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