n", __func__);
4755         nfs_free_seqid(data->arg.open_seqid);
4756         if (data->cancelled != 0) {
4757                 struct rpc_task *task;
4758                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4759                                 data->arg.lock_seqid);
4760                 if (!IS_ERR(task))
4761                         rpc_put_task_async(task);
4762                 dprintk("%s: cancelling lock!\n", __func__);
4763         } else
4764                 nfs_free_seqid(data->arg.lock_seqid);
4765         nfs4_put_lock_state(data->lsp);
4766         put_nfs_open_context(data->ctx);
4767         kfree(data);
4768         dprintk("%s: done!\n", __func__);
4769 }
4770
4771 static const struct rpc_call_ops nfs4_lock_ops = {
4772         .rpc_call_prepare = nfs4_lock_prepare,
4773         .rpc_call_done = nfs4_lock_done,
4774         .rpc_release = nfs4_lock_release,
4775 };
4776
4777 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4778         .rpc_call_prepare = nfs4_recover_lock_prepare,
4779         .rpc_call_done = nfs4_lock_done,
4780         .rpc_release = nfs4_lock_release,
4781 };
4782
4783 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4784 {
4785         switch (error) {
4786         case -NFS4ERR_ADMIN_REVOKED:
4787         case -NFS4ERR_BAD_STATEID:
4788                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4789                 if (new_lock_owner != 0 ||
4790                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4791                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4792                 break;
4793         case -NFS4ERR_STALE_STATEID:
4794                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4795         case -NFS4ERR_EXPIRED:
4796                 nfs4_schedule_lease_recovery(server->nfs_client);
4797         };
4798 }
4799
4800 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4801 {
4802         struct nfs4_lockdata *data;
4803         struct rpc_task *task;
4804         struct rpc_message msg = {
4805                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4806                 .rpc_cred = state->owner->so_cred,
4807         };
4808         struct rpc_task_setup task_setup_data = {
4809                 .rpc_client = NFS_CLIENT(state->inode),
4810                 .rpc_message = &msg,
4811                 .callback_ops = &nfs4_lock_ops,
4812                 .workqueue = nfsiod_workqueue,
4813                 .flags = RPC_TASK_ASYNC,
4814         };
4815         int ret;
4816
4817         dprintk("%s: begin!\n", __func__);
4818         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4819                         fl->fl_u.nfs4_fl.owner,
4820                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4821         if (data == NULL)
4822                 return -ENOMEM;
4823         if (IS_SETLKW(cmd))
4824                 data->arg.block = 1;
4825         if (recovery_type > NFS_LOCK_NEW) {
4826                 if (recovery_type == NFS_LOCK_RECLAIM)
4827                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4828                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4829         }
4830         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4831         msg.rpc_argp = &data->arg;
4832         msg.rpc_resp = &data->res;
4833         task_setup_data.callback_data = data;
4834         task = rpc_run_task(&task_setup_data);
4835         if (IS_ERR(task))
4836                 return PTR_ERR(task);
4837         ret = nfs4_wait_for_completion_rpc_task(task);
4838         if (ret == 0) {
4839                 ret = data->rpc_status;
4840                 if (ret)
4841                         nfs4_handle_setlk_error(data->server, data->lsp,
4842                                         data->arg.new_lock_owner, ret);
4843         } else
4844                 data->cancelled = 1;
4845         rpc_put_task(task);
4846         dprintk("%s: done, ret = %d!\n", __func__, ret);
4847         return ret;
4848 }
4849
4850 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4851 {
4852         struct nfs_server *server = NFS_SERVER(state->inode);
4853         struct nfs4_exception exception = {
4854                 .inode = state->inode,
4855         };
4856         int err;
4857
4858         do {
4859                 /* Cache the lock if possible... */
4860                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4861                         return 0;
4862                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4863                 if (err != -NFS4ERR_DELAY)
4864                         break;
4865                 nfs4_handle_exception(server, err, &exception);
4866         } while (exception.retry);
4867         return err;
4868 }
4869
4870 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4871 {
4872         struct nfs_server *server = NFS_SERVER(state->inode);
4873         struct nfs4_exception exception = {
4874                 .inode = state->inode,
4875         };
4876         int err;
4877
4878         err = nfs4_set_lock_state(state, request);
4879         if (err != 0)
4880                 return err;
4881         do {
4882                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4883                         return 0;
4884                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4885                 switch (err) {
4886                 default:
4887                         goto out;
4888                 case -NFS4ERR_GRACE:
4889                 case -NFS4ERR_DELAY:
4890                         nfs4_handle_exception(server, err, &exception);
4891                         err = 0;
4892                 }
4893         } while (exception.retry);
4894 out:
4895         return err;
4896 }
4897
4898 #if defined(CONFIG_NFS_V4_1)
4899 /**
4900  * nfs41_check_expired_locks - possibly free a lock stateid
4901  *
4902  * @state: NFSv4 state for an inode
4903  *
4904  * Returns NFS_OK if recovery for this stateid is now finished.
4905  * Otherwise a negative NFS4ERR value is returned.
4906  */
4907 static int nfs41_check_expired_locks(struct nfs4_state *state)
4908 {
4909         int status, ret = -NFS4ERR_BAD_STATEID;
4910         struct nfs4_lock_state *lsp;
4911         struct nfs_server *server = NFS_SERVER(state->inode);
4912
4913         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4914                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
4915                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4916                         if (status != NFS_OK) {
4917                                 /* Free the stateid unless the server
4918                                  * informs us the stateid is unrecognized. */
4919                                 if (status != -NFS4ERR_BAD_STATEID)
4920                                         nfs41_free_stateid(server,
4921                                                         &lsp->ls_stateid);
4922                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
4923                                 ret = status;
4924                         }
4925                 }
4926         };
4927
4928         return ret;
4929 }
4930
4931 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4932 {
4933         int status = NFS_OK;
4934
4935         if (test_bit(LK_STATE_IN_USE, &state->flags))
4936                 status = nfs41_check_expired_locks(state);
4937         if (status != NFS_OK)
4938                 status = nfs4_lock_expired(state, request);
4939         return status;
4940 }
4941 #endif
4942
4943 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4944 {
4945         struct nfs_inode *nfsi = NFS_I(state->inode);
4946         unsigned char fl_flags = request->fl_flags;
4947         int status = -ENOLCK;
4948
4949         if ((fl_flags & FL_POSIX) &&
4950                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4951                 goto out;
4952         /* Is this a delegated open? */
4953         status = nfs4_set_lock_state(state, request);
4954         if (status != 0)
4955                 goto out;
4956         request->fl_flags |= FL_ACCESS;
4957         status = do_vfs_lock(request->fl_file, request);
4958         if (status < 0)
4959                 goto out;
4960         down_read(&nfsi->rwsem);
4961         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4962                 /* Yes: cache locks! */
4963                 /* ...but avoid races with delegation recall... */
4964                 request->fl_flags = fl_flags & ~FL_SLEEP;
4965                 status = do_vfs_lock(request->fl_file, request);
4966                 goto out_unlock;
4967         }
4968         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4969         if (status != 0)
4970                 goto out_unlock;
4971         /* Note: we always want to sleep here! */
4972         request->fl_flags = fl_flags | FL_SLEEP;
4973         if (do_vfs_lock(request->fl_file, request) < 0)
4974                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4975                         "manager!\n", __func__);
4976 out_unlock:
4977         up_read(&nfsi->rwsem);
4978 out:
4979         request->fl_flags = fl_flags;
4980         return status;
4981 }
4982
4983 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4984 {
4985         struct nfs4_exception exception = {
4986                 .state = state,
4987                 .inode = state->inode,
4988         };
4989         int err;
4990
4991         do {
4992                 err = _nfs4_proc_setlk(state, cmd, request);
4993                 if (err == -NFS4ERR_DENIED)
4994                         err = -EAGAIN;
4995                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4996                                 err, &exception);
4997         } while (exception.retry);
4998         return err;
4999 }
5000
5001 static int
5002 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5003 {
5004         struct nfs_open_context *ctx;
5005         struct nfs4_state *state;
5006         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5007         int status;
5008
5009         /* verify open state */
5010         ctx = nfs_file_open_context(filp);
5011         state = ctx->state;
5012
5013         if (request->fl_start < 0 || request->fl_end < 0)
5014                 return -EINVAL;
5015
5016         if (IS_GETLK(cmd)) {
5017                 if (state != NULL)
5018                         return nfs4_proc_getlk(state, F_GETLK, request);
5019                 return 0;
5020         }
5021
5022         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5023                 return -EINVAL;
5024
5025         if (request->fl_type == F_UNLCK) {
5026                 if (state != NULL)
5027                         return nfs4_proc_unlck(state, cmd, request);
5028                 return 0;
5029         }
5030
5031         if (state == NULL)
5032                 return -ENOLCK;
5033         /*
5034          * Don't rely on the VFS having checked the file open mode,
5035          * since it won't do this for flock() locks.
5036          */
5037         switch (request->fl_type) {
5038         case F_RDLCK:
5039                 if (!(filp->f_mode & FMODE_READ))
5040                         return -EBADF;
5041                 break;
5042         case F_WRLCK:
5043                 if (!(filp->f_mode & FMODE_WRITE))
5044                         return -EBADF;
5045         }
5046
5047         do {
5048                 status = nfs4_proc_setlk(state, cmd, request);
5049                 if ((status != -EAGAIN) || IS_SETLK(cmd))
5050                         break;
5051                 timeout = nfs4_set_lock_task_retry(timeout);
5052                 status = -ERESTARTSYS;
5053                 if (signalled())
5054                         break;
5055         } while(status < 0);
5056         return status;
5057 }
5058
5059 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
5060 {
5061         struct nfs_server *server = NFS_SERVER(state->inode);
5062         struct nfs4_exception exception = { };
5063         int err;
5064
5065         err = nfs4_set_lock_state(state, fl);
5066         if (err != 0)
5067                 goto out;
5068         do {
5069                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5070                 switch (err) {
5071                         default:
5072                                 printk(KERN_ERR "NFS: %s: unhandled error "
5073                                         "%d.\n", __func__, err);
5074                         case 0:
5075                         case -ESTALE:
5076                                 goto out;
5077                         case -NFS4ERR_EXPIRED:
5078                                 nfs4_schedule_stateid_recovery(server, state);
5079                         case -NFS4ERR_STALE_CLIENTID:
5080                         case -NFS4ERR_STALE_STATEID:
5081                                 nfs4_schedule_lease_recovery(server->nfs_client);
5082                                 goto out;
5083                         case -NFS4ERR_BADSESSION:
5084                         case -NFS4ERR_BADSLOT:
5085                         case -NFS4ERR_BAD_HIGH_SLOT:
5086                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
5087                         case -NFS4ERR_DEADSESSION:
5088                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
5089                                 goto out;
5090                         case -ERESTARTSYS:
5091                                 /*
5092                                  * The show must go on: exit, but mark the
5093                                  * stateid as needing recovery.
5094                                  */
5095                         case -NFS4ERR_DELEG_REVOKED:
5096                         case -NFS4ERR_ADMIN_REVOKED:
5097                         case -NFS4ERR_BAD_STATEID:
5098                         case -NFS4ERR_OPENMODE:
5099                                 nfs4_schedule_stateid_recovery(server, state);
5100                                 err = 0;
5101                                 goto out;
5102                         case -EKEYEXPIRED:
5103                                 /*
5104                                  * User RPCSEC_GSS context has expired.
5105                                  * We cannot recover this stateid now, so
5106                                  * skip it and allow recovery thread to
5107                                  * proceed.
5108                                  */
5109                                 err = 0;
5110                                 goto out;
5111                         case -ENOMEM:
5112                         case -NFS4ERR_DENIED:
5113                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
5114                                 err = 0;
5115                                 goto out;
5116                         case -NFS4ERR_DELAY:
5117                                 break;
5118                 }
5119                 err = nfs4_handle_exception(server, err, &exception);
5120         } while (exception.retry);
5121 out:
5122         return err;
5123 }
5124
5125 struct nfs_release_lockowner_data {
5126         struct nfs4_lock_state *lsp;
5127         struct nfs_server *server;
5128         struct nfs_release_lockowner_args args;
5129 };
5130
5131 static void nfs4_release_lockowner_release(void *calldata)
5132 {
5133         struct nfs_release_lockowner_data *data = calldata;
5134         nfs4_free_lock_state(data->server, data->lsp);
5135         kfree(calldata);
5136 }
5137
5138 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5139         .rpc_release = nfs4_release_lockowner_release,
5140 };
5141
5142 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
5143 {
5144         struct nfs_server *server = lsp->ls_state->owner->so_server;
5145         struct nfs_release_lockowner_data *data;
5146         struct rpc_message msg = {
5147                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5148         };
5149
5150         if (server->nfs_client->cl_mvops->minor_version != 0)
5151                 return -EINVAL;
5152         data = kmalloc(sizeof(*data), GFP_NOFS);
5153         if (!data)
5154                 return -ENOMEM;
5155         data->lsp = lsp;
5156         data->server = server;
5157         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5158         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5159         data->args.lock_owner.s_dev = server->s_dev;
5160         msg.rpc_argp = &data->args;
5161         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5162         return 0;
5163 }
5164
5165 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5166
5167 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5168                                    const void *buf, size_t buflen,
5169                                    int flags, int type)
5170 {
5171         if (strcmp(key, "") != 0)
5172                 return -EINVAL;
5173
5174         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5175 }
5176
5177 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5178                                    void *buf, size_t buflen, int type)
5179 {
5180         if (strcmp(key, "") != 0)
5181                 return -EINVAL;
5182
5183         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5184 }
5185
5186 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5187                                        size_t list_len, const char *name,
5188                                        size_t name_len, int type)
5189 {
5190         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5191
5192         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5193                 return 0;
5194
5195         if (list && len <= list_len)
5196                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5197         return len;
5198 }
5199
5200 /*
5201  * nfs_fhget will use either the mounted_on_fileid or the fileid
5202  */
5203 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5204 {
5205         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5206                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5207               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5208               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5209                 return;
5210
5211         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5212                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5213         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5214         fattr->nlink = 2;
5215 }
5216
5217 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5218                                    const struct qstr *name,
5219                                    struct nfs4_fs_locations *fs_locations,
5220                                    struct page *page)
5221 {
5222         struct nfs_server *server = NFS_SERVER(dir);
5223         u32 bitmask[2] = {
5224                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5225         };
5226         struct nfs4_fs_locations_arg args = {
5227                 .dir_fh = NFS_FH(dir),
5228                 .name = name,
5229                 .page = page,
5230                 .bitmask = bitmask,
5231         };
5232         struct nfs4_fs_locations_res res = {
5233                 .fs_locations = fs_locations,
5234         };
5235         struct rpc_message msg = {
5236                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5237                 .rpc_argp = &args,
5238                 .rpc_resp = &res,
5239         };
5240         int status;
5241
5242         dprintk("%s: start\n", __func__);
5243
5244         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5245          * is not supported */
5246         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5247                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5248         else
5249                 bitmask[0] |= FATTR4_WORD0_FILEID;
5250
5251         nfs_fattr_init(&fs_locations->fattr);
5252         fs_locations->server = server;
5253         fs_locations->nlocations = 0;
5254         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5255         dprintk("%s: returned status = %d\n", __func__, status);
5256         return status;
5257 }
5258
5259 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5260                            const struct qstr *name,
5261                            struct nfs4_fs_locations *fs_locations,
5262                            struct page *page)
5263 {
5264         struct nfs4_exception exception = { };
5265         int err;
5266         do {
5267                 err = nfs4_handle_exception(NFS_SERVER(dir),
5268                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5269                                 &exception);
5270         } while (exception.retry);
5271         return err;
5272 }
5273
5274 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5275 {
5276         int status;
5277         struct nfs4_secinfo_arg args = {
5278                 .dir_fh = NFS_FH(dir),
5279                 .name   = name,
5280         };
5281         struct nfs4_secinfo_res res = {
5282                 .flavors     = flavors,
5283         };
5284         struct rpc_message msg = {
5285                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5286                 .rpc_argp = &args,
5287                 .rpc_resp = &res,
5288         };
5289
5290         dprintk("NFS call  secinfo %s\n", name->name);
5291         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5292         dprintk("NFS reply  secinfo: %d\n", status);
5293         return status;
5294 }
5295
5296 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5297                       struct nfs4_secinfo_flavors *flavors)
5298 {
5299         struct nfs4_exception exception = { };
5300         int err;
5301         do {
5302                 err = nfs4_handle_exception(NFS_SERVER(dir),
5303                                 _nfs4_proc_secinfo(dir, name, flavors),
5304                                 &exception);
5305         } while (exception.retry);
5306         return err;
5307 }
5308
5309 #ifdef CONFIG_NFS_V4_1
5310 /*
5311  * Check the exchange flags returned by the server for invalid flags, having
5312  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5313  * DS flags set.
5314  */
5315 static int nfs4_check_cl_exchange_flags(u32 flags)
5316 {
5317         if (flags & ~EXCHGID4_FLAG_MASK_R)
5318                 goto out_inval;
5319         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5320             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5321                 goto out_inval;
5322         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5323                 goto out_inval;
5324         return NFS_OK;
5325 out_inval:
5326         return -NFS4ERR_INVAL;
5327 }
5328
5329 static bool
5330 nfs41_same_server_scope(struct nfs41_server_scope *a,
5331                         struct nfs41_server_scope *b)
5332 {
5333         if (a->server_scope_sz == b->server_scope_sz &&
5334             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5335                 return true;
5336
5337         return false;
5338 }
5339
5340 /*
5341  * nfs4_proc_bind_conn_to_session()
5342  *
5343  * The 4.1 client currently uses the same TCP connection for the
5344  * fore and backchannel.
5345  */
5346 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5347 {
5348         int status;
5349         struct nfs41_bind_conn_to_session_res res;
5350         struct rpc_message msg = {
5351                 .rpc_proc =
5352                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5353                 .rpc_argp = clp,
5354                 .rpc_resp = &res,
5355                 .rpc_cred = cred,
5356         };
5357
5358         dprintk("--> %s\n", __func__);
5359
5360         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5361         if (unlikely(res.session == NULL)) {
5362                 status = -ENOMEM;
5363                 goto out;
5364         }
5365
5366         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5367         if (status == 0) {
5368                 if (memcmp(res.session->sess_id.data,
5369                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5370                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5371                         status = -EIO;
5372                         goto out_session;
5373                 }
5374                 if (res.dir != NFS4_CDFS4_BOTH) {
5375                         dprintk("NFS: %s: Unexpected direction from server\n",
5376                                 __func__);
5377                         status = -EIO;
5378                         goto out_session;
5379                 }
5380                 if (res.use_conn_in_rdma_mode) {
5381                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5382                                 __func__);
5383                         status = -EIO;
5384                         goto out_session;
5385                 }
5386         }
5387 out_session:
5388         kfree(res.session);
5389 out:
5390         dprintk("<-- %s status= %d\n", __func__, status);
5391         return status;
5392 }
5393
5394 /*
5395  * nfs4_proc_exchange_id()
5396  *
5397  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5398  *
5399  * Since the clientid has expired, all compounds using sessions
5400  * associated with the stale clientid will be returning
5401  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5402  * be in some phase of session reset.
5403  */
5404 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5405 {
5406         nfs4_verifier verifier;
5407         struct nfs41_exchange_id_args args = {
5408                 .verifier = &verifier,
5409                 .client = clp,
5410                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5411         };
5412         struct nfs41_exchange_id_res res = {
5413                 0
5414         };
5415         int status;
5416         struct rpc_message msg = {
5417                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5418                 .rpc_argp = &args,
5419                 .rpc_resp = &res,
5420                 .rpc_cred = cred,
5421         };
5422
5423         nfs4_init_boot_verifier(clp, &verifier);
5424         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5425                                                         sizeof(args.id));
5426         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5427                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5428                 args.id_len, args.id);
5429
5430         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5431                                         GFP_NOFS);
5432         if (unlikely(res.server_owner == NULL)) {
5433                 status = -ENOMEM;
5434                 goto out;
5435         }
5436
5437         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5438                                         GFP_NOFS);
5439         if (unlikely(res.server_scope == NULL)) {
5440                 status = -ENOMEM;
5441                 goto out_server_owner;
5442         }
5443
5444         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5445         if (unlikely(res.impl_id == NULL)) {
5446                 status = -ENOMEM;
5447                 goto out_server_scope;
5448         }
5449
5450         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5451         if (status == 0)
5452                 status = nfs4_check_cl_exchange_flags(res.flags);
5453
5454         if (status == 0) {
5455                 clp->cl_clientid = res.clientid;
5456                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5457                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5458                         clp->cl_seqid = res.seqid;
5459
5460                 kfree(clp->cl_serverowner);
5461                 clp->cl_serverowner = res.server_owner;
5462                 res.server_owner = NULL;
5463
5464                 /* use the most recent implementation id */
5465                 kfree(clp->cl_implid);
5466                 clp->cl_implid = res.impl_id;
5467
5468                 if (clp->cl_serverscope != NULL &&
5469                     !nfs41_same_server_scope(clp->cl_serverscope,
5470                                              res.server_scope)) {
5471                         dprintk("%s: server_scope mismatch detected\n",
5472                                 __func__);
5473                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5474                         kfree(clp->cl_serverscope);
5475                         clp->cl_serverscope = NULL;
5476                 }
5477
5478                 if (clp->cl_serverscope == NULL) {
5479                         clp->cl_serverscope = res.server_scope;
5480                         goto out;
5481                 }
5482         } else
5483                 kfree(res.impl_id);
5484
5485 out_server_owner:
5486         kfree(res.server_owner);
5487 out_server_scope:
5488         kfree(res.server_scope);
5489 out:
5490         if (clp->cl_implid != NULL)
5491                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5492                         "domain: %s, name: %s, date: %llu,%u\n",
5493                         clp->cl_implid->domain, clp->cl_implid->name,
5494                         clp->cl_implid->date.seconds,
5495                         clp->cl_implid->date.nseconds);
5496         dprintk("NFS reply exchange_id: %d\n", status);
5497         return status;
5498 }
5499
5500 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5501                 struct rpc_cred *cred)
5502 {
5503         struct rpc_message msg = {
5504                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5505                 .rpc_argp = clp,
5506                 .rpc_cred = cred,
5507         };
5508         int status;
5509
5510         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5511         if (status)
5512                 dprintk("NFS: Got error %d from the server %s on "
5513                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5514         return status;
5515 }
5516
5517 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5518                 struct rpc_cred *cred)
5519 {
5520         unsigned int loop;
5521         int ret;
5522
5523         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5524                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5525                 switch (ret) {
5526                 case -NFS4ERR_DELAY:
5527                 case -NFS4ERR_CLIENTID_BUSY:
5528                         ssleep(1);
5529                         break;
5530                 default:
5531                         return ret;
5532                 }
5533         }
5534         return 0;
5535 }
5536
5537 int nfs4_destroy_clientid(struct nfs_client *clp)
5538 {
5539         struct rpc_cred *cred;
5540         int ret = 0;
5541
5542         if (clp->cl_mvops->minor_version < 1)
5543                 goto out;
5544         if (clp->cl_exchange_flags == 0)
5545                 goto out;
5546         if (clp->cl_preserve_clid)
5547                 goto out;
5548         cred = nfs4_get_exchange_id_cred(clp);
5549         ret = nfs4_proc_destroy_clientid(clp, cred);
5550         if (cred)
5551                 put_rpccred(cred);
5552         switch (ret) {
5553         case 0:
5554         case -NFS4ERR_STALE_CLIENTID:
5555                 clp->cl_exchange_flags = 0;
5556         }
5557 out:
5558         return ret;
5559 }
5560
5561 struct nfs4_get_lease_time_data {
5562         struct nfs4_get_lease_time_args *args;
5563         struct nfs4_get_lease_time_res *res;
5564         struct nfs_client *clp;
5565 };
5566
5567 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5568                                         void *calldata)
5569 {
5570         int ret;
5571         struct nfs4_get_lease_time_data *data =
5572                         (struct nfs4_get_lease_time_data *)calldata;
5573
5574         dprintk("--> %s\n", __func__);
5575         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5576         /* just setup sequence, do not trigger session recovery
5577            since we're invoked within one */
5578         ret = nfs41_setup_sequence(data->clp->cl_session,
5579                                    &data->args->la_seq_args,
5580                                    &data->res->lr_seq_res, task);
5581
5582         if (ret != -EAGAIN)
5583                 rpc_call_start(task);
5584         dprintk("<-- %s\n", __func__);
5585 }
5586
5587 /*
5588  * Called from nfs4_state_manager thread for session setup, so don't recover
5589  * from sequence operation or clientid errors.
5590  */
5591 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5592 {
5593         struct nfs4_get_lease_time_data *data =
5594                         (struct nfs4_get_lease_time_data *)calldata;
5595
5596         dprintk("--> %s\n", __func__);
5597         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5598                 return;
5599         switch (task->tk_status) {
5600         case -NFS4ERR_DELAY:
5601         case -NFS4ERR_GRACE:
5602                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5603                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5604                 task->tk_status = 0;
5605                 /* fall through */
5606         case -NFS4ERR_RETRY_UNCACHED_REP:
5607                 rpc_restart_call_prepare(task);
5608                 return;
5609         }
5610         dprintk("<-- %s\n", __func__);
5611 }
5612
5613 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5614         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5615         .rpc_call_done = nfs4_get_lease_time_done,
5616 };
5617
5618 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5619 {
5620         struct rpc_task *task;
5621         struct nfs4_get_lease_time_args args;
5622         struct nfs4_get_lease_time_res res = {
5623                 .lr_fsinfo = fsinfo,
5624         };
5625         struct nfs4_get_lease_time_data data = {
5626                 .args = &args,
5627                 .res = &res,
5628                 .clp = clp,
5629         };
5630         struct rpc_message msg = {
5631                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5632                 .rpc_argp = &args,
5633                 .rpc_resp = &res,
5634         };
5635         struct rpc_task_setup task_setup = {
5636                 .rpc_client = clp->cl_rpcclient,
5637                 .rpc_message = &msg,
5638                 .callback_ops = &nfs4_get_lease_time_ops,
5639                 .callback_data = &data,
5640                 .flags = RPC_TASK_TIMEOUT,
5641         };
5642         int status;
5643
5644         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5645         dprintk("--> %s\n", __func__);
5646         task = rpc_run_task(&task_setup);
5647
5648         if (IS_ERR(task))
5649                 status = PTR_ERR(task);
5650         else {
5651                 status = task->tk_status;
5652                 rpc_put_task(task);
5653         }
5654         dprintk("<-- %s return %d\n", __func__, status);
5655
5656         return status;
5657 }
5658
5659 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5660 {
5661         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5662 }
5663
5664 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5665                 struct nfs4_slot *new,
5666                 u32 max_slots,
5667                 u32 ivalue)
5668 {
5669         struct nfs4_slot *old = NULL;
5670         u32 i;
5671
5672         spin_lock(&tbl->slot_tbl_lock);
5673         if (new) {
5674                 old = tbl->slots;
5675                 tbl->slots = new;
5676                 tbl->max_slots = max_slots;
5677         }
5678         tbl->highest_used_slotid = NFS4_NO_SLOT;
5679         for (i = 0; i < tbl->max_slots; i++)
5680                 tbl->slots[i].seq_nr = ivalue;
5681         spin_unlock(&tbl->slot_tbl_lock);
5682         kfree(old);
5683 }
5684
5685 /*
5686  * (re)Initialise a slot table
5687  */
5688 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5689                                  u32 ivalue)
5690 {
5691         struct nfs4_slot *new = NULL;
5692         int ret = -ENOMEM;
5693
5694         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5695                 max_reqs, tbl->max_slots);
5696
5697         /* Does the newly negotiated max_reqs match the existing slot table? */
5698         if (max_reqs != tbl->max_slots) {
5699                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5700                 if (!new)
5701                         goto out;
5702         }
5703         ret = 0;
5704
5705         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5706         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5707                 tbl, tbl->slots, tbl->max_slots);
5708 out:
5709         dprintk("<-- %s: return %d\n", __func__, ret);
5710         return ret;
5711 }
5712
5713 /* Destroy the slot table */
5714 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5715 {
5716         if (session->fc_slot_table.slots != NULL) {
5717                 kfree(session->fc_slot_table.slots);
5718                 session->fc_slot_table.slots = NULL;
5719         }
5720         if (session->bc_slot_table.slots != NULL) {
5721                 kfree(session->bc_slot_table.slots);
5722                 session->bc_slot_table.slots = NULL;
5723         }
5724         return;
5725 }
5726
5727 /*
5728  * Initialize or reset the forechannel and backchannel tables
5729  */
5730 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5731 {
5732         struct nfs4_slot_table *tbl;
5733         int status;
5734
5735         dprintk("--> %s\n", __func__);
5736         /* Fore channel */
5737         tbl = &ses->fc_slot_table;
5738         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5739         if (status) /* -ENOMEM */
5740                 return status;
5741         /* Back channel */
5742         tbl = &ses->bc_slot_table;
5743         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5744         if (status && tbl->slots == NULL)
5745                 /* Fore and back channel share a connection so get
5746                  * both slot tables or neither */
5747                 nfs4_destroy_slot_tables(ses);
5748         return status;
5749 }
5750
5751 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5752 {
5753         struct nfs4_session *session;
5754         struct nfs4_slot_table *tbl;
5755
5756         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5757         if (!session)
5758                 return NULL;
5759
5760         tbl = &session->fc_slot_table;
5761         tbl->highest_used_slotid = NFS4_NO_SLOT;
5762         spin_lock_init(&tbl->slot_tbl_lock);
5763         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5764         init_completion(&tbl->complete);
5765
5766         tbl = &session->bc_slot_table;
5767         tbl->highest_used_slotid = NFS4_NO_SLOT;
5768         spin_lock_init(&tbl->slot_tbl_lock);
5769         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5770         init_completion(&tbl->complete);
5771
5772         session->session_state = 1<<NFS4_SESSION_INITING;
5773
5774         session->clp = clp;
5775         return session;
5776 }
5777
5778 void nfs4_destroy_session(struct nfs4_session *session)
5779 {
5780         struct rpc_xprt *xprt;
5781         struct rpc_cred *cred;
5782
5783         cred = nfs4_get_exchange_id_cred(session->clp);
5784         nfs4_proc_destroy_session(session, cred);
5785         if (cred)
5786                 put_rpccred(cred);
5787
5788         rcu_read_lock();
5789         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5790         rcu_read_unlock();
5791         dprintk("%s Destroy backchannel for xprt %p\n",
5792                 __func__, xprt);
5793         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5794         nfs4_destroy_slot_tables(session);
5795         kfree(session);
5796 }
5797
5798 /*
5799  * Initialize the values to be used by the client in CREATE_SESSION
5800  * If nfs4_init_session set the fore channel request and response sizes,
5801  * use them.
5802  *
5803  * Set the back channel max_resp_sz_cached to zero to force the client to
5804  * always set csa_cachethis to FALSE because the current implementation
5805  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5806  */
5807 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5808 {
5809         struct nfs4_session *session = args->client->cl_session;
5810         unsigned int mxrqst_sz = session->fc_target_max_rqst_sz,
5811                      mxresp_sz = session->fc_target_max_resp_sz;
5812
5813         if (mxrqst_sz == 0)
5814                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5815         if (mxresp_sz == 0)
5816                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5817         /* Fore channel attributes */
5818         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5819         args->fc_attrs.max_resp_sz = mxresp_sz;
5820         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5821         args->fc_attrs.max_reqs = max_session_slots;
5822
5823         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5824                 "max_ops=%u max_reqs=%u\n",
5825                 __func__,
5826                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5827                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5828
5829         /* Back channel attributes */
5830         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5831         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5832         args->bc_attrs.max_resp_sz_cached = 0;
5833         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5834         args->bc_attrs.max_reqs = 1;
5835
5836         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5837                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5838                 __func__,
5839                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5840                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5841                 args->bc_attrs.max_reqs);
5842 }
5843
5844 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5845 {
5846         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5847         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5848
5849         if (rcvd->max_resp_sz > sent->max_resp_sz)
5850                 return -EINVAL;
5851         /*
5852          * Our requested max_ops is the minimum we need; we're not
5853          * prepared to break up compounds into smaller pieces than that.
5854          * So, no point even trying to continue if the server won't
5855          * cooperate:
5856          */
5857         if (rcvd->max_ops < sent->max_ops)
5858                 return -EINVAL;
5859         if (rcvd->max_reqs == 0)
5860                 return -EINVAL;
5861         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5862                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5863         return 0;
5864 }
5865
5866 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5867 {
5868         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5869         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5870
5871         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5872                 return -EINVAL;
5873         if (rcvd->max_resp_sz < sent->max_resp_sz)
5874                 return -EINVAL;
5875         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5876                 return -EINVAL;
5877         /* These would render the backchannel useless: */
5878         if (rcvd->max_ops != sent->max_ops)
5879                 return -EINVAL;
5880         if (rcvd->max_reqs != sent->max_reqs)
5881                 return -EINVAL;
5882         return 0;
5883 }
5884
5885 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5886                                      struct nfs4_session *session)
5887 {
5888         int ret;
5889
5890         ret = nfs4_verify_fore_channel_attrs(args, session);
5891         if (ret)
5892                 return ret;
5893         return nfs4_verify_back_channel_attrs(args, session);
5894 }
5895
5896 static int _nfs4_proc_create_session(struct nfs_client *clp,
5897                 struct rpc_cred *cred)
5898 {
5899         struct nfs4_session *session = clp->cl_session;
5900         struct nfs41_create_session_args args = {
5901                 .client = clp,
5902                 .cb_program = NFS4_CALLBACK,
5903         };
5904         struct nfs41_create_session_res res = {
5905                 .client = clp,
5906         };
5907         struct rpc_message msg = {
5908                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5909                 .rpc_argp = &args,
5910                 .rpc_resp = &res,
5911                 .rpc_cred = cred,
5912         };
5913         int status;
5914
5915         nfs4_init_channel_attrs(&args);
5916         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5917
5918         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5919
5920         if (!status) {
5921                 /* Verify the session's negotiated channel_attrs values */
5922                 status = nfs4_verify_channel_attrs(&args, session);
5923                 /* Increment the clientid slot sequence id */
5924                 clp->cl_seqid++;
5925         }
5926
5927         return status;
5928 }
5929
5930 /*
5931  * Issues a CREATE_SESSION operation to the server.
5932  * It is the responsibility of the caller to verify the session is
5933  * expired before calling this routine.
5934  */
5935 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5936 {
5937         int status;
5938         unsigned *ptr;
5939         struct nfs4_session *session = clp->cl_session;
5940
5941         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5942
5943         status = _nfs4_proc_create_session(clp, cred);
5944         if (status)
5945                 goto out;
5946
5947         /* Init or reset the session slot tables */
5948         status = nfs4_setup_session_slot_tables(session);
5949         dprintk("slot table setup returned %d\n", status);
5950         if (status)
5951                 goto out;
5952
5953         ptr = (unsigned *)&session->sess_id.data[0];
5954         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5955                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5956 out:
5957         dprintk("<-- %s\n", __func__);
5958         return status;
5959 }
5960
5961 /*
5962  * Issue the over-the-wire RPC DESTROY_SESSION.
5963  * The caller must serialize access to this routine.
5964  */
5965 int nfs4_proc_destroy_session(struct nfs4_session *session,
5966                 struct rpc_cred *cred)
5967 {
5968         struct rpc_message msg = {
5969                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5970                 .rpc_argp = session,
5971                 .rpc_cred = cred,
5972         };
5973         int status = 0;
5974
5975         dprintk("--> nfs4_proc_destroy_session\n");
5976
5977         /* session is still being setup */
5978         if (session->clp->cl_cons_state != NFS_CS_READY)
5979                 return status;
5980
5981         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5982
5983         if (status)
5984                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5985                         "Session has been destroyed regardless...\n", status);
5986
5987         dprintk("<-- nfs4_proc_destroy_session\n");
5988         return status;
5989 }
5990
5991 /*
5992  * With sessions, the client is not marked ready until after a
5993  * successful EXCHANGE_ID and CREATE_SESSION.
5994  *
5995  * Map errors cl_cons_state errors to EPROTONOSUPPORT to indicate
5996  * other versions of NFS can be tried.
5997  */
5998 static int nfs41_check_session_ready(struct nfs_client *clp)
5999 {
6000         int ret;
6001
6002         if (clp->cl_cons_state == NFS_CS_SESSION_INITING) {
6003                 ret = nfs4_client_recover_expired_lease(clp);
6004                 if (ret)
6005                         return ret;
6006         }
6007         if (clp->cl_cons_state < NFS_CS_READY)
6008                 return -EPROTONOSUPPORT;
6009         smp_rmb();
6010         return 0;
6011 }
6012
6013 int nfs4_init_session(struct nfs_server *server)
6014 {
6015         struct nfs_client *clp = server->nfs_client;
6016         struct nfs4_session *session;
6017         unsigned int target_max_rqst_sz = NFS_MAX_FILE_IO_SIZE;
6018         unsigned int target_max_resp_sz = NFS_MAX_FILE_IO_SIZE;
6019
6020         if (!nfs4_has_session(clp))
6021                 return 0;
6022
6023         if (server->rsize != 0)
6024                 target_max_resp_sz = server->rsize;
6025         target_max_resp_sz += nfs41_maxread_overhead;
6026
6027         if (server->wsize != 0)
6028                 target_max_rqst_sz = server->wsize;
6029         target_max_rqst_sz += nfs41_maxwrite_overhead;
6030
6031         session = clp->cl_session;
6032         spin_lock(&clp->cl_lock);
6033         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
6034                 /* Initialise targets and channel attributes */
6035                 session->fc_target_max_rqst_sz = target_max_rqst_sz;
6036                 session->fc_attrs.max_rqst_sz = target_max_rqst_sz;
6037                 session->fc_target_max_resp_sz = target_max_resp_sz;
6038                 session->fc_attrs.max_resp_sz = target_max_resp_sz;
6039         } else {
6040                 /* Just adjust the targets */
6041                 if (target_max_rqst_sz > session->fc_target_max_rqst_sz) {
6042                         session->fc_target_max_rqst_sz = target_max_rqst_sz;
6043                         set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
6044                 }
6045                 if (target_max_resp_sz > session->fc_target_max_resp_sz) {
6046                         session->fc_target_max_resp_sz = target_max_resp_sz;
6047                         set_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state);
6048                 }
6049         }
6050         spin_unlock(&clp->cl_lock);
6051
6052         if (test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
6053                 nfs4_schedule_lease_recovery(clp);
6054
6055         return nfs41_check_session_ready(clp);
6056 }
6057
6058 int nfs4_init_ds_session(struct nfs_client *clp, unsigned long lease_time)
6059 {
6060         struct nfs4_session *session = clp->cl_session;
6061         int ret;
6062
6063         spin_lock(&clp->cl_lock);
6064         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
6065                 /*
6066                  * Do not set NFS_CS_CHECK_LEASE_TIME instead set the
6067                  * DS lease to be equal to the MDS lease.
6068                  */
6069                 clp->cl_lease_time = lease_time;
6070                 clp->cl_last_renewal = jiffies;
6071         }
6072         spin_unlock(&clp->cl_lock);
6073
6074         ret = nfs41_check_session_ready(clp);
6075         if (ret)
6076                 return ret;
6077         /* Test for the DS role */
6078         if (!is_ds_client(clp))
6079                 return -ENODEV;
6080         return 0;
6081 }
6082 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
6083
6084
6085 /*
6086  * Renew the cl_session lease.
6087  */
6088 struct nfs4_sequence_data {
6089         struct nfs_client *clp;
6090         struct nfs4_sequence_args args;
6091         struct nfs4_sequence_res res;
6092 };
6093
6094 static void nfs41_sequence_release(void *data)
6095 {
6096         struct nfs4_sequence_data *calldata = data;
6097         struct nfs_client *clp = calldata->clp;
6098
6099         if (atomic_read(&clp->cl_count) > 1)
6100                 nfs4_schedule_state_renewal(clp);
6101         nfs_put_client(clp);
6102         kfree(calldata);
6103 }
6104
6105 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6106 {
6107         switch(task->tk_status) {
6108         case -NFS4ERR_DELAY:
6109                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6110                 return -EAGAIN;
6111         default:
6112                 nfs4_schedule_lease_recovery(clp);
6113         }
6114         return 0;
6115 }
6116
6117 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
6118 {
6119         struct nfs4_sequence_data *calldata = data;
6120         struct nfs_client *clp = calldata->clp;
6121
6122         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
6123                 return;
6124
6125         if (task->tk_status < 0) {
6126                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
6127                 if (atomic_read(&clp->cl_count) == 1)
6128                         goto out;
6129
6130                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
6131                         rpc_restart_call_prepare(task);
6132                         return;
6133                 }
6134         }
6135         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
6136 out:
6137         dprintk("<-- %s\n", __func__);
6138 }
6139
6140 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
6141 {
6142         struct nfs4_sequence_data *calldata = data;
6143         struct nfs_client *clp = calldata->clp;
6144         struct nfs4_sequence_args *args;
6145         struct nfs4_sequence_res *res;
6146
6147         args = task->tk_msg.rpc_argp;
6148         res = task->tk_msg.rpc_resp;
6149
6150         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
6151                 return;
6152         rpc_call_start(task);
6153 }
6154
6155 static void nfs41_sequence_prepare_privileged(struct rpc_task *task, void *data)
6156 {
6157         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
6158         nfs41_sequence_prepare(task, data);
6159 }
6160
6161 static const struct rpc_call_ops nfs41_sequence_ops = {
6162         .rpc_call_done = nfs41_sequence_call_done,
6163         .rpc_call_prepare = nfs41_sequence_prepare,
6164         .rpc_release = nfs41_sequence_release,
6165 };
6166
6167 static const struct rpc_call_ops nfs41_sequence_privileged_ops = {
6168         .rpc_call_done = nfs41_sequence_call_done,
6169         .rpc_call_prepare = nfs41_sequence_prepare_privileged,
6170         .rpc_release = nfs41_sequence_release,
6171 };
6172
6173 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred,
6174                                              const struct rpc_call_ops *seq_ops)
6175 {
6176         struct nfs4_sequence_data *calldata;
6177         struct rpc_message msg = {
6178                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
6179                 .rpc_cred = cred,
6180         };
6181         struct rpc_task_setup task_setup_data = {
6182                 .rpc_client = clp->cl_rpcclient,
6183                 .rpc_message = &msg,
6184                 .callback_ops = seq_ops,
6185                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
6186         };
6187
6188         if (!atomic_inc_not_zero(&clp->cl_count))
6189                 return ERR_PTR(-EIO);
6190         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6191         if (calldata == NULL) {
6192                 nfs_put_client(clp);
6193                 return ERR_PTR(-ENOMEM);
6194         }
6195         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6196         msg.rpc_argp = &calldata->args;
6197         msg.rpc_resp = &calldata->res;
6198         calldata->clp = clp;
6199         task_setup_data.callback_data = calldata;
6200
6201         return rpc_run_task(&task_setup_data);
6202 }
6203
6204 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6205 {
6206         struct rpc_task *task;
6207         int ret = 0;
6208
6209         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6210                 return 0;
6211         task = _nfs41_proc_sequence(clp, cred, &nfs41_sequence_ops);
6212         if (IS_ERR(task))
6213                 ret = PTR_ERR(task);
6214         else
6215                 rpc_put_task_async(task);
6216         dprintk("<-- %s status=%d\n", __func__, ret);
6217         return ret;
6218 }
6219
6220 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6221 {
6222         struct rpc_task *task;
6223         int ret;
6224
6225         task = _nfs41_proc_sequence(clp, cred, &nfs41_sequence_privileged_ops);
6226         if (IS_ERR(task)) {
6227                 ret = PTR_ERR(task);
6228                 goto out;
6229         }
6230         ret = rpc_wait_for_completion_task(task);
6231         if (!ret) {
6232                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6233
6234                 if (task->tk_status == 0)
6235                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6236                 ret = task->tk_status;
6237         }
6238         rpc_put_task(task);
6239 out:
6240         dprintk("<-- %s status=%d\n", __func__, ret);
6241         return ret;
6242 }
6243
6244 struct nfs4_reclaim_complete_data {
6245         struct nfs_client *clp;
6246         struct nfs41_reclaim_complete_args arg;
6247         struct nfs41_reclaim_complete_res res;
6248 };
6249
6250 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6251 {
6252         struct nfs4_reclaim_complete_data *calldata = data;
6253
6254         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
6255         if (nfs41_setup_sequence(calldata->clp->cl_session,
6256                                 &calldata->arg.seq_args,
6257                                 &calldata->res.seq_res, task))
6258                 return;
6259
6260         rpc_call_start(task);
6261 }
6262
6263 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6264 {
6265         switch(task->tk_status) {
6266         case 0:
6267         case -NFS4ERR_COMPLETE_ALREADY:
6268         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6269                 break;
6270         case -NFS4ERR_DELAY:
6271                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6272                 /* fall through */
6273         case -NFS4ERR_RETRY_UNCACHED_REP:
6274                 return -EAGAIN;
6275         default:
6276                 nfs4_schedule_lease_recovery(clp);
6277         }
6278         return 0;
6279 }
6280
6281 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6282 {
6283         struct nfs4_reclaim_complete_data *calldata = data;
6284         struct nfs_client *clp = calldata->clp;
6285         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6286
6287         dprintk("--> %s\n", __func__);
6288         if (!nfs41_sequence_done(task, res))
6289                 return;
6290
6291         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6292                 rpc_restart_call_prepare(task);
6293                 return;
6294         }
6295         dprintk("<-- %s\n", __func__);
6296 }
6297
6298 static void nfs4_free_reclaim_complete_data(void *data)
6299 {
6300         struct nfs4_reclaim_complete_data *calldata = data;
6301
6302         kfree(calldata);
6303 }
6304
6305 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6306         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6307         .rpc_call_done = nfs4_reclaim_complete_done,
6308         .rpc_release = nfs4_free_reclaim_complete_data,
6309 };
6310
6311 /*
6312  * Issue a global reclaim complete.
6313  */
6314 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6315 {
6316         struct nfs4_reclaim_complete_data *calldata;
6317         struct rpc_task *task;
6318         struct rpc_message msg = {
6319                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6320         };
6321         struct rpc_task_setup task_setup_data = {
6322                 .rpc_client = clp->cl_rpcclient,
6323                 .rpc_message = &msg,
6324                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6325                 .flags = RPC_TASK_ASYNC,
6326         };
6327         int status = -ENOMEM;
6328
6329         dprintk("--> %s\n", __func__);
6330         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6331         if (calldata == NULL)
6332                 goto out;
6333         calldata->clp = clp;
6334         calldata->arg.one_fs = 0;
6335
6336         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6337         msg.rpc_argp = &calldata->arg;
6338         msg.rpc_resp = &calldata->res;
6339         task_setup_data.callback_data = calldata;
6340         task = rpc_run_task(&task_setup_data);
6341         if (IS_ERR(task)) {
6342                 status = PTR_ERR(task);
6343                 goto out;
6344         }
6345         status = nfs4_wait_for_completion_rpc_task(task);
6346         if (status == 0)
6347                 status = task->tk_status;
6348         rpc_put_task(task);
6349         return 0;
6350 out:
6351         dprintk("<-- %s status=%d\n", __func__, status);
6352         return status;
6353 }
6354
6355 static void
6356 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6357 {
6358         struct nfs4_layoutget *lgp = calldata;
6359         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6360
6361         dprintk("--> %s\n", __func__);
6362         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6363          * right now covering the LAYOUTGET we are about to send.
6364          * However, that is not so catastrophic, and there seems
6365          * to be no way to prevent it completely.
6366          */
6367         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
6368                                 &lgp->res.seq_res, task))
6369                 return;
6370         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6371                                           NFS_I(lgp->args.inode)->layout,
6372                                           lgp->args.ctx->state)) {
6373                 rpc_exit(task, NFS4_OK);
6374                 return;
6375         }
6376         rpc_call_start(task);
6377 }
6378
6379 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6380 {
6381         struct nfs4_layoutget *lgp = calldata;
6382         struct inode *inode = lgp->args.inode;
6383         struct nfs_server *server = NFS_SERVER(inode);
6384         struct pnfs_layout_hdr *lo;
6385         struct nfs4_state *state = NULL;
6386
6387         dprintk("--> %s\n", __func__);
6388
6389         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
6390                 goto out;
6391
6392         switch (task->tk_status) {
6393         case 0:
6394                 goto out;
6395         case -NFS4ERR_LAYOUTTRYLATER:
6396         case -NFS4ERR_RECALLCONFLICT:
6397                 task->tk_status = -NFS4ERR_DELAY;
6398                 break;
6399         case -NFS4ERR_EXPIRED:
6400         case -NFS4ERR_BAD_STATEID:
6401                 spin_lock(&inode->i_lock);
6402                 lo = NFS_I(inode)->layout;
6403                 if (!lo || list_empty(&lo->plh_segs)) {
6404                         spin_unlock(&inode->i_lock);
6405                         /* If the open stateid was bad, then recover it. */
6406                         state = lgp->args.ctx->state;
6407                 } else {
6408                         LIST_HEAD(head);
6409
6410                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6411                         spin_unlock(&inode->i_lock);
6412                         /* Mark the bad layout state as invalid, then
6413                          * retry using the open stateid. */
6414                         pnfs_free_lseg_list(&head);
6415                 }
6416         }
6417         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6418                 rpc_restart_call_prepare(task);
6419 out:
6420         dprintk("<-- %s\n", __func__);
6421 }
6422
6423 static size_t max_response_pages(struct nfs_server *server)
6424 {
6425         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6426         return nfs_page_array_len(0, max_resp_sz);
6427 }
6428
6429 static void nfs4_free_pages(struct page **pages, size_t size)
6430 {
6431         int i;
6432
6433         if (!pages)
6434                 return;
6435
6436         for (i = 0; i < size; i++) {
6437                 if (!pages[i])
6438                         break;
6439                 __free_page(pages[i]);
6440         }
6441         kfree(pages);
6442 }
6443
6444 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6445 {
6446         struct page **pages;
6447         int i;
6448
6449         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6450         if (!pages) {
6451                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6452                 return NULL;
6453         }
6454
6455         for (i = 0; i < size; i++) {
6456                 pages[i] = alloc_page(gfp_flags);
6457                 if (!pages[i]) {
6458                         dprintk("%s: failed to allocate page\n", __func__);
6459                         nfs4_free_pages(pages, size);
6460                         return NULL;
6461                 }
6462         }
6463
6464         return pages;
6465 }
6466
6467 static void nfs4_layoutget_release(void *calldata)
6468 {
6469         struct nfs4_layoutget *lgp = calldata;
6470         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6471         size_t max_pages = max_response_pages(server);
6472
6473         dprintk("--> %s\n", __func__);
6474         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6475         put_nfs_open_context(lgp->args.ctx);
6476         kfree(calldata);
6477         dprintk("<-- %s\n", __func__);
6478 }
6479
6480 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6481         .rpc_call_prepare = nfs4_layoutget_prepare,
6482         .rpc_call_done = nfs4_layoutget_done,
6483         .rpc_release = nfs4_layoutget_release,
6484 };
6485
6486 struct pnfs_layout_segment *
6487 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6488 {
6489         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6490         size_t max_pages = max_response_pages(server);
6491         struct rpc_task *task;
6492         struct rpc_message msg = {
6493                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6494                 .rpc_argp = &lgp->args,
6495                 .rpc_resp = &lgp->res,
6496         };
6497         struct rpc_task_setup task_setup_data = {
6498                 .rpc_client = server->client,
6499                 .rpc_message = &msg,
6500                 .callback_ops = &nfs4_layoutget_call_ops,
6501                 .callback_data = lgp,
6502                 .flags = RPC_TASK_ASYNC,
6503         };
6504         struct pnfs_layout_segment *lseg = NULL;
6505         int status = 0;
6506
6507         dprintk("--> %s\n", __func__);
6508
6509         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6510         if (!lgp->args.layout.pages) {
6511                 nfs4_layoutget_release(lgp);
6512                 return ERR_PTR(-ENOMEM);
6513         }
6514         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6515
6516         lgp->res.layoutp = &lgp->args.layout;
6517         lgp->res.seq_res.sr_slot = NULL;
6518         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6519         task = rpc_run_task(&task_setup_data);
6520         if (IS_ERR(task))
6521                 return ERR_CAST(task);
6522         status = nfs4_wait_for_completion_rpc_task(task);
6523         if (status == 0)
6524                 status = task->tk_status;
6525         if (status == 0)
6526                 lseg = pnfs_layout_process(lgp);
6527         rpc_put_task(task);
6528         dprintk("<-- %s status=%d\n", __func__, status);
6529         if (status)
6530                 return ERR_PTR(status);
6531         return lseg;
6532 }
6533
6534 static void
6535 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6536 {
6537         struct nfs4_layoutreturn *lrp = calldata;
6538
6539         dprintk("--> %s\n", __func__);
6540         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
6541                                 &lrp->res.seq_res, task))
6542                 return;
6543         rpc_call_start(task);
6544 }
6545
6546 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6547 {
6548         struct nfs4_layoutreturn *lrp = calldata;
6549         struct nfs_server *server;
6550
6551         dprintk("--> %s\n", __func__);
6552
6553         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
6554                 return;
6555
6556         server = NFS_SERVER(lrp->args.inode);
6557         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6558                 rpc_restart_call_prepare(task);
6559                 return;
6560         }
6561         dprintk("<-- %s\n", __func__);
6562 }
6563
6564 static void nfs4_layoutreturn_release(void *calldata)
6565 {
6566         struct nfs4_layoutreturn *lrp = calldata;
6567         struct pnfs_layout_hdr *lo = lrp->args.layout;
6568
6569         dprintk("--> %s\n", __func__);
6570         spin_lock(&lo->plh_inode->i_lock);
6571         if (lrp->res.lrs_present)
6572                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6573         lo->plh_block_lgets--;
6574         spin_unlock(&lo->plh_inode->i_lock);
6575         pnfs_put_layout_hdr(lrp->args.layout);
6576         kfree(calldata);
6577         dprintk("<-- %s\n", __func__);
6578 }
6579
6580 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6581         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6582         .rpc_call_done = nfs4_layoutreturn_done,
6583         .rpc_release = nfs4_layoutreturn_release,
6584 };
6585
6586 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6587 {
6588         struct rpc_task *task;
6589         struct rpc_message msg = {
6590                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6591                 .rpc_argp = &lrp->args,
6592                 .rpc_resp = &lrp->res,
6593         };
6594         struct rpc_task_setup task_setup_data = {
6595                 .rpc_client = lrp->clp->cl_rpcclient,
6596                 .rpc_message = &msg,
6597                 .callback_ops = &nfs4_layoutreturn_call_ops,
6598                 .callback_data = lrp,
6599         };
6600         int status;
6601
6602         dprintk("--> %s\n", __func__);
6603         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6604         task = rpc_run_task(&task_setup_data);
6605         if (IS_ERR(task))
6606                 return PTR_ERR(task);
6607         status = task->tk_status;
6608         dprintk("<-- %s status=%d\n", __func__, status);
6609         rpc_put_task(task);
6610         return status;
6611 }
6612
6613 /*
6614  * Retrieve the list of Data Server devices from the MDS.
6615  */
6616 static int _nfs4_getdevicelist(struct nfs_server *server,
6617                                     const struct nfs_fh *fh,
6618                                     struct pnfs_devicelist *devlist)
6619 {
6620         struct nfs4_getdevicelist_args args = {
6621                 .fh = fh,
6622                 .layoutclass = server->pnfs_curr_ld->id,
6623         };
6624         struct nfs4_getdevicelist_res res = {
6625                 .devlist = devlist,
6626         };
6627         struct rpc_message msg = {
6628                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6629                 .rpc_argp = &args,
6630                 .rpc_resp = &res,
6631         };
6632         int status;
6633
6634         dprintk("--> %s\n", __func__);
6635         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6636                                 &res.seq_res, 0);
6637         dprintk("<-- %s status=%d\n", __func__, status);
6638         return status;
6639 }
6640
6641 int nfs4_proc_getdevicelist(struct nfs_server *server,
6642                             const struct nfs_fh *fh,
6643                             struct pnfs_devicelist *devlist)
6644 {
6645         struct nfs4_exception exception = { };
6646         int err;
6647
6648         do {
6649                 err = nfs4_handle_exception(server,
6650                                 _nfs4_getdevicelist(server, fh, devlist),
6651                                 &exception);
6652         } while (exception.retry);
6653
6654         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6655                 err, devlist->num_devs);
6656
6657         return err;
6658 }
6659 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6660
6661 static int
6662 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6663 {
6664         struct nfs4_getdeviceinfo_args args = {
6665                 .pdev = pdev,
6666         };
6667         struct nfs4_getdeviceinfo_res res = {
6668                 .pdev = pdev,
6669         };
6670         struct rpc_message msg = {
6671                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6672                 .rpc_argp = &args,
6673                 .rpc_resp = &res,
6674         };
6675         int status;
6676
6677         dprintk("--> %s\n", __func__);
6678         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6679         dprintk("<-- %s status=%d\n", __func__, status);
6680
6681         return status;
6682 }
6683
6684 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6685 {
6686         struct nfs4_exception exception = { };
6687         int err;
6688
6689         do {
6690                 err = nfs4_handle_exception(server,
6691                                         _nfs4_proc_getdeviceinfo(server, pdev),
6692                                         &exception);
6693         } while (exception.retry);
6694         return err;
6695 }
6696 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6697
6698 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6699 {
6700         struct nfs4_layoutcommit_data *data = calldata;
6701         struct nfs_server *server = NFS_SERVER(data->args.inode);
6702
6703         if (nfs4_setup_sequence(server, &data->args.seq_args,
6704                                 &data->res.seq_res, task))
6705                 return;
6706         rpc_call_start(task);
6707 }
6708
6709 static void
6710 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6711 {
6712         struct nfs4_layoutcommit_data *data = calldata;
6713         struct nfs_server *server = NFS_SERVER(data->args.inode);
6714
6715         if (!nfs4_sequence_done(task, &data->res.seq_res))
6716                 return;
6717
6718         switch (task->tk_status) { /* Just ignore these failures */
6719         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6720         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6721         case -NFS4ERR_BADLAYOUT:     /* no layout */
6722         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6723                 task->tk_status = 0;
6724                 break;
6725         case 0:
6726                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6727                                                    data->res.fattr);
6728                 break;
6729         default:
6730                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6731                         rpc_restart_call_prepare(task);
6732                         return;
6733                 }
6734         }
6735 }
6736
6737 static void nfs4_layoutcommit_release(void *calldata)
6738 {
6739         struct nfs4_layoutcommit_data *data = calldata;
6740         struct pnfs_layout_segment *lseg, *tmp;
6741         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6742
6743         pnfs_cleanup_layoutcommit(data);
6744         /* Matched by references in pnfs_set_layoutcommit */
6745         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6746                 list_del_init(&lseg->pls_lc_list);
6747                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6748                                        &lseg->pls_flags))
6749                         pnfs_put_lseg(lseg);
6750         }
6751
6752         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6753         smp_mb__after_clear_bit();
6754         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6755
6756         put_rpccred(data->cred);
6757         kfree(data);
6758 }
6759
6760 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6761         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6762         .rpc_call_done = nfs4_layoutcommit_done,
6763         .rpc_release = nfs4_layoutcommit_release,
6764 };
6765
6766 int
6767 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6768 {
6769         struct rpc_message msg = {
6770                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6771                 .rpc_argp = &data->args,
6772                 .rpc_resp = &data->res,
6773                 .rpc_cred = data->cred,
6774         };
6775         struct rpc_task_setup task_setup_data = {
6776                 .task = &data->task,
6777                 .rpc_client = NFS_CLIENT(data->args.inode),
6778                 .rpc_message = &msg,
6779                 .callback_ops = &nfs4_layoutcommit_ops,
6780                 .callback_data = data,
6781                 .flags = RPC_TASK_ASYNC,
6782         };
6783         struct rpc_task *task;
6784         int status = 0;
6785
6786         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6787                 "lbw: %llu inode %lu\n",
6788                 data->task.tk_pid, sync,
6789                 data->args.lastbytewritten,
6790                 data->args.inode->i_ino);
6791
6792         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6793         task = rpc_run_task(&task_setup_data);
6794         if (IS_ERR(task))
6795                 return PTR_ERR(task);
6796         if (sync == false)
6797                 goto out;
6798         status = nfs4_wait_for_completion_rpc_task(task);
6799         if (status != 0)
6800                 goto out;
6801         status = task->tk_status;
6802 out:
6803         dprintk("%s: status %d\n", __func__, status);
6804         rpc_put_task(task);
6805         return status;
6806 }
6807
6808 static int
6809 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6810                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6811 {
6812         struct nfs41_secinfo_no_name_args args = {
6813                 .style = SECINFO_STYLE_CURRENT_FH,
6814         };
6815         struct nfs4_secinfo_res res = {
6816                 .flavors = flavors,
6817         };
6818         struct rpc_message msg = {
6819                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6820                 .rpc_argp = &args,
6821                 .rpc_resp = &res,
6822         };
6823         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6824 }
6825
6826 static int
6827 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6828                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6829 {
6830         struct nfs4_exception exception = { };
6831         int err;
6832         do {
6833                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6834                 switch (err) {
6835                 case 0:
6836                 case -NFS4ERR_WRONGSEC:
6837                 case -NFS4ERR_NOTSUPP:
6838                         goto out;
6839                 default:
6840                         err = nfs4_handle_exception(server, err, &exception);
6841                 }
6842         } while (exception.retry);
6843 out:
6844         return err;
6845 }
6846
6847 static int
6848 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6849                     struct nfs_fsinfo *info)
6850 {
6851         int err;
6852         struct page *page;
6853         rpc_authflavor_t flavor;
6854         struct nfs4_secinfo_flavors *flavors;
6855
6856         page = alloc_page(GFP_KERNEL);
6857         if (!page) {
6858                 err = -ENOMEM;
6859                 goto out;
6860         }
6861
6862         flavors = page_address(page);
6863         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6864
6865         /*
6866          * Fall back on "guess and check" method if
6867          * the server doesn't support SECINFO_NO_NAME
6868          */
6869         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6870                 err = nfs4_find_root_sec(server, fhandle, info);
6871                 goto out_freepage;
6872         }
6873         if (err)
6874                 goto out_freepage;
6875
6876         flavor = nfs_find_best_sec(flavors);
6877         if (err == 0)
6878                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6879
6880 out_freepage:
6881         put_page(page);
6882         if (err == -EACCES)
6883                 return -EPERM;
6884 out:
6885         return err;
6886 }
6887
6888 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6889 {
6890         int status;
6891         struct nfs41_test_stateid_args args = {
6892                 .stateid = stateid,
6893         };
6894         struct nfs41_test_stateid_res res;
6895         struct rpc_message msg = {
6896                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6897                 .rpc_argp = &args,
6898                 .rpc_resp = &res,
6899         };
6900
6901         dprintk("NFS call  test_stateid %p\n", stateid);
6902         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6903         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6904         if (status != NFS_OK) {
6905                 dprintk("NFS reply test_stateid: failed, %d\n", status);
6906                 return status;
6907         }
6908         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
6909         return -res.status;
6910 }
6911
6912 /**
6913  * nfs41_test_stateid - perform a TEST_STATEID operation
6914  *
6915  * @server: server / transport on which to perform the operation
6916  * @stateid: state ID to test
6917  *
6918  * Returns NFS_OK if the server recognizes that "stateid" is valid.
6919  * Otherwise a negative NFS4ERR value is returned if the operation
6920  * failed or the state ID is not currently valid.
6921  */
6922 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6923 {
6924         struct nfs4_exception exception = { };
6925         int err;
6926         do {
6927                 err = _nfs41_test_stateid(server, stateid);
6928                 if (err != -NFS4ERR_DELAY)
6929                         break;
6930                 nfs4_handle_exception(server, err, &exception);
6931         } while (exception.retry);
6932         return err;
6933 }
6934
6935 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6936 {
6937         struct nfs41_free_stateid_args args = {
6938                 .stateid = stateid,
6939         };
6940         struct nfs41_free_stateid_res res;
6941         struct rpc_message msg = {
6942                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6943                 .rpc_argp = &args,
6944                 .rpc_resp = &res,
6945         };
6946         int status;
6947
6948         dprintk("NFS call  free_stateid %p\n", stateid);
6949         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6950         status = nfs4_call_sync_sequence(server->client, server, &msg,
6951                                          &args.seq_args, &res.seq_res, 1);
6952         dprintk("NFS reply free_stateid: %d\n", status);
6953         return status;
6954 }
6955
6956 /**
6957  * nfs41_free_stateid - perform a FREE_STATEID operation
6958  *
6959  * @server: server / transport on which to perform the operation
6960  * @stateid: state ID to release
6961  *
6962  * Returns NFS_OK if the server freed "stateid".  Otherwise a
6963  * negative NFS4ERR value is returned.
6964  */
6965 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6966 {
6967         struct nfs4_exception exception = { };
6968         int err;
6969         do {
6970                 err = _nfs4_free_stateid(server, stateid);
6971                 if (err != -NFS4ERR_DELAY)
6972                         break;
6973                 nfs4_handle_exception(server, err, &exception);
6974         } while (exception.retry);
6975         return err;
6976 }
6977
6978 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6979                 const nfs4_stateid *s2)
6980 {
6981         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6982                 return false;
6983
6984         if (s1->seqid == s2->seqid)
6985                 return true;
6986         if (s1->seqid == 0 || s2->seqid == 0)
6987                 return true;
6988
6989         return false;
6990 }
6991
6992 #endif /* CONFIG_NFS_V4_1 */
6993
6994 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6995                 const nfs4_stateid *s2)
6996 {
6997         return nfs4_stateid_match(s1, s2);
6998 }
6999
7000
7001 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
7002         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
7003         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
7004         .recover_open   = nfs4_open_reclaim,
7005         .recover_lock   = nfs4_lock_reclaim,
7006         .establish_clid = nfs4_init_clientid,
7007         .get_clid_cred  = nfs4_get_setclientid_cred,
7008         .detect_trunking = nfs40_discover_server_trunking,
7009 };
7010
7011 #if defined(CONFIG_NFS_V4_1)
7012 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
7013         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
7014         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
7015         .recover_open   = nfs4_open_reclaim,
7016         .recover_lock   = nfs4_lock_reclaim,
7017         .establish_clid = nfs41_init_clientid,
7018         .get_clid_cred  = nfs4_get_exchange_id_cred,
7019         .reclaim_complete = nfs41_proc_reclaim_complete,
7020         .detect_trunking = nfs41_discover_server_trunking,
7021 };
7022 #endif /* CONFIG_NFS_V4_1 */
7023
7024 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
7025         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
7026         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
7027         .recover_open   = nfs4_open_expired,
7028         .recover_lock   = nfs4_lock_expired,
7029         .establish_clid = nfs4_init_clientid,
7030         .get_clid_cred  = nfs4_get_setclientid_cred,
7031 };
7032
7033 #if defined(CONFIG_NFS_V4_1)
7034 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
7035         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
7036         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
7037         .recover_open   = nfs41_open_expired,
7038         .recover_lock   = nfs41_lock_expired,
7039         .establish_clid = nfs41_init_clientid,
7040         .get_clid_cred  = nfs4_get_exchange_id_cred,
7041 };
7042 #endif /* CONFIG_NFS_V4_1 */
7043
7044 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
7045         .sched_state_renewal = nfs4_proc_async_renew,
7046         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
7047         .renew_lease = nfs4_proc_renew,
7048 };
7049
7050 #if defined(CONFIG_NFS_V4_1)
7051 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
7052         .sched_state_renewal = nfs41_proc_async_sequence,
7053         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
7054         .renew_lease = nfs4_proc_sequence,
7055 };
7056 #endif
7057
7058 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
7059         .minor_version = 0,
7060         .call_sync = _nfs4_call_sync,
7061         .match_stateid = nfs4_match_stateid,
7062         .find_root_sec = nfs4_find_root_sec,
7063         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
7064         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
7065         .state_renewal_ops = &nfs40_state_renewal_ops,
7066 };
7067
7068 #if defined(CONFIG_NFS_V4_1)
7069 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
7070         .minor_version = 1,
7071         .call_sync = _nfs4_call_sync_session,
7072         .match_stateid = nfs41_match_stateid,
7073         .find_root_sec = nfs41_find_root_sec,
7074         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
7075         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
7076         .state_renewal_ops = &nfs41_state_renewal_ops,
7077 };
7078 #endif
7079
7080 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
7081         [0] = &nfs_v4_0_minor_ops,
7082 #if defined(CONFIG_NFS_V4_1)
7083         [1] = &nfs_v4_1_minor_ops,
7084 #endif
7085 };
7086
7087 const struct inode_operations nfs4_dir_inode_operations = {
7088         .create         = nfs_create,
7089         .lookup         = nfs_lookup,
7090         .atomic_open    = nfs_atomic_open,
7091         .link           = nfs_link,
7092         .unlink         = nfs_unlink,
7093         .symlink        = nfs_symlink,
7094         .mkdir          = nfs_mkdir,
7095         .rmdir          = nfs_rmdir,
7096         .mknod          = nfs_mknod,
7097         .rename         = nfs_rename,
7098         .permission     = nfs_permission,
7099         .getattr        = nfs_getattr,
7100         .setattr        = nfs_setattr,
7101         .getxattr       = generic_getxattr,
7102         .setxattr       = generic_setxattr,
7103         .listxattr      = generic_listxattr,
7104         .removexattr    = generic_removexattr,
7105 };
7106
7107 static const struct inode_operations nfs4_file_inode_operations = {
7108         .permission     = nfs_permission,
7109         .getattr        = nfs_getattr,
7110         .setattr        = nfs_setattr,
7111         .getxattr       = generic_getxattr,
7112         .setxattr       = generic_setxattr,
7113         .listxattr      = generic_listxattr,
7114         .removexattr    = generic_removexattr,
7115 };
7116
7117 const struct nfs_rpc_ops nfs_v4_clientops = {
7118         .version        = 4,                    /* protocol version */
7119         .dentry_ops     = &nfs4_dentry_operations,
7120         .dir_inode_ops  = &nfs4_dir_inode_operations,
7121         .file_inode_ops = &nfs4_file_inode_operations,
7122         .file_ops       = &nfs4_file_operations,
7123         .getroot        = nfs4_proc_get_root,
7124         .submount       = nfs4_submount,
7125         .try_mount      = nfs4_try_mount,
7126         .getattr        = nfs4_proc_getattr,
7127         .setattr        = nfs4_proc_setattr,
7128         .lookup         = nfs4_proc_lookup,
7129         .access         = nfs4_proc_access,
7130         .readlink       = nfs4_proc_readlink,
7131         .create         = nfs4_proc_create,
7132         .remove         = nfs4_proc_remove,
7133         .unlink_setup   = nfs4_proc_unlink_setup,
7134         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
7135         .unlink_done    = nfs4_proc_unlink_done,
7136         .rename         = nfs4_proc_rename,
7137         .rename_setup   = nfs4_proc_rename_setup,
7138         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
7139         .rename_done    = nfs4_proc_rename_done,
7140         .link           = nfs4_proc_link,
7141         .symlink        = nfs4_proc_symlink,
7142         .mkdir          = nfs4_proc_mkdir,
7143         .rmdir          = nfs4_proc_remove,
7144         .readdir        = nfs4_proc_readdir,
7145         .mknod          = nfs4_proc_mknod,
7146         .statfs         = nfs4_proc_statfs,
7147         .fsinfo         = nfs4_proc_fsinfo,
7148         .pathconf       = nfs4_proc_pathconf,
7149         .set_capabilities = nfs4_server_capabilities,
7150         .decode_dirent  = nfs4_decode_dirent,
7151         .read_setup     = nfs4_proc_read_setup,
7152         .read_pageio_init = pnfs_pageio_init_read,
7153         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
7154         .read_done      = nfs4_read_done,
7155         .write_setup    = nfs4_proc_write_setup,
7156         .write_pageio_init = pnfs_pageio_init_write,
7157         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
7158         .write_done     = nfs4_write_done,
7159         .commit_setup   = nfs4_proc_commit_setup,
7160         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
7161         .commit_done    = nfs4_commit_done,
7162         .lock           = nfs4_proc_lock,
7163         .clear_acl_cache = nfs4_zap_acl_attr,
7164         .close_context  = nfs4_close_context,
7165         .open_context   = nfs4_atomic_open,
7166         .have_delegation = nfs4_have_delegation,
7167         .return_delegation = nfs4_inode_return_delegation,
7168         .alloc_client   = nfs4_alloc_client,
7169         .init_client    = nfs4_init_client,
7170         .free_client    = nfs4_free_client,
7171         .create_server  = nfs4_create_server,
7172         .clone_server   = nfs_clone_server,
7173 };
7174
7175 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7176         .prefix = XATTR_NAME_NFSV4_ACL,
7177         .list   = nfs4_xattr_list_nfs4_acl,
7178         .get    = nfs4_xattr_get_nfs4_acl,
7179         .set    = nfs4_xattr_set_nfs4_acl,
7180 };
7181
7182 const struct xattr_handler *nfs4_xattr_handlers[] = {
7183         &nfs4_xattr_nfs4_acl_handler,
7184         NULL
7185 };
7186
7187 /*
7188  * Local variables:
7189  *  c-basic-offset: 8
7190  * End:
7191  */