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/sunrpc/gss_api.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/nfs_idmap.h>
  56 #include <linux/sunrpc/bc_xprt.h>
  57 #include <linux/xattr.h>
  58 #include <linux/utsname.h>
  59 #include <linux/freezer.h>
  60
  61 #include "nfs4_fs.h"
  62 #include "delegation.h"
  63 #include "internal.h"
  64 #include "iostat.h"
  65 #include "callback.h"
  66 #include "pnfs.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 static unsigned short max_session_slots = NFS4_DEF_SLOT_TABLE_SIZE;
  76
  77 struct nfs4_opendata;
  78 static int _nfs4_proc_open(struct nfs4_opendata *data);
  79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  82 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  85                             struct nfs_fattr *fattr, struct iattr *sattr,
  86                             struct nfs4_state *state);
  87 #ifdef CONFIG_NFS_V4_1
  88 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
  89 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
  90 #endif
  91 /* Prevent leaks of NFSv4 errors into userland */
  92 static int nfs4_map_errors(int err)
  93 {
  94         if (err >= -1000)
  95                 return err;
  96         switch (err) {
  97         case -NFS4ERR_RESOURCE:
  98                 return -EREMOTEIO;
  99         case -NFS4ERR_WRONGSEC:
 100                 return -EPERM;
 101         case -NFS4ERR_BADOWNER:
 102         case -NFS4ERR_BADNAME:
 103                 return -EINVAL;
 104         default:
 105                 dprintk("%s could not handle NFSv4 error %d\n",
 106                                 __func__, -err);
 107                 break;
 108         }
 109         return -EIO;
 110 }
 111
 112 /*
 113  * This is our standard bitmap for GETATTR requests.
 114  */
 115 const u32 nfs4_fattr_bitmap[2] = {
 116         FATTR4_WORD0_TYPE
 117         | FATTR4_WORD0_CHANGE
 118         | FATTR4_WORD0_SIZE
 119         | FATTR4_WORD0_FSID
 120         | FATTR4_WORD0_FILEID,
 121         FATTR4_WORD1_MODE
 122         | FATTR4_WORD1_NUMLINKS
 123         | FATTR4_WORD1_OWNER
 124         | FATTR4_WORD1_OWNER_GROUP
 125         | FATTR4_WORD1_RAWDEV
 126         | FATTR4_WORD1_SPACE_USED
 127         | FATTR4_WORD1_TIME_ACCESS
 128         | FATTR4_WORD1_TIME_METADATA
 129         | FATTR4_WORD1_TIME_MODIFY
 130 };
 131
 132 const u32 nfs4_statfs_bitmap[2] = {
 133         FATTR4_WORD0_FILES_AVAIL
 134         | FATTR4_WORD0_FILES_FREE
 135         | FATTR4_WORD0_FILES_TOTAL,
 136         FATTR4_WORD1_SPACE_AVAIL
 137         | FATTR4_WORD1_SPACE_FREE
 138         | FATTR4_WORD1_SPACE_TOTAL
 139 };
 140
 141 const u32 nfs4_pathconf_bitmap[2] = {
 142         FATTR4_WORD0_MAXLINK
 143         | FATTR4_WORD0_MAXNAME,
 144         0
 145 };
 146
 147 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 148                         | FATTR4_WORD0_MAXREAD
 149                         | FATTR4_WORD0_MAXWRITE
 150                         | FATTR4_WORD0_LEASE_TIME,
 151                         FATTR4_WORD1_TIME_DELTA
 152                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 153                         FATTR4_WORD2_LAYOUT_BLKSIZE
 154 };
 155
 156 const u32 nfs4_fs_locations_bitmap[2] = {
 157         FATTR4_WORD0_TYPE
 158         | FATTR4_WORD0_CHANGE
 159         | FATTR4_WORD0_SIZE
 160         | FATTR4_WORD0_FSID
 161         | FATTR4_WORD0_FILEID
 162         | FATTR4_WORD0_FS_LOCATIONS,
 163         FATTR4_WORD1_MODE
 164         | FATTR4_WORD1_NUMLINKS
 165         | FATTR4_WORD1_OWNER
 166         | FATTR4_WORD1_OWNER_GROUP
 167         | FATTR4_WORD1_RAWDEV
 168         | FATTR4_WORD1_SPACE_USED
 169         | FATTR4_WORD1_TIME_ACCESS
 170         | FATTR4_WORD1_TIME_METADATA
 171         | FATTR4_WORD1_TIME_MODIFY
 172         | FATTR4_WORD1_MOUNTED_ON_FILEID
 173 };
 174
 175 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 176                 struct nfs4_readdir_arg *readdir)
 177 {
 178         __be32 *start, *p;
 179
 180         BUG_ON(readdir->count < 80);
 181         if (cookie > 2) {
 182                 readdir->cookie = cookie;
 183                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 184                 return;
 185         }
 186
 187         readdir->cookie = 0;
 188         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 189         if (cookie == 2)
 190                 return;
 191
 192         /*
 193          * NFSv4 servers do not return entries for '.' and '..'
 194          * Therefore, we fake these entries here.  We let '.'
 195          * have cookie 0 and '..' have cookie 1.  Note that
 196          * when talking to the server, we always send cookie 0
 197          * instead of 1 or 2.
 198          */
 199         start = p = kmap_atomic(*readdir->pages);
 200
 201         if (cookie == 0) {
 202                 *p++ = xdr_one;                                  /* next */
 203                 *p++ = xdr_zero;                   /* cookie, first word */
 204                 *p++ = xdr_one;                   /* cookie, second word */
 205                 *p++ = xdr_one;                             /* entry len */
 206                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 207                 p++;
 208                 *p++ = xdr_one;                         /* bitmap length */
 209                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 210                 *p++ = htonl(8);              /* attribute buffer length */
 211                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 212         }
 213
 214         *p++ = xdr_one;                                  /* next */
 215         *p++ = xdr_zero;                   /* cookie, first word */
 216         *p++ = xdr_two;                   /* cookie, second word */
 217         *p++ = xdr_two;                             /* entry len */
 218         memcpy(p, "..\0\0", 4);                         /* entry */
 219         p++;
 220         *p++ = xdr_one;                         /* bitmap length */
 221         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 222         *p++ = htonl(8);              /* attribute buffer length */
 223         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 224
 225         readdir->pgbase = (char *)p - (char *)start;
 226         readdir->count -= readdir->pgbase;
 227         kunmap_atomic(start);
 228 }
 229
 230 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 231 {
 232         int res;
 233
 234         might_sleep();
 235
 236         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 237                         nfs_wait_bit_killable, TASK_KILLABLE);
 238         return res;
 239 }
 240
 241 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 242 {
 243         int res = 0;
 244
 245         might_sleep();
 246
 247         if (*timeout <= 0)
 248                 *timeout = NFS4_POLL_RETRY_MIN;
 249         if (*timeout > NFS4_POLL_RETRY_MAX)
 250                 *timeout = NFS4_POLL_RETRY_MAX;
 251         freezable_schedule_timeout_killable(*timeout);
 252         if (fatal_signal_pending(current))
 253                 res = -ERESTARTSYS;
 254         *timeout <<= 1;
 255         return res;
 256 }
 257
 258 /* This is the error handling routine for processes that are allowed
 259  * to sleep.
 260  */
 261 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 262 {
 263         struct nfs_client *clp = server->nfs_client;
 264         struct nfs4_state *state = exception->state;
 265         struct inode *inode = exception->inode;
 266         int ret = errorcode;
 267
 268         exception->retry = 0;
 269         switch(errorcode) {
 270                 case 0:
 271                         return 0;
 272                 case -NFS4ERR_OPENMODE:
 273                         if (inode && nfs_have_delegation(inode, FMODE_READ)) {
 274                                 nfs_inode_return_delegation(inode);
 275                                 exception->retry = 1;
 276                                 return 0;
 277                         }
 278                         if (state == NULL)
 279                                 break;
 280                         nfs4_schedule_stateid_recovery(server, state);
 281                         goto wait_on_recovery;
 282                 case -NFS4ERR_DELEG_REVOKED:
 283                 case -NFS4ERR_ADMIN_REVOKED:
 284                 case -NFS4ERR_BAD_STATEID:
 285                         if (state == NULL)
 286                                 break;
 287                         nfs_remove_bad_delegation(state->inode);
 288                         nfs4_schedule_stateid_recovery(server, state);
 289                         goto wait_on_recovery;
 290                 case -NFS4ERR_EXPIRED:
 291                         if (state != NULL)
 292                                 nfs4_schedule_stateid_recovery(server, state);
 293                 case -NFS4ERR_STALE_STATEID:
 294                 case -NFS4ERR_STALE_CLIENTID:
 295                         nfs4_schedule_lease_recovery(clp);
 296                         goto wait_on_recovery;
 297 #if defined(CONFIG_NFS_V4_1)
 298                 case -NFS4ERR_BADSESSION:
 299                 case -NFS4ERR_BADSLOT:
 300                 case -NFS4ERR_BAD_HIGH_SLOT:
 301                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 302                 case -NFS4ERR_DEADSESSION:
 303                 case -NFS4ERR_SEQ_FALSE_RETRY:
 304                 case -NFS4ERR_SEQ_MISORDERED:
 305                         dprintk("%s ERROR: %d Reset session\n", __func__,
 306                                 errorcode);
 307                         nfs4_schedule_session_recovery(clp->cl_session);
 308                         exception->retry = 1;
 309                         break;
 310 #endif /* defined(CONFIG_NFS_V4_1) */
 311                 case -NFS4ERR_FILE_OPEN:
 312                         if (exception->timeout > HZ) {
 313                                 /* We have retried a decent amount, time to
 314                                  * fail
 315                                  */
 316                                 ret = -EBUSY;
 317                                 break;
 318                         }
 319                 case -NFS4ERR_GRACE:
 320                 case -NFS4ERR_DELAY:
 321                 case -EKEYEXPIRED:
 322                         ret = nfs4_delay(server->client, &exception->timeout);
 323                         if (ret != 0)
 324                                 break;
 325                 case -NFS4ERR_RETRY_UNCACHED_REP:
 326                 case -NFS4ERR_OLD_STATEID:
 327                         exception->retry = 1;
 328                         break;
 329                 case -NFS4ERR_BADOWNER:
 330                         /* The following works around a Linux server bug! */
 331                 case -NFS4ERR_BADNAME:
 332                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 333                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 334                                 exception->retry = 1;
 335                                 printk(KERN_WARNING "NFS: v4 server %s "
 336                                                 "does not accept raw "
 337                                                 "uid/gids. "
 338                                                 "Reenabling the idmapper.\n",
 339                                                 server->nfs_client->cl_hostname);
 340                         }
 341         }
 342         /* We failed to handle the error */
 343         return nfs4_map_errors(ret);
 344 wait_on_recovery:
 345         ret = nfs4_wait_clnt_recover(clp);
 346         if (ret == 0)
 347                 exception->retry = 1;
 348         return ret;
 349 }
 350
 351
 352 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 353 {
 354         spin_lock(&clp->cl_lock);
 355         if (time_before(clp->cl_last_renewal,timestamp))
 356                 clp->cl_last_renewal = timestamp;
 357         spin_unlock(&clp->cl_lock);
 358 }
 359
 360 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 361 {
 362         do_renew_lease(server->nfs_client, timestamp);
 363 }
 364
 365 #if defined(CONFIG_NFS_V4_1)
 366
 367 /*
 368  * nfs4_free_slot - free a slot and efficiently update slot table.
 369  *
 370  * freeing a slot is trivially done by clearing its respective bit
 371  * in the bitmap.
 372  * If the freed slotid equals highest_used_slotid we want to update it
 373  * so that the server would be able to size down the slot table if needed,
 374  * otherwise we know that the highest_used_slotid is still in use.
 375  * When updating highest_used_slotid there may be "holes" in the bitmap
 376  * so we need to scan down from highest_used_slotid to 0 looking for the now
 377  * highest slotid in use.
 378  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
 379  *
 380  * Must be called while holding tbl->slot_tbl_lock
 381  */
 382 static void
 383 nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
 384 {
 385         BUG_ON(slotid >= NFS4_MAX_SLOT_TABLE);
 386         /* clear used bit in bitmap */
 387         __clear_bit(slotid, tbl->used_slots);
 388
 389         /* update highest_used_slotid when it is freed */
 390         if (slotid == tbl->highest_used_slotid) {
 391                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 392                 if (slotid < tbl->max_slots)
 393                         tbl->highest_used_slotid = slotid;
 394                 else
 395                         tbl->highest_used_slotid = NFS4_NO_SLOT;
 396         }
 397         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
 398                 slotid, tbl->highest_used_slotid);
 399 }
 400
 401 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
 402 {
 403         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 404         return true;
 405 }
 406
 407 /*
 408  * Signal state manager thread if session fore channel is drained
 409  */
 410 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 411 {
 412         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 413                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
 414                                 nfs4_set_task_privileged, NULL);
 415                 return;
 416         }
 417
 418         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 419                 return;
 420
 421         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 422         complete(&ses->fc_slot_table.complete);
 423 }
 424
 425 /*
 426  * Signal state manager thread if session back channel is drained
 427  */
 428 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 429 {
 430         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 431             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 432                 return;
 433         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 434         complete(&ses->bc_slot_table.complete);
 435 }
 436
 437 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 438 {
 439         struct nfs4_slot_table *tbl;
 440
 441         tbl = &res->sr_session->fc_slot_table;
 442         if (!res->sr_slot) {
 443                 /* just wake up the next guy waiting since
 444                  * we may have not consumed a slot after all */
 445                 dprintk("%s: No slot\n", __func__);
 446                 return;
 447         }
 448
 449         spin_lock(&tbl->slot_tbl_lock);
 450         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 451         nfs4_check_drain_fc_complete(res->sr_session);
 452         spin_unlock(&tbl->slot_tbl_lock);
 453         res->sr_slot = NULL;
 454 }
 455
 456 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 457 {
 458         unsigned long timestamp;
 459         struct nfs_client *clp;
 460
 461         /*
 462          * sr_status remains 1 if an RPC level error occurred. The server
 463          * may or may not have processed the sequence operation..
 464          * Proceed as if the server received and processed the sequence
 465          * operation.
 466          */
 467         if (res->sr_status == 1)
 468                 res->sr_status = NFS_OK;
 469
 470         /* don't increment the sequence number if the task wasn't sent */
 471         if (!RPC_WAS_SENT(task))
 472                 goto out;
 473
 474         /* Check the SEQUENCE operation status */
 475         switch (res->sr_status) {
 476         case 0:
 477                 /* Update the slot's sequence and clientid lease timer */
 478                 ++res->sr_slot->seq_nr;
 479                 timestamp = res->sr_renewal_time;
 480                 clp = res->sr_session->clp;
 481                 do_renew_lease(clp, timestamp);
 482                 /* Check sequence flags */
 483                 if (res->sr_status_flags != 0)
 484                         nfs4_schedule_lease_recovery(clp);
 485                 break;
 486         case -NFS4ERR_DELAY:
 487                 /* The server detected a resend of the RPC call and
 488                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 489                  * of RFC5661.
 490                  */
 491                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 492                         __func__,
 493                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 494                         res->sr_slot->seq_nr);
 495                 goto out_retry;
 496         default:
 497                 /* Just update the slot sequence no. */
 498                 ++res->sr_slot->seq_nr;
 499         }
 500 out:
 501         /* The session may be reset by one of the error handlers. */
 502         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 503         nfs41_sequence_free_slot(res);
 504         return 1;
 505 out_retry:
 506         if (!rpc_restart_call(task))
 507                 goto out;
 508         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 509         return 0;
 510 }
 511
 512 static int nfs4_sequence_done(struct rpc_task *task,
 513                                struct nfs4_sequence_res *res)
 514 {
 515         if (res->sr_session == NULL)
 516                 return 1;
 517         return nfs41_sequence_done(task, res);
 518 }
 519
 520 /*
 521  * nfs4_find_slot - efficiently look for a free slot
 522  *
 523  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 524  * If found, we mark the slot as used, update the highest_used_slotid,
 525  * and respectively set up the sequence operation args.
 526  * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
 527  *
 528  * Note: must be called with under the slot_tbl_lock.
 529  */
 530 static u32
 531 nfs4_find_slot(struct nfs4_slot_table *tbl)
 532 {
 533         u32 slotid;
 534         u32 ret_id = NFS4_NO_SLOT;
 535
 536         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
 537                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 538                 tbl->max_slots);
 539         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 540         if (slotid >= tbl->max_slots)
 541                 goto out;
 542         __set_bit(slotid, tbl->used_slots);
 543         if (slotid > tbl->highest_used_slotid ||
 544                         tbl->highest_used_slotid == NFS4_NO_SLOT)
 545                 tbl->highest_used_slotid = slotid;
 546         ret_id = slotid;
 547 out:
 548         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 549                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 550         return ret_id;
 551 }
 552
 553 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 554                 struct nfs4_sequence_res *res, int cache_reply)
 555 {
 556         args->sa_session = NULL;
 557         args->sa_cache_this = 0;
 558         if (cache_reply)
 559                 args->sa_cache_this = 1;
 560         res->sr_session = NULL;
 561         res->sr_slot = NULL;
 562 }
 563
 564 int nfs41_setup_sequence(struct nfs4_session *session,
 565                                 struct nfs4_sequence_args *args,
 566                                 struct nfs4_sequence_res *res,
 567                                 struct rpc_task *task)
 568 {
 569         struct nfs4_slot *slot;
 570         struct nfs4_slot_table *tbl;
 571         u32 slotid;
 572
 573         dprintk("--> %s\n", __func__);
 574         /* slot already allocated? */
 575         if (res->sr_slot != NULL)
 576                 return 0;
 577
 578         tbl = &session->fc_slot_table;
 579
 580         spin_lock(&tbl->slot_tbl_lock);
 581         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 582             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 583                 /* The state manager will wait until the slot table is empty */
 584                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 585                 spin_unlock(&tbl->slot_tbl_lock);
 586                 dprintk("%s session is draining\n", __func__);
 587                 return -EAGAIN;
 588         }
 589
 590         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 591             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 592                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 593                 spin_unlock(&tbl->slot_tbl_lock);
 594                 dprintk("%s enforce FIFO order\n", __func__);
 595                 return -EAGAIN;
 596         }
 597
 598         slotid = nfs4_find_slot(tbl);
 599         if (slotid == NFS4_NO_SLOT) {
 600                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 601                 spin_unlock(&tbl->slot_tbl_lock);
 602                 dprintk("<-- %s: no free slots\n", __func__);
 603                 return -EAGAIN;
 604         }
 605         spin_unlock(&tbl->slot_tbl_lock);
 606
 607         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 608         slot = tbl->slots + slotid;
 609         args->sa_session = session;
 610         args->sa_slotid = slotid;
 611
 612         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 613
 614         res->sr_session = session;
 615         res->sr_slot = slot;
 616         res->sr_renewal_time = jiffies;
 617         res->sr_status_flags = 0;
 618         /*
 619          * sr_status is only set in decode_sequence, and so will remain
 620          * set to 1 if an rpc level failure occurs.
 621          */
 622         res->sr_status = 1;
 623         return 0;
 624 }
 625 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 626
 627 int nfs4_setup_sequence(const struct nfs_server *server,
 628                         struct nfs4_sequence_args *args,
 629                         struct nfs4_sequence_res *res,
 630                         struct rpc_task *task)
 631 {
 632         struct nfs4_session *session = nfs4_get_session(server);
 633         int ret = 0;
 634
 635         if (session == NULL)
 636                 goto out;
 637
 638         dprintk("--> %s clp %p session %p sr_slot %td\n",
 639                 __func__, session->clp, session, res->sr_slot ?
 640                         res->sr_slot - session->fc_slot_table.slots : -1);
 641
 642         ret = nfs41_setup_sequence(session, args, res, task);
 643 out:
 644         dprintk("<-- %s status=%d\n", __func__, ret);
 645         return ret;
 646 }
 647
 648 struct nfs41_call_sync_data {
 649         const struct nfs_server *seq_server;
 650         struct nfs4_sequence_args *seq_args;
 651         struct nfs4_sequence_res *seq_res;
 652 };
 653
 654 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 655 {
 656         struct nfs41_call_sync_data *data = calldata;
 657
 658         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 659
 660         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 661                                 data->seq_res, task))
 662                 return;
 663         rpc_call_start(task);
 664 }
 665
 666 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 667 {
 668         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 669         nfs41_call_sync_prepare(task, calldata);
 670 }
 671
 672 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 673 {
 674         struct nfs41_call_sync_data *data = calldata;
 675
 676         nfs41_sequence_done(task, data->seq_res);
 677 }
 678
 679 static const struct rpc_call_ops nfs41_call_sync_ops = {
 680         .rpc_call_prepare = nfs41_call_sync_prepare,
 681         .rpc_call_done = nfs41_call_sync_done,
 682 };
 683
 684 static const struct rpc_call_ops nfs41_call_priv_sync_ops = {
 685         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 686         .rpc_call_done = nfs41_call_sync_done,
 687 };
 688
 689 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 690                                    struct nfs_server *server,
 691                                    struct rpc_message *msg,
 692                                    struct nfs4_sequence_args *args,
 693                                    struct nfs4_sequence_res *res,
 694                                    int privileged)
 695 {
 696         int ret;
 697         struct rpc_task *task;
 698         struct nfs41_call_sync_data data = {
 699                 .seq_server = server,
 700                 .seq_args = args,
 701                 .seq_res = res,
 702         };
 703         struct rpc_task_setup task_setup = {
 704                 .rpc_client = clnt,
 705                 .rpc_message = msg,
 706                 .callback_ops = &nfs41_call_sync_ops,
 707                 .callback_data = &data
 708         };
 709
 710         if (privileged)
 711                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 712         task = rpc_run_task(&task_setup);
 713         if (IS_ERR(task))
 714                 ret = PTR_ERR(task);
 715         else {
 716                 ret = task->tk_status;
 717                 rpc_put_task(task);
 718         }
 719         return ret;
 720 }
 721
 722 int _nfs4_call_sync_session(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 cache_reply)
 728 {
 729         nfs41_init_sequence(args, res, cache_reply);
 730         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
 731 }
 732
 733 #else
 734 static inline
 735 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 736                 struct nfs4_sequence_res *res, int cache_reply)
 737 {
 738 }
 739
 740 static int nfs4_sequence_done(struct rpc_task *task,
 741                                struct nfs4_sequence_res *res)
 742 {
 743         return 1;
 744 }
 745 #endif /* CONFIG_NFS_V4_1 */
 746
 747 int _nfs4_call_sync(struct rpc_clnt *clnt,
 748                     struct nfs_server *server,
 749                     struct rpc_message *msg,
 750                     struct nfs4_sequence_args *args,
 751                     struct nfs4_sequence_res *res,
 752                     int cache_reply)
 753 {
 754         nfs41_init_sequence(args, res, cache_reply);
 755         return rpc_call_sync(clnt, msg, 0);
 756 }
 757
 758 static inline
 759 int nfs4_call_sync(struct rpc_clnt *clnt,
 760                    struct nfs_server *server,
 761                    struct rpc_message *msg,
 762                    struct nfs4_sequence_args *args,
 763                    struct nfs4_sequence_res *res,
 764                    int cache_reply)
 765 {
 766         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 767                                                 args, res, cache_reply);
 768 }
 769
 770 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 771 {
 772         struct nfs_inode *nfsi = NFS_I(dir);
 773
 774         spin_lock(&dir->i_lock);
 775         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 776         if (!cinfo->atomic || cinfo->before != dir->i_version)
 777                 nfs_force_lookup_revalidate(dir);
 778         dir->i_version = cinfo->after;
 779         spin_unlock(&dir->i_lock);
 780 }
 781
 782 struct nfs4_opendata {
 783         struct kref kref;
 784         struct nfs_openargs o_arg;
 785         struct nfs_openres o_res;
 786         struct nfs_open_confirmargs c_arg;
 787         struct nfs_open_confirmres c_res;
 788         struct nfs4_string owner_name;
 789         struct nfs4_string group_name;
 790         struct nfs_fattr f_attr;
 791         struct nfs_fattr dir_attr;
 792         struct dentry *dir;
 793         struct dentry *dentry;
 794         struct nfs4_state_owner *owner;
 795         struct nfs4_state *state;
 796         struct iattr attrs;
 797         unsigned long timestamp;
 798         unsigned int rpc_done : 1;
 799         int rpc_status;
 800         int cancelled;
 801 };
 802
 803
 804 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 805 {
 806         p->o_res.f_attr = &p->f_attr;
 807         p->o_res.dir_attr = &p->dir_attr;
 808         p->o_res.seqid = p->o_arg.seqid;
 809         p->c_res.seqid = p->c_arg.seqid;
 810         p->o_res.server = p->o_arg.server;
 811         nfs_fattr_init(&p->f_attr);
 812         nfs_fattr_init(&p->dir_attr);
 813         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 814 }
 815
 816 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 817                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 818                 const struct iattr *attrs,
 819                 gfp_t gfp_mask)
 820 {
 821         struct dentry *parent = dget_parent(dentry);
 822         struct inode *dir = parent->d_inode;
 823         struct nfs_server *server = NFS_SERVER(dir);
 824         struct nfs4_opendata *p;
 825
 826         p = kzalloc(sizeof(*p), gfp_mask);
 827         if (p == NULL)
 828                 goto err;
 829         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 830         if (p->o_arg.seqid == NULL)
 831                 goto err_free;
 832         nfs_sb_active(dentry->d_sb);
 833         p->dentry = dget(dentry);
 834         p->dir = parent;
 835         p->owner = sp;
 836         atomic_inc(&sp->so_count);
 837         p->o_arg.fh = NFS_FH(dir);
 838         p->o_arg.open_flags = flags;
 839         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 840         p->o_arg.clientid = server->nfs_client->cl_clientid;
 841         p->o_arg.id = sp->so_seqid.owner_id;
 842         p->o_arg.name = &dentry->d_name;
 843         p->o_arg.server = server;
 844         p->o_arg.bitmask = server->attr_bitmask;
 845         p->o_arg.dir_bitmask = server->cache_consistency_bitmask;
 846         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 847         if (attrs != NULL && attrs->ia_valid != 0) {
 848                 __be32 verf[2];
 849
 850                 p->o_arg.u.attrs = &p->attrs;
 851                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 852
 853                 verf[0] = jiffies;
 854                 verf[1] = current->pid;
 855                 memcpy(p->o_arg.u.verifier.data, verf,
 856                                 sizeof(p->o_arg.u.verifier.data));
 857         }
 858         p->c_arg.fh = &p->o_res.fh;
 859         p->c_arg.stateid = &p->o_res.stateid;
 860         p->c_arg.seqid = p->o_arg.seqid;
 861         nfs4_init_opendata_res(p);
 862         kref_init(&p->kref);
 863         return p;
 864 err_free:
 865         kfree(p);
 866 err:
 867         dput(parent);
 868         return NULL;
 869 }
 870
 871 static void nfs4_opendata_free(struct kref *kref)
 872 {
 873         struct nfs4_opendata *p = container_of(kref,
 874                         struct nfs4_opendata, kref);
 875         struct super_block *sb = p->dentry->d_sb;
 876
 877         nfs_free_seqid(p->o_arg.seqid);
 878         if (p->state != NULL)
 879                 nfs4_put_open_state(p->state);
 880         nfs4_put_state_owner(p->owner);
 881         dput(p->dir);
 882         dput(p->dentry);
 883         nfs_sb_deactive(sb);
 884         nfs_fattr_free_names(&p->f_attr);
 885         kfree(p);
 886 }
 887
 888 static void nfs4_opendata_put(struct nfs4_opendata *p)
 889 {
 890         if (p != NULL)
 891                 kref_put(&p->kref, nfs4_opendata_free);
 892 }
 893
 894 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 895 {
 896         int ret;
 897
 898         ret = rpc_wait_for_completion_task(task);
 899         return ret;
 900 }
 901
 902 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 903 {
 904         int ret = 0;
 905
 906         if (open_mode & (O_EXCL|O_TRUNC))
 907                 goto out;
 908         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 909                 case FMODE_READ:
 910                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 911                                 && state->n_rdonly != 0;
 912                         break;
 913                 case FMODE_WRITE:
 914                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 915                                 && state->n_wronly != 0;
 916                         break;
 917                 case FMODE_READ|FMODE_WRITE:
 918                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 919                                 && state->n_rdwr != 0;
 920         }
 921 out:
 922         return ret;
 923 }
 924
 925 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 926 {
 927         if (delegation == NULL)
 928                 return 0;
 929         if ((delegation->type & fmode) != fmode)
 930                 return 0;
 931         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 932                 return 0;
 933         nfs_mark_delegation_referenced(delegation);
 934         return 1;
 935 }
 936
 937 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 938 {
 939         switch (fmode) {
 940                 case FMODE_WRITE:
 941                         state->n_wronly++;
 942                         break;
 943                 case FMODE_READ:
 944                         state->n_rdonly++;
 945                         break;
 946                 case FMODE_READ|FMODE_WRITE:
 947                         state->n_rdwr++;
 948         }
 949         nfs4_state_set_mode_locked(state, state->state | fmode);
 950 }
 951
 952 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 953 {
 954         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 955                 nfs4_stateid_copy(&state->stateid, stateid);
 956         nfs4_stateid_copy(&state->open_stateid, stateid);
 957         switch (fmode) {
 958                 case FMODE_READ:
 959                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 960                         break;
 961                 case FMODE_WRITE:
 962                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 963                         break;
 964                 case FMODE_READ|FMODE_WRITE:
 965                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 966         }
 967 }
 968
 969 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 970 {
 971         write_seqlock(&state->seqlock);
 972         nfs_set_open_stateid_locked(state, stateid, fmode);
 973         write_sequnlock(&state->seqlock);
 974 }
 975
 976 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 977 {
 978         /*
 979          * Protect the call to nfs4_state_set_mode_locked and
 980          * serialise the stateid update
 981          */
 982         write_seqlock(&state->seqlock);
 983         if (deleg_stateid != NULL) {
 984                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
 985                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 986         }
 987         if (open_stateid != NULL)
 988                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 989         write_sequnlock(&state->seqlock);
 990         spin_lock(&state->owner->so_lock);
 991         update_open_stateflags(state, fmode);
 992         spin_unlock(&state->owner->so_lock);
 993 }
 994
 995 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 996 {
 997         struct nfs_inode *nfsi = NFS_I(state->inode);
 998         struct nfs_delegation *deleg_cur;
 999         int ret = 0;
1000
1001         fmode &= (FMODE_READ|FMODE_WRITE);
1002
1003         rcu_read_lock();
1004         deleg_cur = rcu_dereference(nfsi->delegation);
1005         if (deleg_cur == NULL)
1006                 goto no_delegation;
1007
1008         spin_lock(&deleg_cur->lock);
1009         if (nfsi->delegation != deleg_cur ||
1010             (deleg_cur->type & fmode) != fmode)
1011                 goto no_delegation_unlock;
1012
1013         if (delegation == NULL)
1014                 delegation = &deleg_cur->stateid;
1015         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1016                 goto no_delegation_unlock;
1017
1018         nfs_mark_delegation_referenced(deleg_cur);
1019         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1020         ret = 1;
1021 no_delegation_unlock:
1022         spin_unlock(&deleg_cur->lock);
1023 no_delegation:
1024         rcu_read_unlock();
1025
1026         if (!ret && open_stateid != NULL) {
1027                 __update_open_stateid(state, open_stateid, NULL, fmode);
1028                 ret = 1;
1029         }
1030
1031         return ret;
1032 }
1033
1034
1035 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1036 {
1037         struct nfs_delegation *delegation;
1038
1039         rcu_read_lock();
1040         delegation = rcu_dereference(NFS_I(inode)->delegation);
1041         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1042                 rcu_read_unlock();
1043                 return;
1044         }
1045         rcu_read_unlock();
1046         nfs_inode_return_delegation(inode);
1047 }
1048
1049 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1050 {
1051         struct nfs4_state *state = opendata->state;
1052         struct nfs_inode *nfsi = NFS_I(state->inode);
1053         struct nfs_delegation *delegation;
1054         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1055         fmode_t fmode = opendata->o_arg.fmode;
1056         nfs4_stateid stateid;
1057         int ret = -EAGAIN;
1058
1059         for (;;) {
1060                 if (can_open_cached(state, fmode, open_mode)) {
1061                         spin_lock(&state->owner->so_lock);
1062                         if (can_open_cached(state, fmode, open_mode)) {
1063                                 update_open_stateflags(state, fmode);
1064                                 spin_unlock(&state->owner->so_lock);
1065                                 goto out_return_state;
1066                         }
1067                         spin_unlock(&state->owner->so_lock);
1068                 }
1069                 rcu_read_lock();
1070                 delegation = rcu_dereference(nfsi->delegation);
1071                 if (!can_open_delegated(delegation, fmode)) {
1072                         rcu_read_unlock();
1073                         break;
1074                 }
1075                 /* Save the delegation */
1076                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1077                 rcu_read_unlock();
1078                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1079                 if (ret != 0)
1080                         goto out;
1081                 ret = -EAGAIN;
1082
1083                 /* Try to update the stateid using the delegation */
1084                 if (update_open_stateid(state, NULL, &stateid, fmode))
1085                         goto out_return_state;
1086         }
1087 out:
1088         return ERR_PTR(ret);
1089 out_return_state:
1090         atomic_inc(&state->count);
1091         return state;
1092 }
1093
1094 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1095 {
1096         struct inode *inode;
1097         struct nfs4_state *state = NULL;
1098         struct nfs_delegation *delegation;
1099         int ret;
1100
1101         if (!data->rpc_done) {
1102                 state = nfs4_try_open_cached(data);
1103                 goto out;
1104         }
1105
1106         ret = -EAGAIN;
1107         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1108                 goto err;
1109         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1110         ret = PTR_ERR(inode);
1111         if (IS_ERR(inode))
1112                 goto err;
1113         ret = -ENOMEM;
1114         state = nfs4_get_open_state(inode, data->owner);
1115         if (state == NULL)
1116                 goto err_put_inode;
1117         if (data->o_res.delegation_type != 0) {
1118                 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1119                 int delegation_flags = 0;
1120
1121                 rcu_read_lock();
1122                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1123                 if (delegation)
1124                         delegation_flags = delegation->flags;
1125                 rcu_read_unlock();
1126                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1127                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1128                                         "returning a delegation for "
1129                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1130                                         clp->cl_hostname);
1131                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1132                         nfs_inode_set_delegation(state->inode,
1133                                         data->owner->so_cred,
1134                                         &data->o_res);
1135                 else
1136                         nfs_inode_reclaim_delegation(state->inode,
1137                                         data->owner->so_cred,
1138                                         &data->o_res);
1139         }
1140
1141         update_open_stateid(state, &data->o_res.stateid, NULL,
1142                         data->o_arg.fmode);
1143         iput(inode);
1144 out:
1145         return state;
1146 err_put_inode:
1147         iput(inode);
1148 err:
1149         return ERR_PTR(ret);
1150 }
1151
1152 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1153 {
1154         struct nfs_inode *nfsi = NFS_I(state->inode);
1155         struct nfs_open_context *ctx;
1156
1157         spin_lock(&state->inode->i_lock);
1158         list_for_each_entry(ctx, &nfsi->open_files, list) {
1159                 if (ctx->state != state)
1160                         continue;
1161                 get_nfs_open_context(ctx);
1162                 spin_unlock(&state->inode->i_lock);
1163                 return ctx;
1164         }
1165         spin_unlock(&state->inode->i_lock);
1166         return ERR_PTR(-ENOENT);
1167 }
1168
1169 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1170 {
1171         struct nfs4_opendata *opendata;
1172
1173         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1174         if (opendata == NULL)
1175                 return ERR_PTR(-ENOMEM);
1176         opendata->state = state;
1177         atomic_inc(&state->count);
1178         return opendata;
1179 }
1180
1181 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1182 {
1183         struct nfs4_state *newstate;
1184         int ret;
1185
1186         opendata->o_arg.open_flags = 0;
1187         opendata->o_arg.fmode = fmode;
1188         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1189         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1190         nfs4_init_opendata_res(opendata);
1191         ret = _nfs4_recover_proc_open(opendata);
1192         if (ret != 0)
1193                 return ret;
1194         newstate = nfs4_opendata_to_nfs4_state(opendata);
1195         if (IS_ERR(newstate))
1196                 return PTR_ERR(newstate);
1197         nfs4_close_state(newstate, fmode);
1198         *res = newstate;
1199         return 0;
1200 }
1201
1202 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1203 {
1204         struct nfs4_state *newstate;
1205         int ret;
1206
1207         /* memory barrier prior to reading state->n_* */
1208         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1209         smp_rmb();
1210         if (state->n_rdwr != 0) {
1211                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1212                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1213                 if (ret != 0)
1214                         return ret;
1215                 if (newstate != state)
1216                         return -ESTALE;
1217         }
1218         if (state->n_wronly != 0) {
1219                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1220                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1221                 if (ret != 0)
1222                         return ret;
1223                 if (newstate != state)
1224                         return -ESTALE;
1225         }
1226         if (state->n_rdonly != 0) {
1227                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1228                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1229                 if (ret != 0)
1230                         return ret;
1231                 if (newstate != state)
1232                         return -ESTALE;
1233         }
1234         /*
1235          * We may have performed cached opens for all three recoveries.
1236          * Check if we need to update the current stateid.
1237          */
1238         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1239             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1240                 write_seqlock(&state->seqlock);
1241                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1242                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1243                 write_sequnlock(&state->seqlock);
1244         }
1245         return 0;
1246 }
1247
1248 /*
1249  * OPEN_RECLAIM:
1250  *      reclaim state on the server after a reboot.
1251  */
1252 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1253 {
1254         struct nfs_delegation *delegation;
1255         struct nfs4_opendata *opendata;
1256         fmode_t delegation_type = 0;
1257         int status;
1258
1259         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1260         if (IS_ERR(opendata))
1261                 return PTR_ERR(opendata);
1262         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1263         opendata->o_arg.fh = NFS_FH(state->inode);
1264         rcu_read_lock();
1265         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1266         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1267                 delegation_type = delegation->type;
1268         rcu_read_unlock();
1269         opendata->o_arg.u.delegation_type = delegation_type;
1270         status = nfs4_open_recover(opendata, state);
1271         nfs4_opendata_put(opendata);
1272         return status;
1273 }
1274
1275 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1276 {
1277         struct nfs_server *server = NFS_SERVER(state->inode);
1278         struct nfs4_exception exception = { };
1279         int err;
1280         do {
1281                 err = _nfs4_do_open_reclaim(ctx, state);
1282                 if (err != -NFS4ERR_DELAY)
1283                         break;
1284                 nfs4_handle_exception(server, err, &exception);
1285         } while (exception.retry);
1286         return err;
1287 }
1288
1289 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1290 {
1291         struct nfs_open_context *ctx;
1292         int ret;
1293
1294         ctx = nfs4_state_find_open_context(state);
1295         if (IS_ERR(ctx))
1296                 return PTR_ERR(ctx);
1297         ret = nfs4_do_open_reclaim(ctx, state);
1298         put_nfs_open_context(ctx);
1299         return ret;
1300 }
1301
1302 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1303 {
1304         struct nfs4_opendata *opendata;
1305         int ret;
1306
1307         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1308         if (IS_ERR(opendata))
1309                 return PTR_ERR(opendata);
1310         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1311         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1312         ret = nfs4_open_recover(opendata, state);
1313         nfs4_opendata_put(opendata);
1314         return ret;
1315 }
1316
1317 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1318 {
1319         struct nfs4_exception exception = { };
1320         struct nfs_server *server = NFS_SERVER(state->inode);
1321         int err;
1322         do {
1323                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1324                 switch (err) {
1325                         case 0:
1326                         case -ENOENT:
1327                         case -ESTALE:
1328                                 goto out;
1329                         case -NFS4ERR_BADSESSION:
1330                         case -NFS4ERR_BADSLOT:
1331                         case -NFS4ERR_BAD_HIGH_SLOT:
1332                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1333                         case -NFS4ERR_DEADSESSION:
1334                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1335                                 goto out;
1336                         case -NFS4ERR_STALE_CLIENTID:
1337                         case -NFS4ERR_STALE_STATEID:
1338                         case -NFS4ERR_EXPIRED:
1339                                 /* Don't recall a delegation if it was lost */
1340                                 nfs4_schedule_lease_recovery(server->nfs_client);
1341                                 goto out;
1342                         case -ERESTARTSYS:
1343                                 /*
1344                                  * The show must go on: exit, but mark the
1345                                  * stateid as needing recovery.
1346                                  */
1347                         case -NFS4ERR_DELEG_REVOKED:
1348                         case -NFS4ERR_ADMIN_REVOKED:
1349                         case -NFS4ERR_BAD_STATEID:
1350                                 nfs_inode_find_state_and_recover(state->inode,
1351                                                 stateid);
1352                                 nfs4_schedule_stateid_recovery(server, state);
1353                         case -EKEYEXPIRED:
1354                                 /*
1355                                  * User RPCSEC_GSS context has expired.
1356                                  * We cannot recover this stateid now, so
1357                                  * skip it and allow recovery thread to
1358                                  * proceed.
1359                                  */
1360                         case -ENOMEM:
1361                                 err = 0;
1362                                 goto out;
1363                 }
1364                 err = nfs4_handle_exception(server, err, &exception);
1365         } while (exception.retry);
1366 out:
1367         return err;
1368 }
1369
1370 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1371 {
1372         struct nfs4_opendata *data = calldata;
1373
1374         data->rpc_status = task->tk_status;
1375         if (data->rpc_status == 0) {
1376                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1377                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1378                 renew_lease(data->o_res.server, data->timestamp);
1379                 data->rpc_done = 1;
1380         }
1381 }
1382
1383 static void nfs4_open_confirm_release(void *calldata)
1384 {
1385         struct nfs4_opendata *data = calldata;
1386         struct nfs4_state *state = NULL;
1387
1388         /* If this request hasn't been cancelled, do nothing */
1389         if (data->cancelled == 0)
1390                 goto out_free;
1391         /* In case of error, no cleanup! */
1392         if (!data->rpc_done)
1393                 goto out_free;
1394         state = nfs4_opendata_to_nfs4_state(data);
1395         if (!IS_ERR(state))
1396                 nfs4_close_state(state, data->o_arg.fmode);
1397 out_free:
1398         nfs4_opendata_put(data);
1399 }
1400
1401 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1402         .rpc_call_done = nfs4_open_confirm_done,
1403         .rpc_release = nfs4_open_confirm_release,
1404 };
1405
1406 /*
1407  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1408  */
1409 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1410 {
1411         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1412         struct rpc_task *task;
1413         struct  rpc_message msg = {
1414                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1415                 .rpc_argp = &data->c_arg,
1416                 .rpc_resp = &data->c_res,
1417                 .rpc_cred = data->owner->so_cred,
1418         };
1419         struct rpc_task_setup task_setup_data = {
1420                 .rpc_client = server->client,
1421                 .rpc_message = &msg,
1422                 .callback_ops = &nfs4_open_confirm_ops,
1423                 .callback_data = data,
1424                 .workqueue = nfsiod_workqueue,
1425                 .flags = RPC_TASK_ASYNC,
1426         };
1427         int status;
1428
1429         kref_get(&data->kref);
1430         data->rpc_done = 0;
1431         data->rpc_status = 0;
1432         data->timestamp = jiffies;
1433         task = rpc_run_task(&task_setup_data);
1434         if (IS_ERR(task))
1435                 return PTR_ERR(task);
1436         status = nfs4_wait_for_completion_rpc_task(task);
1437         if (status != 0) {
1438                 data->cancelled = 1;
1439                 smp_wmb();
1440         } else
1441                 status = data->rpc_status;
1442         rpc_put_task(task);
1443         return status;
1444 }
1445
1446 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1447 {
1448         struct nfs4_opendata *data = calldata;
1449         struct nfs4_state_owner *sp = data->owner;
1450
1451         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1452                 return;
1453         /*
1454          * Check if we still need to send an OPEN call, or if we can use
1455          * a delegation instead.
1456          */
1457         if (data->state != NULL) {
1458                 struct nfs_delegation *delegation;
1459
1460                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1461                         goto out_no_action;
1462                 rcu_read_lock();
1463                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1464                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1465                     can_open_delegated(delegation, data->o_arg.fmode))
1466                         goto unlock_no_action;
1467                 rcu_read_unlock();
1468         }
1469         /* Update sequence id. */
1470         data->o_arg.id = sp->so_seqid.owner_id;
1471         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1472         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1473                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1474                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1475         }
1476         data->timestamp = jiffies;
1477         if (nfs4_setup_sequence(data->o_arg.server,
1478                                 &data->o_arg.seq_args,
1479                                 &data->o_res.seq_res, task))
1480                 return;
1481         rpc_call_start(task);
1482         return;
1483 unlock_no_action:
1484         rcu_read_unlock();
1485 out_no_action:
1486         task->tk_action = NULL;
1487
1488 }
1489
1490 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1491 {
1492         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1493         nfs4_open_prepare(task, calldata);
1494 }
1495
1496 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1497 {
1498         struct nfs4_opendata *data = calldata;
1499
1500         data->rpc_status = task->tk_status;
1501
1502         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1503                 return;
1504
1505         if (task->tk_status == 0) {
1506                 switch (data->o_res.f_attr->mode & S_IFMT) {
1507                         case S_IFREG:
1508                                 break;
1509                         case S_IFLNK:
1510                                 data->rpc_status = -ELOOP;
1511                                 break;
1512                         case S_IFDIR:
1513                                 data->rpc_status = -EISDIR;
1514                                 break;
1515                         default:
1516                                 data->rpc_status = -ENOTDIR;
1517                 }
1518                 renew_lease(data->o_res.server, data->timestamp);
1519                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1520                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1521         }
1522         data->rpc_done = 1;
1523 }
1524
1525 static void nfs4_open_release(void *calldata)
1526 {
1527         struct nfs4_opendata *data = calldata;
1528         struct nfs4_state *state = NULL;
1529
1530         /* If this request hasn't been cancelled, do nothing */
1531         if (data->cancelled == 0)
1532                 goto out_free;
1533         /* In case of error, no cleanup! */
1534         if (data->rpc_status != 0 || !data->rpc_done)
1535                 goto out_free;
1536         /* In case we need an open_confirm, no cleanup! */
1537         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1538                 goto out_free;
1539         state = nfs4_opendata_to_nfs4_state(data);
1540         if (!IS_ERR(state))
1541                 nfs4_close_state(state, data->o_arg.fmode);
1542 out_free:
1543         nfs4_opendata_put(data);
1544 }
1545
1546 static const struct rpc_call_ops nfs4_open_ops = {
1547         .rpc_call_prepare = nfs4_open_prepare,
1548         .rpc_call_done = nfs4_open_done,
1549         .rpc_release = nfs4_open_release,
1550 };
1551
1552 static const struct rpc_call_ops nfs4_recover_open_ops = {
1553         .rpc_call_prepare = nfs4_recover_open_prepare,
1554         .rpc_call_done = nfs4_open_done,
1555         .rpc_release = nfs4_open_release,
1556 };
1557
1558 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1559 {
1560         struct inode *dir = data->dir->d_inode;
1561         struct nfs_server *server = NFS_SERVER(dir);
1562         struct nfs_openargs *o_arg = &data->o_arg;
1563         struct nfs_openres *o_res = &data->o_res;
1564         struct rpc_task *task;
1565         struct rpc_message msg = {
1566                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1567                 .rpc_argp = o_arg,
1568                 .rpc_resp = o_res,
1569                 .rpc_cred = data->owner->so_cred,
1570         };
1571         struct rpc_task_setup task_setup_data = {
1572                 .rpc_client = server->client,
1573                 .rpc_message = &msg,
1574                 .callback_ops = &nfs4_open_ops,
1575                 .callback_data = data,
1576                 .workqueue = nfsiod_workqueue,
1577                 .flags = RPC_TASK_ASYNC,
1578         };
1579         int status;
1580
1581         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1582         kref_get(&data->kref);
1583         data->rpc_done = 0;
1584         data->rpc_status = 0;
1585         data->cancelled = 0;
1586         if (isrecover)
1587                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1588         task = rpc_run_task(&task_setup_data);
1589         if (IS_ERR(task))
1590                 return PTR_ERR(task);
1591         status = nfs4_wait_for_completion_rpc_task(task);
1592         if (status != 0) {
1593                 data->cancelled = 1;
1594                 smp_wmb();
1595         } else
1596                 status = data->rpc_status;
1597         rpc_put_task(task);
1598
1599         return status;
1600 }
1601
1602 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1603 {
1604         struct inode *dir = data->dir->d_inode;
1605         struct nfs_openres *o_res = &data->o_res;
1606         int status;
1607
1608         status = nfs4_run_open_task(data, 1);
1609         if (status != 0 || !data->rpc_done)
1610                 return status;
1611
1612         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1613
1614         nfs_refresh_inode(dir, o_res->dir_attr);
1615
1616         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1617                 status = _nfs4_proc_open_confirm(data);
1618                 if (status != 0)
1619                         return status;
1620         }
1621
1622         return status;
1623 }
1624
1625 /*
1626  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1627  */
1628 static int _nfs4_proc_open(struct nfs4_opendata *data)
1629 {
1630         struct inode *dir = data->dir->d_inode;
1631         struct nfs_server *server = NFS_SERVER(dir);
1632         struct nfs_openargs *o_arg = &data->o_arg;
1633         struct nfs_openres *o_res = &data->o_res;
1634         int status;
1635
1636         status = nfs4_run_open_task(data, 0);
1637         if (!data->rpc_done)
1638                 return status;
1639         if (status != 0) {
1640                 if (status == -NFS4ERR_BADNAME &&
1641                                 !(o_arg->open_flags & O_CREAT))
1642                         return -ENOENT;
1643                 return status;
1644         }
1645
1646         nfs_fattr_map_and_free_names(server, &data->f_attr);
1647
1648         if (o_arg->open_flags & O_CREAT) {
1649                 update_changeattr(dir, &o_res->cinfo);
1650                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1651         } else
1652                 nfs_refresh_inode(dir, o_res->dir_attr);
1653         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1654                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1655         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1656                 status = _nfs4_proc_open_confirm(data);
1657                 if (status != 0)
1658                         return status;
1659         }
1660         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1661                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1662         return 0;
1663 }
1664
1665 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1666 {
1667         unsigned int loop;
1668         int ret;
1669
1670         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1671                 ret = nfs4_wait_clnt_recover(clp);
1672                 if (ret != 0)
1673                         break;
1674                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1675                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1676                         break;
1677                 nfs4_schedule_state_manager(clp);
1678                 ret = -EIO;
1679         }
1680         return ret;
1681 }
1682
1683 static int nfs4_recover_expired_lease(struct nfs_server *server)
1684 {
1685         return nfs4_client_recover_expired_lease(server->nfs_client);
1686 }
1687
1688 /*
1689  * OPEN_EXPIRED:
1690  *      reclaim state on the server after a network partition.
1691  *      Assumes caller holds the appropriate lock
1692  */
1693 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1694 {
1695         struct nfs4_opendata *opendata;
1696         int ret;
1697
1698         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1699         if (IS_ERR(opendata))
1700                 return PTR_ERR(opendata);
1701         ret = nfs4_open_recover(opendata, state);
1702         if (ret == -ESTALE)
1703                 d_drop(ctx->dentry);
1704         nfs4_opendata_put(opendata);
1705         return ret;
1706 }
1707
1708 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1709 {
1710         struct nfs_server *server = NFS_SERVER(state->inode);
1711         struct nfs4_exception exception = { };
1712         int err;
1713
1714         do {
1715                 err = _nfs4_open_expired(ctx, state);
1716                 switch (err) {
1717                 default:
1718                         goto out;
1719                 case -NFS4ERR_GRACE:
1720                 case -NFS4ERR_DELAY:
1721                         nfs4_handle_exception(server, err, &exception);
1722                         err = 0;
1723                 }
1724         } while (exception.retry);
1725 out:
1726         return err;
1727 }
1728
1729 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1730 {
1731         struct nfs_open_context *ctx;
1732         int ret;
1733
1734         ctx = nfs4_state_find_open_context(state);
1735         if (IS_ERR(ctx))
1736                 return PTR_ERR(ctx);
1737         ret = nfs4_do_open_expired(ctx, state);
1738         put_nfs_open_context(ctx);
1739         return ret;
1740 }
1741
1742 #if defined(CONFIG_NFS_V4_1)
1743 static int nfs41_check_expired_stateid(struct nfs4_state *state, nfs4_stateid *stateid, unsigned int flags)
1744 {
1745         int status = NFS_OK;
1746         struct nfs_server *server = NFS_SERVER(state->inode);
1747
1748         if (state->flags & flags) {
1749                 status = nfs41_test_stateid(server, stateid);
1750                 if (status != NFS_OK) {
1751                         nfs41_free_stateid(server, stateid);
1752                         state->flags &= ~flags;
1753                 }
1754         }
1755         return status;
1756 }
1757
1758 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1759 {
1760         int deleg_status, open_status;
1761         int deleg_flags = 1 << NFS_DELEGATED_STATE;
1762         int open_flags = (1 << NFS_O_RDONLY_STATE) | (1 << NFS_O_WRONLY_STATE) | (1 << NFS_O_RDWR_STATE);
1763
1764         deleg_status = nfs41_check_expired_stateid(state, &state->stateid, deleg_flags);
1765         open_status = nfs41_check_expired_stateid(state,  &state->open_stateid, open_flags);
1766
1767         if ((deleg_status == NFS_OK) && (open_status == NFS_OK))
1768                 return NFS_OK;
1769         return nfs4_open_expired(sp, state);
1770 }
1771 #endif
1772
1773 /*
1774  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1775  * fields corresponding to attributes that were used to store the verifier.
1776  * Make sure we clobber those fields in the later setattr call
1777  */
1778 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1779 {
1780         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1781             !(sattr->ia_valid & ATTR_ATIME_SET))
1782                 sattr->ia_valid |= ATTR_ATIME;
1783
1784         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1785             !(sattr->ia_valid & ATTR_MTIME_SET))
1786                 sattr->ia_valid |= ATTR_MTIME;
1787 }
1788
1789 /*
1790  * Returns a referenced nfs4_state
1791  */
1792 static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1793 {
1794         struct nfs4_state_owner  *sp;
1795         struct nfs4_state     *state = NULL;
1796         struct nfs_server       *server = NFS_SERVER(dir);
1797         struct nfs4_opendata *opendata;
1798         int status;
1799
1800         /* Protect against reboot recovery conflicts */
1801         status = -ENOMEM;
1802         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1803         if (sp == NULL) {
1804                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1805                 goto out_err;
1806         }
1807         status = nfs4_recover_expired_lease(server);
1808         if (status != 0)
1809                 goto err_put_state_owner;
1810         if (dentry->d_inode != NULL)
1811                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1812         status = -ENOMEM;
1813         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1814         if (opendata == NULL)
1815                 goto err_put_state_owner;
1816
1817         if (dentry->d_inode != NULL)
1818                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1819
1820         status = _nfs4_proc_open(opendata);
1821         if (status != 0)
1822                 goto err_opendata_put;
1823
1824         state = nfs4_opendata_to_nfs4_state(opendata);
1825         status = PTR_ERR(state);
1826         if (IS_ERR(state))
1827                 goto err_opendata_put;
1828         if (server->caps & NFS_CAP_POSIX_LOCK)
1829                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1830
1831         if (opendata->o_arg.open_flags & O_EXCL) {
1832                 nfs4_exclusive_attrset(opendata, sattr);
1833
1834                 nfs_fattr_init(opendata->o_res.f_attr);
1835                 status = nfs4_do_setattr(state->inode, cred,
1836                                 opendata->o_res.f_attr, sattr,
1837                                 state);
1838                 if (status == 0)
1839                         nfs_setattr_update_inode(state->inode, sattr);
1840                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1841         }
1842         nfs4_opendata_put(opendata);
1843         nfs4_put_state_owner(sp);
1844         *res = state;
1845         return 0;
1846 err_opendata_put:
1847         nfs4_opendata_put(opendata);
1848 err_put_state_owner:
1849         nfs4_put_state_owner(sp);
1850 out_err:
1851         *res = NULL;
1852         return status;
1853 }
1854
1855
1856 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1857 {
1858         struct nfs4_exception exception = { };
1859         struct nfs4_state *res;
1860         int status;
1861
1862         do {
1863                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred, &res);
1864                 if (status == 0)
1865                         break;
1866                 /* NOTE: BAD_SEQID means the server and client disagree about the
1867                  * book-keeping w.r.t. state-changing operations
1868                  * (OPEN/CLOSE/LOCK/LOCKU...)
1869                  * It is actually a sign of a bug on the client or on the server.
1870                  *
1871                  * If we receive a BAD_SEQID error in the particular case of
1872                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1873                  * have unhashed the old state_owner for us, and that we can
1874                  * therefore safely retry using a new one. We should still warn
1875                  * the user though...
1876                  */
1877                 if (status == -NFS4ERR_BAD_SEQID) {
1878                         pr_warn_ratelimited("NFS: v4 server %s "
1879                                         " returned a bad sequence-id error!\n",
1880                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1881                         exception.retry = 1;
1882                         continue;
1883                 }
1884                 /*
1885                  * BAD_STATEID on OPEN means that the server cancelled our
1886                  * state before it received the OPEN_CONFIRM.
1887                  * Recover by retrying the request as per the discussion
1888                  * on Page 181 of RFC3530.
1889                  */
1890                 if (status == -NFS4ERR_BAD_STATEID) {
1891                         exception.retry = 1;
1892                         continue;
1893                 }
1894                 if (status == -EAGAIN) {
1895                         /* We must have found a delegation */
1896                         exception.retry = 1;
1897                         continue;
1898                 }
1899                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1900                                         status, &exception));
1901         } while (exception.retry);
1902         return res;
1903 }
1904
1905 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1906                             struct nfs_fattr *fattr, struct iattr *sattr,
1907                             struct nfs4_state *state)
1908 {
1909         struct nfs_server *server = NFS_SERVER(inode);
1910         struct nfs_setattrargs  arg = {
1911                 .fh             = NFS_FH(inode),
1912                 .iap            = sattr,
1913                 .server         = server,
1914                 .bitmask = server->attr_bitmask,
1915         };
1916         struct nfs_setattrres  res = {
1917                 .fattr          = fattr,
1918                 .server         = server,
1919         };
1920         struct rpc_message msg = {
1921                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1922                 .rpc_argp       = &arg,
1923                 .rpc_resp       = &res,
1924                 .rpc_cred       = cred,
1925         };
1926         unsigned long timestamp = jiffies;
1927         int status;
1928
1929         nfs_fattr_init(fattr);
1930
1931         if (state != NULL) {
1932                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
1933                                 current->files, current->tgid);
1934         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
1935                                 FMODE_WRITE)) {
1936                 /* Use that stateid */
1937         } else
1938                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
1939
1940         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1941         if (status == 0 && state != NULL)
1942                 renew_lease(server, timestamp);
1943         return status;
1944 }
1945
1946 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1947                            struct nfs_fattr *fattr, struct iattr *sattr,
1948                            struct nfs4_state *state)
1949 {
1950         struct nfs_server *server = NFS_SERVER(inode);
1951         struct nfs4_exception exception = {
1952                 .state = state,
1953                 .inode = inode,
1954         };
1955         int err;
1956         do {
1957                 err = nfs4_handle_exception(server,
1958                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1959                                 &exception);
1960         } while (exception.retry);
1961         return err;
1962 }
1963
1964 struct nfs4_closedata {
1965         struct inode *inode;
1966         struct nfs4_state *state;
1967         struct nfs_closeargs arg;
1968         struct nfs_closeres res;
1969         struct nfs_fattr fattr;
1970         unsigned long timestamp;
1971         bool roc;
1972         u32 roc_barrier;
1973 };
1974
1975 static void nfs4_free_closedata(void *data)
1976 {
1977         struct nfs4_closedata *calldata = data;
1978         struct nfs4_state_owner *sp = calldata->state->owner;
1979         struct super_block *sb = calldata->state->inode->i_sb;
1980
1981         if (calldata->roc)
1982                 pnfs_roc_release(calldata->state->inode);
1983         nfs4_put_open_state(calldata->state);
1984         nfs_free_seqid(calldata->arg.seqid);
1985         nfs4_put_state_owner(sp);
1986         nfs_sb_deactive(sb);
1987         kfree(calldata);
1988 }
1989
1990 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1991                 fmode_t fmode)
1992 {
1993         spin_lock(&state->owner->so_lock);
1994         if (!(fmode & FMODE_READ))
1995                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1996         if (!(fmode & FMODE_WRITE))
1997                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1998         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1999         spin_unlock(&state->owner->so_lock);
2000 }
2001
2002 static void nfs4_close_done(struct rpc_task *task, void *data)
2003 {
2004         struct nfs4_closedata *calldata = data;
2005         struct nfs4_state *state = calldata->state;
2006         struct nfs_server *server = NFS_SERVER(calldata->inode);
2007
2008         dprintk("%s: begin!\n", __func__);
2009         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2010                 return;
2011         /* hmm. we are done with the inode, and in the process of freeing
2012          * the state_owner. we keep this around to process errors
2013          */
2014         switch (task->tk_status) {
2015                 case 0:
2016                         if (calldata->roc)
2017                                 pnfs_roc_set_barrier(state->inode,
2018                                                      calldata->roc_barrier);
2019                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2020                         renew_lease(server, calldata->timestamp);
2021                         nfs4_close_clear_stateid_flags(state,
2022                                         calldata->arg.fmode);
2023                         break;
2024                 case -NFS4ERR_STALE_STATEID:
2025                 case -NFS4ERR_OLD_STATEID:
2026                 case -NFS4ERR_BAD_STATEID:
2027                 case -NFS4ERR_EXPIRED:
2028                         if (calldata->arg.fmode == 0)
2029                                 break;
2030                 default:
2031                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2032                                 rpc_restart_call_prepare(task);
2033         }
2034         nfs_release_seqid(calldata->arg.seqid);
2035         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2036         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2037 }
2038
2039 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2040 {
2041         struct nfs4_closedata *calldata = data;
2042         struct nfs4_state *state = calldata->state;
2043         int call_close = 0;
2044
2045         dprintk("%s: begin!\n", __func__);
2046         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2047                 return;
2048
2049         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2050         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2051         spin_lock(&state->owner->so_lock);
2052         /* Calculate the change in open mode */
2053         if (state->n_rdwr == 0) {
2054                 if (state->n_rdonly == 0) {
2055                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2056                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2057                         calldata->arg.fmode &= ~FMODE_READ;
2058                 }
2059                 if (state->n_wronly == 0) {
2060                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2061                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2062                         calldata->arg.fmode &= ~FMODE_WRITE;
2063                 }
2064         }
2065         spin_unlock(&state->owner->so_lock);
2066
2067         if (!call_close) {
2068                 /* Note: exit _without_ calling nfs4_close_done */
2069                 task->tk_action = NULL;
2070                 goto out;
2071         }
2072
2073         if (calldata->arg.fmode == 0) {
2074                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2075                 if (calldata->roc &&
2076                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2077                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2078                                      task, NULL);
2079                         goto out;
2080                 }
2081         }
2082
2083         nfs_fattr_init(calldata->res.fattr);
2084         calldata->timestamp = jiffies;
2085         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2086                                 &calldata->arg.seq_args,
2087                                 &calldata->res.seq_res,
2088                                 task))
2089                 goto out;
2090         rpc_call_start(task);
2091 out:
2092         dprintk("%s: done!\n", __func__);
2093 }
2094
2095 static const struct rpc_call_ops nfs4_close_ops = {
2096         .rpc_call_prepare = nfs4_close_prepare,
2097         .rpc_call_done = nfs4_close_done,
2098         .rpc_release = nfs4_free_closedata,
2099 };
2100
2101 /*
2102  * It is possible for data to be read/written from a mem-mapped file
2103  * after the sys_close call (which hits the vfs layer as a flush).
2104  * This means that we can't safely call nfsv4 close on a file until
2105  * the inode is cleared. This in turn means that we are not good
2106  * NFSv4 citizens - we do not indicate to the server to update the file's
2107  * share state even when we are done with one of the three share
2108  * stateid's in the inode.
2109  *
2110  * NOTE: Caller must be holding the sp->so_owner semaphore!
2111  */
2112 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2113 {
2114         struct nfs_server *server = NFS_SERVER(state->inode);
2115         struct nfs4_closedata *calldata;
2116         struct nfs4_state_owner *sp = state->owner;
2117         struct rpc_task *task;
2118         struct rpc_message msg = {
2119                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2120                 .rpc_cred = state->owner->so_cred,
2121         };
2122         struct rpc_task_setup task_setup_data = {
2123                 .rpc_client = server->client,
2124                 .rpc_message = &msg,
2125                 .callback_ops = &nfs4_close_ops,
2126                 .workqueue = nfsiod_workqueue,
2127                 .flags = RPC_TASK_ASYNC,
2128         };
2129         int status = -ENOMEM;
2130
2131         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2132         if (calldata == NULL)
2133                 goto out;
2134         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2135         calldata->inode = state->inode;
2136         calldata->state = state;
2137         calldata->arg.fh = NFS_FH(state->inode);
2138         calldata->arg.stateid = &state->open_stateid;
2139         /* Serialization for the sequence id */
2140         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2141         if (calldata->arg.seqid == NULL)
2142                 goto out_free_calldata;
2143         calldata->arg.fmode = 0;
2144         calldata->arg.bitmask = server->cache_consistency_bitmask;
2145         calldata->res.fattr = &calldata->fattr;
2146         calldata->res.seqid = calldata->arg.seqid;
2147         calldata->res.server = server;
2148         calldata->roc = roc;
2149         nfs_sb_active(calldata->inode->i_sb);
2150
2151         msg.rpc_argp = &calldata->arg;
2152         msg.rpc_resp = &calldata->res;
2153         task_setup_data.callback_data = calldata;
2154         task = rpc_run_task(&task_setup_data);
2155         if (IS_ERR(task))
2156                 return PTR_ERR(task);
2157         status = 0;
2158         if (wait)
2159                 status = rpc_wait_for_completion_task(task);
2160         rpc_put_task(task);
2161         return status;
2162 out_free_calldata:
2163         kfree(calldata);
2164 out:
2165         if (roc)
2166                 pnfs_roc_release(state->inode);
2167         nfs4_put_open_state(state);
2168         nfs4_put_state_owner(sp);
2169         return status;
2170 }
2171
2172 static struct inode *
2173 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2174 {
2175         struct nfs4_state *state;
2176
2177         /* Protect against concurrent sillydeletes */
2178         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2179         if (IS_ERR(state))
2180                 return ERR_CAST(state);
2181         ctx->state = state;
2182         return igrab(state->inode);
2183 }
2184
2185 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2186 {
2187         if (ctx->state == NULL)
2188                 return;
2189         if (is_sync)
2190                 nfs4_close_sync(ctx->state, ctx->mode);
2191         else
2192                 nfs4_close_state(ctx->state, ctx->mode);
2193 }
2194
2195 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2196 {
2197         struct nfs4_server_caps_arg args = {
2198                 .fhandle = fhandle,
2199         };
2200         struct nfs4_server_caps_res res = {};
2201         struct rpc_message msg = {
2202                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2203                 .rpc_argp = &args,
2204                 .rpc_resp = &res,
2205         };
2206         int status;
2207
2208         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2209         if (status == 0) {
2210                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2211                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2212                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2213                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2214                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2215                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2216                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2217                         server->caps |= NFS_CAP_ACLS;
2218                 if (res.has_links != 0)
2219                         server->caps |= NFS_CAP_HARDLINKS;
2220                 if (res.has_symlinks != 0)
2221                         server->caps |= NFS_CAP_SYMLINKS;
2222                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2223                         server->caps |= NFS_CAP_FILEID;
2224                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2225                         server->caps |= NFS_CAP_MODE;
2226                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2227                         server->caps |= NFS_CAP_NLINK;
2228                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2229                         server->caps |= NFS_CAP_OWNER;
2230                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2231                         server->caps |= NFS_CAP_OWNER_GROUP;
2232                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2233                         server->caps |= NFS_CAP_ATIME;
2234                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2235                         server->caps |= NFS_CAP_CTIME;
2236                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2237                         server->caps |= NFS_CAP_MTIME;
2238
2239                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2240                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2241                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2242                 server->acl_bitmask = res.acl_bitmask;
2243                 server->fh_expire_type = res.fh_expire_type;
2244         }
2245
2246         return status;
2247 }
2248
2249 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2250 {
2251         struct nfs4_exception exception = { };
2252         int err;
2253         do {
2254                 err = nfs4_handle_exception(server,
2255                                 _nfs4_server_capabilities(server, fhandle),
2256                                 &exception);
2257         } while (exception.retry);
2258         return err;
2259 }
2260
2261 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2262                 struct nfs_fsinfo *info)
2263 {
2264         struct nfs4_lookup_root_arg args = {
2265                 .bitmask = nfs4_fattr_bitmap,
2266         };
2267         struct nfs4_lookup_res res = {
2268                 .server = server,
2269                 .fattr = info->fattr,
2270                 .fh = fhandle,
2271         };
2272         struct rpc_message msg = {
2273                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2274                 .rpc_argp = &args,
2275                 .rpc_resp = &res,
2276         };
2277
2278         nfs_fattr_init(info->fattr);
2279         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2280 }
2281
2282 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2283                 struct nfs_fsinfo *info)
2284 {
2285         struct nfs4_exception exception = { };
2286         int err;
2287         do {
2288                 err = _nfs4_lookup_root(server, fhandle, info);
2289                 switch (err) {
2290                 case 0:
2291                 case -NFS4ERR_WRONGSEC:
2292                         goto out;
2293                 default:
2294                         err = nfs4_handle_exception(server, err, &exception);
2295                 }
2296         } while (exception.retry);
2297 out:
2298         return err;
2299 }
2300
2301 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2302                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2303 {
2304         struct rpc_auth *auth;
2305         int ret;
2306
2307         auth = rpcauth_create(flavor, server->client);
2308         if (!auth) {
2309                 ret = -EIO;
2310                 goto out;
2311         }
2312         ret = nfs4_lookup_root(server, fhandle, info);
2313 out:
2314         return ret;
2315 }
2316
2317 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2318                               struct nfs_fsinfo *info)
2319 {
2320         int i, len, status = 0;
2321         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2322
2323         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2324         flav_array[len] = RPC_AUTH_NULL;
2325         len += 1;
2326
2327         for (i = 0; i < len; i++) {
2328                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2329                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2330                         continue;
2331                 break;
2332         }
2333         /*
2334          * -EACCESS could mean that the user doesn't have correct permissions
2335          * to access the mount.  It could also mean that we tried to mount
2336          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2337          * existing mount programs don't handle -EACCES very well so it should
2338          * be mapped to -EPERM instead.
2339          */
2340         if (status == -EACCES)
2341                 status = -EPERM;
2342         return status;
2343 }
2344
2345 /*
2346  * get the file handle for the "/" directory on the server
2347  */
2348 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2349                               struct nfs_fsinfo *info)
2350 {
2351         int minor_version = server->nfs_client->cl_minorversion;
2352         int status = nfs4_lookup_root(server, fhandle, info);
2353         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2354                 /*
2355                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2356                  * by nfs4_map_errors() as this function exits.
2357                  */
2358                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2359         if (status == 0)
2360                 status = nfs4_server_capabilities(server, fhandle);
2361         if (status == 0)
2362                 status = nfs4_do_fsinfo(server, fhandle, info);
2363         return nfs4_map_errors(status);
2364 }
2365
2366 /*
2367  * Get locations and (maybe) other attributes of a referral.
2368  * Note that we'll actually follow the referral later when
2369  * we detect fsid mismatch in inode revalidation
2370  */
2371 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2372                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2373 {
2374         int status = -ENOMEM;
2375         struct page *page = NULL;
2376         struct nfs4_fs_locations *locations = NULL;
2377
2378         page = alloc_page(GFP_KERNEL);
2379         if (page == NULL)
2380                 goto out;
2381         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2382         if (locations == NULL)
2383                 goto out;
2384
2385         status = nfs4_proc_fs_locations(dir, name, locations, page);
2386         if (status != 0)
2387                 goto out;
2388         /* Make sure server returned a different fsid for the referral */
2389         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2390                 dprintk("%s: server did not return a different fsid for"
2391                         " a referral at %s\n", __func__, name->name);
2392                 status = -EIO;
2393                 goto out;
2394         }
2395         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2396         nfs_fixup_referral_attributes(&locations->fattr);
2397
2398         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2399         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2400         memset(fhandle, 0, sizeof(struct nfs_fh));
2401 out:
2402         if (page)
2403                 __free_page(page);
2404         kfree(locations);
2405         return status;
2406 }
2407
2408 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2409 {
2410         struct nfs4_getattr_arg args = {
2411                 .fh = fhandle,
2412                 .bitmask = server->attr_bitmask,
2413         };
2414         struct nfs4_getattr_res res = {
2415                 .fattr = fattr,
2416                 .server = server,
2417         };
2418         struct rpc_message msg = {
2419                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2420                 .rpc_argp = &args,
2421                 .rpc_resp = &res,
2422         };
2423
2424         nfs_fattr_init(fattr);
2425         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2426 }
2427
2428 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2429 {
2430         struct nfs4_exception exception = { };
2431         int err;
2432         do {
2433                 err = nfs4_handle_exception(server,
2434                                 _nfs4_proc_getattr(server, fhandle, fattr),
2435                                 &exception);
2436         } while (exception.retry);
2437         return err;
2438 }
2439
2440 /*
2441  * The file is not closed if it is opened due to the a request to change
2442  * the size of the file. The open call will not be needed once the
2443  * VFS layer lookup-intents are implemented.
2444  *
2445  * Close is called when the inode is destroyed.
2446  * If we haven't opened the file for O_WRONLY, we
2447  * need to in the size_change case to obtain a stateid.
2448  *
2449  * Got race?
2450  * Because OPEN is always done by name in nfsv4, it is
2451  * possible that we opened a different file by the same
2452  * name.  We can recognize this race condition, but we
2453  * can't do anything about it besides returning an error.
2454  *
2455  * This will be fixed with VFS changes (lookup-intent).
2456  */
2457 static int
2458 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2459                   struct iattr *sattr)
2460 {
2461         struct inode *inode = dentry->d_inode;
2462         struct rpc_cred *cred = NULL;
2463         struct nfs4_state *state = NULL;
2464         int status;
2465
2466         if (pnfs_ld_layoutret_on_setattr(inode))
2467                 pnfs_return_layout(inode);
2468
2469         nfs_fattr_init(fattr);
2470
2471         /* Search for an existing open(O_WRITE) file */
2472         if (sattr->ia_valid & ATTR_FILE) {
2473                 struct nfs_open_context *ctx;
2474
2475                 ctx = nfs_file_open_context(sattr->ia_file);
2476                 if (ctx) {
2477                         cred = ctx->cred;
2478                         state = ctx->state;
2479                 }
2480         }
2481
2482         /* Deal with open(O_TRUNC) */
2483         if (sattr->ia_valid & ATTR_OPEN)
2484                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2485
2486         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2487         if (status == 0)
2488                 nfs_setattr_update_inode(inode, sattr);
2489         return status;
2490 }
2491
2492 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2493                 const struct qstr *name, struct nfs_fh *fhandle,
2494                 struct nfs_fattr *fattr)
2495 {
2496         struct nfs_server *server = NFS_SERVER(dir);
2497         int                    status;
2498         struct nfs4_lookup_arg args = {
2499                 .bitmask = server->attr_bitmask,
2500                 .dir_fh = NFS_FH(dir),
2501                 .name = name,
2502         };
2503         struct nfs4_lookup_res res = {
2504                 .server = server,
2505                 .fattr = fattr,
2506                 .fh = fhandle,
2507         };
2508         struct rpc_message msg = {
2509                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2510                 .rpc_argp = &args,
2511                 .rpc_resp = &res,
2512         };
2513
2514         nfs_fattr_init(fattr);
2515
2516         dprintk("NFS call  lookup %s\n", name->name);
2517         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2518         dprintk("NFS reply lookup: %d\n", status);
2519         return status;
2520 }
2521
2522 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2523 {
2524         memset(fh, 0, sizeof(struct nfs_fh));
2525         fattr->fsid.major = 1;
2526         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2527                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2528         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2529         fattr->nlink = 2;
2530 }
2531
2532 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2533                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2534 {
2535         struct nfs4_exception exception = { };
2536         int err;
2537         do {
2538                 int status;
2539
2540                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2541                 switch (status) {
2542                 case -NFS4ERR_BADNAME:
2543                         return -ENOENT;
2544                 case -NFS4ERR_MOVED:
2545                         return nfs4_get_referral(dir, name, fattr, fhandle);
2546                 case -NFS4ERR_WRONGSEC:
2547                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2548                 }
2549                 err = nfs4_handle_exception(NFS_SERVER(dir),
2550                                 status, &exception);
2551         } while (exception.retry);
2552         return err;
2553 }
2554
2555 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2556 {
2557         struct nfs_server *server = NFS_SERVER(inode);
2558         struct nfs4_accessargs args = {
2559                 .fh = NFS_FH(inode),
2560                 .bitmask = server->cache_consistency_bitmask,
2561         };
2562         struct nfs4_accessres res = {
2563                 .server = server,
2564         };
2565         struct rpc_message msg = {
2566                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2567                 .rpc_argp = &args,
2568                 .rpc_resp = &res,
2569                 .rpc_cred = entry->cred,
2570         };
2571         int mode = entry->mask;
2572         int status;
2573
2574         /*
2575          * Determine which access bits we want to ask for...
2576          */
2577         if (mode & MAY_READ)
2578                 args.access |= NFS4_ACCESS_READ;
2579         if (S_ISDIR(inode->i_mode)) {
2580                 if (mode & MAY_WRITE)
2581                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2582                 if (mode & MAY_EXEC)
2583                         args.access |= NFS4_ACCESS_LOOKUP;
2584         } else {
2585                 if (mode & MAY_WRITE)
2586                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2587                 if (mode & MAY_EXEC)
2588                         args.access |= NFS4_ACCESS_EXECUTE;
2589         }
2590
2591         res.fattr = nfs_alloc_fattr();
2592         if (res.fattr == NULL)
2593                 return -ENOMEM;
2594
2595         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2596         if (!status) {
2597                 entry->mask = 0;
2598                 if (res.access & NFS4_ACCESS_READ)
2599                         entry->mask |= MAY_READ;
2600                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2601                         entry->mask |= MAY_WRITE;
2602                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2603                         entry->mask |= MAY_EXEC;
2604                 nfs_refresh_inode(inode, res.fattr);
2605         }
2606         nfs_free_fattr(res.fattr);
2607         return status;
2608 }
2609
2610 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2611 {
2612         struct nfs4_exception exception = { };
2613         int err;
2614         do {
2615                 err = nfs4_handle_exception(NFS_SERVER(inode),
2616                                 _nfs4_proc_access(inode, entry),
2617                                 &exception);
2618         } while (exception.retry);
2619         return err;
2620 }
2621
2622 /*
2623  * TODO: For the time being, we don't try to get any attributes
2624  * along with any of the zero-copy operations READ, READDIR,
2625  * READLINK, WRITE.
2626  *
2627  * In the case of the first three, we want to put the GETATTR
2628  * after the read-type operation -- this is because it is hard
2629  * to predict the length of a GETATTR response in v4, and thus
2630  * align the READ data correctly.  This means that the GETATTR
2631  * may end up partially falling into the page cache, and we should
2632  * shift it into the 'tail' of the xdr_buf before processing.
2633  * To do this efficiently, we need to know the total length
2634  * of data received, which doesn't seem to be available outside
2635  * of the RPC layer.
2636  *
2637  * In the case of WRITE, we also want to put the GETATTR after
2638  * the operation -- in this case because we want to make sure
2639  * we get the post-operation mtime and size.  This means that
2640  * we can't use xdr_encode_pages() as written: we need a variant
2641  * of it which would leave room in the 'tail' iovec.
2642  *
2643  * Both of these changes to the XDR layer would in fact be quite
2644  * minor, but I decided to leave them for a subsequent patch.
2645  */
2646 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2647                 unsigned int pgbase, unsigned int pglen)
2648 {
2649         struct nfs4_readlink args = {
2650                 .fh       = NFS_FH(inode),
2651                 .pgbase   = pgbase,
2652                 .pglen    = pglen,
2653                 .pages    = &page,
2654         };
2655         struct nfs4_readlink_res res;
2656         struct rpc_message msg = {
2657                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2658                 .rpc_argp = &args,
2659                 .rpc_resp = &res,
2660         };
2661
2662         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2663 }
2664
2665 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2666                 unsigned int pgbase, unsigned int pglen)
2667 {
2668         struct nfs4_exception exception = { };
2669         int err;
2670         do {
2671                 err = nfs4_handle_exception(NFS_SERVER(inode),
2672                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2673                                 &exception);
2674         } while (exception.retry);
2675         return err;
2676 }
2677
2678 /*
2679  * Got race?
2680  * We will need to arrange for the VFS layer to provide an atomic open.
2681  * Until then, this create/open method is prone to inefficiency and race
2682  * conditions due to the lookup, create, and open VFS calls from sys_open()
2683  * placed on the wire.
2684  *
2685  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2686  * The file will be opened again in the subsequent VFS open call
2687  * (nfs4_proc_file_open).
2688  *
2689  * The open for read will just hang around to be used by any process that
2690  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2691  */
2692
2693 static int
2694 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2695                  int flags, struct nfs_open_context *ctx)
2696 {
2697         struct dentry *de = dentry;
2698         struct nfs4_state *state;
2699         struct rpc_cred *cred = NULL;
2700         fmode_t fmode = 0;
2701         int status = 0;
2702
2703         if (ctx != NULL) {
2704                 cred = ctx->cred;
2705                 de = ctx->dentry;
2706                 fmode = ctx->mode;
2707         }
2708         sattr->ia_mode &= ~current_umask();
2709         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2710         d_drop(dentry);
2711         if (IS_ERR(state)) {
2712                 status = PTR_ERR(state);
2713                 goto out;
2714         }
2715         d_add(dentry, igrab(state->inode));
2716         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2717         if (ctx != NULL)
2718                 ctx->state = state;
2719         else
2720                 nfs4_close_sync(state, fmode);
2721 out:
2722         return status;
2723 }
2724
2725 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2726 {
2727         struct nfs_server *server = NFS_SERVER(dir);
2728         struct nfs_removeargs args = {
2729                 .fh = NFS_FH(dir),
2730                 .name.len = name->len,
2731                 .name.name = name->name,
2732                 .bitmask = server->attr_bitmask,
2733         };
2734         struct nfs_removeres res = {
2735                 .server = server,
2736         };
2737         struct rpc_message msg = {
2738                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2739                 .rpc_argp = &args,
2740                 .rpc_resp = &res,
2741         };
2742         int status = -ENOMEM;
2743
2744         res.dir_attr = nfs_alloc_fattr();
2745         if (res.dir_attr == NULL)
2746                 goto out;
2747
2748         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2749         if (status == 0) {
2750                 update_changeattr(dir, &res.cinfo);
2751                 nfs_post_op_update_inode(dir, res.dir_attr);
2752         }
2753         nfs_free_fattr(res.dir_attr);
2754 out:
2755         return status;
2756 }
2757
2758 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2759 {
2760         struct nfs4_exception exception = { };
2761         int err;
2762         do {
2763                 err = nfs4_handle_exception(NFS_SERVER(dir),
2764                                 _nfs4_proc_remove(dir, name),
2765                                 &exception);
2766         } while (exception.retry);
2767         return err;
2768 }
2769
2770 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2771 {
2772         struct nfs_server *server = NFS_SERVER(dir);
2773         struct nfs_removeargs *args = msg->rpc_argp;
2774         struct nfs_removeres *res = msg->rpc_resp;
2775
2776         args->bitmask = server->cache_consistency_bitmask;
2777         res->server = server;
2778         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2779         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2780 }
2781
2782 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
2783 {
2784         if (nfs4_setup_sequence(NFS_SERVER(data->dir),
2785                                 &data->args.seq_args,
2786                                 &data->res.seq_res,
2787                                 task))
2788                 return;
2789         rpc_call_start(task);
2790 }
2791
2792 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2793 {
2794         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2795
2796         if (!nfs4_sequence_done(task, &res->seq_res))
2797                 return 0;
2798         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2799                 return 0;
2800         update_changeattr(dir, &res->cinfo);
2801         nfs_post_op_update_inode(dir, res->dir_attr);
2802         return 1;
2803 }
2804
2805 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2806 {
2807         struct nfs_server *server = NFS_SERVER(dir);
2808         struct nfs_renameargs *arg = msg->rpc_argp;
2809         struct nfs_renameres *res = msg->rpc_resp;
2810
2811         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2812         arg->bitmask = server->attr_bitmask;
2813         res->server = server;
2814         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2815 }
2816
2817 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
2818 {
2819         if (nfs4_setup_sequence(NFS_SERVER(data->old_dir),
2820                                 &data->args.seq_args,
2821                                 &data->res.seq_res,
2822                                 task))
2823                 return;
2824         rpc_call_start(task);
2825 }
2826
2827 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2828                                  struct inode *new_dir)
2829 {
2830         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2831
2832         if (!nfs4_sequence_done(task, &res->seq_res))
2833                 return 0;
2834         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2835                 return 0;
2836
2837         update_changeattr(old_dir, &res->old_cinfo);
2838         nfs_post_op_update_inode(old_dir, res->old_fattr);
2839         update_changeattr(new_dir, &res->new_cinfo);
2840         nfs_post_op_update_inode(new_dir, res->new_fattr);
2841         return 1;
2842 }
2843
2844 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2845                 struct inode *new_dir, struct qstr *new_name)
2846 {
2847         struct nfs_server *server = NFS_SERVER(old_dir);
2848         struct nfs_renameargs arg = {
2849                 .old_dir = NFS_FH(old_dir),
2850                 .new_dir = NFS_FH(new_dir),
2851                 .old_name = old_name,
2852                 .new_name = new_name,
2853                 .bitmask = server->attr_bitmask,
2854         };
2855         struct nfs_renameres res = {
2856                 .server = server,
2857         };
2858         struct rpc_message msg = {
2859                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2860                 .rpc_argp = &arg,
2861                 .rpc_resp = &res,
2862         };
2863         int status = -ENOMEM;
2864
2865         res.old_fattr = nfs_alloc_fattr();
2866         res.new_fattr = nfs_alloc_fattr();
2867         if (res.old_fattr == NULL || res.new_fattr == NULL)
2868                 goto out;
2869
2870         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2871         if (!status) {
2872                 update_changeattr(old_dir, &res.old_cinfo);
2873                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2874                 update_changeattr(new_dir, &res.new_cinfo);
2875                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2876         }
2877 out:
2878         nfs_free_fattr(res.new_fattr);
2879         nfs_free_fattr(res.old_fattr);
2880         return status;
2881 }
2882
2883 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2884                 struct inode *new_dir, struct qstr *new_name)
2885 {
2886         struct nfs4_exception exception = { };
2887         int err;
2888         do {
2889                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2890                                 _nfs4_proc_rename(old_dir, old_name,
2891                                         new_dir, new_name),
2892                                 &exception);
2893         } while (exception.retry);
2894         return err;
2895 }
2896
2897 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2898 {
2899         struct nfs_server *server = NFS_SERVER(inode);
2900         struct nfs4_link_arg arg = {
2901                 .fh     = NFS_FH(inode),
2902                 .dir_fh = NFS_FH(dir),
2903                 .name   = name,
2904                 .bitmask = server->attr_bitmask,
2905         };
2906         struct nfs4_link_res res = {
2907                 .server = server,
2908         };
2909         struct rpc_message msg = {
2910                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2911                 .rpc_argp = &arg,
2912                 .rpc_resp = &res,
2913         };
2914         int status = -ENOMEM;
2915
2916         res.fattr = nfs_alloc_fattr();
2917         res.dir_attr = nfs_alloc_fattr();
2918         if (res.fattr == NULL || res.dir_attr == NULL)
2919                 goto out;
2920
2921         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2922         if (!status) {
2923                 update_changeattr(dir, &res.cinfo);
2924                 nfs_post_op_update_inode(dir, res.dir_attr);
2925                 nfs_post_op_update_inode(inode, res.fattr);
2926         }
2927 out:
2928         nfs_free_fattr(res.dir_attr);
2929         nfs_free_fattr(res.fattr);
2930         return status;
2931 }
2932
2933 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2934 {
2935         struct nfs4_exception exception = { };
2936         int err;
2937         do {
2938                 err = nfs4_handle_exception(NFS_SERVER(inode),
2939                                 _nfs4_proc_link(inode, dir, name),
2940                                 &exception);
2941         } while (exception.retry);
2942         return err;
2943 }
2944
2945 struct nfs4_createdata {
2946         struct rpc_message msg;
2947         struct nfs4_create_arg arg;
2948         struct nfs4_create_res res;
2949         struct nfs_fh fh;
2950         struct nfs_fattr fattr;
2951         struct nfs_fattr dir_fattr;
2952 };
2953
2954 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2955                 struct qstr *name, struct iattr *sattr, u32 ftype)
2956 {
2957         struct nfs4_createdata *data;
2958
2959         data = kzalloc(sizeof(*data), GFP_KERNEL);
2960         if (data != NULL) {
2961                 struct nfs_server *server = NFS_SERVER(dir);
2962
2963                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2964                 data->msg.rpc_argp = &data->arg;
2965                 data->msg.rpc_resp = &data->res;
2966                 data->arg.dir_fh = NFS_FH(dir);
2967                 data->arg.server = server;
2968                 data->arg.name = name;
2969                 data->arg.attrs = sattr;
2970                 data->arg.ftype = ftype;
2971                 data->arg.bitmask = server->attr_bitmask;
2972                 data->res.server = server;
2973                 data->res.fh = &data->fh;
2974                 data->res.fattr = &data->fattr;
2975                 data->res.dir_fattr = &data->dir_fattr;
2976                 nfs_fattr_init(data->res.fattr);
2977                 nfs_fattr_init(data->res.dir_fattr);
2978         }
2979         return data;
2980 }
2981
2982 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2983 {
2984         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2985                                     &data->arg.seq_args, &data->res.seq_res, 1);
2986         if (status == 0) {
2987                 update_changeattr(dir, &data->res.dir_cinfo);
2988                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2989                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2990         }
2991         return status;
2992 }
2993
2994 static void nfs4_free_createdata(struct nfs4_createdata *data)
2995 {
2996         kfree(data);
2997 }
2998
2999 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3000                 struct page *page, unsigned int len, struct iattr *sattr)
3001 {
3002         struct nfs4_createdata *data;
3003         int status = -ENAMETOOLONG;
3004
3005         if (len > NFS4_MAXPATHLEN)
3006                 goto out;
3007
3008         status = -ENOMEM;
3009         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3010         if (data == NULL)
3011                 goto out;
3012
3013         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3014         data->arg.u.symlink.pages = &page;
3015         data->arg.u.symlink.len = len;
3016
3017         status = nfs4_do_create(dir, dentry, data);
3018
3019         nfs4_free_createdata(data);
3020 out:
3021         return status;
3022 }
3023
3024 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3025                 struct page *page, unsigned int len, struct iattr *sattr)
3026 {
3027         struct nfs4_exception exception = { };
3028         int err;
3029         do {
3030                 err = nfs4_handle_exception(NFS_SERVER(dir),
3031                                 _nfs4_proc_symlink(dir, dentry, page,
3032                                                         len, sattr),
3033                                 &exception);
3034         } while (exception.retry);
3035         return err;
3036 }
3037
3038 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3039                 struct iattr *sattr)
3040 {
3041         struct nfs4_createdata *data;
3042         int status = -ENOMEM;
3043
3044         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3045         if (data == NULL)
3046                 goto out;
3047
3048         status = nfs4_do_create(dir, dentry, data);
3049
3050         nfs4_free_createdata(data);
3051 out:
3052         return status;
3053 }
3054
3055 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3056                 struct iattr *sattr)
3057 {
3058         struct nfs4_exception exception = { };
3059         int err;
3060
3061         sattr->ia_mode &= ~current_umask();
3062         do {
3063                 err = nfs4_handle_exception(NFS_SERVER(dir),
3064                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3065                                 &exception);
3066         } while (exception.retry);
3067         return err;
3068 }
3069
3070 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3071                 u64 cookie, struct page **pages, unsigned int count, int plus)
3072 {
3073         struct inode            *dir = dentry->d_inode;
3074         struct nfs4_readdir_arg args = {
3075                 .fh = NFS_FH(dir),
3076                 .pages = pages,
3077                 .pgbase = 0,
3078                 .count = count,
3079                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3080                 .plus = plus,
3081         };
3082         struct nfs4_readdir_res res;
3083         struct rpc_message msg = {
3084                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3085                 .rpc_argp = &args,
3086                 .rpc_resp = &res,
3087                 .rpc_cred = cred,
3088         };
3089         int                     status;
3090
3091         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3092                         dentry->d_parent->d_name.name,
3093                         dentry->d_name.name,
3094                         (unsigned long long)cookie);
3095         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3096         res.pgbase = args.pgbase;
3097         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3098         if (status >= 0) {
3099                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3100                 status += args.pgbase;
3101         }
3102
3103         nfs_invalidate_atime(dir);
3104
3105         dprintk("%s: returns %d\n", __func__, status);
3106         return status;
3107 }
3108
3109 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3110                 u64 cookie, struct page **pages, unsigned int count, int plus)
3111 {
3112         struct nfs4_exception exception = { };
3113         int err;
3114         do {
3115                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3116                                 _nfs4_proc_readdir(dentry, cred, cookie,
3117                                         pages, count, plus),
3118                                 &exception);
3119         } while (exception.retry);
3120         return err;
3121 }
3122
3123 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3124                 struct iattr *sattr, dev_t rdev)
3125 {
3126         struct nfs4_createdata *data;
3127         int mode = sattr->ia_mode;
3128         int status = -ENOMEM;
3129
3130         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3131         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3132
3133         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3134         if (data == NULL)
3135                 goto out;
3136
3137         if (S_ISFIFO(mode))
3138                 data->arg.ftype = NF4FIFO;
3139         else if (S_ISBLK(mode)) {
3140                 data->arg.ftype = NF4BLK;
3141                 data->arg.u.device.specdata1 = MAJOR(rdev);
3142                 data->arg.u.device.specdata2 = MINOR(rdev);
3143         }
3144         else if (S_ISCHR(mode)) {
3145                 data->arg.ftype = NF4CHR;
3146                 data->arg.u.device.specdata1 = MAJOR(rdev);
3147                 data->arg.u.device.specdata2 = MINOR(rdev);
3148         }
3149
3150         status = nfs4_do_create(dir, dentry, data);
3151
3152         nfs4_free_createdata(data);
3153 out:
3154         return status;
3155 }
3156
3157 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3158                 struct iattr *sattr, dev_t rdev)
3159 {
3160         struct nfs4_exception exception = { };
3161         int err;
3162
3163         sattr->ia_mode &= ~current_umask();
3164         do {
3165                 err = nfs4_handle_exception(NFS_SERVER(dir),
3166                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3167                                 &exception);
3168         } while (exception.retry);
3169         return err;
3170 }
3171
3172 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3173                  struct nfs_fsstat *fsstat)
3174 {
3175         struct nfs4_statfs_arg args = {
3176                 .fh = fhandle,
3177                 .bitmask = server->attr_bitmask,
3178         };
3179         struct nfs4_statfs_res res = {
3180                 .fsstat = fsstat,
3181         };
3182         struct rpc_message msg = {
3183                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3184                 .rpc_argp = &args,
3185                 .rpc_resp = &res,
3186         };
3187
3188         nfs_fattr_init(fsstat->fattr);
3189         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3190 }
3191
3192 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3193 {
3194         struct nfs4_exception exception = { };
3195         int err;
3196         do {
3197                 err = nfs4_handle_exception(server,
3198                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3199                                 &exception);
3200         } while (exception.retry);
3201         return err;
3202 }
3203
3204 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3205                 struct nfs_fsinfo *fsinfo)
3206 {
3207         struct nfs4_fsinfo_arg args = {
3208                 .fh = fhandle,
3209                 .bitmask = server->attr_bitmask,
3210         };
3211         struct nfs4_fsinfo_res res = {
3212                 .fsinfo = fsinfo,
3213         };
3214         struct rpc_message msg = {
3215                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3216                 .rpc_argp = &args,
3217                 .rpc_resp = &res,
3218         };
3219
3220         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3221 }
3222
3223 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3224 {
3225         struct nfs4_exception exception = { };
3226         int err;
3227
3228         do {
3229                 err = nfs4_handle_exception(server,
3230                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3231                                 &exception);
3232         } while (exception.retry);
3233         return err;
3234 }
3235
3236 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3237 {
3238         nfs_fattr_init(fsinfo->fattr);
3239         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3240 }
3241
3242 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3243                 struct nfs_pathconf *pathconf)
3244 {
3245         struct nfs4_pathconf_arg args = {
3246                 .fh = fhandle,
3247                 .bitmask = server->attr_bitmask,
3248         };
3249         struct nfs4_pathconf_res res = {
3250                 .pathconf = pathconf,
3251         };
3252         struct rpc_message msg = {
3253                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3254                 .rpc_argp = &args,
3255                 .rpc_resp = &res,
3256         };
3257
3258         /* None of the pathconf attributes are mandatory to implement */
3259         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3260                 memset(pathconf, 0, sizeof(*pathconf));
3261                 return 0;
3262         }
3263
3264         nfs_fattr_init(pathconf->fattr);
3265         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3266 }
3267
3268 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3269                 struct nfs_pathconf *pathconf)
3270 {
3271         struct nfs4_exception exception = { };
3272         int err;
3273
3274         do {
3275                 err = nfs4_handle_exception(server,
3276                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3277                                 &exception);
3278         } while (exception.retry);
3279         return err;
3280 }
3281
3282 void __nfs4_read_done_cb(struct nfs_read_data *data)
3283 {
3284         nfs_invalidate_atime(data->inode);
3285 }
3286
3287 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3288 {
3289         struct nfs_server *server = NFS_SERVER(data->inode);
3290
3291         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3292                 rpc_restart_call_prepare(task);
3293                 return -EAGAIN;
3294         }
3295
3296         __nfs4_read_done_cb(data);
3297         if (task->tk_status > 0)
3298                 renew_lease(server, data->timestamp);
3299         return 0;
3300 }
3301
3302 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3303 {
3304
3305         dprintk("--> %s\n", __func__);
3306
3307         if (!nfs4_sequence_done(task, &data->res.seq_res))
3308                 return -EAGAIN;
3309
3310         return data->read_done_cb ? data->read_done_cb(task, data) :
3311                                     nfs4_read_done_cb(task, data);
3312 }
3313
3314 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3315 {
3316         data->timestamp   = jiffies;
3317         data->read_done_cb = nfs4_read_done_cb;
3318         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3319         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3320 }
3321
3322 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3323 {
3324         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3325                                 &data->args.seq_args,
3326                                 &data->res.seq_res,
3327                                 task))
3328                 return;
3329         rpc_call_start(task);
3330 }
3331
3332 /* Reset the the nfs_read_data to send the read to the MDS. */
3333 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3334 {
3335         dprintk("%s Reset task for i/o through\n", __func__);
3336         put_lseg(data->lseg);
3337         data->lseg = NULL;
3338         /* offsets will differ in the dense stripe case */
3339         data->args.offset = data->mds_offset;
3340         data->ds_clp = NULL;
3341         data->args.fh     = NFS_FH(data->inode);
3342         data->read_done_cb = nfs4_read_done_cb;
3343         task->tk_ops = data->mds_ops;
3344         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3345 }
3346 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3347
3348 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3349 {
3350         struct inode *inode = data->inode;
3351
3352         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3353                 rpc_restart_call_prepare(task);
3354                 return -EAGAIN;
3355         }
3356         if (task->tk_status >= 0) {
3357                 renew_lease(NFS_SERVER(inode), data->timestamp);
3358                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3359         }
3360         return 0;
3361 }
3362
3363 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3364 {
3365         if (!nfs4_sequence_done(task, &data->res.seq_res))
3366                 return -EAGAIN;
3367         return data->write_done_cb ? data->write_done_cb(task, data) :
3368                 nfs4_write_done_cb(task, data);
3369 }
3370
3371 /* Reset the the nfs_write_data to send the write to the MDS. */
3372 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3373 {
3374         dprintk("%s Reset task for i/o through\n", __func__);
3375         put_lseg(data->lseg);
3376         data->lseg          = NULL;
3377         data->ds_clp        = NULL;
3378         data->write_done_cb = nfs4_write_done_cb;
3379         data->args.fh       = NFS_FH(data->inode);
3380         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3381         data->args.offset   = data->mds_offset;
3382         data->res.fattr     = &data->fattr;
3383         task->tk_ops        = data->mds_ops;
3384         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3385 }
3386 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3387
3388 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3389 {
3390         struct nfs_server *server = NFS_SERVER(data->inode);
3391
3392         if (data->lseg) {
3393                 data->args.bitmask = NULL;
3394                 data->res.fattr = NULL;
3395         } else
3396                 data->args.bitmask = server->cache_consistency_bitmask;
3397         if (!data->write_done_cb)
3398                 data->write_done_cb = nfs4_write_done_cb;
3399         data->res.server = server;
3400         data->timestamp   = jiffies;
3401
3402         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3403         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3404 }
3405
3406 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3407 {
3408         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3409                                 &data->args.seq_args,
3410                                 &data->res.seq_res,
3411                                 task))
3412                 return;
3413         rpc_call_start(task);
3414 }
3415
3416 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3417 {
3418         struct inode *inode = data->inode;
3419
3420         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3421                 rpc_restart_call_prepare(task);
3422                 return -EAGAIN;
3423         }
3424         nfs_refresh_inode(inode, data->res.fattr);
3425         return 0;
3426 }
3427
3428 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3429 {
3430         if (!nfs4_sequence_done(task, &data->res.seq_res))
3431                 return -EAGAIN;
3432         return data->write_done_cb(task, data);
3433 }
3434
3435 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3436 {
3437         struct nfs_server *server = NFS_SERVER(data->inode);
3438
3439         if (data->lseg) {
3440                 data->args.bitmask = NULL;
3441                 data->res.fattr = NULL;
3442         } else
3443                 data->args.bitmask = server->cache_consistency_bitmask;
3444         if (!data->write_done_cb)
3445                 data->write_done_cb = nfs4_commit_done_cb;
3446         data->res.server = server;
3447         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3448         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3449 }
3450
3451 struct nfs4_renewdata {
3452         struct nfs_client       *client;
3453         unsigned long           timestamp;
3454 };
3455
3456 /*
3457  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3458  * standalone procedure for queueing an asynchronous RENEW.
3459  */
3460 static void nfs4_renew_release(void *calldata)
3461 {
3462         struct nfs4_renewdata *data = calldata;
3463         struct nfs_client *clp = data->client;
3464
3465         if (atomic_read(&clp->cl_count) > 1)
3466                 nfs4_schedule_state_renewal(clp);
3467         nfs_put_client(clp);
3468         kfree(data);
3469 }
3470
3471 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3472 {
3473         struct nfs4_renewdata *data = calldata;
3474         struct nfs_client *clp = data->client;
3475         unsigned long timestamp = data->timestamp;
3476
3477         if (task->tk_status < 0) {
3478                 /* Unless we're shutting down, schedule state recovery! */
3479                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3480                         return;
3481                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3482                         nfs4_schedule_lease_recovery(clp);
3483                         return;
3484                 }
3485                 nfs4_schedule_path_down_recovery(clp);
3486         }
3487         do_renew_lease(clp, timestamp);
3488 }
3489
3490 static const struct rpc_call_ops nfs4_renew_ops = {
3491         .rpc_call_done = nfs4_renew_done,
3492         .rpc_release = nfs4_renew_release,
3493 };
3494
3495 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3496 {
3497         struct rpc_message msg = {
3498                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3499                 .rpc_argp       = clp,
3500                 .rpc_cred       = cred,
3501         };
3502         struct nfs4_renewdata *data;
3503
3504         if (renew_flags == 0)
3505                 return 0;
3506         if (!atomic_inc_not_zero(&clp->cl_count))
3507                 return -EIO;
3508         data = kmalloc(sizeof(*data), GFP_NOFS);
3509         if (data == NULL)
3510                 return -ENOMEM;
3511         data->client = clp;
3512         data->timestamp = jiffies;
3513         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3514                         &nfs4_renew_ops, data);
3515 }
3516
3517 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3518 {
3519         struct rpc_message msg = {
3520                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3521                 .rpc_argp       = clp,
3522                 .rpc_cred       = cred,
3523         };
3524         unsigned long now = jiffies;
3525         int status;
3526
3527         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3528         if (status < 0)
3529                 return status;
3530         do_renew_lease(clp, now);
3531         return 0;
3532 }
3533
3534 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3535 {
3536         return (server->caps & NFS_CAP_ACLS)
3537                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3538                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3539 }
3540
3541 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3542  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3543  * the stack.
3544  */
3545 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3546
3547 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3548                 struct page **pages, unsigned int *pgbase)
3549 {
3550         struct page *newpage, **spages;
3551         int rc = 0;
3552         size_t len;
3553         spages = pages;
3554
3555         do {
3556                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3557                 newpage = alloc_page(GFP_KERNEL);
3558
3559                 if (newpage == NULL)
3560                         goto unwind;
3561                 memcpy(page_address(newpage), buf, len);
3562                 buf += len;
3563                 buflen -= len;
3564                 *pages++ = newpage;
3565                 rc++;
3566         } while (buflen != 0);
3567
3568         return rc;
3569
3570 unwind:
3571         for(; rc > 0; rc--)
3572                 __free_page(spages[rc-1]);
3573         return -ENOMEM;
3574 }
3575
3576 struct nfs4_cached_acl {
3577         int cached;
3578         size_t len;
3579         char data[0];
3580 };
3581
3582 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3583 {
3584         struct nfs_inode *nfsi = NFS_I(inode);
3585
3586         spin_lock(&inode->i_lock);
3587         kfree(nfsi->nfs4_acl);
3588         nfsi->nfs4_acl = acl;
3589         spin_unlock(&inode->i_lock);
3590 }
3591
3592 static void nfs4_zap_acl_attr(struct inode *inode)
3593 {
3594         nfs4_set_cached_acl(inode, NULL);
3595 }
3596
3597 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3598 {
3599         struct nfs_inode *nfsi = NFS_I(inode);
3600         struct nfs4_cached_acl *acl;
3601         int ret = -ENOENT;
3602
3603         spin_lock(&inode->i_lock);
3604         acl = nfsi->nfs4_acl;
3605         if (acl == NULL)
3606                 goto out;
3607         if (buf == NULL) /* user is just asking for length */
3608                 goto out_len;
3609         if (acl->cached == 0)
3610                 goto out;
3611         ret = -ERANGE; /* see getxattr(2) man page */
3612         if (acl->len > buflen)
3613                 goto out;
3614         memcpy(buf, acl->data, acl->len);
3615 out_len:
3616         ret = acl->len;
3617 out:
3618         spin_unlock(&inode->i_lock);
3619         return ret;
3620 }
3621
3622 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3623 {
3624         struct nfs4_cached_acl *acl;
3625
3626         if (buf && acl_len <= PAGE_SIZE) {
3627                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3628                 if (acl == NULL)
3629                         goto out;
3630                 acl->cached = 1;
3631                 memcpy(acl->data, buf, acl_len);
3632         } else {
3633                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3634                 if (acl == NULL)
3635                         goto out;
3636                 acl->cached = 0;
3637         }
3638         acl->len = acl_len;
3639 out:
3640         nfs4_set_cached_acl(inode, acl);
3641 }
3642
3643 /*
3644  * The getxattr API returns the required buffer length when called with a
3645  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3646  * the required buf.  On a NULL buf, we send a page of data to the server
3647  * guessing that the ACL request can be serviced by a page. If so, we cache
3648  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3649  * the cache. If not so, we throw away the page, and cache the required
3650  * length. The next getxattr call will then produce another round trip to
3651  * the server, this time with the input buf of the required size.
3652  */
3653 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3654 {
3655         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3656         struct nfs_getaclargs args = {
3657                 .fh = NFS_FH(inode),
3658                 .acl_pages = pages,
3659                 .acl_len = buflen,
3660         };
3661         struct nfs_getaclres res = {
3662                 .acl_len = buflen,
3663         };
3664         void *resp_buf;
3665         struct rpc_message msg = {
3666                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3667                 .rpc_argp = &args,
3668                 .rpc_resp = &res,
3669         };
3670         int ret = -ENOMEM, npages, i, acl_len = 0;
3671
3672         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3673         /* As long as we're doing a round trip to the server anyway,
3674          * let's be prepared for a page of acl data. */
3675         if (npages == 0)
3676                 npages = 1;
3677
3678         for (i = 0; i < npages; i++) {
3679                 pages[i] = alloc_page(GFP_KERNEL);
3680                 if (!pages[i])
3681                         goto out_free;
3682         }
3683         if (npages > 1) {
3684                 /* for decoding across pages */
3685                 res.acl_scratch = alloc_page(GFP_KERNEL);
3686                 if (!res.acl_scratch)
3687                         goto out_free;
3688         }
3689         args.acl_len = npages * PAGE_SIZE;
3690         args.acl_pgbase = 0;
3691         /* Let decode_getfacl know not to fail if the ACL data is larger than
3692          * the page we send as a guess */
3693         if (buf == NULL)
3694                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3695         resp_buf = page_address(pages[0]);
3696
3697         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3698                 __func__, buf, buflen, npages, args.acl_len);
3699         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3700                              &msg, &args.seq_args, &res.seq_res, 0);
3701         if (ret)
3702                 goto out_free;
3703
3704         acl_len = res.acl_len - res.acl_data_offset;
3705         if (acl_len > args.acl_len)
3706                 nfs4_write_cached_acl(inode, NULL, acl_len);
3707         else
3708                 nfs4_write_cached_acl(inode, resp_buf + res.acl_data_offset,
3709                                       acl_len);
3710         if (buf) {
3711                 ret = -ERANGE;
3712                 if (acl_len > buflen)
3713                         goto out_free;
3714                 _copy_from_pages(buf, pages, res.acl_data_offset,
3715                                 acl_len);
3716         }
3717         ret = acl_len;
3718 out_free:
3719         for (i = 0; i < npages; i++)
3720                 if (pages[i])
3721                         __free_page(pages[i]);
3722         if (res.acl_scratch)
3723                 __free_page(res.acl_scratch);
3724         return ret;
3725 }
3726
3727 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3728 {
3729         struct nfs4_exception exception = { };
3730         ssize_t ret;
3731         do {
3732                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3733                 if (ret >= 0)
3734                         break;
3735                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3736         } while (exception.retry);
3737         return ret;
3738 }
3739
3740 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3741 {
3742         struct nfs_server *server = NFS_SERVER(inode);
3743         int ret;
3744
3745         if (!nfs4_server_supports_acls(server))
3746                 return -EOPNOTSUPP;
3747         ret = nfs_revalidate_inode(server, inode);
3748         if (ret < 0)
3749                 return ret;
3750         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3751                 nfs_zap_acl_cache(inode);
3752         ret = nfs4_read_cached_acl(inode, buf, buflen);
3753         if (ret != -ENOENT)
3754                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3755                  * but no cached acl data, just the acl length */
3756                 return ret;
3757         return nfs4_get_acl_uncached(inode, buf, buflen);
3758 }
3759
3760 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3761 {
3762         struct nfs_server *server = NFS_SERVER(inode);
3763         struct page *pages[NFS4ACL_MAXPAGES];
3764         struct nfs_setaclargs arg = {
3765                 .fh             = NFS_FH(inode),
3766                 .acl_pages      = pages,
3767                 .acl_len        = buflen,
3768         };
3769         struct nfs_setaclres res;
3770         struct rpc_message msg = {
3771                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3772                 .rpc_argp       = &arg,
3773                 .rpc_resp       = &res,
3774         };
3775         int ret, i;
3776
3777         if (!nfs4_server_supports_acls(server))
3778                 return -EOPNOTSUPP;
3779         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3780         if (i < 0)
3781                 return i;
3782         nfs_inode_return_delegation(inode);
3783         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3784
3785         /*
3786          * Free each page after tx, so the only ref left is
3787          * held by the network stack
3788          */
3789         for (; i > 0; i--)
3790                 put_page(pages[i-1]);
3791
3792         /*
3793          * Acl update can result in inode attribute update.
3794          * so mark the attribute cache invalid.
3795          */
3796         spin_lock(&inode->i_lock);
3797         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3798         spin_unlock(&inode->i_lock);
3799         nfs_access_zap_cache(inode);
3800         nfs_zap_acl_cache(inode);
3801         return ret;
3802 }
3803
3804 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3805 {
3806         struct nfs4_exception exception = { };
3807         int err;
3808         do {
3809                 err = nfs4_handle_exception(NFS_SERVER(inode),
3810                                 __nfs4_proc_set_acl(inode, buf, buflen),
3811                                 &exception);
3812         } while (exception.retry);
3813         return err;
3814 }
3815
3816 static int
3817 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3818 {
3819         struct nfs_client *clp = server->nfs_client;
3820
3821         if (task->tk_status >= 0)
3822                 return 0;
3823         switch(task->tk_status) {
3824                 case -NFS4ERR_DELEG_REVOKED:
3825                 case -NFS4ERR_ADMIN_REVOKED:
3826                 case -NFS4ERR_BAD_STATEID:
3827                         if (state == NULL)
3828                                 break;
3829                         nfs_remove_bad_delegation(state->inode);
3830                 case -NFS4ERR_OPENMODE:
3831                         if (state == NULL)
3832                                 break;
3833                         nfs4_schedule_stateid_recovery(server, state);
3834                         goto wait_on_recovery;
3835                 case -NFS4ERR_EXPIRED:
3836                         if (state != NULL)
3837                                 nfs4_schedule_stateid_recovery(server, state);
3838                 case -NFS4ERR_STALE_STATEID:
3839                 case -NFS4ERR_STALE_CLIENTID:
3840                         nfs4_schedule_lease_recovery(clp);
3841                         goto wait_on_recovery;
3842 #if defined(CONFIG_NFS_V4_1)
3843                 case -NFS4ERR_BADSESSION:
3844                 case -NFS4ERR_BADSLOT:
3845                 case -NFS4ERR_BAD_HIGH_SLOT:
3846                 case -NFS4ERR_DEADSESSION:
3847                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3848                 case -NFS4ERR_SEQ_FALSE_RETRY:
3849                 case -NFS4ERR_SEQ_MISORDERED:
3850                         dprintk("%s ERROR %d, Reset session\n", __func__,
3851                                 task->tk_status);
3852                         nfs4_schedule_session_recovery(clp->cl_session);
3853                         task->tk_status = 0;
3854                         return -EAGAIN;
3855 #endif /* CONFIG_NFS_V4_1 */
3856                 case -NFS4ERR_DELAY:
3857                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3858                 case -NFS4ERR_GRACE:
3859                 case -EKEYEXPIRED:
3860                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3861                         task->tk_status = 0;
3862                         return -EAGAIN;
3863                 case -NFS4ERR_RETRY_UNCACHED_REP:
3864                 case -NFS4ERR_OLD_STATEID:
3865                         task->tk_status = 0;
3866                         return -EAGAIN;
3867         }
3868         task->tk_status = nfs4_map_errors(task->tk_status);
3869         return 0;
3870 wait_on_recovery:
3871         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3872         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3873                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3874         task->tk_status = 0;
3875         return -EAGAIN;
3876 }
3877
3878 static void nfs4_construct_boot_verifier(struct nfs_client *clp,
3879                                          nfs4_verifier *bootverf)
3880 {
3881         __be32 verf[2];
3882
3883         verf[0] = htonl((u32)clp->cl_boot_time.tv_sec);
3884         verf[1] = htonl((u32)clp->cl_boot_time.tv_nsec);
3885         memcpy(bootverf->data, verf, sizeof(bootverf->data));
3886 }
3887
3888 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3889                 unsigned short port, struct rpc_cred *cred,
3890                 struct nfs4_setclientid_res *res)
3891 {
3892         nfs4_verifier sc_verifier;
3893         struct nfs4_setclientid setclientid = {
3894                 .sc_verifier = &sc_verifier,
3895                 .sc_prog = program,
3896                 .sc_cb_ident = clp->cl_cb_ident,
3897         };
3898         struct rpc_message msg = {
3899                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3900                 .rpc_argp = &setclientid,
3901                 .rpc_resp = res,
3902                 .rpc_cred = cred,
3903         };
3904         int loop = 0;
3905         int status;
3906
3907         nfs4_construct_boot_verifier(clp, &sc_verifier);
3908
3909         for(;;) {
3910                 rcu_read_lock();
3911                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3912                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3913                                 clp->cl_ipaddr,
3914                                 rpc_peeraddr2str(clp->cl_rpcclient,
3915                                                         RPC_DISPLAY_ADDR),
3916                                 rpc_peeraddr2str(clp->cl_rpcclient,
3917                                                         RPC_DISPLAY_PROTO),
3918                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3919                                 clp->cl_id_uniquifier);
3920                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3921                                 sizeof(setclientid.sc_netid),
3922                                 rpc_peeraddr2str(clp->cl_rpcclient,
3923                                                         RPC_DISPLAY_NETID));
3924                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3925                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3926                                 clp->cl_ipaddr, port >> 8, port & 255);
3927                 rcu_read_unlock();
3928
3929                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3930                 if (status != -NFS4ERR_CLID_INUSE)
3931                         break;
3932                 if (loop != 0) {
3933                         ++clp->cl_id_uniquifier;
3934                         break;
3935                 }
3936                 ++loop;
3937                 ssleep(clp->cl_lease_time / HZ + 1);
3938         }
3939         return status;
3940 }
3941
3942 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3943                 struct nfs4_setclientid_res *arg,
3944                 struct rpc_cred *cred)
3945 {
3946         struct nfs_fsinfo fsinfo;
3947         struct rpc_message msg = {
3948                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3949                 .rpc_argp = arg,
3950                 .rpc_resp = &fsinfo,
3951                 .rpc_cred = cred,
3952         };
3953         unsigned long now;
3954         int status;
3955
3956         now = jiffies;
3957         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3958         if (status == 0) {
3959                 spin_lock(&clp->cl_lock);
3960                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3961                 clp->cl_last_renewal = now;
3962                 spin_unlock(&clp->cl_lock);
3963         }
3964         return status;
3965 }
3966
3967 struct nfs4_delegreturndata {
3968         struct nfs4_delegreturnargs args;
3969         struct nfs4_delegreturnres res;
3970         struct nfs_fh fh;
3971         nfs4_stateid stateid;
3972         unsigned long timestamp;
3973         struct nfs_fattr fattr;
3974         int rpc_status;
3975 };
3976
3977 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3978 {
3979         struct nfs4_delegreturndata *data = calldata;
3980
3981         if (!nfs4_sequence_done(task, &data->res.seq_res))
3982                 return;
3983
3984         switch (task->tk_status) {
3985         case -NFS4ERR_STALE_STATEID:
3986         case -NFS4ERR_EXPIRED:
3987         case 0:
3988                 renew_lease(data->res.server, data->timestamp);
3989                 break;
3990         default:
3991                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3992                                 -EAGAIN) {
3993                         rpc_restart_call_prepare(task);
3994                         return;
3995                 }
3996         }
3997         data->rpc_status = task->tk_status;
3998 }
3999
4000 static void nfs4_delegreturn_release(void *calldata)
4001 {
4002         kfree(calldata);
4003 }
4004
4005 #if defined(CONFIG_NFS_V4_1)
4006 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4007 {
4008         struct nfs4_delegreturndata *d_data;
4009
4010         d_data = (struct nfs4_delegreturndata *)data;
4011
4012         if (nfs4_setup_sequence(d_data->res.server,
4013                                 &d_data->args.seq_args,
4014                                 &d_data->res.seq_res, task))
4015                 return;
4016         rpc_call_start(task);
4017 }
4018 #endif /* CONFIG_NFS_V4_1 */
4019
4020 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4021 #if defined(CONFIG_NFS_V4_1)
4022         .rpc_call_prepare = nfs4_delegreturn_prepare,
4023 #endif /* CONFIG_NFS_V4_1 */
4024         .rpc_call_done = nfs4_delegreturn_done,
4025         .rpc_release = nfs4_delegreturn_release,
4026 };
4027
4028 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4029 {
4030         struct nfs4_delegreturndata *data;
4031         struct nfs_server *server = NFS_SERVER(inode);
4032         struct rpc_task *task;
4033         struct rpc_message msg = {
4034                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4035                 .rpc_cred = cred,
4036         };
4037         struct rpc_task_setup task_setup_data = {
4038                 .rpc_client = server->client,
4039                 .rpc_message = &msg,
4040                 .callback_ops = &nfs4_delegreturn_ops,
4041                 .flags = RPC_TASK_ASYNC,
4042         };
4043         int status = 0;
4044
4045         data = kzalloc(sizeof(*data), GFP_NOFS);
4046         if (data == NULL)
4047                 return -ENOMEM;
4048         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4049         data->args.fhandle = &data->fh;
4050         data->args.stateid = &data->stateid;
4051         data->args.bitmask = server->attr_bitmask;
4052         nfs_copy_fh(&data->fh, NFS_FH(inode));
4053         nfs4_stateid_copy(&data->stateid, stateid);
4054         data->res.fattr = &data->fattr;
4055         data->res.server = server;
4056         nfs_fattr_init(data->res.fattr);
4057         data->timestamp = jiffies;
4058         data->rpc_status = 0;
4059
4060         task_setup_data.callback_data = data;
4061         msg.rpc_argp = &data->args;
4062         msg.rpc_resp = &data->res;
4063         task = rpc_run_task(&task_setup_data);
4064         if (IS_ERR(task))
4065                 return PTR_ERR(task);
4066         if (!issync)
4067                 goto out;
4068         status = nfs4_wait_for_completion_rpc_task(task);
4069         if (status != 0)
4070                 goto out;
4071         status = data->rpc_status;
4072         if (status != 0)
4073                 goto out;
4074         nfs_refresh_inode(inode, &data->fattr);
4075 out:
4076         rpc_put_task(task);
4077         return status;
4078 }
4079
4080 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4081 {
4082         struct nfs_server *server = NFS_SERVER(inode);
4083         struct nfs4_exception exception = { };
4084         int err;
4085         do {
4086                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4087                 switch (err) {
4088                         case -NFS4ERR_STALE_STATEID:
4089                         case -NFS4ERR_EXPIRED:
4090                         case 0:
4091                                 return 0;
4092                 }
4093                 err = nfs4_handle_exception(server, err, &exception);
4094         } while (exception.retry);
4095         return err;
4096 }
4097
4098 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4099 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4100
4101 /*
4102  * sleep, with exponential backoff, and retry the LOCK operation.
4103  */
4104 static unsigned long
4105 nfs4_set_lock_task_retry(unsigned long timeout)
4106 {
4107         freezable_schedule_timeout_killable(timeout);
4108         timeout <<= 1;
4109         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4110                 return NFS4_LOCK_MAXTIMEOUT;
4111         return timeout;
4112 }
4113
4114 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4115 {
4116         struct inode *inode = state->inode;
4117         struct nfs_server *server = NFS_SERVER(inode);
4118         struct nfs_client *clp = server->nfs_client;
4119         struct nfs_lockt_args arg = {
4120                 .fh = NFS_FH(inode),
4121                 .fl = request,
4122         };
4123         struct nfs_lockt_res res = {
4124                 .denied = request,
4125         };
4126         struct rpc_message msg = {
4127                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4128                 .rpc_argp       = &arg,
4129                 .rpc_resp       = &res,
4130                 .rpc_cred       = state->owner->so_cred,
4131         };
4132         struct nfs4_lock_state *lsp;
4133         int status;
4134
4135         arg.lock_owner.clientid = clp->cl_clientid;
4136         status = nfs4_set_lock_state(state, request);
4137         if (status != 0)
4138                 goto out;
4139         lsp = request->fl_u.nfs4_fl.owner;
4140         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4141         arg.lock_owner.s_dev = server->s_dev;
4142         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4143         switch (status) {
4144                 case 0:
4145                         request->fl_type = F_UNLCK;
4146                         break;
4147                 case -NFS4ERR_DENIED:
4148                         status = 0;
4149         }
4150         request->fl_ops->fl_release_private(request);
4151 out:
4152         return status;
4153 }
4154
4155 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4156 {
4157         struct nfs4_exception exception = { };
4158         int err;
4159
4160         do {
4161                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4162                                 _nfs4_proc_getlk(state, cmd, request),
4163                                 &exception);
4164         } while (exception.retry);
4165         return err;
4166 }
4167
4168 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4169 {
4170         int res = 0;
4171         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4172                 case FL_POSIX:
4173                         res = posix_lock_file_wait(file, fl);
4174                         break;
4175                 case FL_FLOCK:
4176                         res = flock_lock_file_wait(file, fl);
4177                         break;
4178                 default:
4179                         BUG();
4180         }
4181         return res;
4182 }
4183
4184 struct nfs4_unlockdata {
4185         struct nfs_locku_args arg;
4186         struct nfs_locku_res res;
4187         struct nfs4_lock_state *lsp;
4188         struct nfs_open_context *ctx;
4189         struct file_lock fl;
4190         const struct nfs_server *server;
4191         unsigned long timestamp;
4192 };
4193
4194 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4195                 struct nfs_open_context *ctx,
4196                 struct nfs4_lock_state *lsp,
4197                 struct nfs_seqid *seqid)
4198 {
4199         struct nfs4_unlockdata *p;
4200         struct inode *inode = lsp->ls_state->inode;
4201
4202         p = kzalloc(sizeof(*p), GFP_NOFS);
4203         if (p == NULL)
4204                 return NULL;
4205         p->arg.fh = NFS_FH(inode);
4206         p->arg.fl = &p->fl;
4207         p->arg.seqid = seqid;
4208         p->res.seqid = seqid;
4209         p->arg.stateid = &lsp->ls_stateid;
4210         p->lsp = lsp;
4211         atomic_inc(&lsp->ls_count);
4212         /* Ensure we don't close file until we're done freeing locks! */
4213         p->ctx = get_nfs_open_context(ctx);
4214         memcpy(&p->fl, fl, sizeof(p->fl));
4215         p->server = NFS_SERVER(inode);
4216         return p;
4217 }
4218
4219 static void nfs4_locku_release_calldata(void *data)
4220 {
4221         struct nfs4_unlockdata *calldata = data;
4222         nfs_free_seqid(calldata->arg.seqid);
4223         nfs4_put_lock_state(calldata->lsp);
4224         put_nfs_open_context(calldata->ctx);
4225         kfree(calldata);
4226 }
4227
4228 static void nfs4_locku_done(struct rpc_task *task, void *data)
4229 {
4230         struct nfs4_unlockdata *calldata = data;
4231
4232         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4233                 return;
4234         switch (task->tk_status) {
4235                 case 0:
4236                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4237                                         &calldata->res.stateid);
4238                         renew_lease(calldata->server, calldata->timestamp);
4239                         break;
4240                 case -NFS4ERR_BAD_STATEID:
4241                 case -NFS4ERR_OLD_STATEID:
4242                 case -NFS4ERR_STALE_STATEID:
4243                 case -NFS4ERR_EXPIRED:
4244                         break;
4245                 default:
4246                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4247                                 rpc_restart_call_prepare(task);
4248         }
4249 }
4250
4251 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4252 {
4253         struct nfs4_unlockdata *calldata = data;
4254
4255         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4256                 return;
4257         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4258                 /* Note: exit _without_ running nfs4_locku_done */
4259                 task->tk_action = NULL;
4260                 return;
4261         }
4262         calldata->timestamp = jiffies;
4263         if (nfs4_setup_sequence(calldata->server,
4264                                 &calldata->arg.seq_args,
4265                                 &calldata->res.seq_res, task))
4266                 return;
4267         rpc_call_start(task);
4268 }
4269
4270 static const struct rpc_call_ops nfs4_locku_ops = {
4271         .rpc_call_prepare = nfs4_locku_prepare,
4272         .rpc_call_done = nfs4_locku_done,
4273         .rpc_release = nfs4_locku_release_calldata,
4274 };
4275
4276 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4277                 struct nfs_open_context *ctx,
4278                 struct nfs4_lock_state *lsp,
4279                 struct nfs_seqid *seqid)
4280 {
4281         struct nfs4_unlockdata *data;
4282         struct rpc_message msg = {
4283                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4284                 .rpc_cred = ctx->cred,
4285         };
4286         struct rpc_task_setup task_setup_data = {
4287                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4288                 .rpc_message = &msg,
4289                 .callback_ops = &nfs4_locku_ops,
4290                 .workqueue = nfsiod_workqueue,
4291                 .flags = RPC_TASK_ASYNC,
4292         };
4293
4294         /* Ensure this is an unlock - when canceling a lock, the
4295          * canceled lock is passed in, and it won't be an unlock.
4296          */
4297         fl->fl_type = F_UNLCK;
4298
4299         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4300         if (data == NULL) {
4301                 nfs_free_seqid(seqid);
4302                 return ERR_PTR(-ENOMEM);
4303         }
4304
4305         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4306         msg.rpc_argp = &data->arg;
4307         msg.rpc_resp = &data->res;
4308         task_setup_data.callback_data = data;
4309         return rpc_run_task(&task_setup_data);
4310 }
4311
4312 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4313 {
4314         struct nfs_inode *nfsi = NFS_I(state->inode);
4315         struct nfs_seqid *seqid;
4316         struct nfs4_lock_state *lsp;
4317         struct rpc_task *task;
4318         int status = 0;
4319         unsigned char fl_flags = request->fl_flags;
4320
4321         status = nfs4_set_lock_state(state, request);
4322         /* Unlock _before_ we do the RPC call */
4323         request->fl_flags |= FL_EXISTS;
4324         down_read(&nfsi->rwsem);
4325         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4326                 up_read(&nfsi->rwsem);
4327                 goto out;
4328         }
4329         up_read(&nfsi->rwsem);
4330         if (status != 0)
4331                 goto out;
4332         /* Is this a delegated lock? */
4333         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4334                 goto out;
4335         lsp = request->fl_u.nfs4_fl.owner;
4336         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4337         status = -ENOMEM;
4338         if (seqid == NULL)
4339                 goto out;
4340         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4341         status = PTR_ERR(task);
4342         if (IS_ERR(task))
4343                 goto out;
4344         status = nfs4_wait_for_completion_rpc_task(task);
4345         rpc_put_task(task);
4346 out:
4347         request->fl_flags = fl_flags;
4348         return status;
4349 }
4350
4351 struct nfs4_lockdata {
4352         struct nfs_lock_args arg;
4353         struct nfs_lock_res res;
4354         struct nfs4_lock_state *lsp;
4355         struct nfs_open_context *ctx;
4356         struct file_lock fl;
4357         unsigned long timestamp;
4358         int rpc_status;
4359         int cancelled;
4360         struct nfs_server *server;
4361 };
4362
4363 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4364                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4365                 gfp_t gfp_mask)
4366 {
4367         struct nfs4_lockdata *p;
4368         struct inode *inode = lsp->ls_state->inode;
4369         struct nfs_server *server = NFS_SERVER(inode);
4370
4371         p = kzalloc(sizeof(*p), gfp_mask);
4372         if (p == NULL)
4373                 return NULL;
4374
4375         p->arg.fh = NFS_FH(inode);
4376         p->arg.fl = &p->fl;
4377         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4378         if (p->arg.open_seqid == NULL)
4379                 goto out_free;
4380         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4381         if (p->arg.lock_seqid == NULL)
4382                 goto out_free_seqid;
4383         p->arg.lock_stateid = &lsp->ls_stateid;
4384         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4385         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4386         p->arg.lock_owner.s_dev = server->s_dev;
4387         p->res.lock_seqid = p->arg.lock_seqid;
4388         p->lsp = lsp;
4389         p->server = server;
4390         atomic_inc(&lsp->ls_count);
4391         p->ctx = get_nfs_open_context(ctx);
4392         memcpy(&p->fl, fl, sizeof(p->fl));
4393         return p;
4394 out_free_seqid:
4395         nfs_free_seqid(p->arg.open_seqid);
4396 out_free:
4397         kfree(p);
4398         return NULL;
4399 }
4400
4401 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4402 {
4403         struct nfs4_lockdata *data = calldata;
4404         struct nfs4_state *state = data->lsp->ls_state;
4405
4406         dprintk("%s: begin!\n", __func__);
4407         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4408                 return;
4409         /* Do we need to do an open_to_lock_owner? */
4410         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4411                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4412                         return;
4413                 data->arg.open_stateid = &state->stateid;
4414                 data->arg.new_lock_owner = 1;
4415                 data->res.open_seqid = data->arg.open_seqid;
4416         } else
4417                 data->arg.new_lock_owner = 0;
4418         data->timestamp = jiffies;
4419         if (nfs4_setup_sequence(data->server,
4420                                 &data->arg.seq_args,
4421                                 &data->res.seq_res, task))
4422                 return;
4423         rpc_call_start(task);
4424         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4425 }
4426
4427 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4428 {
4429         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4430         nfs4_lock_prepare(task, calldata);
4431 }
4432
4433 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4434 {
4435         struct nfs4_lockdata *data = calldata;
4436
4437         dprintk("%s: begin!\n", __func__);
4438
4439         if (!nfs4_sequence_done(task, &data->res.seq_res))
4440                 return;
4441
4442         data->rpc_status = task->tk_status;
4443         if (data->arg.new_lock_owner != 0) {
4444                 if (data->rpc_status == 0)
4445                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4446                 else
4447                         goto out;
4448         }
4449         if (data->rpc_status == 0) {
4450                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4451                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4452                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4453         }
4454 out:
4455         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4456 }
4457
4458 static void nfs4_lock_release(void *calldata)
4459 {
4460         struct nfs4_lockdata *data = calldata;
4461
4462         dprintk("%s: begin!\n", __func__);
4463         nfs_free_seqid(data->arg.open_seqid);
4464         if (data->cancelled != 0) {
4465                 struct rpc_task *task;
4466                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4467                                 data->arg.lock_seqid);
4468                 if (!IS_ERR(task))
4469                         rpc_put_task_async(task);
4470                 dprintk("%s: cancelling lock!\n", __func__);
4471         } else
4472                 nfs_free_seqid(data->arg.lock_seqid);
4473         nfs4_put_lock_state(data->lsp);
4474         put_nfs_open_context(data->ctx);
4475         kfree(data);
4476         dprintk("%s: done!\n", __func__);
4477 }
4478
4479 static const struct rpc_call_ops nfs4_lock_ops = {
4480         .rpc_call_prepare = nfs4_lock_prepare,
4481         .rpc_call_done = nfs4_lock_done,
4482         .rpc_release = nfs4_lock_release,
4483 };
4484
4485 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4486         .rpc_call_prepare = nfs4_recover_lock_prepare,
4487         .rpc_call_done = nfs4_lock_done,
4488         .rpc_release = nfs4_lock_release,
4489 };
4490
4491 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4492 {
4493         switch (error) {
4494         case -NFS4ERR_ADMIN_REVOKED:
4495         case -NFS4ERR_BAD_STATEID:
4496                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4497                 if (new_lock_owner != 0 ||
4498                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4499                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4500                 break;
4501         case -NFS4ERR_STALE_STATEID:
4502                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4503         case -NFS4ERR_EXPIRED:
4504                 nfs4_schedule_lease_recovery(server->nfs_client);
4505         };
4506 }
4507
4508 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4509 {
4510         struct nfs4_lockdata *data;
4511         struct rpc_task *task;
4512         struct rpc_message msg = {
4513                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4514                 .rpc_cred = state->owner->so_cred,
4515         };
4516         struct rpc_task_setup task_setup_data = {
4517                 .rpc_client = NFS_CLIENT(state->inode),
4518                 .rpc_message = &msg,
4519                 .callback_ops = &nfs4_lock_ops,
4520                 .workqueue = nfsiod_workqueue,
4521                 .flags = RPC_TASK_ASYNC,
4522         };
4523         int ret;
4524
4525         dprintk("%s: begin!\n", __func__);
4526         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4527                         fl->fl_u.nfs4_fl.owner,
4528                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4529         if (data == NULL)
4530                 return -ENOMEM;
4531         if (IS_SETLKW(cmd))
4532                 data->arg.block = 1;
4533         if (recovery_type > NFS_LOCK_NEW) {
4534                 if (recovery_type == NFS_LOCK_RECLAIM)
4535                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4536                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4537         }
4538         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4539         msg.rpc_argp = &data->arg;
4540         msg.rpc_resp = &data->res;
4541         task_setup_data.callback_data = data;
4542         task = rpc_run_task(&task_setup_data);
4543         if (IS_ERR(task))
4544                 return PTR_ERR(task);
4545         ret = nfs4_wait_for_completion_rpc_task(task);
4546         if (ret == 0) {
4547                 ret = data->rpc_status;
4548                 if (ret)
4549                         nfs4_handle_setlk_error(data->server, data->lsp,
4550                                         data->arg.new_lock_owner, ret);
4551         } else
4552                 data->cancelled = 1;
4553         rpc_put_task(task);
4554         dprintk("%s: done, ret = %d!\n", __func__, ret);
4555         return ret;
4556 }
4557
4558 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4559 {
4560         struct nfs_server *server = NFS_SERVER(state->inode);
4561         struct nfs4_exception exception = {
4562                 .inode = state->inode,
4563         };
4564         int err;
4565
4566         do {
4567                 /* Cache the lock if possible... */
4568                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4569                         return 0;
4570                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4571                 if (err != -NFS4ERR_DELAY)
4572                         break;
4573                 nfs4_handle_exception(server, err, &exception);
4574         } while (exception.retry);
4575         return err;
4576 }
4577
4578 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4579 {
4580         struct nfs_server *server = NFS_SERVER(state->inode);
4581         struct nfs4_exception exception = {
4582                 .inode = state->inode,
4583         };
4584         int err;
4585
4586         err = nfs4_set_lock_state(state, request);
4587         if (err != 0)
4588                 return err;
4589         do {
4590                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4591                         return 0;
4592                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4593                 switch (err) {
4594                 default:
4595                         goto out;
4596                 case -NFS4ERR_GRACE:
4597                 case -NFS4ERR_DELAY:
4598                         nfs4_handle_exception(server, err, &exception);
4599                         err = 0;
4600                 }
4601         } while (exception.retry);
4602 out:
4603         return err;
4604 }
4605
4606 #if defined(CONFIG_NFS_V4_1)
4607 static int nfs41_check_expired_locks(struct nfs4_state *state)
4608 {
4609         int status, ret = NFS_OK;
4610         struct nfs4_lock_state *lsp;
4611         struct nfs_server *server = NFS_SERVER(state->inode);
4612
4613         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4614                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4615                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4616                         if (status != NFS_OK) {
4617                                 nfs41_free_stateid(server, &lsp->ls_stateid);
4618                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4619                                 ret = status;
4620                         }
4621                 }
4622         };
4623
4624         return ret;
4625 }
4626
4627 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4628 {
4629         int status = NFS_OK;
4630
4631         if (test_bit(LK_STATE_IN_USE, &state->flags))
4632                 status = nfs41_check_expired_locks(state);
4633         if (status == NFS_OK)
4634                 return status;
4635         return nfs4_lock_expired(state, request);
4636 }
4637 #endif
4638
4639 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4640 {
4641         struct nfs_inode *nfsi = NFS_I(state->inode);
4642         unsigned char fl_flags = request->fl_flags;
4643         int status = -ENOLCK;
4644
4645         if ((fl_flags & FL_POSIX) &&
4646                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4647                 goto out;
4648         /* Is this a delegated open? */
4649         status = nfs4_set_lock_state(state, request);
4650         if (status != 0)
4651                 goto out;
4652         request->fl_flags |= FL_ACCESS;
4653         status = do_vfs_lock(request->fl_file, request);
4654         if (status < 0)
4655                 goto out;
4656         down_read(&nfsi->rwsem);
4657         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4658                 /* Yes: cache locks! */
4659                 /* ...but avoid races with delegation recall... */
4660                 request->fl_flags = fl_flags & ~FL_SLEEP;
4661                 status = do_vfs_lock(request->fl_file, request);
4662                 goto out_unlock;
4663         }
4664         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4665         if (status != 0)
4666                 goto out_unlock;
4667         /* Note: we always want to sleep here! */
4668         request->fl_flags = fl_flags | FL_SLEEP;
4669         if (do_vfs_lock(request->fl_file, request) < 0)
4670                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4671                         "manager!\n", __func__);
4672 out_unlock:
4673         up_read(&nfsi->rwsem);
4674 out:
4675         request->fl_flags = fl_flags;
4676         return status;
4677 }
4678
4679 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4680 {
4681         struct nfs4_exception exception = {
4682                 .state = state,
4683                 .inode = state->inode,
4684         };
4685         int err;
4686
4687         do {
4688                 err = _nfs4_proc_setlk(state, cmd, request);
4689                 if (err == -NFS4ERR_DENIED)
4690                         err = -EAGAIN;
4691                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4692                                 err, &exception);
4693         } while (exception.retry);
4694         return err;
4695 }
4696
4697 static int
4698 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4699 {
4700         struct nfs_open_context *ctx;
4701         struct nfs4_state *state;
4702         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4703         int status;
4704
4705         /* verify open state */
4706         ctx = nfs_file_open_context(filp);
4707         state = ctx->state;
4708
4709         if (request->fl_start < 0 || request->fl_end < 0)
4710                 return -EINVAL;
4711
4712         if (IS_GETLK(cmd)) {
4713                 if (state != NULL)
4714                         return nfs4_proc_getlk(state, F_GETLK, request);
4715                 return 0;
4716         }
4717
4718         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4719                 return -EINVAL;
4720
4721         if (request->fl_type == F_UNLCK) {
4722                 if (state != NULL)
4723                         return nfs4_proc_unlck(state, cmd, request);
4724                 return 0;
4725         }
4726
4727         if (state == NULL)
4728                 return -ENOLCK;
4729         /*
4730          * Don't rely on the VFS having checked the file open mode,
4731          * since it won't do this for flock() locks.
4732          */
4733         switch (request->fl_type & (F_RDLCK|F_WRLCK|F_UNLCK)) {
4734         case F_RDLCK:
4735                 if (!(filp->f_mode & FMODE_READ))
4736                         return -EBADF;
4737                 break;
4738         case F_WRLCK:
4739                 if (!(filp->f_mode & FMODE_WRITE))
4740                         return -EBADF;
4741         }
4742
4743         do {
4744                 status = nfs4_proc_setlk(state, cmd, request);
4745                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4746                         break;
4747                 timeout = nfs4_set_lock_task_retry(timeout);
4748                 status = -ERESTARTSYS;
4749                 if (signalled())
4750                         break;
4751         } while(status < 0);
4752         return status;
4753 }
4754
4755 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4756 {
4757         struct nfs_server *server = NFS_SERVER(state->inode);
4758         struct nfs4_exception exception = { };
4759         int err;
4760
4761         err = nfs4_set_lock_state(state, fl);
4762         if (err != 0)
4763                 goto out;
4764         do {
4765                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4766                 switch (err) {
4767                         default:
4768                                 printk(KERN_ERR "NFS: %s: unhandled error "
4769                                         "%d.\n", __func__, err);
4770                         case 0:
4771                         case -ESTALE:
4772                                 goto out;
4773                         case -NFS4ERR_EXPIRED:
4774                                 nfs4_schedule_stateid_recovery(server, state);
4775                         case -NFS4ERR_STALE_CLIENTID:
4776                         case -NFS4ERR_STALE_STATEID:
4777                                 nfs4_schedule_lease_recovery(server->nfs_client);
4778                                 goto out;
4779                         case -NFS4ERR_BADSESSION:
4780                         case -NFS4ERR_BADSLOT:
4781                         case -NFS4ERR_BAD_HIGH_SLOT:
4782                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4783                         case -NFS4ERR_DEADSESSION:
4784                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4785                                 goto out;
4786                         case -ERESTARTSYS:
4787                                 /*
4788                                  * The show must go on: exit, but mark the
4789                                  * stateid as needing recovery.
4790                                  */
4791                         case -NFS4ERR_DELEG_REVOKED:
4792                         case -NFS4ERR_ADMIN_REVOKED:
4793                         case -NFS4ERR_BAD_STATEID:
4794                         case -NFS4ERR_OPENMODE:
4795                                 nfs4_schedule_stateid_recovery(server, state);
4796                                 err = 0;
4797                                 goto out;
4798                         case -EKEYEXPIRED:
4799                                 /*
4800                                  * User RPCSEC_GSS context has expired.
4801                                  * We cannot recover this stateid now, so
4802                                  * skip it and allow recovery thread to
4803                                  * proceed.
4804                                  */
4805                                 err = 0;
4806                                 goto out;
4807                         case -ENOMEM:
4808                         case -NFS4ERR_DENIED:
4809                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4810                                 err = 0;
4811                                 goto out;
4812                         case -NFS4ERR_DELAY:
4813                                 break;
4814                 }
4815                 err = nfs4_handle_exception(server, err, &exception);
4816         } while (exception.retry);
4817 out:
4818         return err;
4819 }
4820
4821 struct nfs_release_lockowner_data {
4822         struct nfs4_lock_state *lsp;
4823         struct nfs_server *server;
4824         struct nfs_release_lockowner_args args;
4825 };
4826
4827 static void nfs4_release_lockowner_release(void *calldata)
4828 {
4829         struct nfs_release_lockowner_data *data = calldata;
4830         nfs4_free_lock_state(data->server, data->lsp);
4831         kfree(calldata);
4832 }
4833
4834 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
4835         .rpc_release = nfs4_release_lockowner_release,
4836 };
4837
4838 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
4839 {
4840         struct nfs_server *server = lsp->ls_state->owner->so_server;
4841         struct nfs_release_lockowner_data *data;
4842         struct rpc_message msg = {
4843                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4844         };
4845
4846         if (server->nfs_client->cl_mvops->minor_version != 0)
4847                 return -EINVAL;
4848         data = kmalloc(sizeof(*data), GFP_NOFS);
4849         if (!data)
4850                 return -ENOMEM;
4851         data->lsp = lsp;
4852         data->server = server;
4853         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
4854         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
4855         data->args.lock_owner.s_dev = server->s_dev;
4856         msg.rpc_argp = &data->args;
4857         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
4858         return 0;
4859 }
4860
4861 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4862
4863 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4864                                    const void *buf, size_t buflen,
4865                                    int flags, int type)
4866 {
4867         if (strcmp(key, "") != 0)
4868                 return -EINVAL;
4869
4870         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4871 }
4872
4873 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4874                                    void *buf, size_t buflen, int type)
4875 {
4876         if (strcmp(key, "") != 0)
4877                 return -EINVAL;
4878
4879         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4880 }
4881
4882 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4883                                        size_t list_len, const char *name,
4884                                        size_t name_len, int type)
4885 {
4886         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4887
4888         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4889                 return 0;
4890
4891         if (list && len <= list_len)
4892                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4893         return len;
4894 }
4895
4896 /*
4897  * nfs_fhget will use either the mounted_on_fileid or the fileid
4898  */
4899 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4900 {
4901         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4902                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4903               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4904               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
4905                 return;
4906
4907         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4908                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
4909         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4910         fattr->nlink = 2;
4911 }
4912
4913 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4914                 struct nfs4_fs_locations *fs_locations, struct page *page)
4915 {
4916         struct nfs_server *server = NFS_SERVER(dir);
4917         u32 bitmask[2] = {
4918                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4919         };
4920         struct nfs4_fs_locations_arg args = {
4921                 .dir_fh = NFS_FH(dir),
4922                 .name = name,
4923                 .page = page,
4924                 .bitmask = bitmask,
4925         };
4926         struct nfs4_fs_locations_res res = {
4927                 .fs_locations = fs_locations,
4928         };
4929         struct rpc_message msg = {
4930                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4931                 .rpc_argp = &args,
4932                 .rpc_resp = &res,
4933         };
4934         int status;
4935
4936         dprintk("%s: start\n", __func__);
4937
4938         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4939          * is not supported */
4940         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4941                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4942         else
4943                 bitmask[0] |= FATTR4_WORD0_FILEID;
4944
4945         nfs_fattr_init(&fs_locations->fattr);
4946         fs_locations->server = server;
4947         fs_locations->nlocations = 0;
4948         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4949         dprintk("%s: returned status = %d\n", __func__, status);
4950         return status;
4951 }
4952
4953 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4954 {
4955         int status;
4956         struct nfs4_secinfo_arg args = {
4957                 .dir_fh = NFS_FH(dir),
4958                 .name   = name,
4959         };
4960         struct nfs4_secinfo_res res = {
4961                 .flavors     = flavors,
4962         };
4963         struct rpc_message msg = {
4964                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4965                 .rpc_argp = &args,
4966                 .rpc_resp = &res,
4967         };
4968
4969         dprintk("NFS call  secinfo %s\n", name->name);
4970         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4971         dprintk("NFS reply  secinfo: %d\n", status);
4972         return status;
4973 }
4974
4975 static int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
4976                 struct nfs4_secinfo_flavors *flavors)
4977 {
4978         struct nfs4_exception exception = { };
4979         int err;
4980         do {
4981                 err = nfs4_handle_exception(NFS_SERVER(dir),
4982                                 _nfs4_proc_secinfo(dir, name, flavors),
4983                                 &exception);
4984         } while (exception.retry);
4985         return err;
4986 }
4987
4988 #ifdef CONFIG_NFS_V4_1
4989 /*
4990  * Check the exchange flags returned by the server for invalid flags, having
4991  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4992  * DS flags set.
4993  */
4994 static int nfs4_check_cl_exchange_flags(u32 flags)
4995 {
4996         if (flags & ~EXCHGID4_FLAG_MASK_R)
4997                 goto out_inval;
4998         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4999             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5000                 goto out_inval;
5001         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5002                 goto out_inval;
5003         return NFS_OK;
5004 out_inval:
5005         return -NFS4ERR_INVAL;
5006 }
5007
5008 static bool
5009 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
5010 {
5011         if (a->server_scope_sz == b->server_scope_sz &&
5012             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5013                 return true;
5014
5015         return false;
5016 }
5017
5018 /*
5019  * nfs4_proc_exchange_id()
5020  *
5021  * Since the clientid has expired, all compounds using sessions
5022  * associated with the stale clientid will be returning
5023  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5024  * be in some phase of session reset.
5025  */
5026 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5027 {
5028         nfs4_verifier verifier;
5029         struct nfs41_exchange_id_args args = {
5030                 .verifier = &verifier,
5031                 .client = clp,
5032                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5033         };
5034         struct nfs41_exchange_id_res res = {
5035                 .client = clp,
5036         };
5037         int status;
5038         struct rpc_message msg = {
5039                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5040                 .rpc_argp = &args,
5041                 .rpc_resp = &res,
5042                 .rpc_cred = cred,
5043         };
5044
5045         dprintk("--> %s\n", __func__);
5046         BUG_ON(clp == NULL);
5047
5048         nfs4_construct_boot_verifier(clp, &verifier);
5049
5050         args.id_len = scnprintf(args.id, sizeof(args.id),
5051                                 "%s/%s.%s/%u",
5052                                 clp->cl_ipaddr,
5053                                 init_utsname()->nodename,
5054                                 init_utsname()->domainname,
5055                                 clp->cl_rpcclient->cl_auth->au_flavor);
5056
5057         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
5058         if (unlikely(!res.server_scope)) {
5059                 status = -ENOMEM;
5060                 goto out;
5061         }
5062
5063         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_KERNEL);
5064         if (unlikely(!res.impl_id)) {
5065                 status = -ENOMEM;
5066                 goto out_server_scope;
5067         }
5068
5069         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5070         if (!status)
5071                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
5072
5073         if (!status) {
5074                 /* use the most recent implementation id */
5075                 kfree(clp->impl_id);
5076                 clp->impl_id = res.impl_id;
5077         } else
5078                 kfree(res.impl_id);
5079
5080         if (!status) {
5081                 if (clp->server_scope &&
5082                     !nfs41_same_server_scope(clp->server_scope,
5083                                              res.server_scope)) {
5084                         dprintk("%s: server_scope mismatch detected\n",
5085                                 __func__);
5086                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5087                         kfree(clp->server_scope);
5088                         clp->server_scope = NULL;
5089                 }
5090
5091                 if (!clp->server_scope) {
5092                         clp->server_scope = res.server_scope;
5093                         goto out;
5094                 }
5095         }
5096
5097 out_server_scope:
5098         kfree(res.server_scope);
5099 out:
5100         if (clp->impl_id)
5101                 dprintk("%s: Server Implementation ID: "
5102                         "domain: %s, name: %s, date: %llu,%u\n",
5103                         __func__, clp->impl_id->domain, clp->impl_id->name,
5104                         clp->impl_id->date.seconds,
5105                         clp->impl_id->date.nseconds);
5106         dprintk("<-- %s status= %d\n", __func__, status);
5107         return status;
5108 }
5109
5110 struct nfs4_get_lease_time_data {
5111         struct nfs4_get_lease_time_args *args;
5112         struct nfs4_get_lease_time_res *res;
5113         struct nfs_client *clp;
5114 };
5115
5116 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5117                                         void *calldata)
5118 {
5119         int ret;
5120         struct nfs4_get_lease_time_data *data =
5121                         (struct nfs4_get_lease_time_data *)calldata;
5122
5123         dprintk("--> %s\n", __func__);
5124         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5125         /* just setup sequence, do not trigger session recovery
5126            since we're invoked within one */
5127         ret = nfs41_setup_sequence(data->clp->cl_session,
5128                                    &data->args->la_seq_args,
5129                                    &data->res->lr_seq_res, task);
5130
5131         BUG_ON(ret == -EAGAIN);
5132         rpc_call_start(task);
5133         dprintk("<-- %s\n", __func__);
5134 }
5135
5136 /*
5137  * Called from nfs4_state_manager thread for session setup, so don't recover
5138  * from sequence operation or clientid errors.
5139  */
5140 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5141 {
5142         struct nfs4_get_lease_time_data *data =
5143                         (struct nfs4_get_lease_time_data *)calldata;
5144
5145         dprintk("--> %s\n", __func__);
5146         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5147                 return;
5148         switch (task->tk_status) {
5149         case -NFS4ERR_DELAY:
5150         case -NFS4ERR_GRACE:
5151                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5152                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5153                 task->tk_status = 0;
5154                 /* fall through */
5155         case -NFS4ERR_RETRY_UNCACHED_REP:
5156                 rpc_restart_call_prepare(task);
5157                 return;
5158         }
5159         dprintk("<-- %s\n", __func__);
5160 }
5161
5162 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5163         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5164         .rpc_call_done = nfs4_get_lease_time_done,
5165 };
5166
5167 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5168 {
5169         struct rpc_task *task;
5170         struct nfs4_get_lease_time_args args;
5171         struct nfs4_get_lease_time_res res = {
5172                 .lr_fsinfo = fsinfo,
5173         };
5174         struct nfs4_get_lease_time_data data = {
5175                 .args = &args,
5176                 .res = &res,
5177                 .clp = clp,
5178         };
5179         struct rpc_message msg = {
5180                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5181                 .rpc_argp = &args,
5182                 .rpc_resp = &res,
5183         };
5184         struct rpc_task_setup task_setup = {
5185                 .rpc_client = clp->cl_rpcclient,
5186                 .rpc_message = &msg,
5187                 .callback_ops = &nfs4_get_lease_time_ops,
5188                 .callback_data = &data,
5189                 .flags = RPC_TASK_TIMEOUT,
5190         };
5191         int status;
5192
5193         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5194         dprintk("--> %s\n", __func__);
5195         task = rpc_run_task(&task_setup);
5196
5197         if (IS_ERR(task))
5198                 status = PTR_ERR(task);
5199         else {
5200                 status = task->tk_status;
5201                 rpc_put_task(task);
5202         }
5203         dprintk("<-- %s return %d\n", __func__, status);
5204
5205         return status;
5206 }
5207
5208 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5209 {
5210         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5211 }
5212
5213 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5214                 struct nfs4_slot *new,
5215                 u32 max_slots,
5216                 u32 ivalue)
5217 {
5218         struct nfs4_slot *old = NULL;
5219         u32 i;
5220
5221         spin_lock(&tbl->slot_tbl_lock);
5222         if (new) {
5223                 old = tbl->slots;
5224                 tbl->slots = new;
5225                 tbl->max_slots = max_slots;
5226         }
5227         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5228         for (i = 0; i < tbl->max_slots; i++)
5229                 tbl->slots[i].seq_nr = ivalue;
5230         spin_unlock(&tbl->slot_tbl_lock);
5231         kfree(old);
5232 }
5233
5234 /*
5235  * (re)Initialise a slot table
5236  */
5237 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5238                                  u32 ivalue)
5239 {
5240         struct nfs4_slot *new = NULL;
5241         int ret = -ENOMEM;
5242
5243         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5244                 max_reqs, tbl->max_slots);
5245
5246         /* Does the newly negotiated max_reqs match the existing slot table? */
5247         if (max_reqs != tbl->max_slots) {
5248                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5249                 if (!new)
5250                         goto out;
5251         }
5252         ret = 0;
5253
5254         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5255         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5256                 tbl, tbl->slots, tbl->max_slots);
5257 out:
5258         dprintk("<-- %s: return %d\n", __func__, ret);
5259         return ret;
5260 }
5261
5262 /* Destroy the slot table */
5263 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5264 {
5265         if (session->fc_slot_table.slots != NULL) {
5266                 kfree(session->fc_slot_table.slots);
5267                 session->fc_slot_table.slots = NULL;
5268         }
5269         if (session->bc_slot_table.slots != NULL) {
5270                 kfree(session->bc_slot_table.slots);
5271                 session->bc_slot_table.slots = NULL;
5272         }
5273         return;
5274 }
5275
5276 /*
5277  * Initialize or reset the forechannel and backchannel tables
5278  */
5279 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5280 {
5281         struct nfs4_slot_table *tbl;
5282         int status;
5283
5284         dprintk("--> %s\n", __func__);
5285         /* Fore channel */
5286         tbl = &ses->fc_slot_table;
5287         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5288         if (status) /* -ENOMEM */
5289                 return status;
5290         /* Back channel */
5291         tbl = &ses->bc_slot_table;
5292         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5293         if (status && tbl->slots == NULL)
5294                 /* Fore and back channel share a connection so get
5295                  * both slot tables or neither */
5296                 nfs4_destroy_slot_tables(ses);
5297         return status;
5298 }
5299
5300 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5301 {
5302         struct nfs4_session *session;
5303         struct nfs4_slot_table *tbl;
5304
5305         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5306         if (!session)
5307                 return NULL;
5308
5309         tbl = &session->fc_slot_table;
5310         tbl->highest_used_slotid = NFS4_NO_SLOT;
5311         spin_lock_init(&tbl->slot_tbl_lock);
5312         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5313         init_completion(&tbl->complete);
5314
5315         tbl = &session->bc_slot_table;
5316         tbl->highest_used_slotid = NFS4_NO_SLOT;
5317         spin_lock_init(&tbl->slot_tbl_lock);
5318         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5319         init_completion(&tbl->complete);
5320
5321         session->session_state = 1<<NFS4_SESSION_INITING;
5322
5323         session->clp = clp;
5324         return session;
5325 }
5326
5327 void nfs4_destroy_session(struct nfs4_session *session)
5328 {
5329         struct rpc_xprt *xprt;
5330
5331         nfs4_proc_destroy_session(session);
5332
5333         rcu_read_lock();
5334         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5335         rcu_read_unlock();
5336         dprintk("%s Destroy backchannel for xprt %p\n",
5337                 __func__, xprt);
5338         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5339         nfs4_destroy_slot_tables(session);
5340         kfree(session);
5341 }
5342
5343 /*
5344  * Initialize the values to be used by the client in CREATE_SESSION
5345  * If nfs4_init_session set the fore channel request and response sizes,
5346  * use them.
5347  *
5348  * Set the back channel max_resp_sz_cached to zero to force the client to
5349  * always set csa_cachethis to FALSE because the current implementation
5350  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5351  */
5352 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5353 {
5354         struct nfs4_session *session = args->client->cl_session;
5355         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5356                      mxresp_sz = session->fc_attrs.max_resp_sz;
5357
5358         if (mxrqst_sz == 0)
5359                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5360         if (mxresp_sz == 0)
5361                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5362         /* Fore channel attributes */
5363         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5364         args->fc_attrs.max_resp_sz = mxresp_sz;
5365         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5366         args->fc_attrs.max_reqs = max_session_slots;
5367
5368         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5369                 "max_ops=%u max_reqs=%u\n",
5370                 __func__,
5371                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5372                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5373
5374         /* Back channel attributes */
5375         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5376         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5377         args->bc_attrs.max_resp_sz_cached = 0;
5378         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5379         args->bc_attrs.max_reqs = 1;
5380
5381         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5382                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5383                 __func__,
5384                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5385                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5386                 args->bc_attrs.max_reqs);
5387 }
5388
5389 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5390 {
5391         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5392         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5393
5394         if (rcvd->max_resp_sz > sent->max_resp_sz)
5395                 return -EINVAL;
5396         /*
5397          * Our requested max_ops is the minimum we need; we're not
5398          * prepared to break up compounds into smaller pieces than that.
5399          * So, no point even trying to continue if the server won't
5400          * cooperate:
5401          */
5402         if (rcvd->max_ops < sent->max_ops)
5403                 return -EINVAL;
5404         if (rcvd->max_reqs == 0)
5405                 return -EINVAL;
5406         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5407                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5408         return 0;
5409 }
5410
5411 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5412 {
5413         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5414         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5415
5416         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5417                 return -EINVAL;
5418         if (rcvd->max_resp_sz < sent->max_resp_sz)
5419                 return -EINVAL;
5420         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5421                 return -EINVAL;
5422         /* These would render the backchannel useless: */
5423         if (rcvd->max_ops != sent->max_ops)
5424                 return -EINVAL;
5425         if (rcvd->max_reqs != sent->max_reqs)
5426                 return -EINVAL;
5427         return 0;
5428 }
5429
5430 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5431                                      struct nfs4_session *session)
5432 {
5433         int ret;
5434
5435         ret = nfs4_verify_fore_channel_attrs(args, session);
5436         if (ret)
5437                 return ret;
5438         return nfs4_verify_back_channel_attrs(args, session);
5439 }
5440
5441 static int _nfs4_proc_create_session(struct nfs_client *clp)
5442 {
5443         struct nfs4_session *session = clp->cl_session;
5444         struct nfs41_create_session_args args = {
5445                 .client = clp,
5446                 .cb_program = NFS4_CALLBACK,
5447         };
5448         struct nfs41_create_session_res res = {
5449                 .client = clp,
5450         };
5451         struct rpc_message msg = {
5452                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5453                 .rpc_argp = &args,
5454                 .rpc_resp = &res,
5455         };
5456         int status;
5457
5458         nfs4_init_channel_attrs(&args);
5459         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5460
5461         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5462
5463         if (!status)
5464                 /* Verify the session's negotiated channel_attrs values */
5465                 status = nfs4_verify_channel_attrs(&args, session);
5466         if (!status) {
5467                 /* Increment the clientid slot sequence id */
5468                 clp->cl_seqid++;
5469         }
5470
5471         return status;
5472 }
5473
5474 /*
5475  * Issues a CREATE_SESSION operation to the server.
5476  * It is the responsibility of the caller to verify the session is
5477  * expired before calling this routine.
5478  */
5479 int nfs4_proc_create_session(struct nfs_client *clp)
5480 {
5481         int status;
5482         unsigned *ptr;
5483         struct nfs4_session *session = clp->cl_session;
5484
5485         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5486
5487         status = _nfs4_proc_create_session(clp);
5488         if (status)
5489                 goto out;
5490
5491         /* Init or reset the session slot tables */
5492         status = nfs4_setup_session_slot_tables(session);
5493         dprintk("slot table setup returned %d\n", status);
5494         if (status)
5495                 goto out;
5496
5497         ptr = (unsigned *)&session->sess_id.data[0];
5498         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5499                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5500 out:
5501         dprintk("<-- %s\n", __func__);
5502         return status;
5503 }
5504
5505 /*
5506  * Issue the over-the-wire RPC DESTROY_SESSION.
5507  * The caller must serialize access to this routine.
5508  */
5509 int nfs4_proc_destroy_session(struct nfs4_session *session)
5510 {
5511         int status = 0;
5512         struct rpc_message msg;
5513
5514         dprintk("--> nfs4_proc_destroy_session\n");
5515
5516         /* session is still being setup */
5517         if (session->clp->cl_cons_state != NFS_CS_READY)
5518                 return status;
5519
5520         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5521         msg.rpc_argp = session;
5522         msg.rpc_resp = NULL;
5523         msg.rpc_cred = NULL;
5524         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5525
5526         if (status)
5527                 printk(KERN_WARNING
5528                         "NFS: Got error %d from the server on DESTROY_SESSION. "
5529                         "Session has been destroyed regardless...\n", status);
5530
5531         dprintk("<-- nfs4_proc_destroy_session\n");
5532         return status;
5533 }
5534
5535 int nfs4_init_session(struct nfs_server *server)
5536 {
5537         struct nfs_client *clp = server->nfs_client;
5538         struct nfs4_session *session;
5539         unsigned int rsize, wsize;
5540         int ret;
5541
5542         if (!nfs4_has_session(clp))
5543                 return 0;
5544
5545         session = clp->cl_session;
5546         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5547                 return 0;
5548
5549         rsize = server->rsize;
5550         if (rsize == 0)
5551                 rsize = NFS_MAX_FILE_IO_SIZE;
5552         wsize = server->wsize;
5553         if (wsize == 0)
5554                 wsize = NFS_MAX_FILE_IO_SIZE;
5555
5556         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5557         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5558
5559         ret = nfs4_recover_expired_lease(server);
5560         if (!ret)
5561                 ret = nfs4_check_client_ready(clp);
5562         return ret;
5563 }
5564
5565 int nfs4_init_ds_session(struct nfs_client *clp)
5566 {
5567         struct nfs4_session *session = clp->cl_session;
5568         int ret;
5569
5570         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5571                 return 0;
5572
5573         ret = nfs4_client_recover_expired_lease(clp);
5574         if (!ret)
5575                 /* Test for the DS role */
5576                 if (!is_ds_client(clp))
5577                         ret = -ENODEV;
5578         if (!ret)
5579                 ret = nfs4_check_client_ready(clp);
5580         return ret;
5581
5582 }
5583 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5584
5585
5586 /*
5587  * Renew the cl_session lease.
5588  */
5589 struct nfs4_sequence_data {
5590         struct nfs_client *clp;
5591         struct nfs4_sequence_args args;
5592         struct nfs4_sequence_res res;
5593 };
5594
5595 static void nfs41_sequence_release(void *data)
5596 {
5597         struct nfs4_sequence_data *calldata = data;
5598         struct nfs_client *clp = calldata->clp;
5599
5600         if (atomic_read(&clp->cl_count) > 1)
5601                 nfs4_schedule_state_renewal(clp);
5602         nfs_put_client(clp);
5603         kfree(calldata);
5604 }
5605
5606 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5607 {
5608         switch(task->tk_status) {
5609         case -NFS4ERR_DELAY:
5610                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5611                 return -EAGAIN;
5612         default:
5613                 nfs4_schedule_lease_recovery(clp);
5614         }
5615         return 0;
5616 }
5617
5618 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5619 {
5620         struct nfs4_sequence_data *calldata = data;
5621         struct nfs_client *clp = calldata->clp;
5622
5623         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5624                 return;
5625
5626         if (task->tk_status < 0) {
5627                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5628                 if (atomic_read(&clp->cl_count) == 1)
5629                         goto out;
5630
5631                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5632                         rpc_restart_call_prepare(task);
5633                         return;
5634                 }
5635         }
5636         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5637 out:
5638         dprintk("<-- %s\n", __func__);
5639 }
5640
5641 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5642 {
5643         struct nfs4_sequence_data *calldata = data;
5644         struct nfs_client *clp = calldata->clp;
5645         struct nfs4_sequence_args *args;
5646         struct nfs4_sequence_res *res;
5647
5648         args = task->tk_msg.rpc_argp;
5649         res = task->tk_msg.rpc_resp;
5650
5651         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5652                 return;
5653         rpc_call_start(task);
5654 }
5655
5656 static const struct rpc_call_ops nfs41_sequence_ops = {
5657         .rpc_call_done = nfs41_sequence_call_done,
5658         .rpc_call_prepare = nfs41_sequence_prepare,
5659         .rpc_release = nfs41_sequence_release,
5660 };
5661
5662 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5663 {
5664         struct nfs4_sequence_data *calldata;
5665         struct rpc_message msg = {
5666                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5667                 .rpc_cred = cred,
5668         };
5669         struct rpc_task_setup task_setup_data = {
5670                 .rpc_client = clp->cl_rpcclient,
5671                 .rpc_message = &msg,
5672                 .callback_ops = &nfs41_sequence_ops,
5673                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5674         };
5675
5676         if (!atomic_inc_not_zero(&clp->cl_count))
5677                 return ERR_PTR(-EIO);
5678         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5679         if (calldata == NULL) {
5680                 nfs_put_client(clp);
5681                 return ERR_PTR(-ENOMEM);
5682         }
5683         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5684         msg.rpc_argp = &calldata->args;
5685         msg.rpc_resp = &calldata->res;
5686         calldata->clp = clp;
5687         task_setup_data.callback_data = calldata;
5688
5689         return rpc_run_task(&task_setup_data);
5690 }
5691
5692 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5693 {
5694         struct rpc_task *task;
5695         int ret = 0;
5696
5697         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5698                 return 0;
5699         task = _nfs41_proc_sequence(clp, cred);
5700         if (IS_ERR(task))
5701                 ret = PTR_ERR(task);
5702         else
5703                 rpc_put_task_async(task);
5704         dprintk("<-- %s status=%d\n", __func__, ret);
5705         return ret;
5706 }
5707
5708 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5709 {
5710         struct rpc_task *task;
5711         int ret;
5712
5713         task = _nfs41_proc_sequence(clp, cred);
5714         if (IS_ERR(task)) {
5715                 ret = PTR_ERR(task);
5716                 goto out;
5717         }
5718         ret = rpc_wait_for_completion_task(task);
5719         if (!ret) {
5720                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5721
5722                 if (task->tk_status == 0)
5723                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5724                 ret = task->tk_status;
5725         }
5726         rpc_put_task(task);
5727 out:
5728         dprintk("<-- %s status=%d\n", __func__, ret);
5729         return ret;
5730 }
5731
5732 struct nfs4_reclaim_complete_data {
5733         struct nfs_client *clp;
5734         struct nfs41_reclaim_complete_args arg;
5735         struct nfs41_reclaim_complete_res res;
5736 };
5737
5738 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5739 {
5740         struct nfs4_reclaim_complete_data *calldata = data;
5741
5742         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5743         if (nfs41_setup_sequence(calldata->clp->cl_session,
5744                                 &calldata->arg.seq_args,
5745                                 &calldata->res.seq_res, task))
5746                 return;
5747
5748         rpc_call_start(task);
5749 }
5750
5751 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5752 {
5753         switch(task->tk_status) {
5754         case 0:
5755         case -NFS4ERR_COMPLETE_ALREADY:
5756         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5757                 break;
5758         case -NFS4ERR_DELAY:
5759                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5760                 /* fall through */
5761         case -NFS4ERR_RETRY_UNCACHED_REP:
5762                 return -EAGAIN;
5763         default:
5764                 nfs4_schedule_lease_recovery(clp);
5765         }
5766         return 0;
5767 }
5768
5769 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5770 {
5771         struct nfs4_reclaim_complete_data *calldata = data;
5772         struct nfs_client *clp = calldata->clp;
5773         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5774
5775         dprintk("--> %s\n", __func__);
5776         if (!nfs41_sequence_done(task, res))
5777                 return;
5778
5779         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5780                 rpc_restart_call_prepare(task);
5781                 return;
5782         }
5783         dprintk("<-- %s\n", __func__);
5784 }
5785
5786 static void nfs4_free_reclaim_complete_data(void *data)
5787 {
5788         struct nfs4_reclaim_complete_data *calldata = data;
5789
5790         kfree(calldata);
5791 }
5792
5793 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5794         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5795         .rpc_call_done = nfs4_reclaim_complete_done,
5796         .rpc_release = nfs4_free_reclaim_complete_data,
5797 };
5798
5799 /*
5800  * Issue a global reclaim complete.
5801  */
5802 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5803 {
5804         struct nfs4_reclaim_complete_data *calldata;
5805         struct rpc_task *task;
5806         struct rpc_message msg = {
5807                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5808         };
5809         struct rpc_task_setup task_setup_data = {
5810                 .rpc_client = clp->cl_rpcclient,
5811                 .rpc_message = &msg,
5812                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5813                 .flags = RPC_TASK_ASYNC,
5814         };
5815         int status = -ENOMEM;
5816
5817         dprintk("--> %s\n", __func__);
5818         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5819         if (calldata == NULL)
5820                 goto out;
5821         calldata->clp = clp;
5822         calldata->arg.one_fs = 0;
5823
5824         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
5825         msg.rpc_argp = &calldata->arg;
5826         msg.rpc_resp = &calldata->res;
5827         task_setup_data.callback_data = calldata;
5828         task = rpc_run_task(&task_setup_data);
5829         if (IS_ERR(task)) {
5830                 status = PTR_ERR(task);
5831                 goto out;
5832         }
5833         status = nfs4_wait_for_completion_rpc_task(task);
5834         if (status == 0)
5835                 status = task->tk_status;
5836         rpc_put_task(task);
5837         return 0;
5838 out:
5839         dprintk("<-- %s status=%d\n", __func__, status);
5840         return status;
5841 }
5842
5843 static void
5844 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5845 {
5846         struct nfs4_layoutget *lgp = calldata;
5847         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5848
5849         dprintk("--> %s\n", __func__);
5850         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5851          * right now covering the LAYOUTGET we are about to send.
5852          * However, that is not so catastrophic, and there seems
5853          * to be no way to prevent it completely.
5854          */
5855         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5856                                 &lgp->res.seq_res, task))
5857                 return;
5858         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5859                                           NFS_I(lgp->args.inode)->layout,
5860                                           lgp->args.ctx->state)) {
5861                 rpc_exit(task, NFS4_OK);
5862                 return;
5863         }
5864         rpc_call_start(task);
5865 }
5866
5867 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5868 {
5869         struct nfs4_layoutget *lgp = calldata;
5870         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5871
5872         dprintk("--> %s\n", __func__);
5873
5874         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5875                 return;
5876
5877         switch (task->tk_status) {
5878         case 0:
5879                 break;
5880         case -NFS4ERR_LAYOUTTRYLATER:
5881         case -NFS4ERR_RECALLCONFLICT:
5882                 task->tk_status = -NFS4ERR_DELAY;
5883                 /* Fall through */
5884         default:
5885                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5886                         rpc_restart_call_prepare(task);
5887                         return;
5888                 }
5889         }
5890         dprintk("<-- %s\n", __func__);
5891 }
5892
5893 static void nfs4_layoutget_release(void *calldata)
5894 {
5895         struct nfs4_layoutget *lgp = calldata;
5896
5897         dprintk("--> %s\n", __func__);
5898         put_nfs_open_context(lgp->args.ctx);
5899         kfree(calldata);
5900         dprintk("<-- %s\n", __func__);
5901 }
5902
5903 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5904         .rpc_call_prepare = nfs4_layoutget_prepare,
5905         .rpc_call_done = nfs4_layoutget_done,
5906         .rpc_release = nfs4_layoutget_release,
5907 };
5908
5909 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5910 {
5911         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5912         struct rpc_task *task;
5913         struct rpc_message msg = {
5914                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5915                 .rpc_argp = &lgp->args,
5916                 .rpc_resp = &lgp->res,
5917         };
5918         struct rpc_task_setup task_setup_data = {
5919                 .rpc_client = server->client,
5920                 .rpc_message = &msg,
5921                 .callback_ops = &nfs4_layoutget_call_ops,
5922                 .callback_data = lgp,
5923                 .flags = RPC_TASK_ASYNC,
5924         };
5925         int status = 0;
5926
5927         dprintk("--> %s\n", __func__);
5928
5929         lgp->res.layoutp = &lgp->args.layout;
5930         lgp->res.seq_res.sr_slot = NULL;
5931         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
5932         task = rpc_run_task(&task_setup_data);
5933         if (IS_ERR(task))
5934                 return PTR_ERR(task);
5935         status = nfs4_wait_for_completion_rpc_task(task);
5936         if (status == 0)
5937                 status = task->tk_status;
5938         if (status == 0)
5939                 status = pnfs_layout_process(lgp);
5940         rpc_put_task(task);
5941         dprintk("<-- %s status=%d\n", __func__, status);
5942         return status;
5943 }
5944
5945 static void
5946 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5947 {
5948         struct nfs4_layoutreturn *lrp = calldata;
5949
5950         dprintk("--> %s\n", __func__);
5951         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5952                                 &lrp->res.seq_res, task))
5953                 return;
5954         rpc_call_start(task);
5955 }
5956
5957 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5958 {
5959         struct nfs4_layoutreturn *lrp = calldata;
5960         struct nfs_server *server;
5961         struct pnfs_layout_hdr *lo = lrp->args.layout;
5962
5963         dprintk("--> %s\n", __func__);
5964
5965         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5966                 return;
5967
5968         server = NFS_SERVER(lrp->args.inode);
5969         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5970                 rpc_restart_call_prepare(task);
5971                 return;
5972         }
5973         spin_lock(&lo->plh_inode->i_lock);
5974         if (task->tk_status == 0) {
5975                 if (lrp->res.lrs_present) {
5976                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5977                 } else
5978                         BUG_ON(!list_empty(&lo->plh_segs));
5979         }
5980         lo->plh_block_lgets--;
5981         spin_unlock(&lo->plh_inode->i_lock);
5982         dprintk("<-- %s\n", __func__);
5983 }
5984
5985 static void nfs4_layoutreturn_release(void *calldata)
5986 {
5987         struct nfs4_layoutreturn *lrp = calldata;
5988
5989         dprintk("--> %s\n", __func__);
5990         put_layout_hdr(lrp->args.layout);
5991         kfree(calldata);
5992         dprintk("<-- %s\n", __func__);
5993 }
5994
5995 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
5996         .rpc_call_prepare = nfs4_layoutreturn_prepare,
5997         .rpc_call_done = nfs4_layoutreturn_done,
5998         .rpc_release = nfs4_layoutreturn_release,
5999 };
6000
6001 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6002 {
6003         struct rpc_task *task;
6004         struct rpc_message msg = {
6005                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6006                 .rpc_argp = &lrp->args,
6007                 .rpc_resp = &lrp->res,
6008         };
6009         struct rpc_task_setup task_setup_data = {
6010                 .rpc_client = lrp->clp->cl_rpcclient,
6011                 .rpc_message = &msg,
6012                 .callback_ops = &nfs4_layoutreturn_call_ops,
6013                 .callback_data = lrp,
6014         };
6015         int status;
6016
6017         dprintk("--> %s\n", __func__);
6018         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6019         task = rpc_run_task(&task_setup_data);
6020         if (IS_ERR(task))
6021                 return PTR_ERR(task);
6022         status = task->tk_status;
6023         dprintk("<-- %s status=%d\n", __func__, status);
6024         rpc_put_task(task);
6025         return status;
6026 }
6027
6028 /*
6029  * Retrieve the list of Data Server devices from the MDS.
6030  */
6031 static int _nfs4_getdevicelist(struct nfs_server *server,
6032                                     const struct nfs_fh *fh,
6033                                     struct pnfs_devicelist *devlist)
6034 {
6035         struct nfs4_getdevicelist_args args = {
6036                 .fh = fh,
6037                 .layoutclass = server->pnfs_curr_ld->id,
6038         };
6039         struct nfs4_getdevicelist_res res = {
6040                 .devlist = devlist,
6041         };
6042         struct rpc_message msg = {
6043                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6044                 .rpc_argp = &args,
6045                 .rpc_resp = &res,
6046         };
6047         int status;
6048
6049         dprintk("--> %s\n", __func__);
6050         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6051                                 &res.seq_res, 0);
6052         dprintk("<-- %s status=%d\n", __func__, status);
6053         return status;
6054 }
6055
6056 int nfs4_proc_getdevicelist(struct nfs_server *server,
6057                             const struct nfs_fh *fh,
6058                             struct pnfs_devicelist *devlist)
6059 {
6060         struct nfs4_exception exception = { };
6061         int err;
6062
6063         do {
6064                 err = nfs4_handle_exception(server,
6065                                 _nfs4_getdevicelist(server, fh, devlist),
6066                                 &exception);
6067         } while (exception.retry);
6068
6069         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6070                 err, devlist->num_devs);
6071
6072         return err;
6073 }
6074 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6075
6076 static int
6077 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6078 {
6079         struct nfs4_getdeviceinfo_args args = {
6080                 .pdev = pdev,
6081         };
6082         struct nfs4_getdeviceinfo_res res = {
6083                 .pdev = pdev,
6084         };
6085         struct rpc_message msg = {
6086                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6087                 .rpc_argp = &args,
6088                 .rpc_resp = &res,
6089         };
6090         int status;
6091
6092         dprintk("--> %s\n", __func__);
6093         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6094         dprintk("<-- %s status=%d\n", __func__, status);
6095
6096         return status;
6097 }
6098
6099 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6100 {
6101         struct nfs4_exception exception = { };
6102         int err;
6103
6104         do {
6105                 err = nfs4_handle_exception(server,
6106                                         _nfs4_proc_getdeviceinfo(server, pdev),
6107                                         &exception);
6108         } while (exception.retry);
6109         return err;
6110 }
6111 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6112
6113 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6114 {
6115         struct nfs4_layoutcommit_data *data = calldata;
6116         struct nfs_server *server = NFS_SERVER(data->args.inode);
6117
6118         if (nfs4_setup_sequence(server, &data->args.seq_args,
6119                                 &data->res.seq_res, task))
6120                 return;
6121         rpc_call_start(task);
6122 }
6123
6124 static void
6125 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6126 {
6127         struct nfs4_layoutcommit_data *data = calldata;
6128         struct nfs_server *server = NFS_SERVER(data->args.inode);
6129
6130         if (!nfs4_sequence_done(task, &data->res.seq_res))
6131                 return;
6132
6133         switch (task->tk_status) { /* Just ignore these failures */
6134         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6135         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6136         case -NFS4ERR_BADLAYOUT:     /* no layout */
6137         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6138                 task->tk_status = 0;
6139                 break;
6140         case 0:
6141                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6142                                                    data->res.fattr);
6143                 break;
6144         default:
6145                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6146                         rpc_restart_call_prepare(task);
6147                         return;
6148                 }
6149         }
6150 }
6151
6152 static void nfs4_layoutcommit_release(void *calldata)
6153 {
6154         struct nfs4_layoutcommit_data *data = calldata;
6155         struct pnfs_layout_segment *lseg, *tmp;
6156         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6157
6158         pnfs_cleanup_layoutcommit(data);
6159         /* Matched by references in pnfs_set_layoutcommit */
6160         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6161                 list_del_init(&lseg->pls_lc_list);
6162                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6163                                        &lseg->pls_flags))
6164                         put_lseg(lseg);
6165         }
6166
6167         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6168         smp_mb__after_clear_bit();
6169         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6170
6171         put_rpccred(data->cred);
6172         kfree(data);
6173 }
6174
6175 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6176         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6177         .rpc_call_done = nfs4_layoutcommit_done,
6178         .rpc_release = nfs4_layoutcommit_release,
6179 };
6180
6181 int
6182 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6183 {
6184         struct rpc_message msg = {
6185                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6186                 .rpc_argp = &data->args,
6187                 .rpc_resp = &data->res,
6188                 .rpc_cred = data->cred,
6189         };
6190         struct rpc_task_setup task_setup_data = {
6191                 .task = &data->task,
6192                 .rpc_client = NFS_CLIENT(data->args.inode),
6193                 .rpc_message = &msg,
6194                 .callback_ops = &nfs4_layoutcommit_ops,
6195                 .callback_data = data,
6196                 .flags = RPC_TASK_ASYNC,
6197         };
6198         struct rpc_task *task;
6199         int status = 0;
6200
6201         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6202                 "lbw: %llu inode %lu\n",
6203                 data->task.tk_pid, sync,
6204                 data->args.lastbytewritten,
6205                 data->args.inode->i_ino);
6206
6207         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6208         task = rpc_run_task(&task_setup_data);
6209         if (IS_ERR(task))
6210                 return PTR_ERR(task);
6211         if (sync == false)
6212                 goto out;
6213         status = nfs4_wait_for_completion_rpc_task(task);
6214         if (status != 0)
6215                 goto out;
6216         status = task->tk_status;
6217 out:
6218         dprintk("%s: status %d\n", __func__, status);
6219         rpc_put_task(task);
6220         return status;
6221 }
6222
6223 static int
6224 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6225                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6226 {
6227         struct nfs41_secinfo_no_name_args args = {
6228                 .style = SECINFO_STYLE_CURRENT_FH,
6229         };
6230         struct nfs4_secinfo_res res = {
6231                 .flavors = flavors,
6232         };
6233         struct rpc_message msg = {
6234                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6235                 .rpc_argp = &args,
6236                 .rpc_resp = &res,
6237         };
6238         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6239 }
6240
6241 static int
6242 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6243                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6244 {
6245         struct nfs4_exception exception = { };
6246         int err;
6247         do {
6248                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6249                 switch (err) {
6250                 case 0:
6251                 case -NFS4ERR_WRONGSEC:
6252                 case -NFS4ERR_NOTSUPP:
6253                         goto out;
6254                 default:
6255                         err = nfs4_handle_exception(server, err, &exception);
6256                 }
6257         } while (exception.retry);
6258 out:
6259         return err;
6260 }
6261
6262 static int
6263 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6264                     struct nfs_fsinfo *info)
6265 {
6266         int err;
6267         struct page *page;
6268         rpc_authflavor_t flavor;
6269         struct nfs4_secinfo_flavors *flavors;
6270
6271         page = alloc_page(GFP_KERNEL);
6272         if (!page) {
6273                 err = -ENOMEM;
6274                 goto out;
6275         }
6276
6277         flavors = page_address(page);
6278         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6279
6280         /*
6281          * Fall back on "guess and check" method if
6282          * the server doesn't support SECINFO_NO_NAME
6283          */
6284         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6285                 err = nfs4_find_root_sec(server, fhandle, info);
6286                 goto out_freepage;
6287         }
6288         if (err)
6289                 goto out_freepage;
6290
6291         flavor = nfs_find_best_sec(flavors);
6292         if (err == 0)
6293                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6294
6295 out_freepage:
6296         put_page(page);
6297         if (err == -EACCES)
6298                 return -EPERM;
6299 out:
6300         return err;
6301 }
6302
6303 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6304 {
6305         int status;
6306         struct nfs41_test_stateid_args args = {
6307                 .stateid = stateid,
6308         };
6309         struct nfs41_test_stateid_res res;
6310         struct rpc_message msg = {
6311                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6312                 .rpc_argp = &args,
6313                 .rpc_resp = &res,
6314         };
6315
6316         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6317         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6318
6319         if (status == NFS_OK)
6320                 return res.status;
6321         return status;
6322 }
6323
6324 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6325 {
6326         struct nfs4_exception exception = { };
6327         int err;
6328         do {
6329                 err = nfs4_handle_exception(server,
6330                                 _nfs41_test_stateid(server, stateid),
6331                                 &exception);
6332         } while (exception.retry);
6333         return err;
6334 }
6335
6336 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6337 {
6338         struct nfs41_free_stateid_args args = {
6339                 .stateid = stateid,
6340         };
6341         struct nfs41_free_stateid_res res;
6342         struct rpc_message msg = {
6343                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6344                 .rpc_argp = &args,
6345                 .rpc_resp = &res,
6346         };
6347
6348         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6349         return nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6350 }
6351
6352 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6353 {
6354         struct nfs4_exception exception = { };
6355         int err;
6356         do {
6357                 err = nfs4_handle_exception(server,
6358                                 _nfs4_free_stateid(server, stateid),
6359                                 &exception);
6360         } while (exception.retry);
6361         return err;
6362 }
6363
6364 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6365                 const nfs4_stateid *s2)
6366 {
6367         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6368                 return false;
6369
6370         if (s1->seqid == s2->seqid)
6371                 return true;
6372         if (s1->seqid == 0 || s2->seqid == 0)
6373                 return true;
6374
6375         return false;
6376 }
6377
6378 #endif /* CONFIG_NFS_V4_1 */
6379
6380 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6381                 const nfs4_stateid *s2)
6382 {
6383         return nfs4_stateid_match(s1, s2);
6384 }
6385
6386
6387 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6388         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6389         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6390         .recover_open   = nfs4_open_reclaim,
6391         .recover_lock   = nfs4_lock_reclaim,
6392         .establish_clid = nfs4_init_clientid,
6393         .get_clid_cred  = nfs4_get_setclientid_cred,
6394 };
6395
6396 #if defined(CONFIG_NFS_V4_1)
6397 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6398         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6399         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6400         .recover_open   = nfs4_open_reclaim,
6401         .recover_lock   = nfs4_lock_reclaim,
6402         .establish_clid = nfs41_init_clientid,
6403         .get_clid_cred  = nfs4_get_exchange_id_cred,
6404         .reclaim_complete = nfs41_proc_reclaim_complete,
6405 };
6406 #endif /* CONFIG_NFS_V4_1 */
6407
6408 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6409         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6410         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6411         .recover_open   = nfs4_open_expired,
6412         .recover_lock   = nfs4_lock_expired,
6413         .establish_clid = nfs4_init_clientid,
6414         .get_clid_cred  = nfs4_get_setclientid_cred,
6415 };
6416
6417 #if defined(CONFIG_NFS_V4_1)
6418 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6419         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6420         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6421         .recover_open   = nfs41_open_expired,
6422         .recover_lock   = nfs41_lock_expired,
6423         .establish_clid = nfs41_init_clientid,
6424         .get_clid_cred  = nfs4_get_exchange_id_cred,
6425 };
6426 #endif /* CONFIG_NFS_V4_1 */
6427
6428 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6429         .sched_state_renewal = nfs4_proc_async_renew,
6430         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6431         .renew_lease = nfs4_proc_renew,
6432 };
6433
6434 #if defined(CONFIG_NFS_V4_1)
6435 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6436         .sched_state_renewal = nfs41_proc_async_sequence,
6437         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6438         .renew_lease = nfs4_proc_sequence,
6439 };
6440 #endif
6441
6442 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6443         .minor_version = 0,
6444         .call_sync = _nfs4_call_sync,
6445         .match_stateid = nfs4_match_stateid,
6446         .find_root_sec = nfs4_find_root_sec,
6447         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6448         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6449         .state_renewal_ops = &nfs40_state_renewal_ops,
6450 };
6451
6452 #if defined(CONFIG_NFS_V4_1)
6453 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6454         .minor_version = 1,
6455         .call_sync = _nfs4_call_sync_session,
6456         .match_stateid = nfs41_match_stateid,
6457         .find_root_sec = nfs41_find_root_sec,
6458         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6459         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6460         .state_renewal_ops = &nfs41_state_renewal_ops,
6461 };
6462 #endif
6463
6464 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6465         [0] = &nfs_v4_0_minor_ops,
6466 #if defined(CONFIG_NFS_V4_1)
6467         [1] = &nfs_v4_1_minor_ops,
6468 #endif
6469 };
6470
6471 static const struct inode_operations nfs4_file_inode_operations = {
6472         .permission     = nfs_permission,
6473         .getattr        = nfs_getattr,
6474         .setattr        = nfs_setattr,
6475         .getxattr       = generic_getxattr,
6476         .setxattr       = generic_setxattr,
6477         .listxattr      = generic_listxattr,
6478         .removexattr    = generic_removexattr,
6479 };
6480
6481 const struct nfs_rpc_ops nfs_v4_clientops = {
6482         .version        = 4,                    /* protocol version */
6483         .dentry_ops     = &nfs4_dentry_operations,
6484         .dir_inode_ops  = &nfs4_dir_inode_operations,
6485         .file_inode_ops = &nfs4_file_inode_operations,
6486         .file_ops       = &nfs4_file_operations,
6487         .getroot        = nfs4_proc_get_root,
6488         .getattr        = nfs4_proc_getattr,
6489         .setattr        = nfs4_proc_setattr,
6490         .lookup         = nfs4_proc_lookup,
6491         .access         = nfs4_proc_access,
6492         .readlink       = nfs4_proc_readlink,
6493         .create         = nfs4_proc_create,
6494         .remove         = nfs4_proc_remove,
6495         .unlink_setup   = nfs4_proc_unlink_setup,
6496         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
6497         .unlink_done    = nfs4_proc_unlink_done,
6498         .rename         = nfs4_proc_rename,
6499         .rename_setup   = nfs4_proc_rename_setup,
6500         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
6501         .rename_done    = nfs4_proc_rename_done,
6502         .link           = nfs4_proc_link,
6503         .symlink        = nfs4_proc_symlink,
6504         .mkdir          = nfs4_proc_mkdir,
6505         .rmdir          = nfs4_proc_remove,
6506         .readdir        = nfs4_proc_readdir,
6507         .mknod          = nfs4_proc_mknod,
6508         .statfs         = nfs4_proc_statfs,
6509         .fsinfo         = nfs4_proc_fsinfo,
6510         .pathconf       = nfs4_proc_pathconf,
6511         .set_capabilities = nfs4_server_capabilities,
6512         .decode_dirent  = nfs4_decode_dirent,
6513         .read_setup     = nfs4_proc_read_setup,
6514         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
6515         .read_done      = nfs4_read_done,
6516         .write_setup    = nfs4_proc_write_setup,
6517         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
6518         .write_done     = nfs4_write_done,
6519         .commit_setup   = nfs4_proc_commit_setup,
6520         .commit_done    = nfs4_commit_done,
6521         .lock           = nfs4_proc_lock,
6522         .clear_acl_cache = nfs4_zap_acl_attr,
6523         .close_context  = nfs4_close_context,
6524         .open_context   = nfs4_atomic_open,
6525         .init_client    = nfs4_init_client,
6526         .secinfo        = nfs4_proc_secinfo,
6527 };
6528
6529 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6530         .prefix = XATTR_NAME_NFSV4_ACL,
6531         .list   = nfs4_xattr_list_nfs4_acl,
6532         .get    = nfs4_xattr_get_nfs4_acl,
6533         .set    = nfs4_xattr_set_nfs4_acl,
6534 };
6535
6536 const struct xattr_handler *nfs4_xattr_handlers[] = {
6537         &nfs4_xattr_nfs4_acl_handler,
6538         NULL
6539 };
6540
6541 module_param(max_session_slots, ushort, 0644);
6542 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
6543                 "requests the client will negotiate");
6544
6545 /*
6546  * Local variables:
6547  *  c-basic-offset: 8
6548  * End:
6549  */